## wp18208

## Source details

**Canonical URL:** [wp18208](https://www.imf.org/-/media/files/publications/wp/2018/wp18208.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2018/wp18208.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2018/wp18208.pdf.json)

---

### Macroeconomic and institutional context
- Rwanda: landlocked East African country with a population of 11.5 million.
- Structural characteristics: large rural population, few natural resources.
- Growth and competitiveness:
  - Annual GDP growth averaged 7.8% between 2008 and 2016.
  - Per capita income doubled between 2008 and 2016 (IMF, 2017a).
  - 2018 World Bank Doing Business: Rwanda ranked 2nd in Africa and 41st in the world for ease of doing business.
  - Ease of getting credit: ranked 3rd in Africa and 11th in the world.
  - 2016–2017 World Economic Forum Global Competitiveness Index: Rwanda ranked 52nd among 138 countries, outperforming the Sub-Saharan Africa average on all dimensions other than market size.
- Vision 2020 reforms accompanied by reductions in poverty and income inequality (IMF, 2017b).

### Banking sector structure and microfinance
- Aggregate and sectoral metrics:
  - Total bank assets rose from 22% to 39% of GDP from 2008 to 2016.
  - Bank credit to the private sector grew at an annual average of 13% in real terms (2008–2016).
  - Commercial banks represent about two-thirds of total banking sector assets.
  - Market concentration: the 3 largest commercial banks (out of 17) account for more than half of total bank assets, loans and deposits.
  - Most banks are foreign-owned; majority of bank funding is domestic and from local deposits.
- Microfinance sector:
  - Total MFIs: 523, including 416 municipal credit cooperatives (U-SACCOs).
  - MFIs account for almost 6% of total bank assets.
  - U-SACCOs: typically one per municipality (416 municipalities); some only provide savings accounts, others also grant loans.

### Financial inclusion trends
- Access to formal financial services increased from 21% to 68% of the adult population between 2008 and 2016.
- Access to formal credit increased from 5% to 17% of the adult population (2008–2016).
- Across 26 FinScope countries, Rwanda ranked second in share of adult population with access to formal financial services.
- Microcredit expansion program (U-SACCOs) is a key policy aimed at expanding financial access for the unbanked.

### Microcredit Expansion (U-SACCO) Program design and outreach
- Implementation:
  - One U-SACCO established in each of Rwanda’s 416 municipalities; program launched in March 2009.
  - U-SACCOs initially focused on savings; different U-SACCOs first extended loans in late 2011.
  - By end-2013, 85% of U-SACCOs were profitable and stopped receiving subsidies (AFI, 2014).
- Outreach and use:
  - Program produced 1.6 million new customers.
  - 91% of Rwandans resided within 3 miles of a U-SACCO branch (AFI, 2014).
  - Loan purpose breakdown: almost half for trade and tourism services; about one-fifth for agricultural activities.
- Regulatory/operational parameters:
  - Liquidity ratio requirement: initially 80%, reduced to 30% after December 2013.
  - Minimum capital requirement: about USD 8,000.
  - U-SACCOs generally hold capital in excess of 30% of total assets.

### Data, sample, and descriptive statistics
- Credit register and sample:
  - Loan-level data from all credit institutions maintained by the Credit Reference Bureau (CRB); monthly reporting with no loan-size threshold.
  - Analysis period: January 2008 to December 2016.
  - Raw dataset: 4.1 million observations on bank-borrower loan exposures (monthly).
  - Final cleaned sample: lending in local currency for banks, U-SACCOs and other MFIs related to 177,583 unique individuals in 336 municipalities.
  - Exclusions: 80 municipalities excluded to ensure correct borrower location; foreign-currency loans excluded (account for less than 1% of total loans).
  - U-SACCOs started granting loans in 297 out of 336 municipalities during the sample period; first two U-SACCOs extended credit as early as November 2011 and the last one in April 2016.
- Per-loan and borrower-level summary statistics:
  - Average loan balance: 2.8 million Rwandan francs (RFW) (approximately USD 3,250).
  - Average interest rate on outstanding loans: 18%.
  - Market share by lender type (loan count): commercial banks 50%, U-SACCOs 24.7%, other MFIs 25.2%.
  - Borrower demographics: more than one third female; 23% younger than 30 years old; 10% government employees.
- Municipality pre-program measures:
  - Average share of working-age individuals (older than 16 years) with an outstanding bank loan before the program: 1%.
  - Median share of urban population in a municipality: 0%; average share: 11.8%.
  - Median poverty headcount ratio: 44%; average poverty headcount ratio: 42%.
  - Night-time luminosity (2011, pre-program) average: 2.6.

### Empirical strategy
- Identification uses staggered implementation timing across municipalities.
- Main estimating equation: linear probability model (OLS) at borrower-municipality-month level.
  - Dependent variable Loan_imt = 1 if individual i in municipality m has an outstanding loan at time t, 0 otherwise.
  - Key regressor Post U-SACCO_mt = 1 after a U-SACCO starts lending in municipality m at time t, 0 before.
  - Controls: time-invariant individual characteristics (gender, marital status, age, sector of employment), municipality fixed effects α_m, time fixed effects (year:month φ_t).
  - Standard errors clustered at municipality level.
- Robustness: municipality-specific time trends; balanced panel frequency checks at yearly and quarterly frequency.

### Baseline results — impact on loan access (main estimates)
- Main estimated effects (preferred specification with municipality-specific time trends):
  - U-SACCO program raised the probability of an individual having an outstanding loan by 3.7 percentage points (statistically significant at the 1% level).
  - Pre-program average share of individuals with an outstanding loan: 9.6%.
- Dynamics:
  - Likelihood of having a loan is higher after program implementation and rises over time; pre-program estimates are close to zero and statistically insignificant.
  - U-SACCO-driven effect: probability of having a loan rises in the first two years and then stabilizes at about 10 percentage points higher than pre-program.
  - Spillovers: starting in the second year, probability of obtaining a loan from a commercial bank increases up to 3.5 percentage points relative to pre-program.
- Branch network response:
  - Commercial banks increased number of branches by 5.8% in high bank presence municipalities and by 11.0% in low bank presence municipalities (high/low defined by the 75th percentile of pre-program share of individuals with an outstanding loan).
- Travel time to banks (FinScope): share of individuals for whom it takes less than 30 minutes to reach the nearest bank increased from 21% in 2012 to 25% in 2016; no comparable effect for other MFIs.

### Spatial and borrower heterogeneity
- Heterogeneity definitions:
  - Low vs high bank presence: split at the 75th percentile of pre-program share of individuals with an outstanding loan.
  - Rural vs urban: rural = entire population resides in rural areas pre-program; urban = nonzero fraction resides in urban areas pre-program.
  - High/low income: split by poverty headcount ratio (20% threshold) and night-time luminosity (75th percentile).
- Key heterogeneous findings:
  - Program’s average effect concentrated in municipalities with lower ex-ante levels of financial and economic development.
  - Post U-SACCO × Low Bank Presence: 0.0432*** (Any Institution); 0.0432*** (U-SACCO); 0.00471 (Bank); 0.00154 (Other MFI).
  - Post U-SACCO × Rural: 0.0417*** (Any Institution); 0.0421*** (U-SACCO); 0.00437 (Bank); 0.00125 (Other MFI).
  - Post U-SACCO × Low Development (nightlights): 0.0427*** (Any Institution); 0.0431*** (U-SACCO).
  - Post U-SACCO × High Poverty: 0.0418*** (Any Institution); 0.0425*** (U-SACCO).
- Borrower-level heterogeneity (U-SACCO uptake):
  - Post U-SACCO × Non-Government Employee: 0.0409***.
  - Post U-SACCO × Young: 0.0332***; Post U-SACCO × Old: 0.0399***.
  - Post U-SACCO × Single: 0.0493***; Post U-SACCO × Married: 0.0356***.
  - Post U-SACCO × Male: 0.0564***; Post U-SACCO × Female: 0.00388 (not significant).
  - Interaction specifics: Post U-SACCO × Single Female: 0.0199***; Post U-SACCO × Married Female: 0.00264.

### Relationship lending (intensive margin)
- Research question: how loan terms evolve with length of lender–borrower relationship; comparison across U-SACCOs, commercial banks, and other MFIs.
- Outcome dummies (conditional on having an outstanding loan):
  - Loan size > 75th percentile.
  - Interest rate < 25th percentile.
  - Loan maturity > 75th percentile.
- Relationship length measured in months since first loan with a given financial institution, interacted with Post-U-SACCO dummy.
- Main findings:
  - Relationship length increases likelihood of large, cheap, and long-term loans across financial intermediaries.
  - U-SACCOs are less likely to grant large, cheap, and long-term loans relative to banks and other MFIs.
  - Magnitude examples:
    - One additional year of relationship with a commercial bank or other MFIs raises likelihood of obtaining a top-quartile loan size by 4%; comparable effect for U-SACCOs is 2.2%.
    - One additional year of relationship raises loan maturity by 10.9% for banks and other MFIs, and by 3.3% for U-SACCOs.
  - Longer relationships associated with greater likelihood of bottom-quartile (low) interest rate at banks, but no such effect for U-SACCOs.
- Interpretation: U-SACCO institutional and balance-sheet constraints (funding structure: mostly callable deposits with average maturity about 5 months and maximum maturity not exceeding 4 years; regulatory ceilings on single-borrower exposures) limit their capacity to improve loan terms as relationships mature.

### Transition to commercial banks — definition, identification, and matching
- Definition of switching loans:
  - New loans granted to individuals who (i) had a borrowing relationship with a financial institution in the previous year; and (ii) established a new borrowing relationship with another lender.
- Identified switching loans: 2,180 switching loans from first-time borrowers at U-SACCOs to commercial banks, corresponding to 10% of first-time U-SACCO borrowers that were granted more than one loan during the sample period.
- Matching and comparison groups:
  - Control group 1: new loans granted by U-SACCOs to similar non-switching borrowers in the same month.
  - Control group 2: new loans extended by banks to similar existing bank borrowers in the same month.
  - Exact matching on month, age group, gender, marital status, employment status, and loan type (mortgage or other).
  - Nearest-neighbour selection based on loan amount, interest rate and/or maturity, and degree of bank presence prior to U-SACCO program in borrower’s municipality.
  - Propensity score matching used as robustness check.

### Transition to commercial banks — main results on switching loans
- Descriptive (Table 6):
  - New Loans from U-SACCOs (switchers, n=2,154): Loan Amount 1.678; Interest Rate 19.63; Maturity 23.16; Female Borrower 0.264.
  - New Loans from Banks to non-switchers (n=155,971): Loan Amount 3.324; Interest Rate 20.41; Maturity 23.74; Female Borrower 0.424.
- Switching loans vs similar U-SACCO loans (matched comparisons, Table 7):
  - Switching Loan − Other Loans (exact matching): Loan Amount +0.396***; Interest Rate −4.224*** (basis points? reported as −4.224*** in table); Maturity +6.596*** (months).
  - Without matching: Loan Amount +1.010***; Interest Rate −5.548***; Maturity +9.052***.
  - Propensity score matching: Loan Amount +0.822***; Interest Rate −6.266***; Maturity +8.643***.
- Switching loans vs existing bank borrowers (Table 8):
  - With matching to same commercial bank: Switching Loan − Other Loans: Loan Amount −0.470***; Interest Rate −0.072; Maturity −0.268.
  - Without matching: Loan Amount −1.647**; Interest Rate −0.779***; Maturity −0.578.
  - Propensity-score matched comparisons yield similar direction and significance in several specifications.
- Interpretation:
  - Commercial banks grant switching borrowers larger, cheaper, and longer loans than the U-SACCOs would have.
  - Switching loans are smaller than typical new bank loans to existing bank borrowers but have comparable interest rates and maturities conditional on matching to the same bank.

### Borrower riskiness and loan performance
- NPL definition: loan in arrears for more than 90 days.
- Performance windows: arrears within 1 year from origination; within 2 years; any time until maturity.
- Findings (Table 9):
  - Switching loans are less likely to become NPLs compared to similar loans extended by U-SACCOs.
  - Switching loans are not riskier than similar loans granted by commercial banks.
- Interpretation: evidence consistent with commercial banks selecting (cream-skimming) lower-risk borrowers from the U-SACCO pool.

### Results for post-switching loans (subsequent loans at commercial banks)
- Exact matching within borrower and bank; subsequent loans grouped by time since first loan: <6 months; 7–12 months; 13–24 months; >24 months.
- Loan size dynamics:
  - Repeat borrowers receive progressively larger loans as relationship length increases.
  - Loans granted more than two years after switching are larger by RWA 0.452 million than the original switching loan.
  - Panel A (Loan Amount): New Loan − Original Switching Loan:
    - 1 to 6 months: 0.015 (ns)
    - 7 to 12 months: 0.011 (ns)
    - 12 to 24 months: 0.098***
    - 24+ months: 0.452***
- Maturity dynamics:
  - Maturity increases over time; positive effect disappears after two years.
  - Panel C (Maturity): 7 to 12 months: 0.336*; 12 to 24 months: 0.891***.
- Interest spreads:
  - No difference between interest rate spread (interest rate minus the repo rate) charged on initial and subsequent loans.
  - Discount enjoyed by switchers on their first commercial bank loan persists during the lending relationship.
- Interpretation: persistence of initial discount and absence of higher spreads later are inconsistent with a hold-up problem; consistent with information sharing mitigating hold-up problems.

### Robustness checks and external validation
- Alternative aggregations:
  - Quarterly and yearly balanced panels produce results robust to baseline monthly-frequency findings (Table A2).
- Sample exclusions:
  - Of 336 municipalities observed, 39 had no loan-granting U-SACCO during the sample; dropping these 39 municipalities does not change results (Table A3).
- Falsification tests:
  - Randomized assignment of program implementation dates to municipalities (100 draws) yields average estimated coefficient on randomized Post U-SACCO very close to zero and statistically insignificant (Table A4).
- External survey validation (FinScope 2012 and 2016):
  - Savings account mean: 0.319 (2012) and 0.364 (2016).
  - Loan mean: 0.046 (2012) and 0.081 (2016).
  - District-level regressions: probability of having a savings account and of being granted a loan increased between 2012 and 2016 relatively more for individuals in districts with ex-ante lower bank presence; effects driven by U-SACCOs not commercial banks (Table A8).
  - Note: using 2012 as pre-program benchmark likely underestimates program effects because implementation started in 2011.

### Overall implications and policy-relevant conclusions
- Access and outreach:
  - U-SACCOs expanded credit access, especially in low bank presence municipalities, rural areas, low-development and high-poverty municipalities.
  - The program raised probability of holding an outstanding loan by 3.7 percentage points overall; effect concentrated and larger in underserved areas (e.g., 4.3 percentage points in low bank presence municipalities).
- Market dynamics and spillovers:
  - Commercial banks expanded branch networks and increased lending about one year after program roll-out.
  - Significant share of first-time U-SACCO borrowers who need a second loan switch to commercial banks and obtain larger, cheaper, longer loans.
  - Spillovers consistent with reduced information asymmetries via borrowers building credit histories and signaling through the credit register.
- Selection, screening, and risks:
  - Commercial banks appear to cream-skim lower-risk borrowers from the microfinance sector; switching borrowers are not worse in ex-post loan performance.
  - U-SACCOs reach relatively riskier borrowers (non-government employees) and improve access for young, old, single and married individuals, and for males and single females.
  - Because least-risky borrowers transition to commercial banks, microlenders may be left with relatively risky borrower pools, potentially undermining their viability and posing financial stability risks.
- Role of information sharing:
  - Well-functioning credit reference bureaus play a crucial role in signaling creditworthiness and facilitating transitions from microlenders to banks, mitigating information frictions in credit markets.

*Source: Excerpts from "The Microcredit Expansion Program and Financial Access" (wp18208).*

### 2.1  Rwandan Economy and Financial Sector . . . . . . . . . . . . . . . . . . . . . . . . .   8

### 2.1  Rwandan Economy and Financial Sector

### Macroeconomic and institutional context
- Rwanda is a landlocked country in East Africa with a population of 11.5 million.
- The country has a large rural population and few natural resources.
- Following business-friendly reforms in the early 2000s, Rwanda experienced gains in competitiveness and strong economic growth.
- Annual GDP growth averaged 7.8% between 2008 and 2016 and per capita income doubled during the same period (IMF, 2017a).
- The 2018 World Bank’s Doing Business survey ranks Rwanda 2nd in Africa and 41st in the world according to the ease of doing business.
- Based on the ease of getting credit, Rwanda is ranked 3rd in Africa and 11th in the world.
- The 2016-2017 World Economic Forum’s Global Competitiveness Index ranks Rwanda 52nd among 138 countries, outperforming the Sub-Saharan Africa (SSA) average on all dimensions other than market size.
- The reforms associated with the “Vision 2020” economic strategy strive to make Rwanda a middle-income country by 2020 and have been accompanied by a reduction in poverty and income inequality (IMF, 2017b).

### Banking sector structure and microfinance
- Total bank assets grew from 22% to 39% of GDP from 2008 to 2016.
- Bank credit to the private sector grew at an annual average of 13% in real terms over the same period (IMF, 2017a).
- Commercial banks represent about two-thirds of total banking sector assets.
- The banking sector is relatively concentrated, with the 3 largest commercial banks (out of 17) accounting for more than half of total bank assets, loans and deposits.
- Most banks are foreign-owned, but the majority of bank funding is domestic and comes from local deposits.
- There are 523 microfinance institutions (MFIs), including 416 municipal credit cooperatives (U-SACCOs) that were set up as part of the microcredit expansion program examined in this paper.
  - One U-SACCO in each of the 416 municipalities, with some only providing savings accounts, and others also granting loans.
- MFIs account for almost 6% of total bank assets.

### Financial inclusion trends and relevance to the microcredit expansion program
- Access to formal financial services increased from 21% to 68% of the adult population between 2008 and 2016.
- Access to formal credit increased from 5% to 17% of the adult population over the same period (FinScope, 2012, 2016).
- Across 26 countries with FinScope surveys, Rwanda is ranked second in terms of the share of adult population with access to formal financial services.
- These developments reflect policies and regulations aimed at expanding financial access for the unbanked population, including the nationwide microcredit expansion program establishing U-SACCOs.

*Source: Excerpt from "The Microcredit Expansion Program and Financial Access" (wp18208), Section 2.1.*

### 2.2  Microcredit Expansion (U-SACCO) Program

### 2.2  Microcredit Expansion (U-SACCO) Program

### Program design and outreach
- One U-SACCO established in each of Rwanda’s 416 municipalities; program launched in March 2009.
- Goal: provide financial services at low transaction costs, especially in rural communities; targeted the unbanked in practice.
- U-SACCOs initially focused on providing access to savings accounts, with different U-SACCOs only extending their first loans in late 2011.
- The program produced 1.6 million new customers and 91% of Rwandans residing within 3 miles of a U-SACCO branch (AFI, 2014).
- Almost half of U-SACCO loans are for trade and tourism services; about one-fifth for agricultural activities, including livestock and fishing (MFR, 2015).
- U-SACCOs are municipality-specific member-owned financial intermediaries formed as microfinance institutions, supervised by the Rwanda Cooperative Agency and the National Bank of Rwanda.
- By end-2013, 85% of U-SACCOs were profitable and stopped receiving subsidies (AFI, 2014).
- Regulatory/operational details mentioned: liquidity ratio requirement initially 80%, reduced to 30% after December 2013; minimum capital requirement about USD 8,000; U-SACCOs generally hold capital in excess of 30% of total assets (MFR, 2015).

### Data and sample
- Credit register: loan-level data from all credit institutions in Rwanda, maintained by the Credit Reference Bureau (CRB) under supervision of the National Bank of Rwanda.
- Reporting institutions provide loan-level information monthly with no loan-size threshold.
- Analysis period: January 2008 to December 2016.
- Dataset: 4.1 million observations on bank-borrower loan exposures on a monthly basis.
- Final cleaned sample: lending activities in local currency for banks, U-SACCOs and other MFIs related to 177,583 unique individuals residing in 336 municipalities.
- Exclusions: 80 municipalities excluded to ensure correct borrower location; loans in foreign currency excluded (account for less than 1% of total loans).
- For each loan: amount in arrears, borrower’s municipality and district, age, gender, marital status, and sector of employment (government or non-government).
- U-SACCOs started granting loans in 297 out of 336 municipalities during the sample period; first two U-SACCOs extended credit as early as November 2011 and the last one in April 2016.

### Summary statistics (sample-level)
- Average loan balance: 2.8 million Rwandan francs (RFW) (approximately USD 3,250).
- Average interest rate on outstanding loans: 18%.
- Market share by lender type (loan count): commercial banks 50%, U-SACCOs 24.7%, other MFIs 25.2%.
- Borrower demographics: more than one third of borrowers are female; 23% are younger than 30 years old; 10% are government employees.
- Municipality-level pre-program measures:
  - Average share of working-age individuals (older than 16 years) with an outstanding bank loan before the program: 1%.
  - Median share of urban population in a municipality: 0%; average share: 11.8%.
  - Median poverty headcount ratio: 44%; average poverty headcount ratio: 42%.
  - Night-time luminosity (2011, pre-program) average: 2.6.

### Empirical strategy
- Identification exploits staggered implementation: different U-SACCOs granted first loans in different months starting late 2011.
- Main specification (linear probability model, OLS) at borrower-municipality-month level:
  - Dependent variable Loan_imt = 1 if individual i in municipality m has an outstanding loan at time t, 0 otherwise.
  - Key regressor Post U-SACCO_mt = 1 after a U-SACCO starts lending in municipality m at time t, 0 before.
  - Controls: time-invariant individual characteristics (gender, marital status, age, sector of employment), municipality fixed effects (α_m), time fixed effects (year:month φ_t); standard errors clustered at municipality level.
- Robustness: specifications include municipality-specific time trends; balanced panel frequency robustness checks at yearly or quarterly frequency.

### Baseline results: impact on loan access
- Main estimated effect: U-SACCO program raised the probability of an individual having an outstanding loan by 3.7 percentage points (statistically significant at the 1% level) in the preferred specification with municipality-specific time trends.
- Context: pre-program average share of individuals with an outstanding loan is 9.6%.
- Dynamics:
  - Likelihood of having a loan is higher after program implementation and rises over time; estimates before the program are close to zero and statistically insignificant (supporting parallel trends).
  - The probability of having a loan due to U-SACCOs rises in the first two years and then stabilizes at about 10 percentage points higher than in the pre-program period (U-SACCO-driven effect).
  - Spillovers: starting in the second year of the program, the probability of obtaining a loan from a commercial bank increases up to 3.5 percentage points more than in the pre-program period.
- Interpretation: U-SACCOs drive the initial expansion in credit access; commercial banks expand outreach with a lag, consistent with reduced information asymmetries, borrowers building credit histories, and signaling creditworthiness via the credit register.
- Branch network response: commercial banks increased the number of branches by 5.8% in high bank presence municipalities and by 11.0% in low bank presence municipalities (high/low defined by the75th percentile of pre-program share of individuals with an outstanding loan).
- Travel time to banks (FinScope): share of individuals for whom it takes less than 30 minutes to reach the nearest bank increased from 21% in 2012 to 25% in 2016; no comparable effect for other MFIs.

### Spatial and borrower heterogeneity
- Hypothesis: program effects larger in municipalities with lower ex-ante financial access, urbanization, and economic development.
- Heterogeneity strategy:
  - Define low vs high bank presence by splitting pre-program share of individuals with an outstanding loan at the75th percentile.
  - Define rural vs urban municipalities: rural = entire population resides in rural areas pre-program; urban = nonzero fraction resides in urban areas pre-program.
  - Define high/low income municipalities using poverty headcount ratio (split at 20%) and night-time luminosity (split at the75th percentile).
- Findings:
  - Program’s average effect concentrated in municipalities with lower ex-ante levels of financial and economic development.
  - Interaction of Post U-SACCO with low bank presence dummy is positive and significant for the overall sample and for loans from U-SACCOs, but insignificant for loans from other MFIs or commercial banks.
  - Point estimates for low bank presence interactions are larger than the stand-alone Post U-SACCO coefficient, indicating larger impacts in underserved areas.

*Source: IMF staff analysis of "Microcredit Expansion (U-SACCO) Program" (chapter 2.2), wp18208.*

### 4.3 percentage points in low bank presence municipalities after the program, a sizeable increase given

### wp18208 - 4.3 percentage points in low bank presence municipalities after the program, a sizeable increase given

### Program impact on access to credit
- The program increased the share of individuals with a loan by 4.3 percentage points in low bank presence municipalities after the program.
- Contextual benchmark: the average share of individuals with a loan in the pre-program period in low bank presence municipalities was 4.6%.
- The result holds in sample splits of rural versus urban municipalities, low-income versus high-income municipalities, and low-poverty versus high-poverty municipalities.
- The loan expansion effect was largely driven by newly established U-SACCOs.

### Heterogeneous borrower effects
- Borrower-level characteristics available in the credit register: age, gender, marital status, sector of employment.
- Main heterogeneous findings:
  - The program mainly increased credit provision (through U-SACCOs) to non-government employees, suggesting reach to relatively riskier borrowers assumed less creditworthy than government employees.
  - U-SACCOs improved access to credit for both young and old borrowers.
  - U-SACCOs improved access to credit for single and married individuals.
  - U-SACCOs improved access to credit for males and single females.
  - No significant program impact found for married female borrowers, possibly because credit tends to be contracted by males on behalf of the household.
- Caveat: limited borrower-level controls and most individuals borrowing from a single bank imply inability to fully control for borrower-level credit demand.

### Robustness and falsification tests
- Alternative data aggregations (quarterly and yearly balanced panel borrower-municipality-time datasets) yield results robust to baseline monthly-frequency findings (see Table A2).
- Sample exclusion: of 336 municipalities observed in the credit register, 39 municipalities had no loan-granting U-SACCO during the sample; dropping these 39 municipalities does not change results (see Table A3).
- External validation using two cross-sections from FinScope surveys (2012 and 2016):
  - Differences from baseline: only two time periods and borrower location available at district level.
  - Dependent variables: dummy for having a savings account; dummy for having a savings account and a loan from a U-SACCO or commercial bank.
  - Findings: probability of having a savings account and of being granted a loan (conditional on having an account) increased between 2012 and 2016 relatively more for individuals in districts with ex-ante lower bank presence; effects driven by U-SACCOs not commercial banks (see Table A8).
- Falsification test: randomized assignment of program implementation dates to municipalities in the interval January 2008–December 2016, repeated 100 times:
  - The average estimated coefficient on randomized Post U-SACCO is very close to zero and statistically insignificant (Table A4), indicating main findings are unlikely due to spurious correlation.

### U-SACCOs and relationship lending (intensive margin)
- Research question: how loan terms (size, interest rate, maturity) evolve with length of lender–borrower relationship; comparison of U-SACCOs versus commercial banks and other MFIs.
- Dependent variable definitions (conditional on having an outstanding loan):
  - Dummy = 1 if loan size > 75th percentile of sample distribution.
  - Dummy = 1 if interest rate < 25th percentile of sample distribution.
  - Dummy = 1 if loan maturity > 75th percentile of sample distribution.
- Key explanatory variable: relationship length (months since first loan with a given financial institution), interacted with Post-U-SACCO dummy.
- Main findings from Table 5:
  - Likelihood of obtaining large, cheap, and long-term loans increases with relationship length across financial intermediaries (consistent with relationship lending literature).
  - U-SACCOs are less likely to grant large, cheap, and long-term loans relative to banks and other MFIs.
  - Magnitude examples:
    - One additional year of relationship with a commercial bank or other MFIs raises likelihood of obtaining a top-quartile loan size by 4%; the comparable effect for U-SACCOs is 2.2%.
    - One additional year of relationship raises loan maturity by 10.9% for banks and other MFIs, and by 3.3% for U-SACCOs.
  - Longer lender–borrower relationship is associated with greater likelihood of obtaining a bottom-quartile (low) interest rate at banks, but no such effect for U-SACCOs.
- Interpretation: institutional and balance sheet constraints at U-SACCOs limit their capacity to improve loan terms as relationships mature; U-SACCO funding structure includes mostly callable deposits with average maturity about 5 months and maximum maturity not exceeding 4 years; regulatory ceilings on single-borrower exposures for MFIs cited.

### Transition to commercial banks — empirical strategy
- Definition of switching loans: new loans granted to individuals who (i) had a borrowing relationship with a financial institution in the previous year; and (ii) established a new borrowing relationship with another lender.
- Identified switching loans: 2,180 switching loans from first-time borrowers at U-SACCOs to commercial banks, corresponding to 10% of first-time U-SACCO borrowers that were granted more than one loan during the sample period.
- Comparison groups for switching loans:
  - Control group 1: new loans granted by U-SACCOs to similar non-switching borrowers in the same month.
  - Control group 2: new loans extended by banks to similar existing bank borrowers in the same month.
- Matching procedure:
  - Exact matching on month, age group, gender, marital status, employment status, and loan type (mortgage or other).
  - Nearest-neighbour selection based on loan amount, interest rate and/or maturity, and degree of bank presence prior to U-SACCO program in borrower’s municipality.
  - Robustness: alternative matching combining exact matching with propensity score matching yields similar results (Tables A5 and A6).

### Transition to commercial banks — results
- Summary statistics (Table 6):
  - Switching loans by commercial banks to borrowers who switched from U-SACCOs are larger, have lower interest rates, and longer maturities compared to new loans by SACCOs to non-switchers.
  - Switching loans are smaller, on average, than new loans to individuals already at banks: RWF 1.678 million (switching loans) vs. RWF 3.324 million (all new commercial bank loans) or RWF 6.813 million (loans to switchers across banks).
- Comparison of switchers vs. non-switchers (Table 7):
  - Switching loans are on average larger by RWF 0.396 million compared to similar non-switcher U-SACCO loans.
  - Switching loans are cheaper by 422 basis points compared to similar non-switcher U-SACCO loans.
  - Switching loans have longer maturity by almost 6.6 months compared to similar non-switcher U-SACCO loans.
  - Results robust when restricting control group to loans granted by the same U-SACCO the switching borrower left.
- Comparison of switchers vs. existing bank borrowers (Table 8):
  - Switching loans have similar interest rates and maturities to new loans granted by the same bank in the same month to borrowers already in commercial banks, but switching loans are considerably smaller.
  - Estimated average size difference: switching loans are smaller by RWF 0.470 million relative to new loans to existing bank borrowers.
  - Narrower control group comparison (U-SACCO-to-bank A switchers vs. bank B-to-bank A switchers) yields qualitatively same and quantitatively stronger results.

### Borrower riskiness and loan performance
- Non-performing loan (NPL) definition: loan in arrears for more than 90 days.
- Windows examined: arrears within 1 year from origination; within 2 years; any time until maturity.
- Findings (Table 9):
  - Switching loans are less likely to become NPLs compared to similar loans extended by U-SACCOs.
  - Switching loans are not riskier than similar loans granted by commercial banks.
- Interpretation: commercial banks appear to engage in selection (cream-skimming) of lower-risk borrowers from the U-SACCO pool; switching borrowers are not associated with worse ex-post loan performance.

### Overall implications and interpretation
- U-SACCOs expanded access to credit, especially in low bank presence municipalities, and reached borrowers less likely to obtain loans from banks (e.g., non-government employees).
- U-SACCOs face balance sheet and institutional constraints that limit their ability to improve loan terms over time relative to commercial banks.
- Commercial banks expanded branch networks and credit after the first year of the program, and there are spillovers from U-SACCO expansion to commercial banks.
- Borrowers whose credit demand exceeded U-SACCO capacity switched to commercial banks and obtained larger, cheaper, and longer loans than they would have at U-SACCOs; these switchers exhibited no worse loan performance than comparable bank borrowers, suggesting selection of lower-risk borrowers by banks.

*Source: wp18208 - section excerpts on program effects, robustness, relationship lending, and transition to commercial banks.*

### 5.3  Results for Post-Switching Loans

### 5.3  Results for Post-Switching Loans

### Analysis approach
- Exact matching of loans within borrower and bank to compare terms of subsequent loans to the first loan granted by the same commercial bank.
- Subsequent loans grouped into time buckets based on date granted relative to the first loan: less than 6 months, 7 to 12 months, 13 to 24 months, and more than 24 months after the first loan.
- Objective: to isolate the effect of credit history on subsequent loan conditions for U-SACCO-to-bank switchers.

### Empirical findings on subsequent loans
- Loan size:
  - Repeat borrowers receive progressively larger loans as the length of the relationship with the bank increases.
  - Loans granted more than two years after switching are larger by RWA 0.452 million than the original switching loan.
- Loan maturity:
  - Maturity of loans increases over time.
  - The positive effect on maturity disappears after two years.
- Interest spreads:
  - No difference between the interest rate spread (interest rate minus the repo rate) charged on initial and subsequent loans.
  - The discount enjoyed by switchers for their first loan at a commercial bank persists during the lending relationship.
- Interpretation:
  - Persistence of the initial discount and absence of higher spreads later is inconsistent with a hold-up problem documented in other contexts.
  - Results are consistent with evidence that information sharing mitigates hold-up problems in the African context.

### Broader conclusions and implications (Section 6)
- Program impact on access:
  - The microfinance expansion program raised the likelihood of access to bank loans for the previously unbanked population, especially in rural and less financially developed municipalities.
  - The overall program effect is driven by the newly set-up savings and credit cooperatives (U-SACCOs).
- Spillovers to commercial banks:
  - Commercial banks increase their lending about one year after the roll-out of the program.
  - A significant share of first-time borrowers at U-SACCOs who need a second loan switch to commercial banks, which grant them larger and cheaper loans than the loans they might have received from U-SACCOs.
- Selection and screening:
  - Banks grant loans to borrowers who are least risky among U-SACCO borrowers and no riskier than existing bank borrowers.
  - Microfinance institutions that target low-income individuals play an important screening role for the unbanked population.
  - Creditworthiness is signaled to commercial banks through the credit reference bureau, which provides lenders with access to individuals’ borrowing and payments history, including defaults.
- Market dynamics and risks:
  - Commercial banks expand branch networks in previously under-served areas and cream-skim low-risk borrowers from the microfinance sector by offering more attractive loan terms.
  - The expansion of the microfinance sector, coupled with well-functioning credit reference bureaus, can mitigate information frictions in credit markets and play a crucial role in financial development.
  - Because of the transition of the least risky borrowers to commercial banks, such programs may have the unintended consequence of leaving microlenders with a pool of risky borrowers, undermining the financial viability of their business model and posing financial stability risks.

*Source: wp18208 - 5.3  Results for Post-Switching Loans*

### Introduction and further steps.American Economic Journal: Applied Economics,7(1), 1–21.

### Introduction and further steps. American Economic Journal: Applied Economics, 7(1), 1–21.

### Main findings on access to credit
- Program effect on probability of having a loan (Table 2, full sample, municipality and time FE, borrower controls, municipality time trends):
  - Post U-SACCO: 0.0283*** (columns 1-3), 0.0370*** (U-SACCO, column 4), 0.00479 (Bank, column 5), 0.00102 (Other MFI, column 6). Standard errors reported in table.
  - Adjusted R-squared: 0.169 (col 1), 0.201 (col 2), 0.206 (col 3), 0.143 (col 4), 0.112 (col 5), 0.155 (col 6).
- Timing of effects (Table A1, dynamics around implementation):
  - Post U-SACCO [t+1,t+12]: 0.0361*** (Any Institution), 0.0370*** (U-SACCO), 0.00389* (Bank), 0.000474 (Other MFI).
  - Post U-SACCO [t+13,t+24]: 0.0961*** (Any Institution), 0.0939*** (U-SACCO), 0.0161*** (Bank), -0.000806 (Other MFI).
  - Post U-SACCO [> t+24]: 0.119*** (Any Institution), 0.102*** (U-SACCO), 0.0350*** (Bank), -0.00260 (Other MFI).

### Cross-sectional heterogeneity (Table 3)
- Panel A — Low vs. High Bank Presence (split at 75th percentile):
  - Post U-SACCO x Low Bank Presence: 0.0432*** (Any Institution), 0.0432*** (U-SACCO), 0.00471 (Bank), 0.00154 (Other MFI).
  - Post U-SACCO x High Bank Presence: 0.0100 (Any Institution), 0.00762 (U-SACCO), 0.00515 (Bank), -0.00142 (Other MFI).
- Panel B — Rural vs. Urban:
  - Post U-SACCO x Rural: 0.0417*** (Any Institution), 0.0421*** (U-SACCO), 0.00437 (Bank), 0.00125 (Other MFI).
  - Post U-SACCO x Urban: 0.0123 (Any Institution), 0.00723 (U-SACCO), 0.00723 (Bank), -0.000300 (Other MFI).
- Panel C — Low vs. High Development (nightlights, split at 75th percentile):
  - Post U-SACCO x Low Development: 0.0427*** (Any Institution), 0.0431*** (U-SACCO), 0.00460 (Bank), 0.00128 (Other MFI).
  - Post U-SACCO x High Development: 0.0193* (Any Institution), 0.0160* (U-SACCO), 0.00542 (Bank), 0.000142 (Other MFI).
- Panel D — Low vs. High Poverty (headcount >/< 20% pre-period):
  - Post U-SACCO x High Poverty: 0.0418*** (Any Institution), 0.0425*** (U-SACCO), 0.00443 (Bank), 0.000936 (Other MFI).
  - Post U-SACCO x Low Poverty: 0.0179 (Any Institution), 0.0126 (U-SACCO), 0.00635 (Bank), 0.00139 (Other MFI).

### Borrower heterogeneity for U-SACCO uptake (Table 4)
- Post U-SACCO × Non-Government Employee: 0.0409***.
- Post U-SACCO × Young: 0.0332***; Post U-SACCO × Old: 0.0399***.
- Post U-SACCO × Single: 0.0493***; Post U-SACCO × Married: 0.0356***.
- Post U-SACCO × Male: 0.0564***; Post U-SACCO × Female: 0.00388 (not significant).
- Interaction specifics:
  - Post U-SACCO × Single Female: 0.0199***.
  - Post U-SACCO × Married Female: 0.00264.
  - Post U-SACCO × Single Male: 0.0598***.
  - Post U-SACCO × Married Male: 0.0559***.
- Adjusted R-squared across specifications: 0.144–0.146.

### Effects on loan terms conditional on borrowing (Table 5)
- Relationship length and interaction with Post U-SACCO:
  - Post U-SACCO x Relationship length: consistently negative and significant across specifications (e.g., -0.00145***, -0.00195***, -0.00558***, -0.00456***, -0.00634***, -0.00806***).
- Direct Post U-SACCO effects on loan size, interest rate, maturity dummies (relative to percentiles):
  - Post U-SACCO: -0.144***, -0.125***, -0.101***, -0.112***, -0.233***, -0.334*** (columns 1–6 respectively; see table for dependent-variable definitions).
- Relationship length alone is positive and significant for larger loan amount, lower interest, and longer maturity indicators (e.g., 0.00333***, 0.00396***, 0.00448***, 0.00279***, 0.00907***, 0.0102***).
- Adjusted R-squared ranges from 0.154 to 0.399 depending on column.

### Switching behavior and borrower outcomes (Tables 6–10)
- Descriptive differences (Table 6) — New loans characteristics:
  - New Loans from U-SACCOs (switchers, n=2,154): Loan Amount 1.678; Interest Rate 19.63; Maturity 23.16; Female Borrower 0.264.
  - New Loans from Banks to non-switchers (n=155,971): Loan Amount 3.324; Interest Rate 20.41; Maturity 23.74; Female Borrower 0.424.
  - New Loans from Banks (U-SACCO → Banks switchers) (n=2,154) vs bank non-switchers differences reported with significance stars in table.
- Switching loan regressions — comparison with U-SACCOs (Table 7):
  - Switching Loan − Other Loans (with exact matching): Loan Amount +0.396***, Interest Rate −4.224***, Maturity +6.596***.
  - Without matching: Loan Amount +1.010***, Interest Rate −5.548***, Maturity +9.052***.
  - Results persist with propensity score matching (Table A5): e.g., Loan Amount +0.822***, Interest Rate −6.266***, Maturity +8.643*** (propensity score matching).
- Switching loan regressions — comparison with commercial banks (Table 8 and A6):
  - With matching to same commercial bank: Switching Loan − Other Loans: Loan Amount −0.470***, Interest Rate −0.072, Maturity −0.268.
  - Without matching: Loan Amount −1.647**, Interest Rate −0.779***, Maturity −0.578.
  - Propensity-score matched comparisons show similar direction and significance in some specifications (e.g., Loan Amount −0.758*** with matching, Table A6).
- Non-performing loans (Table 9):
  - Switching Loan − Other Loans (vs U-SACCOs, with matching): NPL within 1 year −0.005 (ns), within 2 years −0.029***, until maturity −0.036***.
  - Switching Loan − Other Loans (vs commercial banks, without matching): some coefficients positive and significant in alternative specs (see table).
- Subsequent loans for switchers (Table 10):
  - Panel A (Loan Amount): New Loan − Original Switching Loan:
    - 1 to 6 months: 0.015 (ns)
    - 7 to 12 months: 0.011 (ns)
    - 12 to 24 months: 0.098***
    - 24+ months: 0.452***
  - Panel B (Interest Rate Spread): New Loan − Original Switching Loan: coefficients negative but not significant in table.
  - Panel C (Maturity): New Loan − Original Switching Loan:
    - 7 to 12 months: 0.336*; 12 to 24 months: 0.891***; others not significant.

### Robustness, alternative data structures, and survey evidence
- Alternative frequencies (Table A2):
  - Quarterly (columns 1–4): Post U-SACCO 0.0366*** (Any Inst.), 0.0374*** (U-SACCO), 0.00359 (Bank), 0.000976 (Other MFI).
  - Yearly (columns 5–8): Post U-SACCO 0.0337*** (Any Inst.), 0.0414*** (U-SACCO), -0.000769 (Bank), -0.000592 (Other MFI).
- Excluding municipalities where U-SACCOs never operated (Table A3, sample restricted to 297 municipalities):
  - Post U-SACCO: 0.0305*** (Any Inst.), 0.0326*** (U-SACCO), 0.00229 (Bank), 0.000285 (Other MFI).
  - Dynamics still show positive and significant post-implementation effects at t+1 to t+24 and >t+24 for the Any Institution and U-SACCO outcomes.
- Falsification tests (Table A4):
  - Randomly assigned Post U-SACCO (simulation average over 100 draws) yields coefficients near zero and not significant for Post U-SACCO across institution outcomes (e.g., 0.0004 for Any Institution).
- FinScope survey evidence (Appendix A-II, Tables A7–A8):
  - Descriptives (Table A7): Savings Account (Bank, SACCO or MFI) mean 0.319 (2012) and 0.364 (2016); SACCO share 0.192 (2012) and 0.258 (2016); Loan (Bank, SACCO or MFI) mean 0.046 (2012) and 0.081 (2016).
  - Regression evidence (Table A8):
    - Savings account regressions (District-level Low Bank Presence × Post):
      - Bank or SACCO: 1.305 (col 1), SACCO: 0.902** (col 2), Bank: 0.879 (col 3). Post coefficients and controls included; see table for standard errors.
    - Conditional on having a savings account, loan regressions:
      - Bank or SACCO: 1.296***, SACCO: 1.098***, Bank: 0.475 (not significant). Mean dependent variables: 0.055 (Any loan), 0.035 (SACCO loan), 0.022 (Bank loan).
  - Authors note: using 2012 as pre-program benchmark likely underestimates program effects because implementation started in 2011.

### Analysis and implications documented in the source
- The microcredit expansion via U-SACCOs increased the probability that individuals hold outstanding loans, primarily through U-SACCO lending, with larger effects in municipalities with low pre-existing bank presence, rural areas, low-development and high-poverty municipalities (all results presented above with exact coefficients).
- Effects grow over time: modest in the first year after U-SACCO operation, larger in years 2 and beyond (see timing coefficients in Table A1).
- Loan-term analysis indicates that newcomers (shorter relationship length) obtain different loan terms post-expansion: relationship length is strongly associated with larger loan size, lower interest, and longer maturity; the interaction terms imply U-SACCO entry affects these dynamics.
- Switching borrowers from U-SACCOs to commercial banks obtain larger loan amounts, lower interest rates relative to U-SACCO loans (Tables 7–8), and exhibit heterogeneous subsequent loan paths (Table 10).
- Robustness checks (alternative aggregation, exclusion of never-treated municipalities, falsification permutations) and survey-based evidence corroborate the register-based findings.

*Content derived from "wp18208 - Introduction and further steps.American Economic Journal: Applied Economics,7(1), 1–21."*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2018/wp18208.pdf_