77640 Measuring Household Usage of Financial Services: Does it Matter How or Whom You Ask? Robert Cull and Kinnon Scott In recent years, the number of surveys on access to and use of �nancial services has multiplied, but little is known about whether the data generated are comparable Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 across countries or within the same country over time. A randomized experiment in Ghana tested whether the identity of the respondent and the inclusion of product- speci�c cues in questions affect reported rates of use of �nancial services. Rates of household use are almost identical whether the head reports on behalf of the house- hold or whether the rate is tabulated from a full enumeration of household members. A less complete summary of household use of �nancial services results when randomly selected informants (nonheads of household) provide the information. For credit from formal institutions, informal sources of savings, and insurance, reported use is higher when questions are asked about speci�c �nancial products rather than about the respondent’s dealings with types of �nancial institutions. In short, who is asked the questions and how the questions are asked both matter. By now, the link between �nancial sector depth and economic growth is well established.1 Most studies rely on aggregate measures of deposits and loans in the formal �nancial system, predominantly through banks.2 Because aggregate measures, such as the ratio of credit extended to the private sector to GDP, do Robert Cull (corresponding author, rcull@worldbank.org is a lead economist in the Development Economics Development Research Group at the World Bank. Kinnon Scott (kscott1@worldbank.org is a senior economist in the Development Economics Development Research Group at the World Bank. The authors thank the Knowledge for Change program for providing funding for this project and Mircea Tranda�r for excellent research assistance. They thank Asli Demirgu ¨c¸ -Kunt, Pete Lanjouw, Jonathan Morduch, and Colin Xu for insightful comments. They also owe a large debt of gratitude to their collaborators on the survey team at the Ghana Statistical Service, without whom this work could not have been done. Finally, they acknowledge that this project grew out of discussions between World Bank and FinMark staff about harmonizing survey approaches, and thus the authors thank Darrell Begin, Norman Bradburn, Anne Marie Chidzero, Bob Curran, Karen Ellis, Anjali Kumar, Mark Napier, Adam Parsons, and Lorraine Ronchi for setting out the issues. 1. See Beck, Levine, and Loayza 2000; Levine 2005; Levine, Loayza, and Beck 2000; Levine and Zervos 1998; and Rajan and Zingales 1998. 2. See Beck, Demirgu ¨c¸ -Kunt, and Levine (2000) for an overview of measures of �nancial sector depth and their construction. THE WORLD BANK ECONOMIC REVIEW, VOL. 24, NO. 2, pp. 199 –233 doi:10.1093/wber/lhq004 Advance Access Publication April 14, 2010 # The Author 2010. Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / THE WORLD BANK. All rights reserved. For permissions, please e-mail: journals.permissions@oxfordjournals.org 199 200 THE WORLD BANK ECONOMIC REVIEW not provide information about the average size of a loan (or deposit), they give an imperfect sense of the reach of the �nancial sector. A highly concentrated banking sector, in which a small number of relatively wealthy depositors and borrowers are responsible for a large share of banking activity, could score comparatively well on �nancial depth while having limited breadth of outreach. There are reasons to be concerned about breadth of outreach, especially in developing countries. Informational asymmetries, transaction costs, and con- tract enforcement costs lead to market imperfections that disproportionately Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 disadvantage the poor, who tend to lack collateral, credit histories, and connec- tions (Levine 2005; World Bank 2007). And recent studies have established a link between �nancial sector development and poverty alleviation (Beck, Demirgu ¸ -Kunt, and Levine, 2007; Clarke, Xu, and Zou, 2006; Honohan ¨c 2004). Perhaps the major reason why �nancial sector breadth has been understu- died is the dif�culty collecting data.3 Whereas measures of �nancial depth can be derived from the balance sheets of �nancial institutions that already furnish this information to supervisors such as central banks, the same information is not readily available for �nancial sector breadth and certainly not in a consist- ent format across countries. Recent attempts to collect data on �nancial sector breadth have pushed beyond balance sheet information, using both demand- and supply-side approaches. On the supply side, measures of �nancial sector outreach often focus on the number of accounts of providers of �nancial services. For example, Beck, Demirgu ¸ -Kunt, and Martinez Peria (2007) collected information on the aggre- ¨c gate number of deposit and loan accounts from bank regulators in 99 countries. They also collected information on the number of bank branches and automated teller machines (ATMs) in each country as a proxy for physical access to �nan- cial services, even among those who do not use them. A limitation of those data is that they are derived only from information about banks, which, while impor- tant or even dominant providers of �nancial services in many countries, are not the full story. Honohan (2008) combines the commercial bank accounts from Beck, Demirgu ¸ -Kunt, and Martinez Peria with accounts at micro�nance insti- ¨c tutions (from Christen, Jayadeva, and Rosenberg 2004) and at savings banks that are members of World Savings Bank Institute (from Peachey and Roe 2006) to produce the most comprehensive, though admittedly still rough, accounts- based estimates of access to date. While this represents a step forward, the accounts-based approach provides little information about the account holders and thus about �nancial exclusion in a given country. A more satisfying, but costlier, approach is to interview users and potential users of �nancial services through surveys of individuals and households. Broadly speaking, there are two approaches: stand-alone surveys on access to 3. See World Bank (2007) for a discussion. Robert Cull and Kinnon Scott 201 �nancial services, which tend to be relatively expensive but produce rich data sets and a detailed portrait of access, and a small module of questions on �nan- cial access and use that is embedded within a larger survey designed to cover another topic (such as surveys of household expenditures or labor market par- ticipation) or multiple topics (as in the Living Standards Measurement Study (LSMS) surveys). The marginal cost of the modules is much lower than that of stand-alone surveys, but they yield data that are much less rich. Neither approach has produced comparable �nancial use data at regular intervals that could be used to monitor the situation in a given country over Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 time or to compare outreach across many countries. Because the stand-alone surveys are costly, they tend not to be repeated at regular intervals, and when they are eventually repeated, the sampling frame and questions may differ or a different organization may conduct the survey. In surveys designed for a differ- ent purpose, modules of �nancial questions tend not to be given high priority, and comparability of data across surveys occurs largely by chance. A recent summary of the �nancial information generated in the LSMS shows that only a handful of basic questions about accounts and loans are asked in most modules, and those are often asked in different ways, making the validity of comparisons across surveys dubious (Gasparini and others 2004). While the accounts-based and survey-based measures of use of �nancial ser- vices are not substitutes, recent research has found a robust statistical link between them (Beck, Demirgu ¸ -Kunt, and Martinez Peria 2007; Honohan ¨c 2008). Thus, a regression model constructed from the more readily available accounts-based information can be used to generate reasonably accurate esti- mates of the harder to collect survey-based data. Still, the �t of these regressions is not perfect. For example, Honohan (2008) estimates that 16 percent of Ghanaians have an account, whereas the information derived from the survey described below places that �gure at 25 percent. At best, it would appear that the estimates derived from accounts-based information could be used to monitor access between surveys of users. Scaling up data collection on use of �nancial services to ensure accuracy and comparability across countries and over time would therefore require a survey- based approach. While there have been other stand-alone efforts to measure use, the most advanced current one is that by the FinMark Trust, which has deployed its FinScope survey (www.�nscope.co.za) in several developing countries, primarily in Africa.4 FinScope surveys are designed to provide nationally representative information on individuals’ use of �nancial services. The questions are similar to those that might be found in a marketing study, including detailed inquiries about speci�c types of �nancial products. These questions are supplemented by others on respondents’ attitudes toward �nan- cial institutions, risk, and coping strategies in times of economic hardship, among other issues. 4. The FinScope website lists ongoing or completed surveys for 14 African countries and Pakistan. 202 THE WORLD BANK ECONOMIC REVIEW By contrast, the most comprehensive effort to use the modular approach to measure use, the LSMS, tends to ask broad, generic questions about “credit� or “accounts� or dealings with types of �nancial institutions. Another impor- tant difference between the FinScope and LSMS approaches is that the LSMS �nance modules track household use of �nancial services, whereas FinScope randomly selects individuals from the population to provide information only on their own use. In light of these differences, a randomized experiment was devised to test whether measured use of �nancial services is similar when respondents are Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 asked detailed product-based questions (the FinScope approach) or are asked more generic, institution-based questions (the LSMS approach). The two approaches are found to yield similar estimates for basic products such as savings accounts with banks or other formal providers but not for others such as insurance or credit provided by banks and other institutions. These comparisons are potentially important because the expense of stand-alone surveys makes it unlikely that they will be rolled out throughout developing countries any time soon. The results in this article provide guidance on the product- and institution-based questions that yield similar estimates of use, and they suggest ways that generic, institution-based questions used in �nance modules could be modi�ed to produce similar estimates of use for pro- ducts such as insurance and formal credit. For household use of �nancial services, an important consideration is whether the identity of the survey respondent affects the accuracy of the infor- mation received. The most comprehensive approach to measuring household use is a full enumeration: each member of the household reports on personal use of �nancial services, and individual responses are then aggregated to the household level. Other approaches use an informant to provide information on the use of �nancial services by all members of the household, typically either the head of household or a randomly selected adult. Another part of the exper- iment, therefore, tests whether the household �nancial use information pro- vided by the household head or a randomly selected informant is as accurate as that provided by a full enumeration. Because a full enumeration is more time consuming, these results can indicate the services for which informants can provide reliable, cost-effective information. Section I describes the experimental design, and section II compares the characteristics of the sample with that of the full Ghana LSMS—only a subset of households were re-visited, though the sample was designed to be nationally representative. Sample characteristics are also compared across treatment groups. Section III reports rates of use across �nancial products for product- and institution-based questions and household use rates provided through full enumeration and through a randomly selected informant. Section IV introduces regressions to test whether certain types of individuals and households are responsible for the under-reporting of access found for some questionnaire formats. Section V offers concluding remarks. Robert Cull and Kinnon Scott 203 I. THE DESIGN OF THE EXPERIMENT Household surveys vary across multiple dimensions as tradeoffs are made among respondents, data quality, and cost. Choice of Respondents For �nancial (and other) surveys, an important dimension is the choice of respondents. Heads of household, however de�ned, are often selected because they are considered knowledgeable, other elements of the survey require their Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 personal information, and selecting one person to provide data at the house- hold level saves time and resources. LSMS surveys had traditionally collected information this way. However, a review of the surveys suggested that the head of household may not be aware of all the services used by household members and that relying on this one informant could lead to underestimation of some types of service use and of overall household use (Kochar 2000; Scott 2000). More recent LSMS surveys have moved to direct informants (full enumeration) for �nancial information. A third option is to randomly select one adult per household, the sampling technique used in the FinScope surveys, whose goal is to achieve a probability selection of adults in the country. Logistically, it would be easiest to have this informant provide the household-level data if such data were desired. The ques- tion is whether this strategy would provide data of similar quality to that from full enumeration of adults or from the head of household. It is not clear, a priori, that every individual in the household will be equally well informed about other members’ �nancial sector involvement. Aggregation of Questions A second key dimension on which surveys vary is the level of aggregation of questions. A short set of highly aggregated questions can reduce costs, simplify �eldwork, and lessen the burden on respondents. The LSMS surveys have used this method, asking about �nancial service use at an aggregate level with a greater focus on relationships with types of �nancial service providers than on products used. There is a concern that some services might be missed using this approach, however. Research in other areas has shown that such aggregation may lead to accidental omissions or memory lapses, thus lowering reported incidence or use. Experiments in measuring household consumption have cer- tainly shown this to be the case (Joliffe 2001; Pradhan 2001; Steele 1998; STATIN 1994). The opposite approach is to ask respondents about each �nancial service or product available. This approach, taken in the FinScope surveys, should prevent accidental omission of service use. It does, however, increase the burden of the interview, which can lead to lower data quality. It also may pre- clude multitopic surveys from addressing �nancial service use as there simply is not space or time for so many questions. 204 THE WORLD BANK ECONOMIC REVIEW The Ghana Experiment The experiment carried out in Ghana explicitly tests the effect of changing the respondent and changing the set of questions on �nancial service use. The Ghana Statistical Service (GSS) collaborated with the authors in developing and administering a �nancial services survey to a subsample of households in the Ghana Living Standards Study (GLSS5) survey. To augment GSS’s own survey experience, the GSS called on experts and other sources in the country to compile a comprehensive list of �nancial services and service providers. The GSS staff also determined the best terminology and strategies for minimizing Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 translation problems, prepared training materials, and trained interviewers. The �nancial services survey, by revisiting GLSS5 households, was able to take advantage of the rich data already collected from those households. This released many constraints on the experiment survey and allowed a more complex design: interviewers needed only to be trained on �nancial questions and data collection. And more risks could be taken because the government’s national survey was in no way at risk from the work (since households were visited after the GLSS5 was �nished). The original framework for the experiment was a three by two matrix— three types of respondents (head, randomly selected adult, and full enumer- ation) and two types of questionnaires ( product-based and institution-based). This framework was determined to be too complex for ensuring quality in the �eldwork. A simpli�ed, feasible design was drawn up that still allowed com- parison on the two main issues of interest: the quality of household use infor- mation provided by informants compared with a full enumeration and the quality of data obtained using a product-based questionnaire compared with an institution-based one. Physically, three different questionnaires were �elded, with the second and third questionnaires containing more than one treatment (table 1). Households were randomly preassigned to one of three groups with each group being admi- nistered a different questionnaire. In households where one of the treatments was for a randomly selected adult to be interviewed, interviewers used Kish tables to make that selection in the �eld.5 Only individuals ages 15 or older were surveyed. The same information was not obtained across all households. Collecting �nancial services use information for individuals and households differs. An individual respondent can provide information about other household members only insofar as the respondent knows about their �nancial activities. 5. Use of a Kish table enables interviewers with a sample of household addresses (in this case the 15 houses in each enumeration area) to randomly sample individuals on the doorstep by following a simple rule for selecting one household resident to interview. A list of eligible individuals at a particular address is ordered by age, and then one person is selected according to the serial number of the address. All individuals in a household have an equal chance of selection, resulting in a representative sample of all individuals in a population. Robert Cull and Kinnon Scott 205 Knowledge of one’s own use is more a matter of straight reporting and thus offered a cleaner test of whether including product cues in questions yields higher use rates than using institutional questions alone. Overlaying the product/institutional treatments on the household use experiment would have made it more dif�cult to distinguish the effects of question format from those of the quality of the respondent’s knowledge. Once the decision was made to separate the household use and individual use experiments, issues arose about the optimal sequencing of treatments within the same visit to a household. One issue was repetition. In general, it Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 was preferable not to have the same individual respond �rst to institutional and then to product-based questions about their personal use of �nancial ser- vices, for two reasons. Respondents might grow impatient at the repetition, and their answers about �nancial use under one question format might influ- ence their subsequent responses under the other format.6 The time spent inter- viewing a household also affected the design of the experiment. For households in group 1, in which all members answered the longer, product-based ques- tions, visit length was a concern. Also, multiple visits to the same household were often necessary to collect information from all members. For those reasons, institution-based questions were not added to the group 1 question- naire. This also would yield a group of responses for which the sequencing concerns described above would not be relevant. The GLSS5 households included in the �nancial services survey were taken from the last two GLSS5 interviewing cycles, which were closest in time to the �elding of the experiment. This was done to minimize the chances that a household might have changed signi�cantly between surveys, since the study relies on the GLSS5 data for non�nancial information on households and indi- viduals. The selected enumeration areas (and households) were distributed throughout the country (table 2). The instruments were piloted and revised, and the survey took place over October–December 2006. 6. The only time such repetition occurred was for households in group 2 (see table 1), in which all household members were �rst asked about their own use of �nancial services, using the institution-based questions. Then a member of the household was randomly selected to answer the more detailed product-based questions. Members were told that because the product-based questions were more time consuming, it made sense to have only one person answer them. While this seemed to be a natural transition, concerns remained that answering the institutional questions �rst might influence the selected member’s product-based responses. This could be checked by comparing responses with those from group 1, which asked all household members only product-based questions. The product/ institutional comparisons are very similar whether the group 2 product-based responses from the randomly selected household members are included or not. Thus, the results for product-based use questionnaires are reported for both groups 1 and 2. The product-based information from group 3 was excluded, however, in constructing the tests because the sampling procedure within households was not random: �rst the household head was interviewed, and then a randomly selected nonhead was interviewed. Again, however, inclusion of the group 3 observations does not greatly affect the comparisons between the product and institutional question formats. 206 THE WORLD BANK ECONOMIC REVIEW T A B L E 1 . Questionnaire Treatments Questionnaire administered Respondent Product Institutional Head of household Group 3 (n ¼ 659 households) Answers institution-based questions about household use in �rst section; answers product-based questions about individual use in second section. Randomly selected adult Group 2 Group 3 Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 (n ¼ 643 households) (n ¼ 659 households) Answers product-based Answers institution-based questions about individual questions about household use. use in �rst section; answers product-based questions about individual use in second section. All adults (15 and older) Group 1 Group 2 (n ¼ 653 households, 1,570 (n ¼ 643 households, 1,568 individuals) individuals) All household members answer All household members answer product-based questions institution-based questions about their own use. about their own use. Note: Each group represents a different questionnaire. T A B L E 2 . Enumeration Areas from the Ghana Living Standards Study 5 Used for Financial Service Survey Sample Ghana Living Standards Study 5, Ghana Living Standards Study 5, Cycle10 Cycle 11 Urban Rural Urban Rural Region enumeration area enumeration area enumeration area enumeration area Total Northern 1 5 0 6 12 Upper East 1 2 0 3 6 Upper West 0 3 0 3 6 Ashanti 3 6 3 6 18 Eastern 2 4 2 4 12 Brong Ahafo 2 4 0 6 12 Volta 0 6 0 6 12 Western 4 2 3 3 12 Central 6 0 4 2 12 Greater Accra 6 3 8 1 18 Total 25 35 20 40 120 Source: Table prepared by Ghana Statistical Service based on GSS (2006). Robert Cull and Kinnon Scott 207 II. DESCRIPTION OF THE D ATA The �nal sample contained 1,955 households. Efforts were made to ensure that the experiment’s results could be extrapolated to the entire population and did not apply only to the households included in the survey. The sample of house- holds for the �nancial services survey was a random subsample of the GLSS5 sample, itself a probability sample (GSS 2006). This simple random sample of enumeration areas, however, failed to take into account the larger population in urban enumeration areas that was captured in the original probability pro- Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 portional to size sample. This shows up in variables related to location: the �nancial services survey sample deviates slightly from the GLSS sample in being more rural, having more households engaged in agriculture, and being more likely to be located in the Coastal and Forest Zones of the country (table 3). The effects are not strong, but it should be remembered that the sample somewhat overrepresents the rural population. Sample attrition was a second potential area of bias. The �nancial survey is essentially a panel survey. Of the 2,291 households revisited, 336 could not be reinterviewed.7 This is a problem only if there are systematic differences between the households that could and those that could not be reinterviewed. A probit model in which the dependent variable takes a value of one if the household was not reinterviewed and zero otherwise found that rural house- holds and households with older heads were less likely to be lost between rounds. The sample may therefore underrepresent more mobile households, and the attrition reinforces the slight bias toward rural households arising from the original selection of enumeration areas. These tendencies also need to be kept in mind when drawing conclusions from the data. Finally, the allocation of households to questionnaire groups could be a concern. Comparing the means of key variables across the three groups is reas- suring on this point (table 4). The only problem area might be that households in group 2 are slightly smaller than households in the other two groups, on the order of 0.25 fewer people per household. While the difference is statistically signi�cant, it is small and unlikely to have any effect on the results. III. BASIC COMPARISONS ACROSS TR EATM E NTS For the institution-based questions asked of a household informant (either the head of household or a randomly selected adult that is not the head), seven indicators of the use of �nancial services were calculated, listed here with the survey questions from which they are derived: 7. Fifty percent of the nonresponses were due to vacant dwellings (either permanent or temporary) and 40 percent to households that had moved. Refusals to participate represented less than 3 percent of nonresponse. 208 THE WORLD BANK ECONOMIC REVIEW (1) Banked: Some people like to keep their money in an account with a bank. Do you or any member of your household have a bank account?8 (2) Indirect access to an account: Do you or other members of your house- hold perform banking transactions using someone else’s account?9 (3) Formal nonbank savings: Now think of all the ways that you and members of your household save money. We are not talking about investing in a business or buying land, but only about where you or other household members put their money to use later. Have you or anyone in your household used an institution such as a credit union or a Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 savings association to save money in the past 12 months? (4) Formal credit: Many people borrow money to buy things on credit. Have you or any other member of your household used an institution such as a credit union, savings association, or bank to borrow money or to buy on credit in the past 12 months? (5) Informal savings: Have you or any other household member used a Susu,10 welfare scheme, or other savings club to save money in the past 12 months? (6) Informal credit: Have you or any member of your household used a Susu, welfare scheme, or savings club to borrow money in the past 12 months? (7) Insurance: Many people insure themselves and their possessions against unexpected circumstances. Have you or any member of your household used an institution to insure yourselves (life, health) or property (house- hold goods, house, vehicle, and the like) in the past 12 months? That is, do you or anyone in the household have any long- or short-term insur- ance policies with any institution? The same indicators were calculated and the same questions were asked for the full enumeration treatments but for individual use. For example, for the 8. To help respondents distinguish banks from other �nancial service providers, interviewers received a list of the banks operating in Ghana and a glossary of de�nitions of �nancial terms, including for example, “micro�nance: small-scale loans typically given to owners of microenterprises to cover business expenses including small-scale investments, though the loan proceeds can be used for nonbusiness purposes including consumption. Liability for loan repayment can apply only to the borrower (individual-based) or to a solidarity lending group to which the borrower belongs. Under solidarity group lending, group members have strong incentives to ensure that fellow group members repay their loans. Some, but not all, micro�nance institutions in Ghana also provide savings services to their members.� Observations of interviews in the �eld indicated that respondents had little trouble identifying banks. 9. “Someone else� could be either a family member or a nonfamily member. In either case, the household would be considered “banked.� If the response was that neither the respondent nor any member of the household had a bank account but that the respondent (or another household member) did banking transactions through someone else’s account, it could be inferred that the “someone else� was not a family member. This occurred rarely. 10. For a small fee, Susu collectors provide a secure, informal means for Ghanaians to save and access their own money and to gain some limited access to microcredit. Robert Cull and Kinnon Scott 209 T A B L E 3 . Descriptive Statistics: Full Ghana Living Standards Study 5 Sample and Subsample Used in the Financial Services Survey Ghana Living Financial Services t-test of Standards Study 5 Survey subsample equivalence Variable full sample mean mean of means Region Coastal 29.65 33.12 3.21 (45.68) (47.07) (0.00) Forest 40.83 38.87 1.70 (49.15) (48.76) (0.09) Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 Savannah 29.52 28.01 1.41 (45.61) (44.91) (0.16) Rural 58.35 62.35 3.46 (49.30) (48.46) (0.00) Household characteristics Female household head 27.88 28.62 0.70 (44.84) (45.21) (0.48) Head of household literate 47.79 44.35 2.94 (49.95) (49.69) (0.00) Head of household numerate 64.24 64.16 0.06 (47.93) (47.96) (0.95) Age of head of household 45.34 45.51 0.46 (15.63) (15.64) (0.65) Extended family 26.82 27.97 1.10 (44.31) (44.89) (0.27) Household size 4.20 4.22 0.24 (2.83) (2.87) (0.81) Household has agricultural worker(s) 65.10 69.50 3.96 (47.67) (46.05) (0.00) Household has self-employed worker(s) 69.59 70.72 1.06 (46.01) (45.51) (0.29) Household has employed worker(s) 23.56 23.56 0.00 (42.44) (42.45) (1.00) Individual characteristics Age 19.62 24.04 0.66 (19.56) (24.19) (0.51) Male 48.69 49.31 1.08 (49.98) (50.00) (0.28) Note: Numbers in parentheses are standard deviations for sample means and p-values for t-statistics. Source: Authors’ analysis of survey data. banked indicator, the question was simply: “Do you have a bank account?� Reponses are aggregated across all members of the household to arrive at the measure of household use. In other words, if one member of the household reports having a bank account, then the whole household is considered banked for the full enumeration treatments. An element of subjectivity went into the crafting of these questions, and one might worry that slight tinkering with the institution-based questions could 210 THE WORLD BANK ECONOMIC REVIEW T A B L E 4 . Region and Household and Individual Characteristics across Treatment Groups Means t-tests of equivalence of means Group Group Group Groups 1 Groups 2 Groups 3 Treatment group 1 2 3 and 2 and 3 and 1 Region Coastal 30.5 29.6 29.3 0.37 0.12 0.49 (46.1) (45.7) (45.5) (0.71) (0.90) (0.62) Forest 40.6 41.4 40.2 0.30 0.45 0.15 Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 (49.1) (49.3) (49.1) (0.76) (0.65) (0.88) Savannah 28.9 29.0 30.5 0.05 0.60 0.65 (45.3) (45.4) (46.1) (0.96) (0.55) (0.51) Rural 64.4 650 65.7 0.20 0.28 0.49 (47.9) (47.7) (47.5) (0.84) (0.78) (0.63) Household characteristics Female household head 26.4 28.8 29.7 0.97 0.37 1.34 (44.1) (45.3) (45.7) (0.33) (0.71) (0.18) Head of household literate 41.8 41.4 43.9 0.15 0.88 0.74 (49.4) (49.3) (49.7) (0.88) (0.38) (0.46) Head of household numerate 63.5 60.7 62.8 1.03 0.77 0.26 (48.2) (48.9) (48.4) (0.31) (0.44) (0.80) Age of head of household 46.0 46.4 46.5 0.39 0.17 0.57 (15.5) (15.7) (15.1) (0.70) (0.87) (0.57) Extended family 29.6 28.0 30.8 0.63 1.09 0.47 (45.7) (45.0) (46.2) (0.53) (0.27) (0.64) Household size 4.50 4.24 4.59 1.63 2.15 0.56 (2.86) (2.92) (2.98) (0.10) (0.03) (0.57) Household has agricultural worker(s) 72.7 71.2 71.9 0.60 0.30 0.30 (44.6) (45.3) (45.0) (0.55) (0.77) (0.76) Household has self-employed worker(s) 73.3 75.1 74.4 0.74 0.30 0.44 (44.6) (43.3) (43.7) (0.46) (0.76) (0.66) Household has employed worker(s) 22.6 21.7 21.4 0.41 0.11 0.52 (41.9) (41.2) (41.0) (0.68) (0.91) (0.60) Individual characteristics Age 23.7 24.3 23.7 1.10 1.13 0.02 (19.6) (20.0) (19.6) (0.27) (0.26) (0.98) Male 49.4 49.1 49.2 0.19 0.04 0.15 (50.0) (50.0) (50.0) (0.84) (0.97) (0.88) Note: Numbers in parentheses are standard deviations for sample means and p-values for t-statistics. Source: Authors’ analysis of survey data. increase reported levels of use. Considerable work with GSS staff to adapt these questions to the country context and extensive piloting of the questions (both at GSS and in the �eld) helped ensure that the questions were well under- stood by respondents. Thus con�dence is high that these questions represent a reasonable and fair attempt to gather data on �nancial service use in Ghana Robert Cull and Kinnon Scott 211 T A B L E 5 . Percentage of Households That Use Financial Services Formal Indirect nonbank Formal Informal Informal Survey type Banked access saving credit savings credit Insurance Sample means (percent) Head of household 26.5 6.4 3.0 3.3 19.7 4.2 11.3 (n ¼ 638) (1.7) (1.0) (0.7) (0.7) (1.6) (0.8) (1.3) Random household 10.0 3.3 1.7 1.5 17.7 4.2 10.6 Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 member (n ¼ 480) (1.4) (0.8) (0.6) (0.5) (1.7) (0.9) (1.4) Full enumeration 25.5 5.1 2.5 1.9 17.3 4.2 7.9 (n ¼ 643) (1.7) (0.9) (0.6) (0.5) (1.5) (0.8) (1.1) t-tests of equivalence of means Head or random 7.05 2.33 1.41 1.94 0.86 0.05 0.35 household member (0.00) (0.02) (0.16) (0.05) (0.39) (0.96) (0.73) Head or full enumeration 0.40 0.99 0.54 1.61 1.15 0.03 2.04 (0.69) (0.32) (0.59) (0.11) (0.25) (0.98) (0.04) Full enumeration or 6.69 1.46 0.94 0.52 0.19 0.03 1.55 random household member (0.00) (0.14) (0.35) (0.60) (0.85) (0.98) (0.12) Note: Numbers in parentheses are standard deviations for sample means and p-values for t-statistics. Source: Authors’ analysis of survey data. through institution-based questions. Moreover, in the �rst set of comparisons between full enumerations and informants, all respondents were asked the same questions. While use levels might be affected by the speci�cs of those questions, any differences in the data generated by a full enumeration and that gathered from informants are much less likely to be affected. For �ve of the seven indicators—banked, indirect access, formal nonbank savings, informal savings, and informal credit—household use rates are almost identical when the head of household is the informant and when a full enumer- ation is undertaken (table 5). For formal credit, use rates reported by the head of household are slightly higher than those from the full enumeration treat- ments, though the hypothesis that the two rates are equal to one another cannot be rejected. Overall, the head of household reports information that is very similar to that generated by full enumeration. 212 THE WORLD BANK ECONOMIC REVIEW This is good news. Interviewing the head only is much cheaper than inter- viewing all adult members of a household, an issue returned to below. However, some surveys with a �nancial services module, such as labor force participation surveys, are designed to interview all members of a household. The results are good news in those cases, too. The information generated through the full enumeration appears to be a reasonable substitute for that gen- erated by the head of household. Because the household use rates calculated from responses to institution-based questions are comparable using either method, valid comparisons could be made across a much broader set of Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 countries. In contrast, a randomly selected adult from the household (who is not the head) does not provide information that is comparable to that generated by the head or by full enumeration. Randomly selected informants produce use rates that are lower than those for the other two methods and signi�cantly lower for banked, indirect access, and formal credit. This pattern suggests that the random informant has substantially less knowledge about household use of �nancial services than does the head of household. Disparities are greatest for services provided by formal institutions. For both informal savings and infor- mal credit, the use rates produced by random informants are almost identical to those produced by the head of household or through full enumeration. This could be because many informal savings and credit arrangements involve social activities (meetings) that all household members know about. Although the head of household respondents and the full enumeration tend to yield very similar use rates, insurance is an exception. One would expect the full enumeration to provide the most complete information and thus produce the highest use levels. And yet the percentage of households that have insurance is reported at 11.3 percent when information is provided by the head of house- hold and 7.9 percent when it is collected through a full enumeration of individ- ual use. It is conceivable that the head of household has purchased insurance for other household members of which those members are not aware. Another issue, turned to in more detail below, is that the institution-based question is a poor method of collecting information on insurance use, and therefore that none of the estimates for that indicator reported in table 5 is reliable.11 Comparisons of use rates calculated from product- and institution-based questions also reveal stark differences across indicators. The product-based questions are similar to those used in FinScope surveys. For example, a respon- dent who answered yes to any of the following questions was considered banked: 11. Recall that the head of household is asked only about his or her own personal use of insurance products in the full enumerations, and thus it is possible that the full enumeration could yield a smaller average use rate than when the head responds on behalf of the household, for the reason mentioned. However, observations of �eld training suggest that an institution-based question is simply not a good method for collecting reliable information about insurance use. Robert Cull and Kinnon Scott 213 (1) Do you currently have an ATM card? (2) Do you currently have a debit card? (3) Do you currently have a Savings Plus account?12 (4) Do you currently have a current account (checking)? (5) Do you currently have a savings account at a bank? (6) Do you currently have a PostBank account or a post of�ce savings account? (7) Do you currently have a bank loan? (8) Do you currently have a bank overdraft facility? Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 The questions underlying each indicator appear in the appendix. Note that there is no product-based indicator for indirect access since there was only one question on that topic and it was asked in the same way in both the product- based and institution-based questionnaires. That indicator is therefore dropped from subsequent tables. As noted, the focus is on individual use of �nancial services so as not to con- flate the effects of the method of eliciting household use information (infor- mant or full enumeration) with the effects of asking product- or institution-based questions. Again, while some degree of subjectivity entered into the selection of questions underlying the product-based indicators of �nan- cial services use, care was taken to adapt those questions to the country context. And many questions were selected from those used in past FinScope surveys. This should therefore constitute a fair test of the importance of asking product-based questions in the sense that it well represents the most advanced surveys undertaken to date. Product- and institution-based questions produce very similar use rates for basic services, such as banked and formal saving (banks þ nonbanks; table 6). By contrast, the product-based questions yield much higher use rates than do the institution-based questions for formal credit (2.8 percent and 0.8 percent), informal savings (18.8 percent and 8.9 percent), and insurance (16.3 percent and 5.7 percent), and all of the differences are statistically signi�cant. For these arguably more complex �nancial services, product-related cues appear to produce a much more complete picture of use. A surprising result is that reported use of informal credit is higher for insti- tutional than for product-based questions. This is because the product-based question on informal credit was poorly designed. It explicitly mentioned Susu’s, welfare schemes, and savings clubs, which mirrors the institutional question. The single institutional question asked about the past year, while the product-based questions asked about current use and whether such services had ever been used. To be consistent across the product-based indicators, only current use should be considered. However, because the institutional question 12. This is the brand name of a specialized savings account offered by some Ghanaian banks with additional features such as limited checking. 214 THE WORLD BANK ECONOMIC REVIEW T A B L E 6 . Percentage of Individuals Who Use Financial Services, by Product and Institutional Questions Formal saving (banks þ Formal Informal Informal Survey type Banked nonbanks) credit savings credit Insurance Sample means (percent) Questions on use of products 14.3 14.2 2.8 18.8 0.7 16.3 (n ¼ 2,201) (0.7) (0.7) (0.4) (0.8) (0.2) (0.8) Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 Questions on use of institutions 13.3 13.8 0.8 8.9 2.2 5.7 (n ¼ 1,568) (0.9) (0.9) (0.2) (0.7) (0.4) (0.6) t-tests of equivalence of means Products or institutions 0.88 0.39 4.32 8.49 3.97 10.04 (0.38) (0.70) (0.00) (0.00) (0.00) (0.00) Note: Numbers in parentheses are standard deviations for sample means and p-values for t-statistics. Source: Authors’ analysis of survey data. asks about the past year and because users switch in and out of these services regularly, the product-based question produces a lower use rate than the insti- tutional question, which is misleading. The construction of the questions on informal credit does not therefore permit meaningfully comparing product and institutional questions.13 By de�nition, the level of individual use of �nancial services would not be expected to exceed the level of household use. The results show that this is true for all services except for insurance (compare tables 5 and 6). For that indi- cator, the individual use rate based on product-related questions far exceeds the household use rate calculated from the institution-based question. This shows that the institutional insurance question is not a good substitute for a series of product-related questions. 13. At the same time, use rates for semiformal and other informal credit services reveal some interesting patterns. First, 4.7 percent of respondents said that they were currently using a hire purchase or installment credit plan, while an additional 8.1 percent reported that they were using credit facilities other than bank loans, credit cards, hire purchase, or installment plans. These are sizable fractions in a country where only about a quarter of households are banked. It suggests that if institutional questions had targeted the providers of such facilities, a meaningful institutional–product comparison could have been made, one that likely would have tipped in favor of product-based questions. Still, it is hard to identify the providers of such facilities for a survey respondent without also de�ning what those facilities are. Indeed, �eld tests showed that interviewers needed to explain some of these concepts in depth to respondents. For informal credit, therefore, it might not be possible to separate institution- and product-based descriptions suf�ciently to construct a test. Robert Cull and Kinnon Scott 215 T A B L E 7 . Percentage of Individuals Who Use Financial Services, by Product and Institutional Questions and Respondent Type Formal saving (banks þ Formal Informal Informal Survey type Banked nonbanks) credit saving credit Insurance Sample means (percent) Household heads Questions on use of products 22.8 22.7 4.6 21.9 0.7 17.8 Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 (n ¼ 978) (1.3) (1.3) (0.7) (1.3) (0.3) (1.2) Questions on use of institutions 23.8 24.5 1.4 12.7 2.7 7.5 (n ¼ 638) (1.7) (1.7) (0.5) (1.3) (0.6) (1.0) t-tests of equivalence of means Products or institutions 0.48 0.81 3.50 4.70 3.17 5.92 (0.63) (0.42) (0.00) (0.00) (0.00) (0.00) Sample means (percent) Nonhousehold heads Questions on use of products 7.4 7.4 1.4 16.3 0.7 15.0 (n ¼ 1,223) (0.8) (0.8) (0.3) (1.1) (0.2) (1.0) Questions on use of institutions 6.0 6.5 0.4 6.3 1.8 4.4 (n ¼ 930) (0.8) (0.8) (0.2) (0.8) (0.4) (0.7) t-tests of equivalence of means Products or institutions 1.29 0.89 2.25 7.10 2.52 8.11 (0.20) (0.37) (0.02) (0.00) (0.01) (0.00) Note: Numbers in parentheses are standard deviations for sample means and p-values for t-statistics. Source: Authors’ analysis of survey data. Nor does the problem appear to stem from the �nancial knowledge of the respondent. One would expect the head of household to be the most �nancially knowledgeable member of the household, but even when the head is asked about personal use of insurance products, the product-based use rate is much higher than the institution-based measure (table 7). A similar pattern holds for formal credit and informal savings, for both household heads and nonheads, and the differences between the product- and institution-based use rates are statistically signi�cant. The evidence points to across-the-board dif�culties for 216 THE WORLD BANK ECONOMIC REVIEW all respondents in using institution-based questions to gather information on formal credit, informal savings, and insurance. In summary, the preliminary comparisons across treatment groups indicate that the identity of the respondent and the way questions are asked affect reported use of some �nancial services. Full enumerations of all household members produce use rates similar to those reported by the head of household, while interviewing a randomly selected nonhead produces lower levels of household use. Product-related cues appear to be important to fully understand the use of insurance, formal credit, and informal savings but do not appear Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 necessary for more basic services such as bank accounts and formal savings. I V. R E G R E S S I O N S This section reports on tests of whether the differences across treatments described in the previous section hold up in regressions after controlling for other factors that could affect use. Some regressions are also designed to ident- ify the characteristics of the individuals and households that reported lower levels of use on institution-based questions than on product-based questions. Another set of regressions examines the household characteristics of the ran- domly selected informants who reported lower household use rates than those obtained from the head of household or the full enumeration. The hope is to identify the types of respondents who have dif�culty with certain question formats. Household Use: Full Enumeration or Informants To describe household use of �nancial services, the following speci�cation was estimated in a probit regression model: Financei ¼ a þ b1 agei þ b2 rurali þ b3 sizei þ b4 dependent sharei þ b5 female headi þ b6 age of headi þ b7 education of headi þ b8 head numeratei þ b9 share in agriculturei þ b10 share employedi þ b11 share self -employedi þ b12 informant is headi þ b13 random informanti þ ei where �nance is one of the seven indicators of household use of �nancial ser- vices described in section III (banked, indirect access, nonbank saving, informal saving, formal credit, informal credit, and insurance). All those indicators are dummy variables equal to one if any member of household i uses that service. Four variables control for the composition and location of the household. Positive coef�cients are expected for the average age of household members and household size because larger households with older members are more likely to have an individual who uses �nancial services. For the same reason, households with a high dependent share are expected to use fewer �nancial Robert Cull and Kinnon Scott 217 services. Use is expected to be lower in rural areas because �nancial services are less available. Variables for gender, age, education, and numeracy control for characteristics of the head of household. The dummy variable indicating whether the head is female is expected to be negatively linked to use of �nan- cial services if providers exhibit biases against women or perhaps for broader cultural reasons. Age, education, and numeracy are expected to be positively associated with use of �nancial services. Education is controlled for using two dummy variables: one indicating whether the head attended primary school and another indicating whether the head attended upper secondary school.14 Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 Three variables control for the employment composition of the household: share in agriculture, share employed, and share self-employed. Those who are employed are expected to have greater need for �nancial services. Agricultural workers and the self-employed might have different needs or �nd it more dif�cult to obtain �nancial services. Informant is head and random informant are dummy variables that describe the identity of the survey respondent. The informant dummy variables therefore capture the effects on reported household use rates relative to the omitted treatment category, a full enumeration of all adult household members’ individual use of �nancial services. De�nitions and summary statistics for the variables used in the analysis are in table 8; the correlations between variables appear in table 9. The correlations indicate that many household characteristics �t together in predictable ways. For example, rural households tend to be larger and more focused on agricultural activities. The summary statistics and correlations are calculated for the 3,630 obser- vations that enter the regressions that summarize individual use. Very similar summary statistics and correlations are found for the 1,734 observations that enter the household use regressions. To conserve space, only the information from the larger sample of individual use is reported here. The regression results for household use of �nancial services appear in table 10. In the regression with banked as the dependent variable (column 1), many of the control variables are signi�cant and of the expected sign. In par- ticular, household size, age of the head of household, and attendance in upper secondary school (or beyond) are all signi�cantly positively linked to being banked. Rural location, female headship, the share of dependents, and the share of self-employed workers are all negatively linked to being banked. The control variables do a better job of explaining variation in the banked indicator than in the other indicators, as reflected in both the overall �t of the regressions and the number of signi�cant variables. There is also a general tendency for the control variables to explain more variation in the use of services from 14. These dummy variables were chosen because they provide a reasonably large number of respondents in the lowest (no formal schooling) and highest (upper secondary school and beyond) categories. Note also that both dummy variables are equal to one for respondents that attended upper secondary school and beyond. To measure the effects of education on �nancial usage for those respondents, the coef�cients on both of the dummy variables must be summed. 218 THE WORLD BANK ECONOMIC REVIEW T A B L E 8 . Variable Descriptions and Summary Statistics Variable De�nition Mean Minimum Maximum Financial use variables Banked Equals 1 if any member of the 0.140 0 1 household has an account with a bank Formal savings Equals 1 if household has 0.142 0 1 formal non-bank savings Informal savings Equals 1 if household has 0.150 0 1 informal savings Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 Formal credit Equals 1 if household has credit 0.021 0 1 from a formal provider of �nancial services Informal credit Equals 1 if household has credit 0.013 0 1 from informal sources Insurance Equals1 if household has 0.120 0 1 insurance product from a formal provider Household characteristics Age Average age of household 36.864 8 98 members Rural Equals 1 if rural 0.679 0 1 Household size Number of household members 5.453 1 23 Share dependents Percentage of dependents in 0.540 0 1 household Female household head Equals 1 if household head is 0.225 0 1 female Age of household head Age of household head in years 47.436 16 98 Household head attended Equals 1 if household head has 0.604 0 1 primary school attended primary school Household head attended Equals 1if household head has 0.339 0 1 upper-secondary school attended upper secondary school Household head Equals 1 if household head can 0.591 0 1 numerate do written calculations Share agricultural Percentage of agricultural 0.398 0 1 workers workers in household Share employed Percentage of employed 0.081 0 1 members of household Share self-employed Percentage of self-employed 0.246 0 1 members of household Attended primary school Equals 1 if household member 0.614 0 1 has attended primary school Attended Equals 1 if household member 0.242 0 1 upper-secondary school has attended upper-secondary school Numerate Equals 1 if household member 0.594 0 1 can do written calculations Source: Authors’ analysis of survey data. T A B L E 9 . Correlations between Variables Female Age of Formal Informal Formal Informal Household Dependent. household household Variable Banked savings savings credit credit Insurance Age Rural size share head head Banked 1 Formal savings 0.935*** 1 Informal savings 0.132*** 0.140*** 1 Formal credit 0.231*** 0.223*** 0.134*** 1 Informal credit 0.035** 0.034** 0.191*** 0.101*** 1 Insurance 0.255*** 0.227*** 0.146*** 0.143*** 0.0228 1 Age 0.144*** 0.144*** 0.033* 0.058*** 0.0054 0.060*** 1 Rural – 0.191*** – 0.186*** –0.132*** –0.016 –0.053*** – 0.121*** 0.034** 1 Household size – 0.102*** – 0.108*** –0.092*** –0.010 –0.017 – 0.034** – 0.166*** 0.224*** 1 Dependent share – 0.132*** – 0.136*** –0.092*** –0.036** –0.007 – 0.019 0.226*** 0.202*** 0.314*** 1 Female household head – 0.072*** – 0.070*** 0.047*** –0.032* 0.011 – 0.003 0.071*** –0.151*** –0.250*** 0.125*** 1 Age of household head – 0.021 – 0.033** –0.108*** –0.022 –0.016 0.020 0.458*** 0.009 0.064*** 0.391*** 0.104*** 1 Household head attended 0.190*** 0.195*** 0.145*** 0.058*** –0.008 0.145*** – 0.180*** –0.267*** –0.143*** – 0.293*** –0.107*** –0.326*** primary school Household head attended 0.274*** 0.270*** 0.107*** 0.084*** 0.002 0.186*** – 0.057*** –0.308*** –0.097*** – 0.216*** –0.122*** –0.038** upper-secondary school Household head 0.217*** 0.217*** 0.140*** 0.050*** 0.010 0.129*** – 0.147*** –0.296*** –0.150*** – 0.245*** –0.089*** –0.256*** numerate Share agricultural – 0.166*** – 0.156*** –0.095*** –0.027 –0.016 – 0.177*** 0.112*** 0.403*** –0.022 – 0.038** –0.180*** 0.070*** workers Share employed 0.245*** 0.244*** 0.105*** 0.087*** –0.007 0.094*** – 0.036** –0.339*** –0.247*** – 0.380*** –0.050*** –0.126*** Share self-employed – 0.088*** – 0.082*** –0.017 –0.044*** –0.037** – 0.060*** 0.143*** 0.0903*** –0.345*** – 0.123*** 0.047*** 0.021 Attended primary school 0.186*** 0.184*** 0.099*** 0.048*** –0.010 0.134*** – 0.336*** –0.296*** –0.159*** – 0.235*** 0.031* –0.156*** Attended 0.357*** 0.348*** 0.118*** 0.108*** 0.012 0.201*** 0.068*** –0.286*** –0.144*** – 0.226*** –0.023 –0.018 upper-secondary school Robert Cull and Kinnon Scott Numerate 0.208*** 0.207*** 0.106*** 0.0452*** 0.014 0.128*** – 0.289*** –0.319*** –0.167*** – 0.206*** 0.039** –0.122*** Employed 0.067*** 0.074*** 0.110*** 0.073*** 0.033** – 0.050*** 0.152*** 0.163*** –0.028* – 0.105*** –0.121*** –0.183*** 219 (Continued) Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 220 T A B L E 9 . Continued Household head has Household head some has some Household Share Some Some primary upper-secondary head agricultural Share Share primary upper-secondary education education numerate workers employed self-employed education education Numerate Employed Head has some primary 1 education Head has some 0.579*** 1 upper-secondary education THE WORLD BANK ECONOMIC REVIEW Household head numerate 0.730*** 0.573*** 1 Share agricultural workers – 0.281*** –0.338*** – 0.304*** 1 Share employed 0.210*** 0.275*** 0.204*** – 0.212*** 1 Share self-employed – 0.027 –0.110*** – 0.034** 0.412*** –0.297*** 1 Attended primary school 0.651*** 0.436*** 0.533*** – 0.303*** 0.197*** –0.037** 1 Attended upper-secondary 0.394*** 0.641*** 0.389*** – 0.268*** 0.274*** –0.080*** 0.448*** 1 school Numerate 0.511*** 0.445*** 0.691*** – 0.305*** 0.198*** –0.049*** 0.767*** 0.447*** 1 Employed – 0.040** –0.114*** – 0.060*** 0.394*** 0.060*** 0.253*** –0.153*** –0.027 – 0.157*** 1 *Signi�cant at the 10 percent level;**signi�cant at the 5 percent level;*** signi�cant at the 1 percent level. Source: Authors’ analysis of survey data. Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 T A B L E 1 0 . Household Financial Services Use Rate Regressions, Marginal Effects Indirect Formal access to nonbank Informal Expected Banked account savings savings Formal credit Informal credit Insurance Variable signa (1) (2) (3) (4) (5) (6) (7) Average age of household þ 2 0.0015 0.0005 0.0002 0.0001 0.0005* 2 0.0001 0.0004 members (0.0011) (0.0005) (0.0003) (0.0009) (0.0003) (0.0004) (0.0006) Rural 2 2 0.0734*** 0.0229** 2 0.0060 0.0461** 0.0020 2 0.0271*** 0.0187 (0.0223) (0.0116) (0.0057) (0.0216) (0.0059) (0.0102) (0.0153) Household size þ 0.0040*** 0.0020 0.0003 0.0049 0.0028*** 0.0024 0.0044 (0.0040) (0.0021) (0.0010) (0.0040) (0.0009) (0.0017) (0.0027) Share dependents 2 2 0.1324*** 2 0.0063 0.0157 0.0425 2 0.0287** 0.0135 0.0121 (0.0459) (0.0226) (0.0129) (0.0446) (0.0126) (0.0217) (0.0313) Female household head 2 2 0.1243*** 0.0042 2 0.0052 0.0566** 2 0.0092 0.0143 2 0.0028 (0.0201) (0.0131) (0.0060) (0.0256) (0.0057) (0.0125) (0.0170) Age of household head þ 0.0034*** 2 0.0002 2 0.0004 0.0009 2 0.0003 2 0.0003 0.0012* (0.0011) (0.0005) (0.0003) (0.0010) (0.0003) (0.0004) (0.0007) Household head attended þ 2 0.0520 0.0218 2 0.0035 0.0869*** 0.0064 2 0.0070 0.0011 primary school (0.0346) (0.0174) (0.0102) (0.0309) (0.0097) (0.0141) (0.0247) Household head attended þ 0.1767*** 0.0287** 0.0125* 0.0251 0.0056 2 0.0142 0.0616*** upper-secondary school (0.0258) (0.0125) (0.0068) (0.0239) (0.0065) (0.0116) (0.0176) Household head numerate þ 0.0948*** 0.0103 0.0196** 0.0329 0.0038 0.0247* 0.0494** (0.0299) (0.0159) (0.0087) (0.0286) (0.0089) (0.0121) (0.0212) Share agricultural workers 2 – 0.0552 2 0.0013 0.0073 0.0116 2 0.0068 0.0178 2 0.1015*** (0.0368) (0.0189) (0.0107) (0.0355) (0.0084) (0.0168) (0.0273) Share employed þ 0.0420 0.0482* 0.0125 0.0406 0.0157 2 0.0046 0.0275 Robert Cull and Kinnon Scott (0.0508) (0.0251) (0.0127) (0.0527) (0.0129) (0.0240) (0.0381) Share self-employed 2 2 0.1096** 0.0301 0.0146 0.0380 0.0051 2 0.0630** 0.0562* 221 (Continued ) Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 222 TABLE 10. Continued Indirect Formal access to nonbank Informal Expected Banked account savings savings Formal credit Informal credit Insurance Variable signa (1) (2) (3) (4) (5) (6) (7) (0.0474) (0.0234) (0.0144) (0.0452) (0.0126) (0.0238) (0.0334) Informant is head of household ? 0.0233 0.0071 0.0015 0.0197 0.0055 2 0.0004 0.0257 (0.0235) (0.0117) (0.0061) (0.0229) (0.0065) (0.0107) (0.0175) Informant is randomly selected 2 2 0.1655*** 2 0.0149 0.0052 0.0044 2 0.0043 2 0.0015 0.0259 nonhead THE WORLD BANK ECONOMIC REVIEW (0.0190) (0.0120) (0.0061) (0.0250) (0.0063) (0.0113) (0.0197) Number of observations 1734 1734 1734 1734 1734 1734 1734 Log likelihood 2 738.6819 2 329.6634 179.6205 788.4633 2 165.9691 2 287.8245 2 516.1416 Pseudo R2 0.1910 0.0607 0.1083 0.0457 0.1103 0.0491 0.0829 Chi2 head random 47.05*** 2.38 0.89 0.30 1.87 0.01 0.00 p head random 0.0000 0.1231 0.3442 0.5838 0.1712 0.9310 0.9845 *Signi�cant at the 10 percent level;**signi�cant at the 5 percent level;*** signi�cant at the 1 percent level. a. Expected signs are for access to formal �nancial services, especially for being banked. For other services, especially those that are informally pro- vided, these hypothesized relationships might not hold as well, or at all. Note: Numbers in parentheses are standard errors. Source: Authors’ analysis of survey data. Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 Robert Cull and Kinnon Scott 223 formal providers (banked, formal nonbank saving, and formal credit) than from informal providers (informal credit and informal savings). When household characteristics are controlled for in the regressions, the comparisons across the treatment groups remain similar to those in the summary statistics in table 5. There are no signi�cant differences between the head of household as the respondent and a full enumeration, as reflected in the insigni�cant coef�cients for the informant ¼ head variable for all indi- cators. By contrast, the tests of whether coef�cients for the two informants (head and random) are equal, reported at the bottom of the table, reveal signi�- Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 cant differences for the banked indicator. There is no longer a signi�cant differ- ence for formal credit or indirect access, as there was in the summary comparisons, but that could be because there is so little use of those services in the sample. In countries where formal credit or indirect access is more preva- lent, signi�cant differences might emerge.15 In addition, the coef�cient for a randomly selected informant is negative and highly signi�cant for being banked, indicating that the random informant pro- vides less complete information on household use of banking services than does a full enumeration. In short, though the signi�cance levels fall when con- trolling for additional factors that affect use, the same qualitative patterns emerge: the head of household and a full enumeration produce similar house- hold use rates, but a randomly selected (nonhead) informant produces lower use rates for services from formal providers. To get a better understanding of whether particular household character- istics drive the lower use rates reported by random informants, the control vari- ables are interacted with the treatment variables. That is, the explanatory variables are multiplied by head ¼ informant to derive a second set of explana- tory variables and multiplied by random informant to derive a third set. The two new sets of explanatory variables are included in the original regressions in the full-interaction speci�cations: X 3 Financei ¼ ð@t þ bt Xi Þ þ 1i t ¼1 where t refers to the three treatment categories (full enumeration, head of household informant, and random informant) and X is the set covariates from the original regression. This allows the control variables to affect reported use in different ways across treatment categories. For the most part, the determinants of banked, indirect access, and formal credit are similar across the three treatments, as indicated by insigni�cant 15. The signi�cant differences for formal credit and indirect access found in table 5 might reflect the problem of small numbers (where small deviations seem to imply highly signi�cant differences since so few respondents use these services). In any event, signi�cance disappears after controlling for other relevant factors through regression. 224 THE WORLD BANK ECONOMIC REVIEW coef�cients on the interaction variables. To conserve space, these results are not presented here. There are some exceptions for banked that are worth noting, however. For the trials that use a random informant, the share of dependents has a strong negative association with being banked. Note that for the survey, qualifying adults were all household members age 15 or older. Thus, a number of the randomly selected informants were dependents under the de�nition used to construct the dependent share control variable. The nega- tive signi�cant coef�cient for dependent share reflects, at least in part, the dif�- culties that young adults face in responding to institution-based questions Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 about household use of banking services.16 Since the determinants of household use of banking services are similar whether use is reported by a random informant or calculated from a full enu- meration of individuals’ use, and since the constant is not statistically different for those two treatment categories in the full-interaction speci�cation, it appears that younger, poorly informed household members were largely responsible for the relatively low use of banking services reported by random informants and shown in the summary statistics in table 5 and the basic regressions in table 10. The great majority of coef�cients for heads of household are insigni�cant, indicating that the determinants of use are similar for those treatments and the ones that used full enumeration, but again the exceptions are instructive. The �rst is that in households where the head is numerate, use of banking services is signi�cantly greater under full enumeration but not when the head reports on household use. This suggests that numerate heads pass on knowledge to other household members about banking services that increases others’ per- sonal use, but both numerate and innumerate heads have a reasonable grasp of household use of banking services when they are asked. A second difference is that the share of employed household members is positive and signi�cant in the full enumeration speci�cations, presumably because the employed have greater need of banking services, but insigni�cant when the head reports on household use. Like the insigni�cant result for numerate heads of household, the one for share of employed household members suggests that household heads know about the use of banking services by the employed members of their household and are able to report on it when asked the institution-based question. Product or Institutional Questions For the regressions that describe individual use of �nancial services and compare product- and institution-based questions, the following individual 16. When a dummy variable is included indicating that the random respondent is 15– 18 years old, it is negative and signi�cant while the dependent share variable is no longer signi�cant. This provides additional evidence that it is younger respondents who have dif�culty providing accurate information about household use of banking services. Robert Cull and Kinnon Scott 225 characteristics are added to the household characteristics used in the regressions in the previous section: education level of the respondent (dummy variables for attended primary and attended upper secondary school) and dummy variables indicating whether the respondent is numerate and employed. Education level, numeracy, and employment are expected to be positively linked to personal use of �nancial services. The informant dummy variables are replaced with a dummy variable indicating whether the respondent was asked product-based questions. The coef�cient on that variable therefore measures reported use rates relative to the omitted category, respondents who answered institution- Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 based questions. The regressions results appear in table 11.17 The individual characteristics are almost all positive and signi�cant for banked and formal savings (banks þ nonbanks). Employed respondents are signi�cantly more likely to use all types of �nancial services except for insurance, and the marginal effects for the non- insurance indicators are large relative to the average individual use rates reported in table 6. The education level of the respondent is associated with greater use of insurance, however. In all, the individual characteristics explain substantial variation in the �nancial use indicators. That said, household characteristics also explain substantial variation in individual use. As in the household use regressions, average age in the house- hold is positively associated with the indicators. Rural location, female head- ship, and the shares of agricultural workers and self-employed workers are signi�cantly negatively associated with the indicators.18 The overall �t is also better in the individual use regressions than in the household use regressions, as reflected in the pseudo-R2 values. Most important, the dummy variable indicating whether the respondent answered product-based questions is positive and signi�cant for informal savings, formal credit, and insurance, as was true for the summary compari- sons in table 6. The marginal effects of the product-based questions variable are also large in those regressions relative to the levels of personal use of those services reported in table 6. Moreover, the coef�cient for the product-based question format implies a disparity between treatments similar to that implied by the simple bivariate comparisons. For example, table 6 indicates that 17. For households in group 2, all members were �rst asked about their own use of �nancial services using the institution-based questions, and then a member of the household was randomly selected to answer the more detailed product-based questions. In this way, the random respondent provides an observation under both question formats. Although this could have implications for the standard errors, this does not appear to be a cause for major concern: product– institutional comparisons are very similar whether the group 2 product-based responses from the randomly selected household members are included or not. Results are therefore reported for product-based questions for both groups 1 and 2. 18. The age of the head of household is negatively associated with indicators of individual use, whereas it was positively associated with household use. This is because the age of the household head competes with the average age of all household members in the individual use regressions. When one of those variables is dropped, the other is positive and signi�cant in the regressions in table 11. 226 T A B L E 1 1 . Individual Financial Services Use Rate Regressions, Marginal Effects of Product versus Institutional Questions Expected Formal saving signa Banked (banks þ nonbanks) Informal savings Formal credit Informal credit Insurance Variable (1) (2) (3) (4) (5) (6) Average age of household members þ 0.0042*** 0.0045*** 0.0019*** 0.0005*** 2 0.00001 0.0014*** (0.0004) (0.0004) (0.0005) (0.0001) (0.0001) (0.0004) Rural 2 2 0.0386*** 2 0.0385*** 2 0.0533*** 0.0018 2 0.0128*** 2 0.0064 THE WORLD BANK ECONOMIC REVIEW (0.0124) (0.0126) (0.0148) (0.0023) (0.0049) (0.0113) Household size ? 2 0.0021 2 0.0026 2 0.0035 0.0003 – 0.0012** 0.0012 (0.0020) (0.0021) (0.0023) (0.0004) (0.0005) (0.0018) Share dependents 2 2 0.0371 2 0.0368 2 0.0305 2 0.0030 2 0.0036 0.0280 (0.0226) (0.0228) (0.0270) (0.0051) (0.0054) (0.0232) Female household head 2 2 0.0543*** 2 0.0524*** 0.0391*** 2 0.0035 2 0.0021 2 0.0168 (0.0099) (0.0102) (0.0158) (0.0024) (0.0025) (0.0114) Age of household head þ 2 0.0018*** 2 0.0022*** 2 0.0021*** 2 0.0003*** 2 0.00005 0.0002 (0.0005) (0.0005) (0.0005) (0.0001) (0.0001) (0.0004) Household head attended primary þ 2 0.0479** 2 0.0270 0.0482** 0.0039 2 0.0021 0.0449** school (0.0253) (0.0243) (0.0215) (0.0043) (0.0046) (0.0189) Household head attended þ 2 0.0137 0.0172 0.0094 0.0046 2 0.0004 0.0272* upper-secondary school (0.0174) (0.0176) (0.0184) (0.0050) (0.0039) (0.0162) Household head numerate þ 0.0343 0.0200 0.0120 2 0.0076 2 0.0007 2 0.0283 (0.0222) (0.0231) (0.0229) (0.0068) (0.0047) (0.0214) Share agricultural workers 2 2 0.0713*** 2 0.0630*** 2 0.0606*** 2 0.0015 2 0.0026 2 0.1295*** Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 (0.0196) (0.0197) (0.0226) (0.0041) (0.0047) (0.0205) Share employed þ 0.0490* 0.0497* 2 0.0335 0.0025 2 0.0290*** 0.0284 (0.0259) (0.0262) (0.0316) (0.0053) (0.0087) (0.0276) Share self-employed 2 2 0.0754*** 2 0.0754*** 2 0.0684** 2 0.0147** 2 0.0292*** 0.0211 (0.0265) (0.0266) (0.0296) (0.0069) (0.0074) (0.0268) Respondent attended primary þ 0.0458** 0.0350*** 2 0.0081 0.0004 2 0.0076 0.0151 school (0.0215) (0.0220) (0.0233) (0.0048) (0.0059) (0.0203) Respondent attended þ 0.1109*** 0.1008*** 0.0082 0.0048 0.0020 0.0467*** upper-secondary school (0.0230) (0.0225) (0.0187) (0.0053) (0.0044) (0.0175) Respondent numerate þ 0.03700 0.0488** 0.0241 0.0057 0.0046 0.0219 (0.0223) (0.0223) (0.0230) (0.0051) (0.0045) (0.0203) Respondent employed þ 0.0509*** 0.0527*** 0.0991*** 0.0122*** 0.0093*** 2 0.0046 (0.0106) (0.0107) (0.0114) (0.0027) (0.0022) (0.0126) Product þ 0.0122 0.0067 0.0913*** 0.0100*** 2 0.0112*** 0.1019*** (0.0095) (0.0096) (0.0106) (0.0028) (0.0022) (0.0091) Number of observations 3630 3630 3630 3630 3630 3630 Log likelihood 2 1126.0058 2 1145.0915 2 1371.3258 2 303.0288 2 227.5562 2 1152.5292 Pseudo R2 0.2357 0.2294 0.1068 0.1702 0.1235 0.1365 *Signi�cant at the 10 percent level;**signi�cant at the 5 percent level;*** signi�cant at the 1 percent level. a. Expected signs are for access to formal �nancial services, especially for being banked. For other services, especially those that are informally pro- vided, these hypothesized relationships might not hold as well, or at all. Note. Numbers in parentheses are standard errors. Source: Authors’ analysis of survey data. Robert Cull and Kinnon Scott 227 Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 228 THE WORLD BANK ECONOMIC REVIEW product-based questions produce an insurance use rate 10.6 percentage points higher than institution-based questions, while the product-based coef�cient from the regression indicates a 10.2 percentage point difference between treat- ments.19 The regression results therefore reinforce the conclusion that product- based cues help respondents provide a more complete picture of their use of those three �nancial services. For banked and formal savings (banks þ non- banks), the product-based questions dummy variable is insigni�cant, indicating again that product-based cues are less important for those services. For informal credit, the product-based dummy variable is negative and signi�cant, Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 but again the way that product-based question was constructed led to a test that is not very meaningful. To better identify the types of individuals who bene�t most from product-related cues, the explanatory variables in table 11 were interacted with the dummy variable for product-based questions for the three services for which a signi�cant increase in use rates was found for product-based questions compared with institution-based questions. Almost all the coef�cients on the interaction terms are insigni�cant, indicating that the determinants of reported use are similar for the two question formats and suggesting that all respondents bene�t from product-related cues regarding formal credit, informal savings, and insurance. Therefore, those results are not presented. These �ndings reinforce the conclusions drawn from the simple sample breakdown in table 7. Controlling for Supply-Side Effects To ensure that supply-side effects—the presence of providers of �nancial services—are not driving the differences in use across treatments that were reported above, additional regressions were run to control for travel time (in minutes) to the nearest bank as a measure of the local availability of �nancial services and others that include regional or local dummy variables to capture these effects. As in the base regressions, household use is very similar for full enumeration and when the head of household is the informant, and reported use of banking services is signi�cantly lower for the random informant. Individual use rates are signi�cantly higher for product-based questions for informal savings, formal credit, and insurance.20 In short, it seems unlikely that the omission of supply-side variables from the base regressions could be driving the results. 19. The same is true of the household use regressions. For example, the regressions imply that a random respondent is 16.6 percentage points less likely to report that someone in the household is banked than is revealed through a full enumeration. The difference between those two treatments in the bivariate comparison is 15.5 percentage points. 20. The working paper version of this article provides more details about these regressions, including the results (Cull and Scott 2009). Robert Cull and Kinnon Scott 229 V. C O N C L U S I O N S AND I M P L I CAT I O N S FOR FUTURE SURVEYS Measuring the breadth of outreach of �nancial sectors in developing countries remains a challenge, but one that must be met to better understand how �nan- cial services (or their absence) affect the livelihoods of the poor. Surveys of individuals and households about their use of �nancial services hold the most promise for measuring outreach well, but their cost and other logistical hurdles have made it dif�cult to develop a standard method of questioning that would generate comparable �nancial use data across countries and within countries Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 over time. This experimental analysis was designed to contribute to an under- standing of the comparability of �nancial use data generated under different question formats. The main �ndings are straightforward, intuitive, and should be useful for future data gathering efforts. Rates of household use of �nancial services are similar when the head reports on behalf of the household and when the rate is tabulated from a full enumeration of individual use. By contrast, randomly selected informants provide a less complete picture of household use of �nan- cial services than do the other two methods. The comparability of data for the head of household and the full enumeration is potentially important because interviewing only the head is much less costly than interviewing all household members. At the same time, some surveys, such as those measuring labor force participation, are designed to be full enumerations. Using the head of house- hold when possible and a full enumeration when dictated for other reasons should increase the number of countries for which comparable data can be generated. For formal credit, informal savings, and insurance, reported use is higher when questions are asked about speci�c �nancial products rather than about the respondents’ dealings with types of �nancial institutions. Product-related cues for these services appear to be important for all respondents, not just those who might be expected to be less knowledgeable about �nancial matters. The results are therefore similar to those from experiments on measuring household consumption, where inclusion of a longer list of items leads to higher reported consumption than does a shorter list of broader consumption categories. If treatments that yield higher use rates are presumed to be a more accurate depiction of reality (as seems likely), then using techniques such as random respondents and institution-based questions, though certainly less costly, will capture only a fraction of the use of some �nancial services (only half for some services in Ghana). This could make it dif�cult to design appropriate policies to foster �nancial inclusion and to measure their effectiveness. Although product-based and institution-based use were tested only for per- sonal use of �nancial services, it seems likely that product-related cues would also bene�t respondents informing about household use of those services. That implies adapting the institution-based questions used in the �nancial modules 230 THE WORLD BANK ECONOMIC REVIEW T A B L E 1 2 . Time Costs of Administering Financial Services Survey (minutes) Questionnaire type Services and response 1 2 3 Total Banked No 41.52 32.15 35.92 36.50 (22.26) (14.21) (15.61) (18.08) [n ¼ 468] [n ¼ 475] [n ¼ 479] [n ¼ 1,422] Yes 52.75 36.09 42.76 44.32 (34.89) (15.97) (20.50) (26.45) Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 [n ¼ 187] [n ¼ 158] [n ¼ 178] [n ¼ 523] Formal credit No 44.12 33.15 37.50 38.23 (26.30) (14.79) (17.27) (20.51) [n ¼ 613] [n ¼ 621] [n ¼ 634] [n ¼ 1,868] Yes 53.50 32.08 45.22 47.69 (34.20) (13.51) (17.68) (28.41) [n ¼ 42] [n ¼ 12] [n ¼ 23] [n ¼ 77] Insurance No 41.23 32.95 37.02 36.85 (22.46) (14.74) (17.48) (18.57) [n ¼ 495] [n ¼ 582] [n ¼ 579] [n ¼ 1,656] Yes 55.55 35.18 43.32 48.65 (35.51) (14.99) (15.09) (29.37) [n ¼ 160] [n ¼ 51] [n ¼ 78] [n ¼ 289] Total 44.73 33.13 37.77 38.60 (26.94) (14.76) (17.32) (20.95) [n ¼ 655] [n ¼ 633] [n ¼ 657] [n ¼ 1,945] Notes: Numbers in parentheses are standard errors. Source: Authors’ analysis of survey data. of larger, multipurpose surveys to include product-based cues that are appro- priate to the country context. Decisions on future questionnaires will also need to consider the relative costs (in interview time) of implementing the different treatments, which conform to expectations.21 The full enumeration using the product list takes the longest to administer. But full enumeration itself, using either the product- based or institution-based questionnaire, adds considerable time to the inter- views compared with use of a proxy respondent for the household (see results for group 1 in table 12). In other words, the �nding that the head of household is able to provide similar data to that obtained from full enumeration for most products has positive implications for the feasibility of expanding data 21. The time data collected in this survey are, at best, rough approximations of the actual time required. No effort was made to record time at the level of the speci�c product or institution modules. Only a total for the entire household interview, which includes a roster and further questions on attitudes and knowledge of �nance, is available. Also, as groups 2 and 3 contain two different treatments, it is not possible to separate the time costs associated with each one. Robert Cull and Kinnon Scott 231 collection on �nancial service use to other countries. Finally, for survey designers in countries that may have higher levels of �nancial service use, it is important to note how much average interview time rises when household use of �nancial services is higher. For example, the full enumeration product-based format in questionnaire 1 took 20 –30 percent more time to administer when members of the household used banking or insurance services than when they did not (see table 12). Discernible throughout this article is a concern about the ability to general- ize beyond Ghana. While there is a strong undercurrent of common sense to Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 the main �ndings, which are thus likely to be relevant in other countries as well, the article is speci�cally about Ghana. And while Ghana might be an ade- quate reflection of low-income countries in much of Sub-Saharan Africa, it is unlikely to be reflective of the whole developing world. The best that can be done in the context of this article is simply to acknowledge this limitation. In future, however, this type of experiment can be repeated in other countries. A similar study in Timor Leste, where very few respondents use any �nancial ser- vices, found that the differences across treatments were not signi�cant, suggesting that the concerns raised in this analysis are of second order impor- tance in the most �nancially underdeveloped countries. We live in a world of rough approximation when it comes to measuring the outreach of the �nancial systems of developing countries. The reliability of esti- mates from accounts-based approaches and approaches that meld accounts- based and survey-based information through regressions is dif�cult to assess. The hope is that the results here provide some practical guidance on how to generate comparable �nancial use data across countries through surveys, which appear to represent the best vehicle for generating accurate data. APPENDIX. CONSTRUCTION O F I N D I CATO R S FROM PRODUCT-LEVEL QUESTIONS Banked: Q2 ATM card Q4 Debit card Q6 Savings Plus account Q8 Current account Q10 Savings account at bank Q12 PostBank account, post of�ce savings account Q36 Bank loan Q54 Bank overdraft facility Indirect: Q16 Use of someone else’s account 232 THE WORLD BANK ECONOMIC REVIEW Formal savings: Q6 Savings Plus account Q10 Savings account at bank Q12 PostBank account, post of�ce savings account Q14 CDs, treasury bills, notes, money market funds Q22 Savings with regulated micro�nance institution Q24 Savings with credit union Q30 Shares, investment funds Q32 Provident fund Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 Q34 Pensions fund Informal savings: Q26 Susu scheme Q28 Welfare scheme, other savings club (e.g., with religious organization). Formal credit: Q36 Bank loan Q38 Loan from government Q40 Loan from credit union Q42 Loan from micro�nance institution Q44 Loan from employer Informal credit: Q48 Welfare scheme, Susu, savings club Insurance: Q60 Vehicle Q62 Property Q64 Homeowners Q66 Debts Q68 Travel Q70 Life Q72 Debts if you die Q74 Disability from employer Q76 Other disability Q78 Professional Q80 Funeral policy with institution Q84 Health/medical Q86 Children’s education REFERENCES ¨c Beck, Thorsten, Asli Demirgu ¸ -Kunt, and Ross Levine. 2000. “A New Database on the Structure and Development of the Financial Sector.� World Bank Economic Review 14 (3): 597– 605. ———. 2007. “Finance, Inequality, and the Poor.� Journal of Economic Growth 12(1): 27– 49. Beck, Thorsten, Ross Levine, and Norman Loayza. 2000. “Finance and the Sources of Growth.� Journal of Financial Economics 58 (1 –2): 261–300. Robert Cull and Kinnon Scott 233 ¨c Beck, Thorsten, Asli Demirgu ¸ -Kunt, and Maria Soledad Martinez Peria. 2007. “Reaching Out: Access to and Use of Banking Services across Countries.� Journal of Financial Economics 85(1): 234–66. Clarke, George R.G., L. Colin Xu, and Heng-Fu Zou. 2006. “Finance and Income Inequality: What Do the Data Tell Us?� Southern Economic Journal 72(3): 578–96. Cull, Robert, and Kinnon Scott. 2009. “Measuring Household Usage of Financial Services : Does It Matter How or Whom You Ask?� Policy Research Working Paper 5048. World Bank, Washington, D.C. Gasparini, Leonardo, Federico Gutierrez, A Tamola, L. Tornarolli, and Guido Porto. 2004. “Finance and Credit Variables in Household Surveys of Developing Countries.� Centro de Estudios Distributivos, Laborales, y Sociales, Universidad Nacional de La Plata, Argentina, and World Bank Development Economics Research Group, Washington, D.C. Downloaded from wber.oxfordjournals.org at International Monetary Fund on September 15, 2010 GSS (Ghana Statistical Service). 2006. “Sample Design for Round Five of the Ghana Living Standards Survey (GLSS 5).� Accra. Honohan, Patrick. 2004. “Financial Development, Growth, and Poverty: How Close Are the Links?� In Financial Development and Economic Growth: Explaining the Links, ed. C. Goodhart. London: Palgrave. ———. 2008. “Cross-Country Variation in Household Access to Financial Services.� Journal of Banking and Finance 32(11): 2493– 500. Jolliffe, Dean. 2001. “Measuring Absolute and Relative Poverty: The Sensitivity of Estimated Household Consumption to Survey Design.� Journal of Economic and Social Measurement 27(1): 1–23. Kochar, Anjali. 2000. “Savings.� In Designing Household Survey Questionnaires for Developing Countries: Lessons from 15 Years of the Living Standards Measurement Study, ed. Margaret Grosh, and Paul Glewwe. Washington, D.C.: World Bank. Christen, Robert P., Veena Jayadeva, and Richard Rosenberg. 2004. “Financial Institutions with a Double Bottom Line: Implications for the Future of Micro�nance.� CGAP Occasional Paper 8. Consultative Group to Assist the Poor, Washington, D.C. Levine, Ross. 2005. “Finance and Growth: Theory and Evidence.� In Handbook of Economic Growth, ed. Philippe Aghion, and Steven Durlauf. Amsterdam: Elsevier Science. Levine, Ross, Norman Loayza, and Thorsten Beck. 2000. “Financial Intermediation and Growth: Causality and Causes.� Journal of Monetary Economics 46 (1): 31– 77. Levine, Ross, and Sara Zervos. 1998. “Stock Markets, Banks, and Economic Growth.� American Economic Review 88 (3): 537– 58. Peachey, Stephen, and Allan Roe. 2006. “Access to Finance: Measuring the Contributions of Savings Banks.� Working Paper. World Savings Banks Institute, Brussels. Pradhan, Menno. 2001. “Welfare Analysis with a Proxy Consumption Measure: Evidence from a Repeated Experiment In Indonesia.� Cornell Food and Nutrition Policy Program Working Paper 126. Cornell University, Ithaca, N.Y. Rajan, Raghuram G., and Luigi Zingales.1998. “Financial Dependence and Growth.� American Economic Review 88 (3): 559– 86. Scott, Kinnon. 2000. “Credit.� In Designing Household Survey Questionnaires for Developing Countries: Lessons from 15 Years of the Living Standards Measurement Study, ed. Margaret Grosh, and Paul Glewwe. Washington, D.C.: World Bank. STATIN (Statistical Institute and Planning Institute of Jamaica). 1994. “Survey of Living Conditions, 1994.� Kingston. Steele, Diane, 1998. “Ecuador Consumption Items.� World Bank, Development Economics Research Group, Washington, D.C. World Bank. 2007. Finance for All? Policies and Pitfalls in Expanding Access. Policy Research Report. Washington, D.C.: World Bank.