1 / 23

Rural micro-finance in Morocco Lessons from an on-going randomized study

Rural micro-finance in Morocco Lessons from an on-going randomized study. Tanguy Bernard Agence Française de Développement (AFD) Dakar, February 2, 2010. Outline. The planned study Implementation difficulties Some lessons. 1. Planned study. Context.

halona
Télécharger la présentation

Rural micro-finance in Morocco Lessons from an on-going randomized study

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Rural micro-finance in MoroccoLessons from an on-going randomized study Tanguy Bernard Agence Française de Développement (AFD) Dakar, February 2, 2010

  2. Outline • The planned study • Implementation difficulties • Some lessons

  3. 1. Planned study

  4. Context • Al Amana: the largest Morrocan micro-credit institution (495,000 active clients in 2005). • Extension of activities to rural areas. • 2004/2005: peri-urban areas • 2006 - … : remote rural areas • Progressive extension to remote areas: a unique opportunity to assess impact of micro-finance (mostly undocumented thus far): • Interesting for the program itself (~accountability, learning) • Interesting for all such type of actions in developing countries.

  5. Partnership • Operator: • Al Amana, committed to the evaluation study • Willing to affect its operations for IE purpose • Research team: • Top-level experts on the conduct of impact evaluation • J-PAL, PSE, INSEE-CREST • Funder: • AFD, covering all (direct) costs • Interested in learning more from impact evaluation studies. •  A winning combination!

  6. Basic set-up • Al amana extention to rural areas in 2006: • Four consecutive waves to set up 73 antennas throughout the country. • Each antenna reachs ~10 villages • IE design: • In each antenna, draw pairs of comparable villages on basis of distance, size, pop density, altitude etc. • In each pair, randomly assign control status to one village, for one year. • Impact measure: • Compare those who have borrowed in treatment villages, to those who would have borrowed in control group.

  7. Central zone Always served Zone B Zone A Treatment or Control: random allocation

  8. Dealing with imperfect compliance • Power of estimation relies on take-up rates of micro-credit products. • Non-borrowers not informative for Treatment on the Treated (ATT). • If take-up rate low, large data collection effort may still lead to limited capacity to observe impact. • Raise power by surveying those with higher probability to borrow. • Wave 1: Survey 100 hh, wait for 5 months, identify best predictors of participation. • Later waves: run mini-survey (best predictors) on 100 hh, apply parameters from wave 1 model, select 25 hh with higher probability. • A very purposive sampling based on significant data collection effort.

  9. Summary: planned steps • Prepare questionnaires • Study sites from wave 1 and split into three zones • Select tretament and control from peripheric zone • Full questionnaire on 100 hh over 10 sites from wave 1 (2000 hh) • Define predictive model for take-up. • Mini-survey and selection of hh to receive full survey • Full baseline survey on selected hh. • Al Amana offers micro-credit in treatment villages. • Midline survey and intermediary results • Al amana offers micro-credit in contrôle villages. • Final survey and final results

  10. Well-designed randomization

  11. 2. Implementation difficulties

  12. Low take-up rate • Much weaker than expected. In itself an interesting question. Reasons? • Product is ill-adapted to farm economy? Too expensive? Etc. • Product is new, there is learning curve? Depends on others’ behavior? There are other sources of credit

  13. Steps taken • Al Amana: increase take-up rate: • Delay expansion (few months) after wave 1, to better calibrate take-up model. • Keep control zones virgin for two years (instead of one). • Increase sensibilization in survey zones • Weekly follow-up with agents in survey zones. • Raise incentives for agents (transport + per-contract incentive). • Remove women quotas. • Research team, AFD: • Study on take-up rates (qualitative) • increase sample size.

  14. Take-up rate, sample size and MDE

  15. Total number of surveys • Mini-survey (waves 2, 3 and 4), one round only • = 15236 households • Full survey • Baseline: Wave 1 (100/village) + waves 2, 3 and 4 (25/ village) • = 6939 households • Midline: Wave 1 only (100/village) • = 1516 households • Endline: Wave 1 (100/village) + waves 2, 3 and 4 (25+9/village) • = 6318 households

  16. Data quality issues • Significant issues with data entry: quality not satisfactory • Re-enter entire database from baseline • Change data entry operator for endline. • Some issues with data collection • Add controlers external to data collection firm • Random resample of households to assess quality • Raise price paid to data collection firm • Drop some pairs of villages (low data quality (2), low credit take-up (1), contagion (5), misplacement (3)) • Some issues from matching hh database and Al Amana Database • Rounds of data collection initially meant to occur at same time, but could not.

  17. 3. Lessons

  18. What is the treatment? • Treatment varies by length of take-up • Possible to ‘control’ for it, but difficult for interpretation. • Type of treatment varied thoughout implementation • Initially group based, then individuals • Repayment period also extended during the program • Removal of women quotas

  19. What is the population? • At the beginning, representative on all Morocco, then a number of villages were dropped. • Only those villages that have a ‘pair’, not all villages. • Only remote rural areas. • A very particular population selected by model, difficult to replace in general population.

  20. What is the impact on borrowers? • Minimum detectable effect: 20% • Impact on specific populations can be detected? • Impact expected in the short run • Also need outcome indicators that are likely affected in SR,and that will clearly affect poverty in medium run  Need a clear theory of impact mechanisms • Take-up rates interesting per se, worth studying. • Cf: How to give access to the poor (credit, insurance etc.)

  21. Overall… • Doing an impact evaluation can be difficult… • Even with top research team, commited operator and funder • Reports rarely mention these difficulties that are however important determinants of the lessons that can be generated… • This is especially the case for programs with • Low take-up • Not matured ‘products’ • Mature  possibility to assess take up rates and expected impact size through previous M&E, literature, experts knowledge

  22. Operational learning vs General learning? • Operational learning • Projects are complexe. Impact depends on several components and means to implement  Black-box effect • MDE: what is the impact that one finds satisfactory? • Cf Cost-benefit analysis • Cf results from other robust evaluations with similar obectives • Use project as it is implemented • General learning • Interested in identifying mechanisms, how relieving constrains leads to behavioral response that affects final outcome. • Need « simple » treatment: just the active ingredient  Affect design of project • Sometimes difficult to reconcile these diverging expectations. • One needs to know, right from the beginning what the expected lessons are.  Preparation is the key!

  23. Further thoughts… • Use local research centers to conduct impact evaluations • May help refine question and intepret results • Can help support local research capacities • Eventually build mixed research teams • Link up with data collection exercise by national statistical agency • E.g. The Beninese integrated modular survey

More Related