With supports from Microcredit Summit Campaign

Introduction
The task of the Bangladesh Expert Panel is to use microcredit research and national level poverty research to estimate the number of microcredit clients in Bangladesh who were living below US$1 a day at the time of their first loan and who have crossed that threshold, between 1990 and 2007. As noted in the Terms of Reference, the Microcredit Summit (MCS) is not seeking to establish causality between microcredit and poverty alleviation.
 
With the above perspective, the Lead Researcher is expected to review existing data, conduct additional independent research (using secondary data) based on the recommendations received at the first in-person meeting and draft a paper on the results and report findings back to the entire panel and advisors at the second in-person meeting. It is further expected that the Expert Panel will submit to the Microcredit Summit Campaign a paper outlining the panel’s research and estimation of the number of clients in Bangladesh who moved above the US$1 a day threshold (1990-2007).

The following section briefly identifies the various practices in estimating the number of clients crossing the threshold every year. The third section proposes an approach that may be adopted to estimate the time series for Bangladesh. Section 4 identifies various data sources and basic information available to allow estimation of the time series. The note is concluded with an outline on proposed future activity for the exercise.

Current Practices
There are several agencies in the microcredit industry, which attempt to estimate the number of clients crossing some threshold income or fulfill a set of livelihood indicators. We discuss these below before outlining the approach of the MCS group reflected in Mark Schreiner and Emilio Hernández’s writings.
 
Grameen Bank reports on the number of their clients who cross a threshold livelihood level every year; no attempt is made to translate these into US$ 1 a day based graduation. The estimates are based on information they gather from sample surveys administered
 
Brac carried out a longitudinal survey of selected households for the purpose of internal assessment of program impacts. Reportedly, surveys were carried out in three years, 1992, 1996 and 2001. The original sample consisted of 1200 households, among whom 800 were MFI participants and 400 were non-participants. Subsequently, there had been missing households and it is understood that a matched sample of 400 households; unfortunately these were not made accessible.

While ASA reportedly has no such practice, some of the MFIs resort to some crude methods to report on the number of their clients graduating out of $1 a day threshold. One such method is to consider clients borrowing less than Tk. 10,000 per year as poor, and consider the percentage of them borrowing more than the threshold amount in the following year as crossing the threshold of US$ 1 a day. Proshika reported of undertaking occasional surveys where an annual income per household (approximately, Tk. 48,000 in 2002) was chosen as the cut-off point that allowed them to arrive at figures crossing the poverty barrier. It is informally learnt that TMSS also adopts a simple rule for arriving at number of clients going above $ 1 a day every year; but neither Buro Tangail nor SSS has any such practice.

PKSF does not estimate the number of clients crossing a threshold. However in their mapping exercises, PKSF estimates the number of poor clients and percentage of poor households effectively covered by the MFIs. These estimates are derived from a special study undertaken by PKSF in 16 selected upazilas from all (six) divisions, with three sites around a growth point in each upazila. Around 4500 clients were surveyed in 2003 and 2005 to estimate the overlapping membership. The data allowed estimation of ‘effective coverage’ (1).3 Estimation of the number of poor households (2) was obtained from small area poverty estimates of BBS. BIDS study (on PKSF-MES) provided estimates on percentages of clients who are poor, which was used, adjusting for the effective coverage (1), to calculate the number of poor clients and the number of poor client households (3). Thus, the percentage of poor households covered by the MFIs was calculated upon dividing (3) by (2). Since summing of individual MFI achievements will exaggerate the estimates in the presence of overlapping membership, it is important to account for the extent of overlaps. This is accounted for in the exercise undertaken later in this paper.

First paper by Sajjad Zohir