Professional Documents
Culture Documents
79±98, 2000
Ó 2000 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
www.elsevier.com/locate/worlddev 0305-750X/00/$ - see front matter
PII: S0305-750X(99)00119-9
Key words Ð methods, micro®nance, credit, impact assessment, monitoring and evaluation,
poverty reduction
science and better art. The scienti®c improve- 1980s. A set of factors are associated with the
ments relate to improving standards of extreme ``pole'' positions of this continuum and
measurement, sampling and analytical tech- these underpin many of the issues that must be
nique. Econometricians and statisticians are resolved (and personal and institutional
particularly concerned with this ®eld. tensions that arise) when impact assessments
Improving the ``art'' of impact assessment has are being initiated (Figure 1).
at least three strands. One concerns making Behind the shift from ``evaluation'' to ``IA''
more systematic and informed judgements are a number of factors. These are not explored
about the overall design of IAs in relation to in any detail in this paper but they form an
their costs, speci®c objectives and contexts. essential element for the understanding of IA
The second is about what mixes of impact and its potential contributions. Explicitly, IAs
assessment methods are most appropriate for are promoted by both the sponsors and
any given study. The third relates to increas- implementers of programs so that they can
ing our understanding of the ways in which learn what is being achieved and improve the
the results of IA studies in¯uence policy- eectiveness and eciency of their activities.
makers and micro®nance institution (MFI) Implicitly, IAs are a method by which sponsors
managers. seek to get more information about program
eectiveness than is available from the routine
accountability systems of implementing orga-
2. IMPACT ASSESSMENT: OBJECTIVES nizations. IAs are also of signi®cance to aid
agencies in terms of meeting the ever increasing
Impact assessment studies have become accountability demands of their governments
increasingly popular with donor agencies and, (in this era of ``results'' and ``value for money'')
in consequence, have become an increasingly and for contesting the rhetoric of the anti-aid
signi®cant activity for recipient agencies. In lobby. While recipient agencies bene®t from
part this re¯ects a cosmetic change, with the this, they are one stage removed, and many are
term IA simply being substituted for evalua- likely to see donor-initiated IA as an activity
tion. But it has also been associated with a that has limited practical relevance for program
greater focus on the outcomes of interventions, activities. To quote the director of a large Asian
rather than inputs and outputs. While the goals micro®nance institution that has received
of IA studies commonly incorporate both substantial amounts of aid ®nanced IA
``proving'' impacts and ``improving'' interven- consultancy and internal IA-capacity building
tions, IAs are more likely to prioritize the ``...impact assessment studies keep donors
proving goal than did the evaluations of the happy... we donÕt use them very much.''
A ®nal issue to raise in this section is whether Ða model of the impact chain that the study
the expectations of OECD-based agencies is to examine,
about the feasibility of the accurate measure- Ðthe speci®cation of the unit(s), or levels, at
ment of impacts in the dicult contexts of which impacts are assessed, and
developing countries (limited numbers of Ðthe speci®cation of the types of impact
professional researchers, few written records, that are to be assessed.
illiteracy, communication problems etc.) are
higher than in their own countries. My
professional experience of EU-®nanced ``small (a) Models of impact chains
enterprise development'' projects in Manches-
ter has revealed a startling lack of concern with Behind all micro®nance programs, and
impacts: this is in marked contrast to my indeed virtually all aid ®nanced initiatives, 2 is
consultancy work in Bangladesh where donors the assumption that intervention will change
criticize nongovernment organizations (NGOs) human behaviors and practices in ways that
for failing to make impact assessment a prior- lead to the achievement (or raise the probability
ity! If recipients perceive that the IA standards of achievement) of desired outcomes. IAs assess
expected of ``them'' are higher than donors the dierence in the values of key variables
expect of themselves then IA will be seen as an between the outcomes on ``agents'' (individuals,
external imposition rather than a shared enterprises, households, populations, policy-
opportunity. makers, etc.) which have experienced an inter-
vention against the values of those variables
that would have occurred had there been no
intervention. The fact that no agent can both
3. ASSESSING IMPACT: THE CHOICE OF experience an intervention and at the same time
CONCEPTUAL FRAMEWORKS not experience an intervention generates many
methodological problems. All changes are
All impact assessment exercises have a in¯uenced by mediating processes (speci®c
conceptual framework at their heart. In well- characteristics of the agent and of the
planned and well-resourced IAs with long economic, physical, social and political envi-
``lead-in'' times such frameworks are usually ronment) that in¯uence both behavioral chan-
explicitly identi®ed (e.g., Khandker, 1998; ges and the outcomes in ways that are dicult
Sebstad, Neill, Barnes & Chen, 1995; Schuler & to predict (Sebstad et al., 1995).
Hashemi, 1994). By contrast, in many smaller The impact chain is very simply depicted in
scale exercises the framework is implicit and Figure 2. A more detailed conceptualization
may be seen as ``common sense.'' There are would present a complex set of links as each
three main elements to a conceptual frame- ``eect'' becomes a ``cause'' in its own right
work: generating further eects. For example, in a
conventional micro®nance project a package of ing'' of loans (Wiig, 1997) which may threaten
technical assistance and capital changes the the long-term viability of an MFI.
behavior (and products) of a micro®nance While the choice between these two schools
institution (MFI). The MFI subsequently can ultimately be seen as an ideological choice
provides dierent services to a client, most (does one prioritize contributions to improved
commonly in the form of a loan. These services welfare or to more ecient markets?) it is
lead to the client modifying her/his microenter- possible to recognize dierent strengths and
prise activities which in turn leads to increased weaknesses. The intended bene®ciary school
or decreased microenterprise income. The makes fewer assumptions about the impact
change in microenterprise income causes chan- chain and is better able to distinguish ``who''
ges in household income which in turn leads to bene®ts and ``how.'' It is, however, demanding
greater or lesser household economic security. in both methodological and cost terms. The
The modi®ed level of household economic intermediary school usefully incorporates
security leads to changes in the morbidity and notions of sustainability and provides an IA
mortality of household members, in educa- methodological framework that can be oper-
tional and skill levels and in future economic ated largely with pre-existing data. It is,
and social opportunities. Ultimately, perhaps, though, very weak on ``who'' bene®ts and
these changes lead to modi®cations in social and ``how'' (as illustrated by assessments of the
political relations and structures. The USAID-®nanced APPLE program). 6 Possible
complexity of such chains provides the assessor ways of strengthening the intermediary school
with a range of choices about which link (or approach have been suggested by Feinstein
links) to focus on. For micro®nance, it is useful (1997) through the analysis of borrower trans-
to distinguish between two main schools of action costs. He proposes the collection of
thought with regard to which link(s) in the chain longitudinal data on borrowers transaction
to focus on. For convenience, these are termed costs (p. 5) to assess whether an MFI has
the ``intended bene®ciary'' 3 school and the bene®ted borrowers, i.e., has reduced their total
``intermediary'' school. costs for accessing ®nance. This oers a
The intended bene®ciary school, building on potential ``bridge'' between the two main
the ideas of conventional evaluation, seeks to get ``schools,'' if data on ``who'' borrowers are also
as far down the impact chain as is feasible (in collected.
terms of budgets and techniques) and to assess
the impact on intended bene®ciaries (individuals (b) Units of assessment
or households). The intermediary school focuses
purely on the beginning of the chain and in Following on from the design of a model of
particular on changes in the MFI and its oper- the impact path comes the choice of the unit(s)
ations. Its roots are closely associated with the of assessment (or levels of assessment). Com-
Ohio State University SchoolÕs analyses of rural mon units of assessment are the household, the
®nance. Generally, two key variables are enterprise or the institutional environment
focused on: institutional outreach and institu- within which agents operate. Occasionally
tional sustainability 4 (Yaron, Benjamin & studies have attempted to assess impact at an
Piprek, 1997). If both outreach and sustaina- individual level (e.g., Goetz & Sen Gupta, 1996;
bility have been enhanced then the intervention Peace & Hulme, 1994), but this is relatively rare
is judged to have a bene®cial impact as it has and has to take a qualitative focus. More
widened the ®nancial market in a sustainable recently some studies have attempted to assess
fashion. This is based on the assumption that impacts at a number of levels, such as Hulme
such institutional impacts extend the choices of and Mosley (1996) who looked at microenter-
people looking for credit and savings services prise, household, community and institutional
and that this extension of choice ultimately leads levels and USAIDÕs Assessing the Impact of
to improved microenterprise performance and Microenterprise (AIMS) Project. Through a
household economic security. While this household economic portfolio model (HEPM)
assumption can be supported by theoretical the latter seeks to assess impacts at household,
frameworks (if a set of further assumptions are enterprise, individual and community levels
made about perfect competition and other and thus produce a fuller picture of overall
factors) it is an assumption which has proved impacts (Chen & Dunn, 1996).
invalid in a number of experiences. 5 In addi- The relative advantages and disadvantages of
tion, it will not reveal borrower ``cross®nanc- dierent units of assessment are summarized in
IMPACT ASSESSMENT METHODOLOGIES FOR MICROFINANCE 83
Table 1. As can be seen, a focus purely on the Conventionally, economic indicators have
``individual'' or the ``enterprise'' has such dominated micro®nance IAs with assessors
drawbacks that they could be viewed as particularly keen to measure changes in income
discredited. The household economic portfolio despite the enormous problems this presents.
model has much to recommend itÐespecially if Other popular variables have been levels and
institutional impacts are incorporated in the patterns of expenditure, consumption and
community level analysis. It does have the assets. A strong case can be made that assets are
profound disadvantage, though, of making a particularly useful indicator of impact because
assessment demanding in terms of costs, skilled their level does not ¯uctuate as greatly as other
personnel and time. If used with limited economic indicators and is not simply based on
resources it risks sacri®cing depth for breadth an annual estimate (Barnes, 1996, p. v).
of coverage of possible impacts. The social indicators that became popular in
the early 1980s (e.g., educational status, access
(c) Types of impact to health services, nutritional levels, anthropo-
metric measures and contraceptive use) have
An almost in®nite array of variables can be recently been extended into the socio-political
identi®ed to assess impacts on dierent units. arena in an attempt to assess whether micro®-
To be of use these must be able to be de®ned nance can promote empowerment (Mayoux,
with precision and must be measurable. 1997; Goetz & Sen Gupta, 1996; Schuler &
84 WORLD DEVELOPMENT
Hashemi, 1994; Hashemi, Schuler & Riley, reasoned argument supported by theory and
1996). This has led to the measurement of speci®c pieces of evidence. Although the former
individual control over resources, involvement has tended to dominate discussions about
in household and community decision-making, micro®nance impact assessment (see, for
levels of participation in community activities example, the studies reviewed by Gaile &
and social networks and electoral participation. Foster, 1996) the latter tradition is being
The bulk of this work has focused on changes increasingly used by MFIs and researchers
in gender relations, but there are sometimes (Bouman & Hospes, 1994; Ardener & Burman,
partially-formulated assessments of class rela- 1995; Remenyi, 1991; Rutherford, 1999). The
tions within it (Fuglesang & Chandler, 1993). third part of this section explores a recent
These extensions to the types of impact assessed entrant to the ®eldÐparticipatory learning and
permit IAs to be more sophisticated and to action (PLA)Ðwhich oers a radical challenge
shed light on developmental impacts at a time to both conventional IA and to ``science'' itself.
when the goals of development have also been Although these three approaches can be sepa-
extended. They do add, however, to the rated for analytical purposes, in recent practice
complexity of IA work and require the skills of many studies have woven elements of these
assessors who are experienced at making approaches together (see Section 5 for a
judgements on social relations. discussion).
Sebstad et al. (1995) usefully distinguish
between ``domains of change'' (e.g., household (a) Scienti®c method
income) and the speci®c ``markers of change''
(e.g., amount of income, number of income Scienti®c method seeks to ensure that eects
sources and seasonality of income) within each can be attributed to causes through experi-
domain. While not comprehensive, the detailed mentation. 8 A particular stimulus to a partic-
sets of domains and markers, produced in their ular object in a rigorously controlled
paper provide an excellent checklist for impact environment is judged to be the cause of the
assessors to consider at the IA design stage. 7 observed eect. The experimental approach is
Often the exact markers used will be shaped by virtually infeasible in the social sciences,
the methodology that is selected. This can because of the nature of the subject matter, and
cause problems for multi-method IAs which so the approach has been adapted into quasi-
may not be able to apply a single de®nition for experiments (Casely & Lury, 1982). Quasi-ex-
a marker for each of the methods used. In periments seek to compare the outcomes of an
addition, impact assessors should always seek intervention with a simulation of what the
to keep the number of variables they measure outcomes would have been, had their been no
to a manageable number and not be tempted to intervention. One method for this is multiple
go for a comprehensive approach that will regression, but this has rarely been used in
impact adversely on data quality and study micro®nance IA because of its enormous
relevance. demands for data on other possible causal
factors and its assumptions (Mosley, 1997, pp.
2±3). A second approach is the control group
4. THE THREE PARADIGMS OF IMPACT method which has been widely used. This
ASSESSMENT: PROBLEMS OF requires a before and after comparison of a
ATTRIBUTION AND FUNGIBILITY population that received a speci®c treatment 9
(i.e., a micro®nance program) and an identical
The major methodological problems that population (or as near as possible) that did not
confront the IA of micro®nance relate to receive the treatment. While this idea is ele-
attribution and fungibility. At the heart of gantly simple a number of ``elephant-traps''
impact assessment is the attribution of speci®c may befall its user. In particular problems of
eects (i.e., impacts) to speci®c causes (i.e., sample selection bias, misspeci®cation of
interventions). From the vast literature on underlying causal relationships and respondent
micro®nance IA it is possible to draw out three motivation (see later) must be overcome.
very dierent paradigms by which authors seek Selection bias may occur because of:
to demonstrate attribution. The ®rst is the (i) diculties in ®nding a location at which
conventional scienti®c method with its origins the control groupÕs economic, physical and
in the natural sciences. The second has its roots social environment matches that of the treat-
in the humanities and focuses on making a ment group,
IMPACT ASSESSMENT METHODOLOGIES FOR MICROFINANCE 85
(ii) the treatment group systematically pretend that microenterprises are ``®rms''
possessing an ``invisible'' attribute which whose inputs and outputs can be precisely
the control group lacks (most commonly identi®ed and measured but to recognize that
identi®ed as entrepreneurial drive and abil- the impacts of micro®nance must be assessed at
ity), a variety of levels. The assessor attempting to
(iii) receiving any form of intervention may control for fungibility (to prove impact) has
result in a short-term positive response from failed to recognize that fungibility is a process
the treatment group (the Hawthorne eect), to be encouraged (to improve impact)!
(iv) the control group becoming contami- The misspeci®cation of underlying causal
nated by contact with the treatment group relationships arises most commonly because of
(though this could be a long-term program the assumption that causality is a one-way
goal!), and process (Figure 2). This may be a reasonable
(v) the fungibility of the treatment (e.g., assumption in the physical sciences (though it
when a loan is transferred from a borrower does not go unchallenged by contemporary
to someone else or when the loan is not used philosophers of science). For human activity it
in the planned way). is commonly invalid, as causation may also run
Problems (i) and (iv) can be tackled by more from impact back to intervention. Mosley
careful selection of the control group. This (1997, p. 6) illustrates this with the example of a
applies particularly to controlling for access to program whose ®eld sta put pressure on a
infrastructure (which has a key in¯uence on borrower to repay her loan; this may succeed in
input and output prices as well as other vari- the short term but may induce the borrower to
ables) and ensuring that the control group is sell assets (machinery, land, trees) which reduce
located far away from the treatment group. the probability of repayment in the longer term.
Problems (ii) and (iii) are more intractable, but Such reverse causation need not necessarily be
in many cases they can be tackled by using negative and, from the perspective of more
program-accepted ``clients-to-be,'' who have process-oriented analytical frameworks, is
not yet received micro®nance services, as the essential if programs are to continually learn
control group (Hulme & Mosley, 1996, chapter from their experience and improve (rather than
4). It must be noted, however, that this prove) their impact.
approach will not be valid when the take up of Such problems can be overcome by the
micro®nance services is based on diusion adoption of models that conceptualize causa-
through a heterogeneous population. 10 tion as a two-way process by the use of two-
This leaves the problem of loan fungibility. stage least squares technique and regression
This can be seen as an intractable problem as analysis (Mosley, 1997, p. 7). Such an approach
``...no study has successfully controlled for the is enormously demanding in terms of data
fungibility of resources between the household requirements, technical expertise and costs. It
and the assisted enterprise'' (Gaile & Foster, will only be feasible on very rare occasions
1996, p. 24). Using case study materials to (for example, see Khandker, 1998). For most
crosscheck actual loan use against intended researchers adopting the scienti®c method,
loan use and thus estimating ``leakage'' is one reverse causality is a problem to be coped
possible approach to controlling for fungibility with rather than overcome. The main means
(Pulley, 1989; Mosley, 1997). But for all studies of dealing with it are tracing dropouts from
except those that focus exclusively on ``the both the treated and control groups; only
enterprise,'' then a concern about fungibility conducting IAs on relatively mature
may be irrelevant. For studies looking at the programs; interim impact monitoring activities
household, the community or the household to gather qualitative information about the
economic portfolio (see Section 3b) fungibility complexity of causality; and retrospective in-
is not a problem for the assessor, rather it is a depth interviews with clients (Mosley, 1997,
vital strategy for the client. The best investment p. 6).
returns may be on ``consumption'' (in terms of
11
developing or maintaining human capital (b) The humanities tradition
through school fees and doctorsÕ bills, or
buying food at a time of crisis when the credit The broad set of approaches that fall under
terms on ``in-kind'' borrowing from traders this heading have their roots in the humanities.
may be exceptionally high). From this Originally geography and rural sociology were
perspective the task of the assessor is not to the ``lead'' subjects, but over the last 20 years
86 WORLD DEVELOPMENT
anthropology has become most important. Its of program sta-bene®ciary relations and
main features are an inductive approach, a fungibility) and partly because of the wide-
focus on key informants, recording by notes or spread recognition that much IA survey work
image, and the data analyst is usually directly was based on inaccurate information collected
(and heavily) involved in data collection. 12 by questionnaire from biased samples (Cham-
This tradition does not try to ``prove'' impact bers, 1993). Low-budget and low-rigor IAs
within statistically de®nable limits of probabil- claiming to adopt the scienti®c method were at
ity. Rather, it seeks to provide an interpretation best pseudo-science, but more often simply bad
of the processes involved in intervention and of science, despite the sophisticated analytical
the impacts that have a high level of plausibil- tools that were applied to poor datasets.
ity. It recognizes that there are usually dierent, IAs with their roots in the humanities have
and often con¯icting, accounts of what has considerable diculties with regard to the
happened and what has been achieved by a attribution of cause and eect. Such studies
program. The validity of speci®c IAs adopting cannot usually demonstrate the causal link as
this approach has to be judged by the reader on they are not able to generate a ``without
the basis of the logical consistency of the program'' control group (although at times
arguments and materials presented; the some researchers neglect to mention this to the
strength and quality of the evidence provided; reader and simply assume causality). Instead,
the degree of triangulation used to crosscheck causality is inferred from the information about
evidence; the quality of the methodology; and the causal chain collected from intended bene-
the reputation of the researcher(s). Whether ®ciaries and key informants, and by compari-
``standards'' could be speci®ed for such workÐ sons with data from secondary sources about
to help its users appreciate how rigorously changes in out-of-program areas. Problems
designed they areÐis an important issue that also arise because not infrequently the labels
merits attention. 13 Commonly the bulk of data ``rapid appraisal,'' ``mini-survey'' and ``case
generated by such an approach is ``qualitative,'' study'' are applied to work which has been
although at later stages of analysis such work done in an ad hoc manner and does not achieve
often quanti®es some data. The main types of a minimum professional standard in terms of
methods used have been discussed in Section 4 informant selection and the rigor of data
(in particular see Table 2). collection and analysis. Examples of this
Although such work has been common in include: basing data collection only in program
development studies for decades, it is only areas that are performing well, and surveying
during the 1980s that its relevance for IA has best clients; and inferring that the data collec-
been recognized. 14 This recognition has arisen ted in one area apply to all clients without
partly because of the potential contribution of explaining this assumption.
qualitative approaches (especially in under- While such studies cannot provide the degree
standing changes in social relations, the nature of con®dence in their conclusions that a fully
resourced scienti®c method approach can enabled to identify their own indicators,
yield, my personal judgement is that in many establish their own participatory baselines,
cases their conclusions are more valid than monitor change, and evaluate causality...''
survey based IA work that masquerades as (Mayoux, 1997, p. 123). By this means two
science but has not collected data with scien- objectives may be achieved: better impact
ti®c rigor. It is nonetheless becoming increas- assessments, and intended bene®ciaries will be
ingly common to combine ``scienti®c'' and ``...empower[ed] through the research process
``humanities'' approaches so as to check the itself'' (Mayoux, 1997, p. 2). In practice, the art
validity of information and provide added of participatory impact assessment (PIA) is in
con®dence in the ®ndings (e.g., Hashemi et al., its infancy and a pragmatic rather than a purist
1996; Hulme & Mosley, 1996; Schuler & approach has been common. Agencies such as
Hashemi, 1994). In the future, dealing with Proshika in Bangladesh have begun to use PLA
attribution by multimethod approaches seems methods extensively for their assessment and
the way forward. planning exercises. 16
The reliability of participatory methods
(c) Participatory learning and action (PLA) varies enormously, as with ``scienti®c''
surveys, depending ``...largely on the motiva-
In the last ®ve years participatory approa- tion and skills of facilitators and those inves-
ches to development planning and manage- tigated and the ways in which informants''
ment have moved from being a fringe activity perceptions of the consequences of research
to center stage. While many donor agencies are addressed'' (Mayoux, 1997, pp. 12±13).
have simply added a bit of PLA to their Nevertheless, it is argued that ``...a number of
existing procedures, it can be argued that this rigorous comparative studies have shown that,
is inappropriate as conceptually participatory when well-conducted, participatory methods
approaches challenge the validity and utility can be more reliable than conventional
of the scienti®c method as applied to devel- surveys'' (Mayoux, 1997 and also Chambers,
opmental problems (Chambers, 1997). 1997, pp. 141±146).
According to this line of argument the scien- To date the literature on PLA and PIA has
ti®c method fails as: it ignores the complexity, only partially addressed the issue of attribu-
diversity and contingency of winning a liveli- tion. From a scienti®c perspective PIA has
hood; it reduces causality to simple unidirec- grave problems because of the subjectivity of
tional chains, rather than complex webs; it its conceptualizations of impact; the subjec-
measures the irrelevant or pretends to measure tivity of the data used to assess impact; the
the immeasurable; and, it empowers profes- variables and measures used vary from case to
sionals, policy-makers and elites, thus rein- case and do not permit comparison; its
forcing the status quo and directly retarding pluralist approach may lead to a number of
the achievement of development goals. At mutually con¯icting accounts being generated
heart, PLA theorists do not believe that ulti- about causality; and, the assumption that
mately there is one objective reality that must because lots of people are taking part in an
be understood. Rather, there are multiple exercise means that all are able to ``voice''
realities and before any analysis or action is their concerns (so that opinions are represen-
taken the individuals concerned must ask tative) is naive about the nature of local power
themselves, ``whose reality counts?'' (Cham- relations. From the perspective of a ``new
bers, 1997). Their answer is that the perceived professional'' (Chambers, 1997) then such a set
reality of the poor must take pride of place as, of accounts is unproblematic, as it re¯ects the
if development is about ``empowering the complexity and contingency of causality in the
poor'' or ``empowering women'' (as virtually real world. In addition, it can be argued that
all development agencies now say), then the PIA contributes to program goals (perhaps
®rst step toward empowerment is ensuring particularly in terms of empowering women
that ``the poor'' or ``women'' take the lead in (Mayoux, 1997) and the poor) by not facili-
problem identi®cation and analysis and tating the continued dominance of target
knowledge creation. 15 groups by powerful outsiders. Why dwell on
For impact assessment the purist PLA line is issues of attribution when eorts to overcome
damningÐ``...conventional baseline surveys are such problems require the adoption of IA
virtually useless for impact assessments... The methods that actively undermine the attain-
question now is how widely local people can be ment of program goals?
88
14. Potential to contribute to stakeholder Low High Low Medium to low Very high
capacity building
15. Probability of enhancing downwards Low High Medium Medium High
accountability
16. Human resource requirements Specialist super- High skilled Medium skilled Medium skilled High skilled
vision, large practitioners, who practitioners, with practitioners with ctitioners
numbers of less are able to write- good supervision, good supervision
quali®ed ®eld up and analyse who are prepared
workers results to commit for
lengthy period
17. Cost range Very high to High to medium Medium to low Medium to low High to Medium
medium
18. Timescale Very high to Medium to low High High to medium Medium to low
medium
a
Source: Adapted from Montgomery et al. (1996).
IMPACT ASSESSMENT METHODOLOGIES FOR MICROFINANCE 89
much more feasible as less primary data personal introductions that interviewers make
collection is necessary (see Montgomery, prior to interview need to be carefully worked
1996 and later parts of this paper). out so that respondents understand why they
are being interviewed and have an opportunity
(c) Human resources for impact assessment to ask their own questions before the interview
begins.
In many, if not most, developing countries Motivation is a more dicult issue with
recruiting IA personnel who have the skills control groups as, having by de®nition no
and qualities to interview, collate, analyze and connection with a program, they have no
write up ®ndings is a key problem at both incentive to cooperate. In many cases, however,
consultant and ®eldworker levels. Commonly, the novelty and amusement value of being
dierent studies ®nd themselves competing for interviewed is sucient encouragement (though
the same small pool of people which, while it expatriates should note that when they are
may usefully raise payments for scarce skills, working at a ®eld site the willingness of people
puts these individuals under great strain and to be interviewed may be higher than is the
does not appear to stimulate a ``supply-side norm because of the rarity value of foreigners).
response''. It is beyond the capacity of this The problems of response increase signi®cantly
paper to explore this issue, but it must be if longitudinal data are collected, as second and
recognized as a key constraint and that eorts third interviews have much less amusement
to build ``impact assessment'' capacities pro- value. In such cases rewarding interviewees
fessionally and institutionally should be a should be considered to promote data quality
priority for development agencies if they and for ethical reasons (what right have impact
intend to continue to emphasize the need for assessors to assume that the opportunity costs
IA. of an interview, particularly for poor people,
are zero?). This can take the form of a social
(d) Respondents: motivation and representation reward, such as bringing soda water and snacks
to share with respondents (this works well in
A ``rational actor'' confronted by an impact East Africa), or ``bribery'' (Mosley, 1997, p. 8)
assessor asking standard IA questions (``what is where the interviewee is paid cash for surren-
your income? how do you spend your money dering her/his time. 21
on? how do you get on with your husband?'') Program dropouts represent a particular
would soon tell the interviewer where to put problem, and a failure to pursue dropouts may
his/her survey instrument. Fortunately, in the have led to some IAs underestimating the
world of practice, more polite responses are the negative impacts of micro®nance (e.g., Hulme
norm but the issue of how to persuade & Mosley, 1996). When the dropout is trace-
respondents to spare the time for an interview, able then signi®cant eort is merited to obtain
and provide accurate and honest answers, is an an interview/re-interview. Where dropouts
important one that is rarely mentioned in IA cannot be traced, or death has occurred, then a
methodological statements. Dierent strategies replacement respondent sampled at random
are needed for dierent types of respondentÐ from the original population, and preferably
program bene®ciary, control group and from the same stratum, should be interviewed
program drop out. As a rule of thumb many (see Mosley, 1997, pp. 7±8).
researchers suggest that interviews should be Participatory and rapid appraisal methods
concluded within one hour and that one and a that work with groups generally manage to
half hours should be seen as the absolute muster respondents because of the social
maximum for an interview. interaction they create. Care needs to be
Bene®ciaries are the easiest group to taken, however, to observe who has turned up
approach as generally they accept ``answering and, perhaps more signi®cantly, who has not
questions'' as one of the unavoidable transac- come to the meeting (Mayoux, 1997). The
tion costs of being in a program or dealing with assumption that participants in a PLA exer-
an MFI. Motivation can be enhanced by cise represent ``the community'' will
having interviewers introduced by program commonly not be valid (Mosse, 1994). Addi-
ocers: but, this has the danger of linking the tional interviews or focus groups may be
assessor with ®eld-level sta and encouraging necessary to collect information from people
the recounting of ``the right answers''. 20 For who do not turn up for communal PLA or
both data quality and ethical reasons the RRA sessions.
IMPACT ASSESSMENT METHODOLOGIES FOR MICROFINANCE 91
(e) The problem of ``low impact'' impact and discussing interim ®ndings are possible
assessments 22 ways of making in¯uence more probable.
features of the client population. This is 1997) demand rigorous empirical testing to ®nd
compared against a carefully selected control out what is being achieved and how more might
group, so that the number of households be achieved.
surveyed is likely to be between 750 and 1,500. The desire of MFIs, donors and impact
At least three interviews will be conducted with assessors themselves to produce results that will
each household over a period of two to three verify ®ndings about impact at high levels of
years. A much wider set of income and asset statistical con®dence has too often driven the
variables will be measured (Gaile, 1997, p. 23) design of IA studies. This can compromise
and the focus will be on high precision through quality (with small sample surveys claiming
interval measurements. A set of related studies exaggerated levels of representativeness) and
on institutional performance would be conduc- impact (with external, data-extraction approa-
ted, but the heart of the study would be the ches making MFI sta unwilling to use ®ndings
statistical and econometric analysis of survey and contradicting the ``empowerment'' goals of
®ndings. The budgets for such approaches are many programs). This paper has argued that
likely to exceed a million dollars as the survey IA eectiveness should not be automatically
costs are high, data processing and analysis equated with the level of scienti®c ``proof'' that
inevitably generate problems and signi®cant a study can claim. While all studies must pursue
amounts of high-powered and high-cost econo- rigor, and this applies equally to quantitative
metrician and/or statistician time is needed. and qualitative work, the eectiveness of an IA
will depend on how well it achieves a ®t
between its objectives, the ®nancial and human
7. CONCLUSION resources it can command and its context.
There is no optimal model and dierent
In recent years donors have been keen to designsÐcharacterized in this paper as ``low,''
assess the impact of their programs. The initial ``moderate'' and ``complex'' and combining
emphasis on ``scienti®c'' sample surveys and scienti®c, humanities and participatory
statistical analyses has shifted as multimethod approachesÐwill be appropriate for dierent
impact assessment studies and most recently studies. All too often, however, donor desires
participatory approaches have been utilized. for objective and external IAs (to meet their
Micro®nance programs and institutions have domestic accountability requirements) lead to
experienced these shifts and examples of IAs on the neglect of a key alternative: strengthening
this topic provide a resource from which this the impact monitoring capacity of the MFI
paper has sought to draw out lessons for future itself. While striving for technical best practice
practice. Much further work will be needed as should be a key goal for all in this ®eld it would
the claims that micro®nance is a panacea for be foolish not to recognize that IA is a ``bat-
poverty-reduction (most publicly through the tle®eld of knowledge'' (Long & Long, 1992) in
Micro-Credit Summit and its follow up), and which dierent actors seek to in¯uence the
the counterclaims that caution against such knowledge creation process so that it meets
enthusiasm (Rogaly, 1996; Wood & Sharrif, their needs.
NOTES
1. For a discussion of impact assessment for other This is an important point as, (a) some MFIs present
forms of services to microenterprise see Hyman and their client populations as intended bene®ciaries when
Dearden (1998). many clients are known to be nonpoor, and (b) some
agencies (e.g., the Consultative Group to Assist the
2. The impact chain model not only underpins ex post Poorest) present the client populations of MFIs as ``the
assessment but is central to most aid ®nanced activity in poorest'' when they know full well that only a propor-
the form of a variety of ``logical frameworks'' that tion of clients are poor and that few if any are ``the
donors use to design projects and programs. poorest'' i.e. social outcasts, destitute, disabled, refugees,
widows or elderly.
3. I use the term intended bene®ciary here, rather than
client, as most MFIs utilize (or have utilized) aid funds 4. The assessment of institutional sustainability has
that are intended, at least in part, to bene®t poor or been greatly advanced by YaronÕs subsidy dependence
vulnerable people and not purely self-selected clients. index (SDI) which permits the assessor to move beyond
94 WORLD DEVELOPMENT
simple statements of pro®t and loss quali®ed by 13. This issue is raised in Hulme (1997) and was
footnotes about changes in levels of subsidy. debated at the CGAP Impact Assessment Group virtual
meeting.
5. The grandest invalidation is probably the United
Kingdom in the mid to late 1980s when an unprece- 14. As Moris and Copestake (1993, p. 1) point out ``the
dented increase in outreach and pro®tability of institu- much recommended text on data collection by Casely
tionsÐproviding ®nancial services for households and and Lury (1982)... included two cursory paragraphs on
small enterprises created a ``bubble'' which ultimately ``quick and dirty'' techniques... almost half of the World
destroyed many enterprises, led to some losing their Bank's publication that superseded it (Casley & Kumar,
homes and impoverished hundreds of thousands of 1988) is concerned with qualitative methods.'' Studies
households. Also note that Mayoux (1997) has found using this approach include Hulme and Mosley (1996),
that programs performing well in terms of outreach and Rutherford (1993), Fuglesang and Chandler (1986 and
repayment rates can have negative impacts in terms of 1993). For details of many recent examples see Mayoux
womenÕs empowerment. (1997).
6. I cannot locate my copy of a study on the USAID 15. The reader wishing to explore PLA and PRA
Anti-Poverty Lending Program! This study ably (participatory rural appraisal) is referred to Chambers
computed almost every measure of outreach, outreach (1997) and the references he provides as there is not
growth and institutional ®nancial health possible space to more fully explore these ideas in this paper.
however, it failed to contain any information on See Mayoux (1997) for a discussion of empowerment
``who'' clients were, but simply assumed they were ``the as a program goal, with particular reference to
poor.'' gender.
8. This section draws heavily on Mosley (1997) excel- 17. This section draws heavily on the work of Mont-
lent paper. gomery, Davies, Saxena and Ashley (1996).
9. This method is used widely in the medical and 18. For a full discussion see the excellent study by
agricultural sciences. Montgomery et al. (1996).
10. We were not able to use this method for the Thrift
19. The only report to hand that explicitly identi®es
and Credit Cooperatives in Sri Lanka, as the ``early''
absolute costs and relative costs (as a percentage of
members of a cooperative are generally drawn from
program budgets) is Montgomery et al. (1996) mainly in
higher income/assets/status groups while ``late'' joiners
relation to natural resource and social programs. It
are from lower income/assets/status groups. In this case
reports that the 1994 IA of BRACÕs credit program cost
the use of a ``clients-to-be'' control group would have led
US $250,000.
to an exaggerated assessment of the economic impacts of
micro®nance.
20. At one large Asian MFI (Hulme and Mosley,
11. Commonly such work is referred to as ``qualita- 1996, Vol. 2), program ®eld-sta visited villages that
tive,'' but the quantitative/qualitative dichotomy is a had been randomly selected for survey and told
false dichotomy. Most quantitative studies extract borrowers to make sure they gave interviewers ``the
qualitative data from the respondent, have the inter- right answers.'' Fortunately, qualitative research
viewer immediately codify or quantify it and then use revealed this and other villages were selected for
only numerical analysis. Many ``qualitative'' studies survey!
transform their data into quantities at later stages of
analysis (Moris & Copestake, 1993, p. 4). 21. This method is widespread in the United States and
United Kingdom when market researchers convene
12. For a classic example of such work see Rutherford focus groups to test new products. It should be noted
(1999). that once interviewees in an area are paid for interviews
IMPACT ASSESSMENT METHODOLOGIES FOR MICROFINANCE 95
then the likelihood of noncooperation in the future, 24. See Montgomery et al. (1996) for an excellent
unless fees are paid, is greatly increased. discussion of the role of impact monitoring vis
a vis
impact assessment.
22. An anonymous reviewer pointed out that, like
most IA theory and practice, this paper is dominated by 25. For example, for BRAC in 1994 and 1995 vast
supply-side issues and neglects ``use'' i.e., the demand amounts of technically valid data collection and analysis
side. This is a valid criticism especially given my earlier were occuring for the Impact Assessment Study
work which indicated that some development agencies (Mustafa et al., 1996) and the World Bank-BIDS
actively ``do not learn'' (Hulme, 1988) and my continued Bangladesh MFI study. The main source of information
belief that the World Bank has a learning disability and ideas for BRACÕs ®ve-year plan (1995±2000) came
(Hulme, 1992 and 1994). from a rapid informal and focus group research exercise
carried out by research ocers from the Research and
23. Unfortunately donor agencies generally lack the Evaluation Department who were ``pulled out'' of the
courage to reach this decision and consultants (mea technical studies, along with discussions with BRAC
culpa) may have a vested interest in not promoting this sta at headquarters and in the ®eld (F. Abed, personal
option. communication).
REFERENCES
Adams, D., & von Pischke, J. D. (1992). Microenterprise Gaile, G. L. (1997). Highlights and recommendations
credit programs: Deja vu. World Development, 20, from the microenterprise impact virtual meeting.
1463±1470. Washington DC: Management Systems Interna-
Ardener, S., & Burman, S. (1995). Money-go-rounds: the tional.
importance of rotating savings and credit associations Gaile, L., & Foster, J. (1996). Review of methodological
for women. Oxford: Berg. approaches to the study of the impact of microenter-
Barnes, C. (1996). Assets and the impact of microenter- prise credit programs. Washington, DC: Manage-
prise ®nance programs. Washington, DC: Manage- ment Systems International.
ment Systems International. Goetz, A. M., & Sen Gupta, R. (1996). Who takes the
Bouman, F. J. A., & Hospes, O. (1994). Financial credit? gender, power and control over loan use in
landscapes reconstructed. Boulder, Co: Westview. rural credit programs in Bangladesh. World Devel-
Buckley, G. (1997). Micro®nance in Africa: Is it either opment, 24 (1), 45±64.
the problem or the solution? World Development, 25 Hashemi, S. M., Schuler, S. R., & Riley, A. P. (1996).
(7), 1081±1094. Rural credit programs and womenÕs empowerment in
Casely, D., & Lury, D. A (1982). Monitoring and Bangladesh. World Development, 24 (4), 635±654.
evaluation of agricultural and rural development Holcombe, S. (1995). Managing to empower: the Gram-
projects. Baltimore: Johns Hopkins University Press. een BankÕs experience of poverty alleviation. London:
Casley, D., & Kumar, K. (1988). The collections, analysis Zed Press.
and use of monitoring evaluation data. Baltimore: Hossain, M. (1988). Credit for alleviation of rural
John Hopkins University Press. poverty: the Grameen Bank in Bangladesh. Washing-
Chambers, R. (1993). Challenging the professions: fron- ton, DC: IFPRI.
tiers for rural development. London: IT Publications. Hulme, D. (1992). Enhancing organizational eective-
Chambers, R. (1997). Whose reality counts? putting the ness in developing countries: the training and visit
®rst last. London: IT Publications. system revisited. Public Administration and Develop-
Chao-Bero, R. (1997). Impact assessment seen from an ment, 12 (5), 433±445.
NGO practitionerÕs scope mimeo. Washington, DC: Hulme, D. (1994). Does the World Bank have a learning
Consultative Group to Assist the Poorest (CGAP). disability? Public Administration and Development, 14
Chen, M. A., & Dunn, E. (1996). Household economic (1), 93±97.
portfolios. Washington, DC: Management Systems Hulme, D. (1988). Learning and not learning from
International. experience in rural project planning. Public Admin-
Feinstein, O. (1997). A transaction costs approach for istration and Development, 9 (1), 1±17.
micro®nance project assessment mimeo. Washington, Hulme, D. (1997). Impact assessment methodologies for
DC: Consultative Group to Assist the Poorest micro®nance: A review. A paper prepared for the
(CGAP). virtual meeting of CGAP Working Group on Impact
Fuglesang, A., & Chandler, D. (1986). Participation as Assessment Methodologies, April 17±19.
process: what we can learn from the Grameen bank. Hulme, D., & Mosley, P. (1996). Finance against poverty,
Oslo: NORAD. Vols. 1 and 2. London: Routledge.
Fuglesang, A., & Chandler, D. (1993). Participation as Hyman, E., & Dearden, K. (1998). Comprehensive
process ± process as growth. Dhaka: Grameen Trust. impact assessment systems for NGO microenterprise
96 WORLD DEVELOPMENT
development program. World Development, 26 (2), Peace, G., & Hulme, D. (1994). Microenterprise and
261±276. children what are the intra household impacts of
Khandker, S. R. (1998). Fighting poverty with micro- income-generating programmes. Small Enterprise
credit: experience in Bangladesh. Washington, DC: Development, 5 (1), 21±29.
Oxford University Press for the World Bank. Pulley, R. (1989). Making the poor creditworthy: a case
Little, P. (1997). Assessing the impact of micro®nance study of the integrated rural development program in
programs on incomes and assets mimeo. Washington, India. World Bank Discussion Paper 58. Washing-
DC: Consultative Group to Assist the Poorest ton, DC: World Bank.
(CGAP). Remenyi, J. (1991). Where credit is due. London: IT
Long, N., & Long, A. (1992). Battle®elds of knowledge. Publications.
London: Routledge. Rogaly, B. (1996). Micro-®nance evangelism, destitute
Martyn-Johns, E. (1996). Participatory Impact Assess- women and the hard selling of a new anti-poverty
ment: Putting Theory into Practice. MA dissertation, formula. Development in Practice, 6 (2), 100±112.
IDPM, University of Manchester, UK. Rutherford, Stuart (1999). The Poor and Their Money,
Mayoux, L. (1997). Impact assessment and womenÕs Finance and Development Working Paper No. 3.
empowerment in micro®nance programmes: issues for IDPM: University of Manchester.
a participatory action and learning approach mimeo. Rutherford, Stuart (1993). Learning to lend: informal
Washington, DC: Consultative Group to Assist the savings and credit schemes in Bangladesh. SCF
Poorest (CGAP). Working Paper No. 5, London: Save the Children
Montgomery, R. (1996). Disciplining or protecting the Fund.
poor? avoiding the social costs of peer pressure in Schuler, S. R., Hashemi, S. M., & Riley, A. P. (1997).
microcredit schemes. Journal of International Devel- The in¯uence of womenÕs changing roles and status
opment, 8 (2), 289±305. in BangladeshÕs fertility transition: evidence from a
Montgomery, R., Davies, R., Saxena, N.C., & Ashley, S. study of credit programs and contraceptive use.
(1996). Guidance materials for improved project World Development, 25 (4), 563±576.
monitoring and impact review systems in India. Centre Schuler, S. R., & Hashemi, S. M. (1994). Credit
for Development Studies, University College, Swan- programs womenÕs empowerment and contraceptive
sea, UK. use in Rural Bangladesh. Studies in Family Planning,
Moris, J., & Copestake, J. (1993). Qualitative enquiry for 25 (2), 65±76.
rural development: a review. London: IT Publications. Sebstad, J., Neill, C., Barnes, C., & Chen, G. (1995).
Mosely, P., & Hulme, D. (1998). Microenterprise ®nance: Assessing the impacts of microenterprise interventions:
is there a con¯ict between growth and poverty a framework for analysis. Washington, DC: USAID.
alleviation. World Development, 26 (5), 783±790. Wiig, A. (1997). Credit expansion in microcredit pro-
Mosley, P. (1997). The use of control groups in impact grammes: dilemmas and feasible methods for studying
assessment for micro®nance mimeo. Washington, DC: them mimeo. Washington, DC: Consultative Group
Consultative Group to Assist the Poorest (CGAP). to Assist the Poorest (CGAP).
Mosse, D. (1994). Authority gender and knowledge: Wood, G., & Sharrif, S., (1997). Credit where credit is
theoretical re¯ections on the practice of participatory due. Dhaka and London: UPL and Zed Press.
rural appraisal. Development and Change, 25 (3), Yaron, J., Benjamin, M., & Piprek, G. (1997). Rural
497±525. ®nance: issues, design and best practices. Washington,
Mustafa, S. et al.(1996). Beacon of hope: an impact DC: Agriculture and Natural Resources Depart-
assessment study of BRACÕs rural development ment, World Bank.
programme. Dhaka: BRAC. Zadek, S., & Gatward, M. (1996). Transforming the
New Economics Foundation (1996). A social audit Transnational NGOs: Social Auditing or Bust? In
manual for small organizations. London: New M. Edwards (Ed.), Beyond the magic bullet: NGO
Economics Foundation. performance and accountability in the post-war world
Otero, M., & Rhyne, E. (1994). The new world of (pp. 226±240). West Hartford: Kumarian Press and
microenterprise ®nance. London: IT Publications. London: Earthscan.
Sample surveys are appropriate Rapid appraisal and/or PLA are appropriate Participant observation and/or case studies are
when: when: appropriate when:
A project aects large numbers of bene®ciaries The project is adopting or promoting participatory An understanding of the motivations and
principles in (re-)planning, implementation, moni- perceptions of project clientele is a priority
toring and evaluation
Policy-makers require accurate estimates of An understanding of the motivations and perceptions Other methods (surveys and rapid appraisals) are
project impacts of project clientele is a priority unlikely to capture the views of minorities or
women
Statistical comparisons must be made between One of the purposes of the study is to assess whether One of the purposes of the study is to assess
groups over time and/or between locations or not felt needs are being addressed by the project whether or not felt needs are being addressed by
the project
Project delivery/implementation mechanisms are The impact of community-based organizations or The impact of community-based organizations or
operating well, thereby justifying investment in other institution-building activities are of importance other institution building activities are of impor-
the assessment of impacts tance
The target population is heterogeneous and it is There is a need to understand the quality of other There is a need to understand the quality of other
dicult to isolate the in¯uence of factors unre- data collected through surveys data collected through surveys or rapid appraisals
lated to the project (e.g., contextual variables, (e.g. causal processes)
other programs etc.)
There is a need for contextual studies before design- There is a need for contextual studies before
ing more complex monitoring or impact assessment designing more complex monitoring or impact
exercises (e.g., case studies, or surveys) assessment exercises (e.g., before carrying out
rapid appraisals, or before designing a survey)
IMPACT ASSESSMENT METHODOLOGIES FOR MICROFINANCE
TableÐContinued
Sample Surveys are usually not appropriate Rapid Appraisal and/or PLA are not appropriate Participant Observation and/or Case Studies are
when: when: usually not appropriate when:
A project aects small numbers of bene®ciaries Projects are relatively uncomplicated in which The project is small and ``uncomplicated,''
bounded locations are not units of analysis (e.g., providing a speci®c service or limited intervention
Health centers serving a wide catchment area, or large which is unlikely to aect community dynamics
schools serving a variety of communities) beyond a few speci®c eects (e.g. disease-speci®c
health facilities or campaigns)
Policymakers are mainly concerned with project Indicators of project impact are uncontroversial, and Bounded locations are not units of analysis (e.g.,
outcomes. (Was infrastructure completed on time negative impacts are unlikely health centers serving a wide catchment area, or
and within budget? How many people us the large schools serving a variety of communities)
health clinics?)
Project implementation is recent and untested, Standardized and statistically representative general- Indicators of project impact are clear and easily
and it is likely that the way in which the project is izations for a large and diverse project population are measurable or assessable (by survey or rapid
implemented will have little impact at the present regarded as the sole priority appraisals)
time
The purpose of the assessment is to study and Participation of clientele is not a priority (e.g., in Indicators of project impact are uncontroversial,
evaluate complex activities or processes (e.g., the public administration or power sector reform, or an and negative impacts are unlikely
development & operation of community-based organizational change programÐin these types of
WORLD DEVELOPMENT
organizations; qualitative use of skills as a result projects more limited focus group discussions with
of training programs) sta may be more appropriate)
The purpose of the assessment is to document Information is needed quickly, and standardized,
easily observable changes in the physical envi- representative (statistical) generalizations are
ronment or other tangibles (which can be regarded as the sole priority
assessed through simpler, structured visits)
The purpose of the assessment is to understand
whether or not the project is meeting the felt
needs of the project clientele
Source: Adapted from Montgomery et al. (1996).