Docstoc

Initially sociologists postulated that the proliferation of mobile

Document Sample
Initially sociologists postulated that the proliferation of mobile Powered By Docstoc
					Surveying the Socioeconomic Status of Microfinance Clients
Using the WHEEL Assessment Tool


John Klingler
Sivan Yosef
Timothy Stein




The Elliott School of International Affairs
George Washington University

FINCA International

May 5, 2008




For more information on this report, the team can be contacted at yosef.sivan@gmail.com
This paper has been produced by a team of students working with FINCA International as part of an International
Development Studies (IDS) Capstone Project. Neither FINCA nor any other entities thereof assumes any legal
liability or responsibility for the accuracy, completeness or usefulness of information, product or process disclosed
in this document. The views and opinions expressed in this paper are those of its authors and do not necessarily
state or reflect the opinion, policy, or practice of FINCA International or any entities thereof. This document is not
to be reproduced, distributed, or otherwise shared without the express consent of FINCA International.
Table of Contents

Executive Summary ......................................................................................................................... i
I. Introduction ................................................................................................................................. 1
II. Background ................................................................................................................................ 2
    Length ......................................................................................................................................... 2
    Multidimensionality .................................................................................................................... 3
    Nomenclature .............................................................................................................................. 3
    Compatibility .............................................................................................................................. 3
III. Methods & Data ........................................................................................................................ 4
    CRM Pilot Findings .................................................................................................................... 4
    FCAT 2007 Results..................................................................................................................... 4
    Literature Review........................................................................................................................ 5
IV. The WHEEL Survey Instrument .............................................................................................. 6
    Parameters ................................................................................................................................... 6
    Innovations .................................................................................................................................. 6
    Concept ....................................................................................................................................... 7
V. Wealth ...................................................................................................................................... 10
    Designing the Model ................................................................................................................. 10
    Testing the Model ..................................................................................................................... 13
    Using the Model ........................................................................................................................ 14
VI. Health ...................................................................................................................................... 15
    Food Security ............................................................................................................................ 15
    Water and Sanitation ................................................................................................................. 16
    Self-Assessment of Health ........................................................................................................ 18
VII. Education ............................................................................................................................... 18
VIII. Empowerment ...................................................................................................................... 20
    Phone Usage.............................................................................................................................. 21
    Decision-making Power ............................................................................................................ 22
IX. Living Conditions ................................................................................................................... 22
X. Limitations ............................................................................................................................... 23
XI. Recommendations................................................................................................................... 25
    Additional Tests ........................................................................................................................ 25
    CRM Implementation ............................................................................................................... 25
    Panel Data ................................................................................................................................. 25
    WHEEL and FCAT................................................................................................................... 25
    FCAT Accuracy ........................................................................................................................ 26
Appendix A: Other Questions Considered ................................................................................... 27
Appendix B: Alternate 10-Question Version ................................................................................ 28
Works Cited .................................................................................................................................. 29
Executive Summary

The following paper presents the WHEEL: a 16-question, user-friendly living standards
assessment tool. The WHEEL is the culmination of a year-long collaboration between George
Washington University (GWU) and FINCA International that sought to address the question of
how FINCA could best measure the well-being of its microfinance clients as part of a broader
monitoring and evaluation initiative called Customer Relationship Management (CRM). The
paper discusses the evolution of the living standards CRM tool, from its conception to its present
form.

Initially, FINCA developed a pilot socioeconomic survey for the CRM platform in the fall of
2007, which the researchers field-tested in January of the following year in Uganda. Titled as the
Poverty Tool, this questionnaire proved to be unwieldy in length, occasionally inconsistent with
cultural contexts, and overly dependant on financial metrics of well-being. The researchers
therefore endeavored to design a new survey instrument that remedied these problems.

In order to address the length of the survey and make it universally applicable, the researchers
utilized innovative strategies in survey design to yield as much information as possible out of a
short question set. First, regression analysis was used to model daily per capita expenditure
levels based on data from the FINCA Client Assessment Tool (FCAT) administered in 2007. The
paper shows how the original 18-question expenditure section in the FCAT can be reduced to a
seven-question section capable of adequately estimating consumption levels. Secondly, for the
purposes of choosing non-wealth indicators of well-being, the researchers looked to best
practices in the field to find cross-functional questions that can yield information on different
areas of a person‘s life. The source of one‘s drinking water, for example, may serve as an
indicator for living conditions as well as health quality. Finally, the WHEEL has been designed
to take advantage of the longitudinal nature of the CRM. With the same clients interviewed at
multiple points in time, questions will become more instructive.

The WHEEL derives its name from the five components of socioeconomic status that it
measures:

     Wealth: Clients‘ consumption levels.
     Health: Clients‘ physical health, risk for environmental disease, and food security.
     Education: The percentage of children attending school and amount spent per child.
     Empowerment: Elements of clients‘ social capital stocks and control over their lives.
     Living Conditions: The comfort and security of clients‘ dwellings.


The importance of each of these components and the indicators chosen to measure them are
explained in greater detail throughout the report.

The WHEEL shows that absolute and relative socioeconomic levels can be adequately
measured in 16 questions, thus fulfilling a vital part of FINCA‘s CRM strategy. Through its
use, the authors believe that FINCA can gain a more nuanced understanding of how its
services affect client well-being in order to make well-informed management decisions.


                                                i
I. Introduction

Over the last few decades, the microfinance movement has exploded in popularity. Microfinance
institutions (MFIs) can now be found in almost every corner of the globe, offering financial
services to low and middle-income segments of the population. What began as a noble cause to
serve the poor has matured into a hyper-competitive business environment, in which non-
government organizations compete alongside commercial banks for client acquisition and
retention. The result is undeniably positive—more people now have access to credit and savings
instruments than ever before. However, for microfinance institutions, this increased competition
has heightened the need for program efficiency and effectiveness. In order to succeed, MFIs need
to better understand their clientele and tailor products to suit their needs.

Alongside a competitive environment, social accountability has served as an added impetus for
improvement. While the microfinance industry has adequately focused on financial
accountability, the last decade has seen a strong push by MFIs to also measure the social impact
of their services. Many organizations and donors are now seeking to confirm that microfinance is
indeed benefiting the world‘s poor. In 2000, for example, the U.S. Congress passed legislation
that mandated that half of all United States Agency for International Development (USAID)
microfinance funds benefit the very poor. In order to verify that they are in fact reaching this
goal, USAID implementing partners are now required to use social assessment tools to report the
poverty level of their clients. Most MFIs continue to search for the best methods for measuring
the financial and social performance of their clients

Toward this end, FINCA International has begun to implement a Customer Relationship
Management (CRM) platform. The CRM platform is envisioned as a comprehensive monitoring
and evaluation tool that will provide management with increased awareness of client behavior to
better inform business decisions. Through extensive surveying of FINCA clients, the CRM will
collect data in three core areas: a ―client well-being‖ component that evaluates individuals‘
socioeconomic status; a ―client loyalty‖ component to assess clients‘ attachment to the FINCA
brand, as well as their willingness to remain with FINCA when faced with alternative options;
and a ―client retention‖ component to better understand why clients exit the FINCA system.

In the fall of 2007, FINCA began to plan and design the CRM initiative. During this initial stage,
the goals were to draft three short survey instruments—named Poverty, Loyalty, and Retention—
and to test their viability before launching a CRM trial implementation phase in late 2008.
Researchers from the George Washington University (GWU) were contracted to administer the
surveys in a pilot setting, analyze the results, and ultimately refine the client well-being
component (i.e. the Poverty survey instrument) to ensure successful future implementation.

During the pilot administration, the researchers found the initial Poverty survey to be
incompatible with the goals of the CRM initiative and in need of a redesign. The team therefore
set about constructing an improved survey instrument that would better suit the needs of FINCA.
The result is a novel approach to a socio-economic survey, dubbed the WHEEL. Its conception,
methodology, and utility are presented in the following pages.




                                                1
The paper is structured as follows: Section 2 describes the conception of the original Poverty tool,
its testing in the field, and the lessons learned from the experience; Section 3 explains the
methodology employed in designing the new survey instrument; Section 4 presents the survey
instrument itself, and a brief introduction into its application; Sections 5-9 break the WHEEL up
into its component parts—Wealth, Health, Education, Empowerment, and Living Conditions—
and provide a question-by-question explanation as to why each indicator was chosen; finally, the
paper concludes with limitations and recommendations for using the WHEEL in the future.



II. Background

The preliminary version of the CRM survey instrument to measure client welfare was developed
internally by FINCA in the fall of 2007. Initially called the Poverty Tool, this instrument
consisted of approximately 50 questions about clients‘ living standards and income levels.
Roughly half of the questions elicited information about the household, including the full
household roster (the sex, age, and level of education of every member of the household). The
second half of the survey elicited information about consumption, including a lengthy section on
household expenditures. Two additional questions asked about household illness and computer
ownership.

From January 7-18, 2008, the members of the GWU Capstone Team traveled to Uganda to
participate in the CRM pilot. Over these two weeks, the researchers completed 121 interviews
using the Poverty tool. In addition, the researchers conducted a focus group with clients who had
taken the Poverty survey, in order to follow up on confusing questions and receive feedback on
the instrument itself.

From the field tests and focus group, the research team concluded that the Poverty tool was
poorly designed for the purposes of the CRM. Its great length quickly fatigued both the
surveyors and the interviewed clients. Additionally, many of the questions contained in the
poverty tool were redundant, occasionally inconsistent with particular cultural contexts, or did
not yield sufficiently rich data at the analysis phase.

The GWU team proposed to create a re-conceptualized survey instrument that would remedy the
shortcomings identified in the pilot. The new survey instrument would make improvements in
the following four areas:


Length
The survey tested in the field took, on average, 30 minutes to conduct. In order for the survey to
be implemented as a component of the CRM, its length would have to be substantially shortened
so that it would not drain resources or trouble clients and staff. In consultation with FINCA staff,
it was decided that the survey instrument should take no longer than 10 minutes to administer.




                                                 2
Multidimensionality
The pilot poverty tool concentrated heavily on clients‘ consumption levels, at the expense of
other aspects of one‘s standard of living—such as vulnerability, food security, and social capital.
Since FINCA‘s products and services potentially impact various aspects of clients‘ livelihoods, it
was decided that the tool must attempt to capture other dimensions of socioeconomic status. The
revised tool would have to pivot away from assessing ―poverty‖ – which often simply denotes
financial income and wealth – and instead move towards a more comprehensive notion of well-
being.

Nomenclature
In the field, the researchers abstained from referring to the Poverty questionnaire by its actual
name. Conversations with other field testers and feedback from the focus group confirmed the
team‘s suspicion that it may be derogatory to ask clients about their ―poverty‖ levels. Besides the
fact that many clients would not be considered poor relative to their neighbors, the term poverty
itself is unnecessarily negative. It was therefore decided that the survey instrument should have a
new name reflective of its purpose.

Compatibility
FINCA clients live in 21 different countries. Therefore, if the CRM is to be administered across
the portfolio, the new survey instrument would need to focus on universal characteristics of
socioeconomic status that can be found in most countries and cultures. The pilot tool, for
example, tested if clients owned a computer. Although such a question may have been relevant
and insightful to ask in Kyrgyzstan, it was an unsuitable question for clients in Uganda, who
lacked the financial means to own a computer. The improved survey instrument would have to
assess indicators of well-being that were applicable to all countries and contexts, an especially
challenging task considering FINCA operates in four distinct regions.




                                                 3
III. Methods & Data

To design a more efficient survey instrument, the researchers drew from three sources: 1)
quantitative and qualitative findings from the CRM pilot testing, 2) results from FINCA‘s 2007
FCAT survey, and 3) an expansive literature review.

CRM Pilot Findings
The team used both qualitative and quantitative data from the CRM pilot in Uganda. While in-
country, the researchers conducted a focus group of FINCA clients who had taken the pilot
survey. To supplement these qualitative findings, the researchers also performed quantitative
analysis on the CRM pilot data to examine correlations between questions. Primarily, the team
ran OLS regression analysis to examine the relationship between various status indicators and a
client‘s time with FINCA. The approach, however, was ultimately abandoned for two reasons.
First, it was difficult to establish causality in these relationships. Second, analysis was hampered
by the manner in which the pilot surveys were administered. Different tools were administered to
different clients, hindering the team‘s ability to link the quality of life of clients to their
relationship with FINCA. The most informative data ended up being simple descriptive statistics
on loan use and program awareness.

FCAT 2007 Results
FINCA‘s Client Assessment Tool (FCAT) proved to be a valuable resource in the design of the
improved socioeconomic tool. The FCAT is a comprehensive survey that has been administered
to a sample of FINCA clients on a bi-annual basis for over a decade. The 2007 version of the tool
consisted of 119 questions, capturing both individual- and household-level data. The researchers
undertook the large task of compiling and cleaning a dataset of eight countries—Ecuador,
Guatemala, Malawi, Mexico, Nicaragua, Russia, Uganda, and Zambia—where the FCAT was
administered in 2007. Due to the presence of extreme outliers, the researchers limited the
analysis to the middle 90% of the distribution for each country, as measured by daily per capita
expenditure.

Using this dataset, the team tested the hypothesis that one can estimate a client‘s total yearly
expenditures based on a subset of other questions. The goal was to generate a model to predict
consumption using fewer questions than found in the FCAT. A similar strategy was employed by
Prisma, who created a short poverty questionnaire by selecting predictive variables from an
established benchmark, in this case the 2002 Living Standards Measurement Survey (LSMS) in
Bosnia-Herzegovina (Schreiner, et al, 2005). The use of regression analysis to derive and
calibrate indicators against a benchmark of absolute poverty was suggested by the literature
review found in AMAP (2004).

In addition to OLS regression, the researchers conducted a Principal Component Analysis (PCA)
as a secondary statistical tool. PCA is a commonly used wealth-ranking method that compresses
large datasets by eliminating redundant variables that measure similar phenomena. Essentially,
the first-principal component is a linear index of a limited number of variables (10 to 15) with


                                                 4
the largest amount of information common to all the variables. During the PCA process, all of
the FCAT variables were measured for both their level of correlation and strength of their linear
association with the first component, the relative size of their communalities, and their
contribution to the Kaiser-Meyer-Olkin index. The resulting linear index of the winning 15
variables was converted into a relative poverty index that could potentially rank clients from
lowest to highest socioeconomic level. Unfortunately, the resulting index showed little
correlation with total expenditures or time with FINCA, and hence was discarded. The
researchers believe though that once CRM data from the trial implementation phase has been
collected, it would be a useful exercise to revisit this issue and see whether an index created
through PCA would be a viable tool in the future (see Zeller, et al, 2001.)

Literature Review

To help select the indicators to use in the study, the researchers augmented the statistical analysis
with a comprehensive literature review of which indicators are widely used in the field and why.
Some oft-used resources included publications from USAID, the World Bank, WHO, UNICEF,
Povertytools.org, IFPRI, as well as past reports and publications from FINCA. These references
are indicated in the Works Cited section of the paper. Appropriate well-being indicators were
judged based on their place in best practices, their appropriateness for FINCA clients and
countries, their ease of use by FINCA field staff, their ease of comprehension by FINCA clients,
and the usefulness of the data they would yield for CRM as a whole. Where possible in this
paper, the authors have explained why a certain indicator was chosen over another.




                                                  5
IV. The WHEEL Survey Instrument

To fulfill the mandate of the CRM, the researchers from GWU have designed a compact
socioeconomic survey that can be administered without undue stress on staff resources. The team
has given the instrument the name WHEEL for its potential to provide rich information in five
categories: Wealth, Health, Education, Empowerment, and Living Conditions. The following
section explains the methodological choices made in the creation of this instrument and the
innovations that allow for a short survey to capture useful data.

Parameters
The design of an improved, socioeconomic survey instrument began by establishing the
parameters that needed to be observed to ensure compatibility with the CRM initiative. Based on
field research and consultations with staff, the team deemed there were two crucial requirements
that had to be met:

      The survey must take 10 minutes or less to administer.
      The survey must be applicable to all country contexts.

These requirements prevented the WHEEL from utilizing several strategies commonly found in
household surveys. First, the researchers could not include a ―household roster‖ in the redesigned
survey; this would have required interviewers to ask about the age, sex, and education level of
every member in the household. In developing countries where household size is rather large,
this alone could take more than 10 minutes. Second, the team could not include a full
expenditure section. The minimum set of questions in a full expenditure breakdown includes 18
items, so this was deemed unfeasible. Finally, the team could not include an expansive asset
section. Asking about client assets is an effective way to gauge well-being. However, assets as
indicators of socio-economic status are not compatible across countries—a rice cooker may be
significant in Uganda, for example, but not in Russia.

Innovations
To overcome these limitations, the researchers decided that every attempt would be made to
obtain as much information as possible out of every question on the survey. In particular, three
innovative strategies allowed the researchers to make a small survey more powerful.

   1. Consumption levels can be modeled with reasonable accuracy. Using statistical
      analysis, the researchers discovered that it is possible to determine a client‘s total yearly
      expenditures using a sub-set of expenditure questions found on the FCAT. This, in turn,
      allows the WHEEL survey instrument to serve as a predictor of a client‘s daily per capita
      expenditure level. Once adjusted for purchasing power parity (PPP), this figure allows for
      comparisons of client income across countries (the details of these findings are presented
      in the following section).




                                                6
   2. Indicators can serve double functions. Although one question will produce one answer
      and, thus, one data point, some single answers can be interpreted in multiple ways to
      inform about different topics. For example, a question about the nearest water supply is
      both an indicator of living conditions and an indicator of health status (since disease is
      often carried through unclean water.) Wherever possible, the researchers sought to
      include questions that can be reused in subsequent analysis, thus increasing the
      questionnaire‘s utility.

   3. Certain indicators become more valuable than others over time. It is envisioned that
      the CRM will involve interviews with the same clients at multiple points in time. This
      offers an opportunity to capture changes in clients‘ welfare in a new way, and alters what
      should be asked. Questions that might have limited utility in a cross-sectional, one-time
      study could possibly be very useful when tracked over time. For example, a question
      rating a person‘s level of empowerment will probably not yield very good results if asked
      just once, as there are too many cultural considerations that must be accounted for in
      order to understand the response. However, the same question asked at multiple points in
      time will show rises and falls, indicating relative improvements or setbacks in welfare.
      Linking a rise or fall in a client‘s self-assessment with the availability or lack of certain
      services can allow an MFI to evaluate the role of its products vis-à-vis a client‘s level of
      well-being.

Concept
Using these three strategies, the GWU researchers designed a survey that effectively addresses
the multiple facets of socioeconomic status in just 16 questions. The team named the tool the
WHEEL for the five components of welfare it captures:

      Wealth: Clients‘ consumption levels.
      Health: Clients‘ self-assessment of their physical health, their level of risk for
       environmental disease, and their level of food security.
      Education: The number of school-age children attending school, and the amount of
       money spent on education.
      Empowerment: Elements of clients‘ social capital stocks and personal control.
      Living Conditions: The comfort and security of clients‘ dwellings.

The full text of the survey can be found in the following pages. Following its presentation, the
researchers discuss each of the components in turn, explaining which questions on the survey
relate to the concept, and why each was chosen.




                                                 7
Pages 8-9 left intentionally blank. The actual WHEEL Assessment Tool is considered to
                 be the proprietary information of FINCA International.




                                         8
Pages 8-9 left intentionally blank. The actual WHEEL Assessment Tool is considered to
                 be the proprietary information of FINCA International.




                                         9
                                                   V. Wealth

An essential component of an individual‘s well-being is his or her capacity to purchase the goods
and services necessary for a full life. In this paper, the researchers use wealth to refer to the level
of financial capital at one‘s disposal; it may be thought of as the amount of money somebody
has, or their stock of financial capital.

Most assessments of economic welfare either measure income (the amount one earns) or
consumption (the amount one spends.) Sound theoretical arguments can be made for both
approaches; however, in developing countries, consumption-based surveys are often the more
practical approach (Deaton and Grosh, 2000). First, the concept of an expenditure—giving
money for a good or service—is easier to understand than income, especially for agricultural or
self-employed workers who often combine personal and business accounts. Secondly, for
seasonal workers, income can be variable, while consumption tends to be smoothed over the
season, making consumption the better measure for a periodic survey. Lastly, individuals are
more reluctant to disclose their income, especially in a public setting. Hence, people are less
likely to lie about their consumption.

Traditional consumption-based surveys ask individuals how much the household spends during a
set period of time (weekly, monthly, or annually) on a long list of consumption items (food, rent,
clothes, etc.)1 All of these expenditures are summed and then divided by the number of people in
the house and the number of days in a year. The resulting figure is daily per capita expenditures,
or DPCE.

DPCE is the mostly widely accepted and used
                                                                  Wealth Indicators:
metric of individual wealth within the literature.
                                                                    Q7: Food Expenditures
National and regional ―poverty lines‖ within the
                                                                    Q8: Transport Expenditures
developing world are most often expressed in
                                                                    Q9: Utilities Expenditures
DPCE terms, such as the oft-heard US $1 per day
                                                                    Q10: Value of Food Production
level. Governments routinely conduct household
                                                                    Q11: Home Expenditures
surveys to calculate this figure, as does the World
                                                                    Q14: Clothing & Shoes Expenditures
Bank through its exhaustive Living Standards
                                                                    Q15: Education Expenditures
Measurement Survey (LSMS).

Designing the Model
Though the utility of having DPCE estimates for FINCA clients was apparent from the start, the
researchers faced the challenge of how to accomplish this in a short-form questionnaire.
Household consumption surveys are difficult to administer and lengthy in nature—in direct
contrast to the speed and ease envisioned for this instrument. For example, the FCAT calculates
a client‘s consumption level by asking for 18 distinct expenditure estimates. If the WHEEL had
included all of these expenditure questions, the survey would have exceeded its mandated length
without even capturing one non-economic measure of well-being.

1
 The potential difficulty of recalling such expenditures is a challenge. Zeller (2001) notes that this problem can be
minimized through extensive training of interviewers and the selection of appropriate recall periods for the survey.


                                                          10
Realizing this constraint, the researchers attempted to see if the expenditure section of a
household survey could be reduced with minimum loss to its ability to estimate DPCE. The team
used the FCAT datasets as the benchmark for the analysis, since the surveys had been recently
conducted and represented numerous countries in different regions.

The 18 expenditure questions included in each of the FCAT surveys were:

           Charity                             Food Produced                       Personal Products
           Clothing & Shoes                    Furniture                           Rent
           Cooking Fuel                        Health Special                      Events
           Education                           Home Improvement                    Taxes & Bribes
           Family & Friends                    Leisure                             Transport
           Food Purchased                      Other                               Utilities

The team began by examining the correlations between each of these expenditure components
and the total yearly expenditure sum. All figures were adjusted for purchasing power parity
(PPP) to enable comparisons across countries. Additionally, the researchers examined the
correlations between non-expenditure questions—such as loan size, access to water, and asset
ownership—and total expenditures to see if there existed any discernable relationships.

The researchers then used OLS regression to build consumption models using the components
identified as highly predictive. Hundreds of regressions were run with different sets of variables,
continuously adding or replacing variables to find the best overall model-fit and the highest
levels of statistical significance on the coefficients.

In the end, the ―winning‖ model consisted of seven expenditure questions: home improvement,
utilities, food purchases, food production, clothing, education, and transport.2 The results of the
regressions are presented in Table 1. As shown, the team ran regressions for each country using
this consumption model, as well as one ―universal‖ regression against the complete dataset with
dummy variables representing each country.

As expected, the majority of coefficients are around one or above, and positive. This can be
interpreted to mean that when one‘s expenditure increases, there is a corresponding increase in
total yearly expenditures. For example, looking at the coefficient in the top-left corner, one can
say: In Ecuador, a $1 increase in home expenditures correlates with a $1.20 increase in total
yearly expenditures, holding all else constant.

The model displays a high R2 indicating good model-fit, and most coefficients are statistically
significant at the .001 level. However, neither of these two measures are sufficient proof that the
model is useful. For this purpose, the team needed to ascertain whether the seven expenditure
questions included in the model can adequately predict the DPCE level of individual clients.




2
    These findings are consistent with other international tools. See China Statistics Census (2005) and UN (2005).


                                                           11
Table 1: OLS Regressions with Yearly Household Expenditures as the Dependent Variable

                                   Ecuador         Guatemala           Malawi          Mexico          Nicaragua          Russia          Uganda           Zambia          8 Countries
Home Expenditures                  1.199***         1.116***          1.022***         1.184***          1.02***         0.889***         1.167***         1.130***          0.971***
                                    (6.49)           (9.52)           (23.25)          (14.59)          (11.17)          (19.09)          (16.83)           (6.56)           (24.52)
Utilities Expenditures             1.632***         1.196***          1.421***         1.522***         1.289***         0.320***         1.626***         1.202***          1.286***
                                    (9.74)           (8.67)            (8.66)            (9.5)           (8.49)           (1.12)           (9.62)          (11.23)           (13.08)
Food Expenditures                  1.022***         1.089***          1.136***         1.130***         1.040***         1.346***         1.226***         1.049***          1.216***
                                   (18.07)          (22.16)           (12.62)          (24.45)          (23.83)           (9.99)          (13.72)          (13.42)           (19.61)
Household Food Production          0.966***          0.878            1.112***         0.911***         0.983***          0.064           1.142***         0.726**           1.021***
                                    (3.81)           (1.73)            (13.8)          (14.89)          (11.64)           (0.17)          (17.09)           (2.59)           (13.89)
Clothing Expenditures              1.713***         2.141***          3.189***         2.193***         2.746***         1.633***         1.368***         3.007***          1.715***
                                    (8.66)           (7.05)            (4.82)           (6.89)           (3.81)           (7.04)          (16.38)           (4.67)           (10.01)
Education Expenditures             1.021***         1.121***          1.255***         1.135***         1.289***         0.946***         1.065***         1.401***          1.139***
                                    (8.07)           (5.24)            (6.58)          (11.79)           (8.31)           (2.88)          (23.39)          (12.11)           (23.49)
Transport Expenditures             1.097***         1.243***          1.045***         1.072***         1.187***         1.126***         1.132***         1.344***          1.147***
                                   (12.45)          (13.19)           (16.08)          (23.93)          (13.21)           (5.94)           (8.32)          (10.05)           (16.85)
Guatemala                                                                                                                                                                  505.931***
                                                                                                                                                                              (3.58)
Malawi                                                                                                                                                                     664.246***
                                                                                                                                                                              (4.23)
Mexico                                                                                                                                                                      358.872**
                                                                                                                                                                              (3.15)
Nicaragua                                                                                                                                                                   632.456*
                                                                                                                                                                              (2.38)
Russia                                                                                                                                                                     6166.772***
                                                                                                                                                                             (10.16)
Uganda                                                                                                                                                                     1217.674***
                                                                                                                                                                               (5.4)
Zambia                                                                                                                                                                      341.985*
                                                                                                                                                                              (1.99)
Constant                          850.591***       1251.495***       701.354**        652.401***      1612.772***      7876.664***      1127.105***       626.166***         347.213
                                    (3.87)           (5.29)            (2.66)           (3.32)          (4.31)           (6.51)           (3.31)            (4.63)            (1.51)
N                                    297               288              275              715              300              270              291              272               2708
R-squared                           0.852             0.906            0.963            0.907            0.846            0.746            0.946            0.933             0.899

Notes: T-statistic shown in parenthesis, calculated using robust standard errors. All expenditures are PPP adjusted and expressed as yearly values. The seven country variables are
dummies (1 = is the country, 0 = not the country); Ecuador is left out.
* Significant at the 0.05 level.
** Significant at the 0.01 level.
*** Significant at the 0.001 level.



                                                                                         12
Testing the Model
To test the predictive power of the model, the team needed to see how closely the predicted
DPCE figures match the ―real‖ DPCE figures. For each client in the dataset, the researchers ran
the figures for the seven expenditure questions through the model to generate the predicted total
yearly value and then divided this figure by the number of household members and 365 (for days
in a year) to generate the predicted DPCE. They then set about comparing these estimated values
against what they knew to be the actual DPCE values calculated using all 18 of the original
expenditure figures.

For the first test, the researchers examined if the estimated distribution of clients matched the
actual distribution of clients. They split DPCE levels into six groups: $0-$1.99, $2-$3.99, $4-
$5.99, $6-$7.99, $8-$9.99, and $10 and above. (These groups were recommended by FINCA
staff to be consistent with previous reporting.) The team then counted the predicted frequency of
members and compared it against the actual frequency of members.

Table 2: Predicted Group Dispersion vs. "Actual" Group Dispersion in Sample
                            Ecuador           Guatemala                  Malawi                Mexico
                     Sample     Predicted   Sample    Predicted   Sample    Predicted   Sample    Predicted
   DPCE               Freq        Freq       Freq       Freq       Freq       Freq       Freq       Freq
   $0 - $1.99          11             4       1             3       44            35      4             4
   $2 - $3.99         104         113         71            61      93            100    180        178
   $4 - $5.99          90             90      83            90      54            63     239        234
   $6 - $7.99          48             54      60            64      35            31     136        129
   $8 - $9.99          25             18      39            33      22            12      66            96
   $10 and up          19             18      34            37      27            34      90            74



                        Nicaragua                  Russia                Uganda                Zambia
                     Sample     Predicted   Sample    Predicted   Sample    Predicted   Sample    Predicted
   DPCE               Freq        Freq       Freq       Freq       Freq       Freq       Freq       Freq
   $0 - $1.99           0             0       0             0       13            5      135        134
   $2 - $3.99          29             15      0             0       73            79      89            88
   $4 - $5.99          59             69      0             0       52            49      33            35
   $6 - $7.99          52             61      0             1       55            52      12            14
   $8 - $9.99          53             46      8             3       25            31      3             1
   $10 and up         107         109        262            266     73            75      0             0



As shown, the model predictions of the DPCE dispersion are generally accurate. For example, in
Ecuador, it is predicted there will be 104 clients between $2-$3.99; there are 113. It is predicted
there will be 90 clients between $4-$5.99; there are exactly 90. It is predicted there will be 48
clients between $6-$7.99; there are 54.

However, the short-form model‘s ability to predict the dispersion of clients may be specious.
After all, if a client from $2-$3.99 was wrongly placed in $4-$5.99, but a client from $4-$5.99


                                                     13
was also wrongly placed in $2-3.99, then the frequencies of each group would still be right
although the model had been twice incorrect.

To remove this possibility and to test just how well the model can predict expenditure levels, the
team predicted the DPCE grouping for each client using the seven-question form and compared
this DPCE grouping using the full 18-question form. If the model‘s estimation was the same as
the actual grouping, this was considered a successful prediction. If the model‘s estimation
differed from the actual grouping, this was considered a failed prediction. The table below
presents the results:

Table 3: Success Rate of Predicting an Individual's DPCE Group
                         Successful Predictions         Total               Success Rate
    Ecuador                       222                    297                   74.75%
    Guatemala                     200                    288                   69.44%
    Malawi                        204                    275                   74.18%
    Mexico                        507                    715                   70.91%
    Nicaragua                     211                    300                   70.33%
    Russia                        259                    270                   95.93%
    Uganda                        211                    291                   72.51%
    Zambia                        243                    272                   89.34%
    Total                        2057                   2708                   75.96%

Rounding up, it can be said that the seven-expenditure question model can predict DPCE levels
with 76% accuracy. This is a very positive result considering that the expenditure section was
originally 18 questions. In other words, the question set has been reduced by 61%, but the
accuracy only by 24%.

Using the Model
Through this exercise, the researchers have shown that a household questionnaire can be
shortened to its essential components using statistical analysis. Only seven questions are
necessary to estimate total yearly expenditures, and the OLS regressions have yielded equations
that can now be used in the field to calculate DPCE.

When conducting surveys, FINCA researchers enter client responses into Palm Pilots. These
Palm Pilots can be easily programmed to compute DPCE using the seven-question model
proposed in this paper. For example, using the OLS regression results provided in Table 1, the
equation for Ecuador would be:

DPCE = [850.591 + 1.199 (Home Expenditures) + 1.632 (Utilities) + 1.022 (Food Purchase) +
       0.966 (Food Production) + 1.713 (Clothing) + 1.021 (Education) + 1.097 (Transport)]
       / (Household Members) / 365




                                                  14
VI. Health

Health is an invaluable indicator of well-being that is essential for the CRM to measure. It is
defined by the World Health Organization as ―a state of complete physical, mental and social
well-being and not merely the absence of disease or infirmity‖ (World Health Organization,
2008). It is among the most basic and essential elements of well-being and, to many, a
fundamental human right. In the broadest sense, the WHEEL is in fact a tool that measures one‘s
overall level of health, or well-being. The ‗health‘ segment of the WHEEL, however, is more
narrowly concerned with one‘s physical well-being, including the level of risk for environmental
disease and illness.

Quality of health has been shown to be a critically
significant factor in socioeconomic development        Health Indicators:
(Sen 1998) and can lead to stronger economic               Q5: Drinking Water Source
productivity (Strauss and Thomas 1998). By                 Q6: Toilet Facility
coupling the Loyalty and WHEEL survey, the                 Q7: Food Expenditures
CRM would allow FINCA to observe, over                     Q11: Food Production
various points in time, if changes in client               Q15: Health Self-Assessment
business performance coincide with changes in
health performance. Healthier clients are more likely to be productive clients, and it would be
helpful to observe if loan performance is affected by this aspect of well-being.

In addition, FINCA potentially affords clients the opportunity to achieve better health status,
either directly through health insurance plans or indirectly through primary credit schemes. For
example, FINCA‘s products and services may provide the financial means for clients to access
better and cleaner sources of water, or to receive proper medical attention. Through the
measurement of health indicators, the effect of these types of services can be evaluated.

Food Security

The World Food Summit of 1996 defined food security as existing ―when all people at all times
have access to sufficient, safe, nutritious food to maintain a healthy and active life‖ (WHO 2008).
Clearly, food security is a vital component of overall health. In terms of food security, the
literature reviewed makes a distinction between process indicators, which pertain to food supply
and food access, and outcome indicators, which describe food consumption. In many studies, the
latter is favored because little correlation has been found between area-level food production and
food security at the household level (Hoddinott 1999). Additionally, many food security
indicators that are related to household supply and access to food, such as individual intakes,
household caloric acquisition, and dietary diversity, require a great amount of staff time and
resources to measure. For example, IFPRI reports that in order to measure food security
accurately, one would need to ask a household about a minimum of 40 different foods (Smith
and Subandaro 2007, 18). For this reason, these indicators were eliminated from consideration.
Another indicator, household coping strategies when faced with food shortage (such as
consuming less preferred foods or skipping meals) was actually concluded to be appropriate for




                                                15
this survey. However, while coping strategies may have highly predictive power, the accuracy of
this assertion is uncertain across countries (Christiaensen and Boisvert 2000, 19).

Although household food expenditures is a component of the WHEEL‘s absolute income
measurement, it can also be used as a measure of food security. The strength of food
expenditures as an indicator of food security is confirmed by a generally recognized inverse
relationship between food expenditures and food insecurity (Kennedy 2002). In this case, food
insecurity was measured more comprehensively and then graphed against the simpler measure of
food expenditures. The food expenditure question can also be interpreted as the percentage of
total household expenditures spent on food. IFPRI indicates that households that spend a large
portion of their income on food, in the range of 75% or more, are vulnerable to food insecurity
because ―regardless of their current food consumption status, if they were to experience a
reduction in income, it would likely be accompanied by a reduction in food consumption or the
quality of food eaten‖ (Smith and Subandaro 2007)

Asking clients about their household food expenditures poses some limitations. Within a country,
rural areas are characterized by the production of food within the household, rather than its
exchange or purchase in the marketplace. Thus, rural clients‘ monetary expenditure on food is
likely to be low, regardless of their level of food security. However, the WHEEL accounts for
this urban bias by also asking about household production of food. Thus, all food sources are
accounted for. Another limitation is that expenditures on food may vary across households due to
quality of food or the amount of food purchased at one time in bulk (ibid). While some surveys
attempt to overcome this obstacle by only asking about the price of foods that are fixed across
households, such as sugar and milk, the types of foods that are fixed differs greatly across
countries. Finally, it must be noted that there is no definitive relationship between food security
and nutritional outcomes. That is, that a household is food secure does not necessarily mean that
its members, especially children, do not suffer from nutritional deficiency (Bhattacharya and
Currie 2004).

Water and Sanitation
Improvements in water and sanitation also have a direct effect on health. Safe and accessible
water reduces the presence of pathogens in the environment, reducing the frequency of such
diseases as diarrhea, guinea worm, and skin diseases, and improving nutritional status (Billig et
al. 1999). Similarly, sanitation improvements, such as the move from pit latrine to flush toilets,
disrupt the fecal-oral transmission of disease, thereby dramatically improving a household‘s
health and productivity, especially in areas with low levels of education (Ibid). These types of
findings have been replicated in numerous settings (Bateman and Smith 1991).

While it is commonly believed that the primary benefit to overall health is water quality, water
quantity is just as important. Communities that consistently use larger quantities of water have
better health as they see increased opportunities for hand washing, food washing, and household
cleaning (Esrey et al 1991). Of special relevance to FINCA clients, increased water use can also
signify the use of water for income-generation or food production activities (i.e. gardening), and
the ease of access to water saves women time that is then used to prepare food, care for their
children, or improve their businesses (Bergeron and Esrey 1993; ICRW 1996).


                                                16
Major development organizations currently use various indicators to measure water and
sanitation. The World Health Survey uses ‗sustainable access to drinking water‘ and ‗sustainable
access to sanitation‘. The Demographic Health and Surveys (DHS) utilizes these whilst adding
‗time to water source‘ as an additional indicator, while UNICEF distinguishes between drinking
and other-use water, and includes ‗time to fetch water‘, ‗distance to water‘, and ‗water treatment‘
(UNICEF, MICS 2005).

The Joint Monitoring Program (JMP), a collaboration of these various organizations, represents
an effort to harmonize water supply and sanitation questions across commonly used surveys.
JMP recommends, and the WHO and DHS have accepted these terms, that access to improved
drinking water and sanitation sources be among the uniform questions asked in all health surveys
(JMP, 2008). The categories of improved and unimproved are presented below:

Improved Water Sources                            Unimproved Water Sources
Piped into dwelling                               Bottled water
Piped into yard or plot                           Unprotected dug well
Public tap                                        Unprotected spring
Tubewell/borehole with pump                       Pond, river or stream
Protected dug well                                Tanker-truck, vendor
Protected spring
Rainwater collection

Improved Sanitation Sources                       Unimproved Sanitation Sources
Flush to sewage system or septic tank             Open pit
Pour flush latrine (water seal type)              Bucket
Improved pit latrine                              No facilities or bush or field
Traditional pit latrine

There are two supplementary notes. First, bottled water is considered unimproved for the
aforementioned reason that while quality is satisfactory, quantity is considered to be limited
(Ibid). Second, although the Millennium Development Goals currently set the water benchmark
as being within the household (i.e. improved water sources are inside the dwelling) and the
sanitation benchmark outside the household, recent debate argues that the sanitation benchmark
should be set within the household as well.

The volume of water per capita per day collected by or delivered to the household for domestic
use was an indicator considered for the WHEEL assessment. While this indicator reflects the
importance of water quantity to health, and can yield accurate data for households that utilize
collected water, the FCAT 2007 data showed that 52% of FINCA clients use water that is piped
directly into the dwelling—for most clients, then, estimates of water use would be highly
inaccurate.

The main obstacle to collecting data on water and sanitation regards the nature of the CRM tools
as cluster surveys. Many improvements in water and sanitation are instituted at the community
level—as such, improvements in these areas must be understood as an improvement in the


                                                17
quality of life of the household, not necessarily as a reflection of a rise in income. This caveat,
however, does not detract from the importance of water and sanitation as vital components of
health in the WHEEL assessment tool. What is more, the FCAT 2007 data showed ‗toilet type‘
to be correlated strongly and significantly with DPCE. Future analysis of the CRM data should
take these observations into account.


Self-Assessment of Health
Admittedly, the most accurate way to measure one‘s health status is through observational
analyses (Gertler et al. 2000, 186). The WHEEL, however, is a limited, concise tool that is
unqualified to undertake such measures. Therefore, like most household surveys, the WHEEL
has respondents themselves report about their health status.

Idler and Benyamini (1997) examined 27 health studies worldwide and found that in all but four,
self-rated health is an independent predictor of mortality even after controlling for numerous
other variables. ―The findings are consistent and effect sizes are quite large; self-ratings of health,
which take only seconds to obtain in a survey interview, reliably predict survival in populations
even when known health risk factors have been accounted for.‖ (Ibid, 21)

The question eliciting the self-rating differs slightly from study to study, but all are approximate
variations of ―In general, how is your health at this time?‖ The WHEEL‘s space constraints do
not permit an extended self-assessment of various health indicators and therefore, this singular
question is used as a proxy for one‘s overall level of health. This is not the most precise
measurement, but it is a practical and useful way to get summary measures of health status that
are useful benchmarks for measuring changes in overall health over time (Gertler et al. 2000,
184).



VII. Education

As the World Bank so succinctly puts it, ―Education is development‖. Education comprises two
of the eight Millennium Development Goals. At the country level, education can increase labor
productivity, attract new capital, and substitute for physical capital where it is scarce. At the
household level, education—especially girls‘ education— equalizes opportunities and has a
proven impact on child and reproductive health, in terms of better nutrition, higher immunization
rates, and lower HIV/AIDS rates (World Bank,
2008). Studies have also shown that on the           Education Indicators:
global scale, every year of schooling increases         Q5: Education Expenditures
individual wages by approximately 10% (Ibid).           Q6: Number of School-Aged Kids
                                                        Q7: Number of Kids in School
As part of the team‘s fieldwork in Uganda, the




                                                 18
researchers administered the poverty tool to 121 clients, of whom 101 reported having school-
aged children3 living in the household. Most of these respondents reported spending a substantial
share of household expenditures on educating their children. Furthermore, for many clients, the
goal of loans themselves may be to provide their children with an education. In Uganda, for
example, 78 clients were asked to identify the goal of their loan, with 57 percent responding that
they used their loan to pay for school fees or to care for their children.

                                   “What was the goal of your loan?”

                                                              Ca re for Children
                                                                     17%

                             School Fe e s
                                40%
                                                                     De ve lop Se lf
                                                                         15%




                                             Other           Im prove Busine ss
                                              5%                   23%


                                 Results from CRM pilot in Uganda (n = 78)


Further analysis of the relationship between education-expenditures-per-child and a client‘s
tenure with FINCA suggested that clients who have been with FINCA longer report, on average,
higher expenditures on education per child. Causality in either direction could not be established,
but linking this type of data to the CRM loyalty survey may yield some insights on whether loans
increase the demand for education, conversely create opportunities that require children to help
their families with labor, and so on.

The education questions in the WHEEL, taken together, offer a nice synopsis of the educational
opportunities of a particular household. They determine the education expenditures per child as
well as the number of children who are provided the opportunity to receive an education. These
are the core, most important aspects of childhood education, and they are sufficiently captured by
the WHEEL.

The pilot poverty tool and the FCAT assess education by administering a lengthy sub-form that
asks the age, level of education, and literacy level of every household member. Neither tool,
however, asks whether all children in the household have the opportunity to receive education.
Assessing how many children have the opportunity to go to school does not only provide
valuable information about household education, but is also the simplest way of assessing
educational capabilities.


3
 Number of school-aged children was not a direct question asked on the Poverty survey. This figure was derived by
counting the number of people listed in the household who were between 5 and 18 years of age.


                                                       19
VIII. Empowerment

Empowerment is not limited to financial opportunities, but also the social ties and capabilities
that provide individuals and communities with the resilience to improve their well-being. This
type of empowerment is often known as social capital. Social capital has many definitions, but it
is generally discussed within two broad categories: the resources that one is able to obtain by
virtue of relationships, and the nature and extent of one‘s involvement in various networks and
groups (Grooteart et al. 2004, 3). Social capital can empower individuals and communities and
create new social and business opportunities that may help improve one‘s standard of living.
Thus, it is crucial to observe the connection between social capital and the use of microfinance.

There are numerous studies that back the assumption that ―the networks and organizations to
which people belong…have measurable benefits to these individuals, and lead, directly or
indirectly, to higher levels of well-being,‖ (Ibid, 17). Fafchamps and Minten (2002), for example,
showed how social capital and trade networks – more than working capital, labor, or
management – led to increased financial profits for
agricultural traders in Madagascar; Isham and                Empowerment Indicators:
Kahkoken (2002) found that social capital had a                 Q8: Transport Expenditures
positive effect on the construction and maintenance of          Q10: Phone Expenditures
water systems in Indonesia; and Coleman (1998)                  Q16: Decision-Making Power
illustrated how community cohesion lowered school
dropout rates in the U.S.

For FINCA, the CRM could be a valuable new way of weighing social capital stocks against
loan performance over time. While it would be difficult to establish causality between
microfinance and social capital, engaging in the question of whether social capital leads to better
business and repayment performance, or whether FINCA‘s products and services lead to new or
increased stocks of social capital, would be a worthwhile venture.

Perhaps the most comprehensive social capital assessment tool is the World Bank‘s Integrated
Questionnaire for the Measurement of Social Capital, or SC-IQ (Grootaert et al. 2004).4 The SC-
IQ identifies six dimensions of social capital: (1) Groups and Networks; (2) Trust and Solidarity;
(3) Collective Action and Cooperation; (4) Information and Communication; (5) Social Cohesion
and Inclusion; and (6) Empowerment and Political Action.

A concise, 16-question well-being survey cannot afford to be as all-inclusive as the SC-IQ or
attempt to map every dimension of social capital. Rather, the WHEEL captures just a few core
aspects of social capital that possibly affect, or are affected by, FINCA services. Toward this end,
while trust, collective action, and social cohesion are all important elements of social capital,
their weaker link with microfinance led to their exclusion from the WHEEL. Instead, the
indicators that relate to information and communication, as well as empowerment, were selected
for inclusion.

4
 The SC-IQ is an aggregation of the following surveys: The Tanzania Social Capital Survey (see Narayan and
Pritchett 1999); The Local Level Institutions Survey (see Grootaert 2000); The Social Capital Initiative (Grootaert
and van Bastalaer 2002a, 2002b); The Social Capital Survey (see Narayan and Cassidy 2001); and The Guatemala
Poverty Assessment (see World Bank 2003).


                                                         20
Phone Usage
The SC-IQ observes patterns and sources of information and communication by the following
indicators: the client‘s distance to a post office; the frequency that one listens to the radio; the
frequency one watches television; how often s/he reads the newspaper; how much s/he spends on
the internet; how many phone calls s/he has received in the last month; and the respondent‘s
three most important sources of market information. All of these questions gather important
information about a client‘s access to communication. They do not, however, equitably apply to
all of FINCA‘s clients throughout the world, in both rural and urban areas, nor do they
necessarily capture aspects of well-being beyond social capital.

Phones, on the other hand, are a useful measure of social capital for very reasons. First, they are
less expensive and more easily accessible than other communications technologies, such as a
computer or Internet. All clients do not necessarily own a phone, but the use of phones through
pre-paid phone cards has increased exponentially in developing countries in recent years.5 This
has made phone usage not only universal, but also easily measurable. Second, mobile phones are
user-friendly for poor populations with low levels of education, such as microfinance clientele.
And third, a large portion of FINCA‘s clients live in rural areas with lower population densities,
which, as a result, means there is often less basic infrastructural services (communication,
electricity, roads, sewage) that inhibit clients‘ ability to travel to, and gather information about,
other communities and markets.

Mobile phones help people bridge this divide by serving as a link-up mechanism to previously
unknown individuals and communities. Mobile phones help create a type of ―bridging‖ social
capital that extends the realm of personal relationships and networks (Cummings, Heeks, &
Huysman 2003, 88-101). They also help build up ―bonding‖ social capital by improving and
increasing the amount of interaction between members of a social network (ibid). This increased
stock of social capital offers new opportunities to both access new markets and customers and
improve on existing social networks that could help scale-up business activities. Reduced
informational asymmetries, due to the mobile phone, can enhance the development of trust
between actors in the market and thus facilitates transactions (Overa 2006, 1309).

It is for these reasons that development theorists have repeatedly lauded cellular technology not
only a means to accrue social capital, but also as the technological form best suited to serve the
socioeconomic needs of the world‘s poor (Narayan & Shah, 2000; Baliamoune, 2002; Curtain,
2004), and even as an important tool to combat global poverty.6 Numerous studies have
attributed the mobile phones as a key variable in economic growth both at the micro and macro
level (see Jenson 2007; and Waverman et al. 2005). Jensen (2005) found that the introduction of
the cell phone to Kerala, India created a rise in information flow that led to increased efficiency
in local markets, but also to externalities such as increased educational enrollments and higher
probability of poor individuals using healthcare when sick. In short, mobile phones have a


5
  Due to the simplicity and of tracking cell phone use by the amount of cards one purchases, as well as the difficulty
in recalling how often one actually makes or receives calls, the Well-Being survey, unlike the SC-IQ, asks how
much clients expend on cell phones per week, not how often they have received a phone call over the past month.
6
  See The New York Times, April 13, 2008, ―Can the Cell Phone Help End Global Poverty?‖


                                                         21
double function, in that they lead to increased social capital and economic performance, and have
positive externalities that are an integral part of FINCA‘s goal and mission.

Decision-making Power
The SC-IQ frames various questions regarding empowerment, including asking about one‘s level
of happiness and a person‘s ability to mobilize politically. The happiness indicator was not
chosen because happiness may be independent from financial well-being, and because it is
difficult to measure causality between microfinance outcomes and happiness. Additionally,
political empowerment is not an explicit goal of FINCA, so it cannot usefully be included in the
Well-Being survey.

The power to make important decisions in one‘s life, on the other hand, comprehensively
addresses the broader notion of personal control and the ability to affect one‘s own physical,
social and financial well-being. That is an implicit goal of providing loans and services to
disadvantaged groups: to provide individuals with the means to climb out of their structural
hardships and control their own lives. This is particularly true for female clients who are the
primary beneficiaries of FINCA‘s products, and who often face structural hurdles and do not
have the same financial or social opportunities as men. A higher level of personal empowerment
is a potential outcome of FINCA‘s program and should be studied and monitored by the CRM.



IX. Living Conditions

Shelter and physical security are universal basic needs that are essential conditions for, and
strong indicators of, one‘s well-being. Living conditions are often measured through a wide
range of indicators, including housing size, a household‘s water source, and the structural
condition of the housing. Housing conditions are considered by many to be a strong, albeit not
conclusive, indicator of a household‘s income level. The underlying assumption is that the level
of wealth of a household will be generally reflected in the quality of the family‘s dwelling. Hatch
and Frederick (1998, 16) note that because home improvements are usually done in small
increments over many months or years, they can reflect the economic progress of a household.
The authors also note, however, that housing quality is not a very accurate indicator of short-
term fluctuations in income. Therefore, while living conditions cannot be adequately measured in
a one-time survey, they are extremely valuable when measured over time.

The CGAP Microfinance Assessment Tool (Henry et al. 2003)—a comprehensive social
performance assessment tool—recommends 10 indicators for dwelling conditions that, saving
one (observed structural condition), are in the FCAT: (1) Ownership Status; (2) Number of
Rooms; (3) Type of Roofing Material; (4) Type of Exterior Walls; (5) Type of Flooring; (6)
Observed Structural Condition of Housing; (7) Type of Electric Connection; (8) Type of
Cooking Fuel Used; (9) Source of Drinking Water; and (10) Type of Latrine.

The WHEEL measures living conditions by first asking how many household members live in
the dwelling, an integral question in determining household size and per-capita expenditures. The


                                                22
follow-up question of the number of rooms in        Living Condition Indicators:
the home captures the number of rooms per               Q1: Number of People in the House
capita; a lower people to room ratio denotes            Q4: Number of Rooms in the House
more comfortable living space. Indeed,                  Q5: Drinking Water Source
according to CGAP‘s measures, number of                 Q6: Toilet Facility
rooms per capita shares the highest index               Q9: Utilities Expenditures
score amongst other dwelling indicators by              Q12: Home Expenditures
meeting seven out of nine poverty testing
standards. Other questions are strong and important measurements of living conditions, but they
are not as practical or comprehensive as this question. What is more, this indicator is simple,
time-efficient, and universally valid across countries and regions (Ibid, 177).

The type of water source and type of latrine also share the high index score of living conditions
by the CGAP, and are thus considered important for this section. The other high index scores
were: number of stories, structural condition, roof material, wall material, and vehicle type.
Amongst those, number of stories and structural condition were not considered relevant, because
they were not in the 2007 FCAT data set, not universally applicable across locales, or were
impractical to administer. The rates of vehicle ownership are not consistent across the countries
in which FINCA operates, so this indicator was also problematic. Finally, while roof material
and wall material are excellent proxies for household wealth in locales where common types of
materials can be identified, the well-being survey does not distinguish between rural and urban
areas, and these materials cannot be generalized to all areas of society.




X. Limitations

The WHEEL captures the core aspects of clients‘ well-being in 16 questions. Because of its
brevity, the WHEEL fits well within the framework of CRM and will be useful in obtaining
information about clients that is not captured by the FCAT. This being said, there are a few
limitations to this methodology that need to be recognized. Some of these are inherent limitations
that are embedded within the need of the CRM for a short survey tool; others may eventually be
overcome as the WHEEL is tested and improved in the future.

The most important among them, as stated earlier, is that the WHEEL is not a complete or
thorough assessment of every aspect of client well-being. The WHEEL at times sacrifices
comprehensive knowledge for the sake of obtaining practical and manageable data. The CRM
must strike a balance between being short and easy to administer and obtaining as much
information as possible. There is perhaps room, however, to ask a few more questions without
foregoing the simplicity and utility of the WHEEL. Potential additional questions that FINCA
management may wish to consider when finalizing the CRM are noted in Appendix A.

A second area for further consideration is that the WHEEL measures wealth primarily by
household expenditures (income) and not assets (stock wealth). This is problematic because self-
assessed expenditure questions often create an upward bias and are variable based on the season
in which the question is asked. Many household expenditure questions are difficult for people to


                                               23
interpret and recall, respondents often report higher amounts of expenditures than the reality, and
respondents will often give responses based on the time of year that the question is asked. This
type of variability in income measurements is even more pronounced in rural areas (Hatch and
Frederick 1998, 12). For example, people may recall their household production of food for their
most immediate harvest, but neglect to account for how much was produced six months earlier.

This phenomenon was confirmed by the team‘s observations in the field. A focus group held
with clients who had already taken the survey revealed that respondents interpreted many pilot
survey questions differently, particularly household expenditure questions. For example, the pilot
survey was administered in January, and many clients calculated for ―special events‖ the amount
they had spent on the Christmas holidays and neglected what they had spent on previous
festivities. In addition, it was quite clear that questions relating to leisure expenditures and how
much people spent on family and friends elicited different interpretations and were difficult for
people to recall. In short, the team does not believe that responses to expenditure questions are
consistent nor entirely accurate responses. Expenditures by themselves are not a sufficient
variable to determine absolute levels of DPCE at a given time. If the WHEEL is to be
administered to the same clients over time (and at the same time each year), however, the upward
bias will be constant and the WHEEL could still accurately observe changes in spending.

The final limitation is that the WHEEL is not consistent in whose well-being it is assessing. At
times, it probes into household behavior and well-being, while at other times it asks solely about
the individual client. This is because there are certain questions that individuals cannot
accurately answer for an entire household. Whenever possible, the WHEEL tries to assess
overall household well-being, but this was not always feasible. This is not a significant
shortcoming, however, for many other living standards surveys, including the FCAT, incorporate
this design. However, what this means is that the administration and analysis of the survey
requires diligence. Surveyors must make it clear to respondents who the question is addressing,
and analysts must be cognizant of these slight variations. The subject changes will not affect the
quality of client information, only the ease with which the tool is used.




                                                24
XI. Recommendations

The authors believe the WHEEL survey to be an efficient means of assessing the well-being of
FINCA clients and a valuable addition to the CRM platform. In hopes of maximizing the utility
of the instrument and to ease its installation in the field, we offer the following recommendations.


Additional Tests
The WHEEL represents a significant departure from the survey that was piloted in January of
2007. As such, it should undergo further field tests before its incorporation into the CRM
platform. Particular attention ought to be given to the predictions of total expenditure levels. The
results presented here were obtained from in-sample forecasting only; the accuracy of the models
will need to be confirmed with field-testing.

CRM Implementation
The researchers believe that the WHEEL component should be administered to clients at the
same time as the Loyalty component. Both surveys are brief in length and should not over-
burden clients or staff. The richness of data generated through the coupling of these survey
instruments will more than compensate for any additional resources required for their
administration. Understanding the links between socioeconomic conditions and client
satisfaction is key to analyzing client retention and successful performance.

Panel Data
The CRM should strive to acquire panel data by interviewing the same clients at multiple points
in time. Doing so will allow FINCA to better understand the relationships between indicators
tracked by the CRM. In particular, panel data more clearly shows the cause-and-effect
relationships that make monitoring client behavior instructive. Although more difficult to collect,
data collected over time affords the possibility of richer analysis and more useful findings.

WHEEL and FCAT
The WHEEL survey instrument proposed in this paper is an attempt to fulfill the unique needs of
the CRM initiative; it should not be viewed as a replacement for or an improvement upon the
FCAT. The team deems that the two instruments should have a complementary relationship.
Findings from one should inform the design of the other, and vice versa.

The authors believe that once the CRM is fully operational within the FINCA system, the
administration of the FCAT can be reduced. For example, instead of every two years in each
country, it may be every five or so years. However, the FCAT still needs to serve as the
benchmark that allows the WHEEL to estimate expenditures using a reduced question set; these
models will need to be recalibrated as FINCA acquires new FCAT data.


                                                25
To this end, all WHEEL questions should be included in subsequent applications of the FCAT.
Doing so will allow FINCA analysts to generate stronger expenditure models for each country,
increasing the accuracy of these estimates.

FCAT Accuracy
Though the research team was not originally tasked with analyzing FCAT results, it nonetheless
spent much time immersed in the data, and the researchers would be remiss if they did not
comment on one particular concern. Namely, the expenditure figures in the FCAT data seem to
be upwardly biased and possibly not representative of FINCA clientele. During pilot testing of
the original survey instrument, the research team spent two weeks in Uganda interacting with
clients. The reality on the ground does not match the picture painted by the DPCE calculations
for clients in this country (median DPCE = US$6 / day.) It is likely that the practice of simply
asking clients to report consumption for the entire household results in inflated numbers. This
may be due to the difficulty of the task itself or the bias of respondents trying to impress their
lender. Regardless, the GWU team strongly believes that FINCA should examine best practices
for household consumption surveys and adjust FCAT procedures accordingly.




                                                26
Appendix A: Other Questions Considered

The selection of indicators for a short, 16-question well-being tool proved to be a challenging
task, as space constraints forced many solid questions to be eliminated. The following appendix
lists some of these, in the case that FINCA staff would like to reconsider them in future versions
of the CRM or FCAT. This list is not exhaustive.

“Have any members of your household skipped meals in the last month?”
      a. Never
      b. Rarely (1-2 times)
      c. Occasionally (3-5 times)
      d. Often (6 or more times)

Households‘ coping strategies when faced with food shortage were concluded to be most
appropriate for this survey due to their highly predictive power, but had to be eliminated due to
space constraints (Christiaensen and Boisvert 2000, 19). The ideal version of this method
introduces six separate coping strategies to the respondent, ranging from consuming less
preferred foods (the least severe coping strategy) to skipping meals for a whole day (the most
severe strategy). The respondents‘ answers are then weighted and summed (Hoddinott 1999, 7-
13). In the interest of simplicity, the researchers chose only one of these questions, at the ―more
severe‖ end of the scale, in order to capture the largest possible variation in answers from typical
FINCA clientele. For example, most FINCA clients would probably report having to eat less
preferred foods. Such a question would therefore not yield very rich data. The recall period of 14
days was chosen in order to capture weekly differences in food availability. This was done to
counteract the usual tendency of coping questions to underreport food insecurity across
households when asked with a 7 day recall period (Ibid).

“If you suddenly needed a small amount of money [RURAL: enough to pay for expenses for your
household for one week; URBAN: equal to about one week’s wages], how many people beyond
your immediate household could you turn to who would be willing to provide this money?”
        a. No one
        b. One or two people
        c. Three or four people
        d. Five or more people

This question is a strong indicator of the ―size‖ of one‘s safety net. First, the ability to borrow
from members within a particular group or community mitigates investment risks that are
necessary to make businesses profitable. This is likely an underlying reason that groups with
higher levels of social cohesion have a higher repayment rate (Zeller and Sharma 1998). Second,
the ability to borrow often leads to income and consumption smoothing—or less variance in
income and consumption patterns—that is an often unrecognized benefit microfinance programs
(see Zeller 1999). Consumption smoothing is potentially a direct benefit from FINCA‘s loans
and services, but also, perhaps, an indirect benefit from the Village Banking system. It would be
valuable to examine if Village Banking group members that have stronger group cohesion –
measured by this question – also have better loan ratings or business performance.



                                                 27
Appendix B: Alternate 10-Question Version

If for reasons of practicality or simplicity, FINCA decides that survey needed to be shortened to
ten questions, the authors recommend the following questions:

   1. How much does your household usually spend per week for buying food?
   2. In a typical month, how much does your household spend for all goods and services?‖
   3. How many rooms are there in your home (Include detached rooms in same compound if
       same household)?
   4. How many school-aged (5-18) children live in your household?
   5. How many of those attend school?
   6. In general, how is your health at this time?
   7. What is the main source of drinking water for members of your household?
   8. What kind of toilet facility does your household use?
   9. In the past month, how much have you spent on the use of phones (can include cell
       phones, land phones, or phone booths)?
   10. Do you feel you have the power to make important decisions that change the course of
       your life? Rate yourself on a 1-5 scale.

The selected questions are an equitable distribution of every dimension of well-being. This
modified survey does not consider wealth or any other dimension to be a more accurate indicator
of well-being than others. If the full, accurate wealth-model were incorporated into this survey, it
would leave room for only three other questions to measure other aspects of well-being. From
there, it would have been arbitrary and random to select and discard questions from other
categories.

All of the questions selected here are also in the WHEEL except for ―In a typical month, how
much does your household spend for all goods and services?‖ It is a question that Hatch and
Frederick (1998, 26) cite as a solitary expenditure question that has been used successfully in
other tools. Hatch and Frederick recommend testing this simple and direct single question on
expenditures, including the collection of information on the value of home-produced
consumption. This question is not currently asked in the FCAT, so it was not tested in the
WHEEL model. To strengthen this wealth-indicator on a 10-question survey, the team
recommends adding one more expenditure question. Weekly food expenditures, from the FCAT
data, is the indicator with the highest correlation to daily per capita expenditures, and is therefore
used here to indicate household wealth.




                                                 28
Works Cited

Accelerated Microenterprise Advancement Project (2004). ―Review of Poverty Assessment
       Tools.‖ Developing Poverty Assessment Tools Project, USAID.

Baliamoune, M. N. (2002). ―The new economy and developing countries: Assessing the role of
      ICT diffusion.‖ (Discussion paper No. 2002/77). Jacksonville, Florida: United Nations
      University, World Institute for Development Economics Research.

Bateman, O., and Smith, S. (1991). ―A Comparison of the Health Effects of Water Supply and
      Sanitation in Urban and Rural Guatemala.‖ Water and Sanitation for Health (WASH)
      Project Field Report 352 reprinted by the Environmental Health Project, Arlington, VA.

Bergeron, G. and Esrey, S. (1993). ―Baseline Survey for the Guatemala Highlands Rural Water
       and Sanitation Project.‖ WASH Field Report #403. WASH: Washington DC.

Bhattacharya, J., Currie, J., and Haider, S. (2004). ―Poverty, food security and nutritional
       outcomes.‖ Science Direct. 23(4)

Billig, P., Bendahmane, D., and Swindale, A. (1999). ―Water and Sanitation Indicators
        Measurement Guide.‖ Title 2: Indicators Guide. Food and Nutrition Technical Assistance.

China Statistics Census (2005). <www.allcountries.org/china_statistics/10_25_composition_of
       _per_capita_consumption.html> Accessed 5 May 2008.

Christiaensen, L., Boisvert, R., and Hoddinott, J. (2000). ―Validating Operational Food
        Insecurity Indicators against a Dynamic Benchmark: Evidence from Mali.‖ World Bank
        Policy Research Working Paper.

Cummings, S., Heeks, R., and Huysman (2003), ―M. Knowledge and Learning in Online
     Communities in Development: A Social Capital Perspective.‖ Institute for Development
     Policy and Management, p. 11.

Curtain, R. (2004). ―Information and communications technologies and development: Help or
       hindrance?‖ Melbourne, Australia: Australian Agency for International Development.

Deaton, A., and Grosh, M. (2000) ―Consumption.‖ Designing Household Survey Questionnaires
      for Developing Countries. Eds. Margaret Grosh and Paul Glewwe. World Bank.

Demographic and Health Surveys. <www.measureddhs.com>. Accessed April 2008.

Esrey, S., Potash, L. Roberts, and Shiff, C.(1991). ―Effects of improved water supply and
       sanitation on ascariasis, diarrhea, drancunculiasis, hookworm infection, schistosomiasis
       and trachoma.‖ Bulletin of the World Health Organization 60:609-621.




                                                 29
Fafchamps, M. and Minten, B. (2002), ―Social Capital and the Firm: Evidence from Agricultural
      Traders in Madagascar.‖ In Grootaert, C. and van Bastelaer, T. eds. The Role of Social
      Capital in Development: And Empirical Assessment. New York: Cambridge University
      Press, pp. 125-54.

Gertler, P., Rose, E., and Glewwe, P. (2000). ―Health.‖ Designing Household Survey
        Questionnaires for Developing Countries, Volume 1. Eds. Grosh, Margaret and Paul
        Glewwe. Washington, DC: World Bank.

Geser, H. (2004). ―Towards a Sociological Theory of the Mobile Phone.‖ Soziologisches Institut
       der Universitat Zurich.

Grootaert, C., Narayan, D., Jones, V.N., and Woolcock, M. (2004). ―Measuring Social Capital:
       An Integrated Questionnaire.‖ World Bank working paper, no. 18.

Grootaert, C. and Narayan (2000). ―Local Institutions, Poverty and Household Welfare in
       Bolivia.‖ Local Level Institutions Working Paper 9. World Bank, Social Development
       Department, Washington, D.C.

Grootaert, C. and van Bastalaer, T. (2002a). ―The Role of Social Capital in Development.‖ In
       Development: An Empirical Assessment. New York: Cambridge University Press.

Grootaert, C. and van Bastalaer, T. (2002b). ―Understanding and Measuring Social Capital: A
       Multidisciplinary Tool for Practitioners.‖ Word Bank: Washington, D.C.

Hatch, J., and Frederick, L. (1998). ―FINCA International. Poverty Assessment by Microfinance
       Institutions: A Review of Current Practice,‖ Accelerated Microenterprise Advancement
       Project: USAID.

Henry, C., Sharma, M., Lapenu, C.; and Zeller, M. (2003). ―Microfinance Poverty Assessments
       Tool,‖ Consultative Group to Assist the Poor, Technical Tools Series No. 5.

Hoddinott, J. (1999). ―Choosing Outcome Indicators of Household Food Security,‖ Technical
      Guide #7. IFPRI.

Idler, E. and Yael, B. (1997). "Self-Rated Health and Mortality: A Review of Twenty-Seven
        Community Studies." Journal of Health and Social Belhavior3 8 (1): 21-37.

International Household Survey Network. Question Bank. <www.surveynetwork.org>. Accessed
        April 2008.

Isham, J. and Kahkoken, S. (2002). ―How Do Participation and Social Capital Affect
       Community-Based Water Projects? Evidence from Central Java, Indonesia.‖ In Christean,
       G. and Bastelaer, T., eds. The Role of Social Capital in Development: An Empirical
       Assessment. New York: Cambridge University Press, pp. 155-187.




                                              30
Jensen, R. (2007). ―The Digital Provide: Information (Technology), Market Performance, and
       Welfare in the South Indian Fisheries Sector,‖ Quarterly Journal of Economics. Vol.
       CXXII, 3.

Joint Monitoring Programme for Water Supply and Sanitation. <www.wssinfo.org>. Accessed
       April 2008.

Kennedy, Eileen (2002). ―Measurement and Assessment of Food Deprivation and
      Undernutrition‖ FAO Conference Paper.

Maxwell, D.G. (1995). ―Measuring Food Insecurity: The Frequency and Severity of Coping
     Strategies.,‖ FCND Discussion Paper No. 8. IFPRI.

Narayan, D. and Cassidy, M. (2001) ―A Dimensional Approach to Measuring Social Capital:
      Development and Validation of Social Capital Inventory.‖ Current Sociology 49(2):49-93.

Narayan, D. and Pritchett, L (1999). ―Cents and Sociability: Household Income and Social
      Capital in Rural Tanzania.‖ Economic Development and Cultural Change 47(4):871-97.

Narayan, D. and Shah, T. (2000). ―Connecting the local to the global: Voices of the poor.‖ In
      Workshop on Local to Global Connectivity for Voices of the Poor. Washington, DC:
      World Bank.

Newman, K. (2002). ―Going From Voice to Data to Development.‖ Fletcher School of Law and
     Diplomacy.

Overa, R. (2006). ―Networks, Distance, and Trust: Telecommunications Development and
       Changing Trading Practices in Ghana.‖ World Development, 34(7), p. 1301.

Schreiner, M., Matul, M., Pawlak, E., and Kline, S. (2005). ―The Power of Prizma‘s Poverty
       Scorecard: Lessons for Microfinance.‖ Microfinance Risk Management.

Sen, A. (1998). "Mortality as an Indicator of Economic Success and Failure," Economic Journal,
       Royal Economic Society, vol. 108(446), pages 1-25, January.

Smith, L. and Subandaro, A. (2007). ―Measuring Food Security using Household Expenditure
       Surveys.‖ Food Security in Practice. Number 3. IFPRI.

Strauss and Thomas (1998). ―Health, Nutrition, and Economic Development,‖ Journal of
       Economic Literature, Vol. 36, No. 2, pp. 766-817

UNICEF (2002). Multiple Indicator Cluster Surveys.

United Nations (2005). ―UN Questionnaire on Poverty Measurement.‖ United Nations Statistics
       Division, New York.




                                               31
Waverman, L., Meschi, M., and Fuss, M. (2005). ―The Impact of Telecoms on Economic
     Growth in Developing Countries.‖ Vodafone Policy Paper Series.

Woolcock, M., and Narayan, D. (2000). ―Social Capital: Implications for Development Theory,
      Research, and Policy.‖ World Bank Research Observer 15 (2):225-50.

World Bank (2003). ―Guatemala Poverty Assessment.‖ Washington, D.C.: World Bank.

World Bank (2008). ―Education and Development‖. World Bank homepage.
      <web.worldbank.org> Accessed April 2008.

World Health Organization (2007). ―Core Health Indicators‖. WHO Statistical Information.

Zeller, M. and Sharma, M. (1998). ―Rural finance and poverty alleviation.‖ International Food
        Policy Research Institute.

Zeller, M. (1999). ―The Role of Micro-Finance for Income and Consumption Smoothing‖,
        International Food Policy Research Institute.

Zeller, M., Sharma, M., Henry, C., and Lapenu, C. (2001). ―An Operational Tool For Evaluating
        Poverty Outreach Of Development Policies And Projects.‖ FCND Discussion Paper No.
        111.




The researchers would like to thank the good people at FINCA International and FINCA
Uganda for their support throughout the survey testing and refinement process. In particular,
Laura Cordero, Ben Hermoso, Robert Lule, Camille Selosse, and Katie Torrington were
tremendously helpful and kind. We would also like to thank our advisor, Daniel Morrow, for his
invaluable insights, patience, and encouragement.




                                              32

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:2
posted:10/15/2011
language:English
pages:36
chenleihor chenleihor
About