|
|  |
Description of the Research Study
2.1 Prior Thoughts and the Design of the Field Survey
The
study had three primary objectives: generate a panel data set at the household
level with a view to monitor the performance of the MFIs; use this data to evaluate
the impact of microcredit; and compare between different institutional arrangements
in the field of microfinance in Bangladesh. Thus, much of the study design concentrated
on selection of MFIs and households for the purpose of the repeat surveys. The
initial design aimed at comparing between program and control villages, and between
participants and non-participants in the program villages. Thus, a three-stage
sampling procedure was followed to select the study villages. It included, selection
of POs, selection of thanas under selected POs if there were more than one thana
covered by the PO, and selection of villages. Such selections were sequentially
done; first a sample of POs was selected, followed by PO-thana(s) covered under
the PKSF program, and finally, selection of program and control villages in these
thanas. The details of the selection procedure are outlined below.
Selection
of POs
There are several ways one can distinguish
between program designs. One general approach is in terms of “credit only” and
“credit plus”. It is, however, to be acknowledged that any credit operation involves
a minimum set of activities that include group formation, savings generation and
some degree of training on group discipline. At the other end, “credit plus” programs
may have a wide number of shades. The program may also vary in terms of approaches
to group formation, interest rates charged on lending, methods and time schedule
of repayment, etc. In the absence of a
complete census on POs, we collated basic desk-level information on POs, available
from PKSF, to construct the PO-universe. This provided us with information on
activities, area of operation, current status with PKSF, share of PKSF in PO’s
total loan portfolio, interest rate charged, etc. Thus, information on 138 POs,
broadly grouped into BIPOOL (Big Partner Organizations Operating in Large Areas) and OOSA (Organizations Operating in Small
Areas)
category, was compiled. First category includes 3 POs, viz. Proshika, ASA and
BRAC. There were 135 active Oosa category
POs during 1997. Of these, 18 were not included since some of them were reported
to have had some kind of problems in repayments and/or disbursement had been suspended;
and few others were only recently included as PKSF’s PO with very low coverage.
Based
on the activities that POs engage in, three broad categories were identified;
1.
Those who engage in only credit operation without any additional support to the
clients; and have no involvement in providing non-financial services;
2.
Those who engage in providing supports to income generating activities, such as
in the form of providing relevant skill development training and providing marketing
support and extension services (they may or may not engage in providing other
non-financial services); 3.
Those who engage in providing non-financial services, such as, general education,
health and sanitation, etc., without engaging in direct marketing support.
Of the 13 selected POs, two were purposively
chosen from the Bipool (large) category. Since BRAC has been relatively more frequently
studied than the others, we chose Proshika (a ‘credit plus’ NGO) and ASA (a ‘credit
only’ NGO) in our sample. Capturing diversity
in program packaging has been the main criterion in the selection of eleven other
sample POs from the list of currently active ones (see Table 2.1), whose partnership
with the PKSF are not expected to be withdrawn in the near future.
Table 2.1 Distribution
of Oosha Partner Organizations in terms of their activities
| Category |
Final universe of POs (number) |
Number of selected PO |
|
Engage in credit only | 10 | 3 |
|
Credit and marketing support | 9 | 2 |
|
Credit and other non-financial services | 98 | 6 |
|
Total | 117 | 11 |
Selection
of Villages
Within
a PO area, selection of villages involved visiting the PO local offices and some
key informants were interviewed to prepare a list of all villages in the area
and compile village specific information on type of PO-activities, number of PO
groups/samity, number of PO borrowers, infrastructure condition and on existence
of other NGO activities. Upon obtaining these information, a tentative sample
of villages under each of the selected PO was drawn through stratified random
sampling. All villages in a study area were stratified in two dimensions, The
first was to group them into ‘program’ and ‘control’ (non-program) villages. At
a second level, the ‘program’ villages were stratified further in terms of variations
in NGO activities undertaken. The program villages were further grouped into infrastructurally
developed, moderate (medium) and backward; as well as in terms of broad types
of NGO programs present in the area. Major clusters from the final selection had
been made only upon further verification of the village characteristics at the
time of administering the census.
Thus, the villages were
at first grouped into the following 4 categories;
1.
Villages with no MFI/NGO (control village);
2.
Villages with (PKSF) POs only;
3.
Villages with non-PO MFI/NGOs along with POs; and
4.
Villages with
other MFI/NGOs, but no PO involvement.
We
excluded the fourth category; and for each of the PO area, two control villages
were to be selected from category 1; and 6 program villages from the set defined
by categories 2 & 3. Since control villages could not be found in some of
the study area, only 11 control villages were finally included, and subsequent
revision included 8 program villages in one of the 13 PO areas (TMSS in Bogra).
Selection of Households
In selecting the survey households,
the universe of households in program villages, drawn from the census, was grouped
into four: eligible participants, eligible non-participants, non-eligible participants
and non-eligible non-participants. The criterion for eligibility varies across
MFIs. We chose to define it by the more commonly applied criterion – those owning
50 decimals or less cultivable land. Participation was defined in terms of current
membership – as reported during the census in 1997. Since participants were not
to be found in control villages, “potential participants” were identified to include
those whose heads expressed their willingness to participate in MFI programs,
if available. It is important to note that the survey did not confine to members
of selected POs. The selection of the latter facilitated identification of the
regions – and the study included members of MFIs as participants, irrespective
of their affiliation. The sample size for each of the four categories was pre-determined,
and was not necessarily proportional to the distribution observed in the census.
However, within each group, the sample households were drawn randomly from the
census households.
2.2
After-thoughts and the Method of Data Analysis
A number of caveats in the analytical approach
underlying the survey design gradually got revealed as the study progressed. The
very first was the absence of adequate number of control villages, which made
program-control comparisons less attractive. This was more so once we found that
MFI programs had encroached into areas previously selected as control villages.
A more important problem arose due to an initial premise that participation would
remain stable (except for permanent drop-outs) throughout the survey period. Interestingly,
the three rounds of surveys revealed that there is significant presence of occasional
participants, who join a program in one year, drop out in the next year, and re-joins
in the same or another program in the subsequent year.
These two considerations led the research team to reformulate the sample categories.
Since the survey, administered in three rounds, enabled capturing changes of a
matched sample, the sample households were grouped into six broad categories:
(i) regular participants
(i.e., current participant in all three rounds);
(ii)
complete drop-outs (who reported of previous participation during the first survey,
but were not current participants during any of the three rounds);
(iii) recent drop-outs (who were current participant during the first round,
but not in subsequent rounds);
(iv) new participants
(who did not participate in MF programs previously, but were current participants
during the last two rounds);
(v) non-participants
(those who never participated in any MFI program ever); and
(vi) occasional participants (those who had participated on and off). Distribution
of surveyed households, in terms of these six categories, is presented in Table
2.2 below; and the region-wise information are relegated to Tables 2.A.1 and 2.A.2
in the annexure at the end of this chapter.
Table 2.2
Distribution
of Surveyed Households in terms of Participation Status(Percentage)
Participation
Status | Program
villages |
Control
villages | | Regular
Participants | 40.42 | 1.80 |
| Complete
Drop-outs | 2.90 | - |
| Recent
Drop-outs | 7.89 | - |
| New
Participants | 1.94 | 6.12 |
| Non-participants | 28.30 | 76.62 |
| Occasional
Participants | 18.55 | 15.47 |
| Number
of Households | 2625 | 278 |
Note: Details by regions
are presented in Appendix Tables A.1 and A.2. Presence of Participants in control
villages was noted during the first round survey. | |
|
