## wpiea2022063-print-pdf

## Source details

**Canonical URL:** [wpiea2022063-print-pdf](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022063-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022063-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022063-print-pdf.pdf.json)

---

### I. Introduction and context
- Remittances reached $559 billion to low- and middle-income countries in 2019.
- Due to Covid-19, remittances to low- and middle-income countries dropped to $549 billion in 2020, a marginal decline of 1.7 percent compared to 2020.
- Policy targets highlighted:
  - G8 “5x5 Objective”: reduce remittance cost by five percentage points over five years.
  - SDG 10.C: “by 2030, reduce to less than 3 per cent the transaction costs of migrant remittances and eliminate remittance corridors with costs higher than 5 per cent.”
- Data: up to 365 corridors from 2011 to 2020 from the World Bank’s Remittance Prices Worldwide database, covering 48 sending and 105 receiving countries.

### II. Data coverage and descriptive patterns
- Sample and measurement
  - Corridors covered: up to 365 corridors; balanced subsample of 203 corridors with full time coverage.
  - Providers sampled to cover a minimum aggregated market share of 80 percent within corridors.
  - Median fee per corridor-year is the primary dependent variable.
  - Fee equivalents in dataset: $200 and $500; analysis focuses on the $200 equivalent.
- Cross-corridor variation (2020)
  - Median remittance fee ranges between 1 percent (Russia → Azerbaijan, Georgia, Kazakhstan, Lithuania, Latvia, Moldova and Ukraine corridors) and 25.8 percent (South Africa → China corridor).
  - Examples of variation from same sending country:
    - US → Peru: 2.9 percent.
    - US → Cape Verde: 10.6 percent.
  - India as recipient:
    - From Singapore: 1.6 percent.
    - From Thailand: 16.6 percent.
- Time trends (2011–2020)
  - Median remittance price (203-corridor balanced sample) declined from 7.7 percent in 2011 to 5.7 percent in 2020.
  - Distributional changes: 75th percentile from 11.1 percent to 7.7 percent; 25th percentile from 5.2 percent to 4 percent.
  - Majority of corridors still have median fees above 5 percent in 2020.
- Provider types and market shares over time
  - Median bank fee persistently above median MTO fee.
  - Average share of banks in a corridor decreased from 27 to 17 percent between 2011 and 2020.
  - Share of MTOs increased from 69 to 80 percent over the same period.
- Access points, payment instruments and speed (2011 → 2020)
  - Bank branch access share declined to less than 5 percent in 2020.
  - Internet access share increased from 18 to 33 percent.
  - Agent access share stable between 40 and 45 percent.
  - Cash payments in sending country decreased from 56 to 42 percent.
  - Bank transfer and card payments in 2020: 34 percent and 12 percent, respectively.
  - Transactions arriving within one hour increased from 43 to 56 percent.
  - Same-day transactions increased from 10 to 12 percent.
  - Transactions taking six or more days decreased from 8 to one percent.
  - Mobile-money-based transactions remain very rare.

### III. Methodology and explanatory variables
- Regression frameworks
  - Cross-sectional corridor model for 2018: Pij = β1(Sending country factors)i + β2(Receiving country factors)j + β3(Corridor-specific factors)ij + εij.
  - Pooled panel and fixed-effects models across 2011–2020: Pijt models adding corridor, country, and year fixed effects (μi, νj, λt).
  - Firm-level model: Pfijt = β1(Sending country factors)it + β2(Receiving country factors)jt + β3(Corridor-specific factors)ijt + β4(Firm-specific factors)ijt + εfijt.
  - Median costs per corridor-year as main dependent variable; firm-level regressions use firm costs; standard errors clustered at corridor level for firm regressions.
- Key explanatory variable groups (lagged at country/corridor level; contemporaneous firm-level factors)
  - Socioeconomic: GDP per capita (WDI), rural population share (WDI), bank branch penetration per capita.
    - Sending-country GDP per capita range (2017): 936 USD (Tanzania) to 80,000 USD (Switzerland).
    - Receiving-country GDP per capita range (2017): 340 USD (Malawi) to 19,700 USD (Estonia).
    - Rural population share range: zero (Singapore, Kuwait) to 83 percent (Malawi).
  - Corridor characteristics: bilateral trade volume (WITS; up to 2017; range example: 412,000 between Nepal and Oman to 683 billion between the US and China in 2018), geographic distance between capitals (CEPII; 215 km to 17,676 km), bilateral migrant stock (UN International Migrant Stock; available for 124 corridors in 2017).
  - Market structure proxies: number of providers per corridor (range: 1 to 26 in 2017), share of banks among providers (zero in 139 out of 300 corridors to one in four corridors in 2017).
  - Macroeconomic/regulatory: pegged exchange rate dummy; capital controls dummy (IMF ARER); regulation index (0–5) from Global Payment Systems Survey 2008; Basel Institute AML risk index (0–10, higher = higher risk).
    - In 2018 cross-sectional sample: 15.2 percent of source countries and 13.4 percent of destination countries have fixed exchange rates.
  - Firm/product features: provider type (bank, MTO, post office), payment instrument (cash, bank transfer, card, mobile money), access point (agent, bank branch, post office, Internet, mobile phone), speed categories. Fast indicates same-day or faster.

### IV. Main empirical results (headline findings)
- Five headline results (across corridor, time and provider dimensions)
  1. Higher sending-country GDP and easier geographic access to financial institutions are associated with lower fees, especially for banks.
     - One standard deviation increase in the sending country’s GDP per capita is associated with about half a percentage lower remittance fees (cross-sectional coefficient, column (1) reference).
  2. Scale economies matter: larger market size (closer economic ties, larger migrant population) and shorter distance are associated with lower costs.
     - One standard deviation increase in bilateral trade within a corridor over time is associated with one percentage point lower remittance fees (fixed-effects, column 11).
     - One standard deviation increase in distance is associated with 0.6 percentage points higher remittance costs (column 1).
     - One standard deviation increase in migrant population is associated with 1.3 percentage points lower remittance costs (column 2).
  3. Market structure is important: banks charge higher fees than MTOs; a larger share of banks among providers is associated with higher fees charged by MTOs.
     - One standard deviation increase in the share of banks among market participants is associated with higher remittance costs (significant across regressions at least at the 10 percent level).
     - A standard deviation increase in the number of market participants reduces remittance costs by 0.3 percentage points (column 1).
     - Median bank fee persistently above median MTO fee; share of banks in corridors fell from 27 to 17 percent while share of MTOs rose from 69 to 80 percent (2011–2020).
  4. Exchange rate stability matters: in corridors where the sending country has a pegged exchange rate, both banks and MTOs charge lower fees.
  5. Payment mode and digitalization: cash payments are associated with higher fees; Internet payments are associated with lower fees.
     - Internet access share increased from 18 to 33 percent (2011–2020); cash payments decreased from 56 to 42 percent.
- Additional patterns and nuances
  - Rural population: one standard deviation increase in rural population share of sending country associated with one percentage point higher remittance costs (column 1). An increase in rural population share of a destination country over time is associated with 2.2 percentage points higher remittance costs (fixed-effects).
  - Bank branch penetration: branch penetration in sending country enters negatively and significantly in fixed-effects regressions.
  - Several regulatory variables and destination-country GDP per capita show inconsistent signs across samples and subsamples.
  - Authors caution results are partial correlations, not causal estimates.

### V. Baseline results and robustness (Tables summary)
- Fit measures:
  - R-squared in cross-sectional regressions: 17 percent to 42 percent.
  - Pooled regressions: 23 percent to 40 percent.
  - Fixed-effect regressions: above 80 percent (driven mostly by source country and year-fixed effects).
- Strongest predictors of cross-corridor variation:
  - (i) income per capita and geographic access to financial institutions in the sending country;
  - (ii) economic links between source and destination country;
  - (iii) market structure in the remittance service market;
  - (iv) the exchange rate regime in the sending country.
- Regulatory and macroeconomic indicators (selected patterns)
  - Regulation index 2008 in sending country associated with higher remittance costs in cross-sectional and pooled regressions; regulation index for receiving country enters negatively in pooled regression (column 3).
  - Reporting requirements: destination-country reporting requirements associated with lower remittance costs; source-country reporting requirements associated with higher remittance costs (column 3).
  - AML risk: higher AML risk in sending country associated with higher remittance costs; higher AML risk in receiving country associated with lower remittance costs in some specifications.
  - Caution: coefficient signs for regulatory variables sometimes counter-intuitive and vary across models.

### VI. Provider-specific regressions: Banks, MTOs, Western Union, MoneyGram
- Banks (up to 232 corridors)
  - Ln(GDP per capita) source country coefficients as large as -1.414*** (column 1).
  - Pegged exchange rate source country: -4.204*** (column 11) and -28.10*** (column 15) in some specifications.
  - Number market participants: positive and significant in pooled regressions (e.g., 0.256*** in column 6).
  - General pattern: banks charge higher fees; distance associated with higher costs.
- MTOs (up to 336 corridors)
  - Number of market participants: significant and negative association with remittance costs (competition effect), e.g., -0.175*** in pooled regressions.
  - Share of banks: positive association with MTO remittance costs (e.g., 5.018*** in column 6).
  - Pegged exchange rate source country associated with lower MTO fees (e.g., -2.423*** and -13.67*** in some specs).
  - Larger migrant population associated with lower remittance costs.
- Western Union
  - Bilateral trade tends to predict lower remittance costs (e.g., -0.276* column 1).
  - Number of market participants generally reduces fees for Western Union.
  - Share of banks often associated with higher fees for Western Union (e.g., 3.067** column 5; 3.318*** column 6).
  - Pegged exchange rate source country associated with lower fees (e.g., -3.793*** column 11; -4.641*** column 12).
- MoneyGram
  - Number of market participants negative and significant across specifications (e.g., -0.156*** column 1; -0.145*** column 11).
  - Pegged exchange rate source country strongly associated with lower fees (e.g., -4.921*** column 11; -5.590*** column 12; -12.08*** column 15).
  - AML index source country shows large negative coefficients in some specifications (e.g., -2.365*** column 11).

### VII. Firm-level regressions and product/channel effects (Table 9)
- Specification overview: five specifications including baseline clustered at corridor level (column 1), 2018 cross-section with and without corridor fixed effects (columns 2–3), and panel regressions with and without corridor-year fixed effects (columns 6–8).
- Provider-type effects (exact reported ranges)
  - Banks charge higher fees: in 2018 cross-sectional regressions banks have 1.3 to 1.9 percentage points higher fees than other providers.
  - MTOs charge lower fees: in 2018 cross-sectional regressions MTOs have 0.9 to 1.3 percentage points lower fees than other providers.
  - In panel regressions over time, bank coefficient sizes are larger; MTO coefficients turn insignificant in panel specifications.
- Delivery and channel effects (exact reported ranges)
  - Internet transactions: 1.7 to 2.7 percentage points lower fees (Internet dummy: -2.376***, -2.690***, -2.391*** in columns 1–3).
  - Agent payments: Agent coefficients around -2.157*** to -2.762*** depending on specification.
  - Branch payments: Branch coefficients around -1.213* to -1.340* depending on specification.
  - Cash payments: 0.7 to 0.9 percentage points higher fees in 2018 cross-sectional regressions (Cash payment coefficients: 0.742**, 0.892**, 0.770** in columns 1–3).
  - Fast transfer (same day) dummy: not significant at 5 percent in any regression.
- Corridor- and country-level firm-regression findings
  - Share of banks in corridor associated with higher fees in firm-level regressions (e.g., Share of banks 5.875*** column 1; 2.584** column 2).
  - Number of market participants generally associated with lower fees in firm-level cross-sections (e.g., -0.0811** column 1; -0.0822** column 2; -0.0859*** column 3).
  - Ln(distance) positive and significant in firm-level cross-sections (e.g., 0.558**, 0.592**, 0.545** in columns 1–3); very large positive coefficients in some fixed-effect specs (19.12***, 19.51*** in columns 6–7).

### VIII. Interpretation, policy-relevant implications and conclusions
- Structural constraints
  - Higher transaction costs driven by a more rural population in sending countries and lower scale in some corridors imply limits to how far policy can reduce remittance fees purely via regulatory change.
- Policy-relevant avenues for cost reduction
  - Strengthening competition, particularly from non-bank providers (MTOs), is associated with lower fees.
  - Digitalization and increased Internet-based access (versus cash) are associated with lower fees.
  - Exchange rate stability (or improved hedging possibilities) is associated with lower fees and can reduce provider risk premia.
- Caveats and interpretation
  - Analysis documents partial correlations; causal inference is not claimed.
  - Market structure and regulatory variables may be endogenous to remittance fees.
  - Not all variables are available for all corridors or years; multiple specifications and subsamples are used.
- Summary conclusions (five key results)
  1. Higher GDP per capita in sending countries and easier geographic access to financial institutions (especially for banks) are associated with lower fees.
  2. Scale economies matter: closer economic ties and larger migrant stocks are associated with lower costs; shorter distance also lowers costs.
  3. Market structure is important:
     - Banks charge higher fees than MTOs.
     - A larger share of banks among providers is associated with higher fees charged by MTOs.
     - MTO fees respond to competition; bank fees do not respond similarly.
  4. Pegged exchange rates in the sending country are associated with lower remittance costs for both banks and MTOs.
  5. Product/channel effects: cash payments attract higher fees; Internet payments lower fees.
- Policy-relevant implication: structural country factors may limit further fee reductions, but stronger competition—particularly from non-bank providers—and digitalization can help reduce remittance costs.

### IX. Appendix: variables, definitions and sources (selected)
- Remittance fee: Remittance fee per 200 (500) US dollars. Source: Remittance Prices Worldwide (World Bank).
- Ln (GDP per capita) source/destination: Log of GDP per capita. Source: World Development Indicators.
- Rural population share source/destination: Percentage of population living in rural areas. Source: World Development Indicators.
- Ln (Bilateral trade): Log of bilateral trade. Source: World Integrated Trade Solution (WITS).
- Ln (distance): Log of distance between sending and receiving capitals (km). Source: Distances Database (CEPII).
- Branch penetration source/destination: Branches of commercial banks per capita (100,000 adults). Source: Database on Access to Financial Services (IMF).
- Share of banks: percentage of banks among respondents in a corridor and year. Source: authors’ calculation, based on Remittance Prices Worldwide (World Bank).
- Number market participants: number of respondents in a corridor and year. Source: authors’ calculation, based on Remittance Prices Worldwide (World Bank).
- Ln(migrants) 2017: Log of number of migrants in each corridor. Source: International Migrant Stock Database (2017 revision).
- Regulation (2008) source/destination: Index of importance of banks in remittances. Source: World Bank Global Payment Systems Survey (2008).
- Pegged exchange rate source/destination: Dummy for pegged exchange rate. Source: Exchange Rate Regime Classification (Ilzetzki, Reinhart, and Rogoff).
- AML index source/destination: Index of AML risk. Source: Basel Institute on Governance.
- Capital controls source/destination: Dummy for controls on personal capital transactions. Source: Annual Report on Exchange Arrangements and Restrictions (IMF).
- Provider and product dummies: Bank, MTO, Branch, Agent, Internet, Cash payment, Fast transfer. Source: Remittance Prices Worldwide (World Bank).

*Italic: Content extracted from the provided IMF working paper content unit.*

### REFERENCES  ________________________________________________________________________ 35

### wpiea2022063-print-pdf - REFERENCES  ________________________________________________________________________ 35

### Figures
- 1. Corridors with highest fees, 2020 ___________________________________________________________ 9
- 2. Corridors with lowest fees, 2020 ___________________________________________________________ 9
- 3. Average price of remittances for corridors with US as sending country ____________________________ 10
- 4. Average price of remittances for corridors with India as receiving country __________________________ 10
- 5. Median price across corridors over time _____________________________________________________ 11
- 6. Median price over time for banks and MTOs _________________________________________________ 12

### Tables
- 3. Baseline Results _______________________________________________________________________ 19
- 4. Developing to advanced corridors only _____________________________________________________ 22
- 5. Remittance Costs across Corridors – Banks only ______________________________________________ 23
- 6. Remittance Costs across Corridors – MTOs only ______________________________________________ 25
- 7. Remittance Costs across Corridors – Western Union ___________________________________________ 27
- 8. Remittance Costs across Corridors – Moneygram _____________________________________________ 28
- 9. Remittance Costs across Corridors – Firm-level Variation ______________________________________ 30

### Appendix Tables
- 1. Summary statistics for regression sample ____________________________________________________ 32

*Source: wpiea2022063-print-pdf - REFERENCES  ________________________________________________________________________ 35*

### 2. Correlation table _______________________________________________________________________ 33

### 2. Correlation table

### I. Introduction and context
- Remittances reached $559 billion to low- and middle-income countries in 2019.
- Due to Covid-19, remittances to low- and middle-income countries dropped to $549 billion in 2020, a marginal decline of 1.7 percent compared to 2020.
- Policy targets: the G8 “5x5 Objective” (reduce remittance cost by five percentage points over five years) and SDG 10.C: “by 2030, reduce to less than 3 per cent the transaction costs of migrant remittances and eliminate remittance corridors with costs higher than 5 per cent.”
- Paper uses data for up to 365 corridors from 2011 to 2020 from the World Bank’s Remittance Prices Worldwide database, covering 48 sending and 105 receiving countries.

### II. Data coverage and descriptive patterns
- Sample and measurement
  - Corridors covered: up to 365 corridors; balanced subsample of 203 corridors with full time coverage.
  - Providers sampled to cover a minimum aggregated market share of 80 percent within corridors; median fee per corridor-year is the primary dependent variable.
  - Fee amounts in the dataset: equivalents of $200 and $500; analysis focuses on the $200 equivalent.
- Cross-corridor variation (2020)
  - Median remittance fee ranges between 1 percent (Russia → Azerbaijan, Georgia, Kazakhstan, Lithuania, Latvia, Moldova and Ukraine corridors) and 25.8 percent (South Africa → China corridor).
  - Considerable variation even for same sending country: US → Peru 2.9 percent; US → Cape Verde 10.6 percent.
  - India as recipient: 1.6 percent if sent from Singapore; 16.6 percent if sent from Thailand.
- Time trends (2011–2020)
  - Median remittance price (203-corridor balanced sample) declined from 7.7 percent in 2011 to 5.7 percent in 2020.
  - Distributional changes: 75th percentile from 11.1 percent to 7.7 percent; 25th percentile from 5.2 percent to 4 percent.
  - Majority of corridors still have median fees above 5 percent in 2020.
- Provider types and market shares over time
  - Median bank fee persistently above median MTO fee.
  - Average share of banks in a corridor decreased from 27 to 17 percent between 2011 and 2020.
  - Share of MTOs increased from 69 to 80 percent over the same period.
- Access points, payment instruments and speed (2011 → 2020)
  - Bank branch access share declined to less than 5 percent in 2020.
  - Internet access share increased from 18 to 33 percent.
  - Agent access share stable between 40 and 45 percent.
  - Cash payments in sending country decreased from 56 to 42 percent.
  - Bank transfer and card payments in 2020: 34 percent and 12 percent, respectively.
  - Transactions arriving within one hour increased from 43 to 56 percent.
  - Same-day transactions increased from 10 to 12 percent.
  - Transactions taking six or more days decreased from 8 to one percent.
  - Mobile-money-based transactions remain very rare.

### III. Methodology and explanatory variables
- Regression frameworks
  - Cross-sectional corridor model for 2018: Pij = β1(Sending country factors)i + β2(Receiving country factors)j + β3(Corridor-specific factors)ij + εij.
  - Pooled panel and fixed-effects models across 2011–2020: Pijt models that add corridor, country, and year fixed effects (μi, νj, λt).
  - Firm-level model: Pfijt = β1(Sending country factors)it + β2(Receiving country factors)jt + β3(Corridor-specific factors)ijt + β4(Firm-specific factors)ijt + εfijt.
  - Median costs per corridor-year as main dependent variable; firm-level regressions use firm costs; standard errors clustered at corridor level for firm regressions.
- Key explanatory variable groups (lagged at country/corridor level; contemporaneous firm-level factors)
  - Socioeconomic: GDP per capita (WDI), rural population share (WDI), bank branch penetration per capita (robustness).
    - GDP per capita (sending countries range): 936 USD (Tanzania) to 80,000 USD (Switzerland) among sending countries (2017).
    - GDP per capita (receiving countries range): 340 USD (Malawi) to 19,700 USD (Estonia) among receiving countries (2017).
    - Rural population share ranges from zero (Singapore, Kuwait) to 83 percent (Malawi).
  - Corridor characteristics: bilateral trade volume (WITS, available up to 2017; ranges from 412,000 between Nepal and Oman to 683 billion between the US and China in 2018), geographic distance between capitals (CEPII; 215 km to 17,676 km), bilateral migrant stock (UN International Migrant Stock; available for 124 corridors in 2017).
  - Market structure proxies: number of providers per corridor (ranges: 1 to 26 in 2017), share of banks among providers (zero in 139 out of 300 corridors to one in four corridors in 2017).
  - Macroeconomic/regulatory: dummy for pegged exchange rates (including currency boards, de facto pegs, no separate legal tender); capital controls dummy (IMF ARER); regulation index (0–5) from Global Payment Systems Survey 2008; Basel Institute AML risk index (0–10, higher = higher risk).
    - In 2018 cross-sectional sample: 15.2 percent of source countries and 13.4 percent of destination countries have fixed exchange rates.
  - Firm/product features: provider type (bank, MTO, post office), payment instrument (cash, bank transfer, card, mobile money), access point (agent, bank branch, post office, Internet, mobile phone), speed categories (less than an hour; same-day; next-day; two days; three to five days; six or more days). Constructed variable Fast indicates same-day or faster.

### IV. Main empirical results (high-level findings)
- Five headline results from regression analysis (across corridor, time and provider dimensions)
  1. Higher sending-country GDP and easier geographic access to financial institutions are associated with lower fees, especially for banks.
     - An increase of one standard deviation in the sending country’s GDP per capita is associated with about half a percentage lower remittance fees (cross-sectional coefficient, column (1) reference).
  2. Scale economies matter: larger market size (closer economic ties, larger migrant population) and shorter distance are associated with lower costs.
     - One standard deviation increase in bilateral trade within a corridor over time is associated with one percentage point lower remittance fees (fixed-effects, column 11).
     - One standard deviation increase in distance is associated with 0.6 percentage points higher remittance costs (column 1).
     - One standard deviation increase in migrant population is associated with 1.3 percentage points lower remittance costs (column 2).
  3. Market structure is important: banks charge higher fees than money transfer operators (MTOs); a larger share of banks among providers is associated with higher fees charged by MTOs.
     - One standard deviation increase in the share of banks among market participants is associated with higher remittance costs (significant across regressions at least at the 10 percent level).
     - A standard deviation increase in the number of market participants reduces remittance costs by 0.3 percentage points (column 1).
     - Median bank fee is persistently above median MTO fee; share of banks in corridors fell from 27 to 17 percent while share of MTOs rose from 69 to 80 percent (2011–2020).
  4. Exchange rate stability matters: in corridors where the sending country has a pegged exchange rate, both banks and MTOs charge lower fees (negative relationship).
  5. Payment mode and digitalization: cash payments are associated with higher fees; Internet payments are associated with lower fees.
     - Internet access share increased from 18 to 33 percent (2011–2020); cash payments decreased from 56 to 42 percent.
- Additional regression nuances
  - Rural population: one standard deviation increase in rural population share of sending country associated with one percentage point higher remittance costs (column 1). An increase in rural population share of a destination country over time is associated with 2.2 percentage points higher remittance costs (fixed-effects).
  - Bank branch penetration: results vary by specification; branch penetration in sending country enters negatively and significantly in fixed-effects regressions (consistent with rural population interpretation).
  - Several regulatory variables and GDP per capita of destination country show inconsistent signs across samples and subsamples (e.g., negative in overall and bank sub-sample, positive in MTO sub-sample).
  - Many explanatory variables are correlated with each other; results are partial correlations, not causal estimates. Authors explicitly caution against causal interpretation.

### V. Interpretation and policy-relevant implications (from results)
- Structural constraints:
  - Higher transaction costs driven by a more rural population in sending countries and lower scale in some corridors imply limits to how far policy can reduce remittance fees purely via regulatory change.
- Policy-relevant avenues for cost reduction:
  - Strengthening competition, particularly from non-bank providers (MTOs), is associated with lower fees.
  - Digitalization and increased Internet-based access (versus cash) are associated with lower fees.
  - Exchange rate stability (or improved hedging possibilities) is associated with lower fees and can reduce provider risk premia.
- Caveats:
  - Analysis documents partial correlations; causal inference is not claimed.
  - Market structure and regulatory variables may be endogenous to remittance fees.
  - Not all variables are available for all corridors or years; multiple specifications and subsamples are used.

*Source: IMF Working Paper (sections II–IV as provided in the supplied content).*

### 0.7 percentage points higher remittance costs (column 1). While crude indicators of market

### wpiea2022063-print-pdf - 0.7 percentage points higher remittance costs (column 1). While crude indicators of market

### Baseline results (Tables 3–4)
- R-squared in cross-sectional regressions varies between 17 percent and 42 percent; in pooled regressions between 23 percent and 40 percent.  
- Fixed-effect regressions produce R-squared above 80 percent, driven mostly by source country and year-fixed effects.  
- Strongest predictors of cross-corridor variation in remittance costs:
  - (i) income per capita and geographic access to financial institutions in the sending country;
  - (ii) economic links between source and destination country;
  - (iii) market structure in the remittance service market;
  - (iv) the exchange rate regime in the sending country.
- Correlational findings:
  - Lower fees in corridors with sending countries with higher GDP per capita.
  - Lower fees in corridors with a lower rural population share in sending country.
  - Lower fees in corridors with a fixed exchange rate in sending country.
  - Lower fees in corridors with stronger trade and migration links and shorter distance.
  - Lower fees in corridors with more players and fewer banks.

### Regulatory and macroeconomic indicators (selected coefficients and patterns)
- Broader regulation of remittance service providers in the sending country (Regulation index 2008) is associated with higher remittance costs in cross-sectional and pooled regressions; regulation index for the receiving country enters negatively in pooled regression (column 3).
- If the currency of the sending country is pegged, remittance costs are lower (significant in several specifications).
- Reporting requirements: evidence that reporting requirements in destination country are associated with lower remittance costs, while such requirements in the source country are associated with higher remittance costs (column 3).
- AML risk: higher AML risk in the sending country is associated with higher remittance costs; higher AML risk in the receiving country associated with lower remittance costs in some specifications (column 4).
- Caution: coefficient signs for regulatory variables are sometimes counter-intuitive and vary across models; interpretation is careful due to potential endogeneity and omitted factors.

### Corridor-level regressions: Banks vs. MTOs (Tables 5–6)
- Sample sizes:
  - Banks: up to 232 corridors.
  - MTOs: up to 336 corridors.
- Banks (Table 5):
  - GDP per capita of sending and receiving country negatively associated with remittance cost in cross-section; within-country (fixed effects) estimates turn positive and significant in several specifications.
  - Rural population share in sending country associated with higher remittance costs in cross-sectional and pooled regressions; receiving-country rural share positive in fixed-effects regressions.
  - Higher trade volume associated with lower remittance costs primarily in fixed-effects regressions; in cross-sectional and pooled regressions trade sometimes enters positively.
  - Higher distance associated with higher remittance costs.
  - Larger number of remittance service providers associated with higher remittance costs (pooled and fixed effects); share of banks associated with higher remittance costs (pooled regressions).
  - Pegged exchange rates in sending country associated with lower remittance costs.
  - Observed coefficient examples (banks-only regressions):
    - Ln (GDP per capita) source country: coefficients as large as -1.414*** (column 1).
    - Pegged exchange rate source country: -4.204*** (column 11) and -28.10*** (column 15) in some specifications.
    - Number market participants: positive and significant in pooled regressions (e.g., 0.256*** in column 6).
- MTOs (Table 6):
  - GDP per capita in sending and receiving country often positively and significantly associated with remittance prices (not always consistent).
  - Rural population share in sending country positively associated with remittance prices; receiving-country rural share negatively associated (primarily cross-sectional and pooled).
  - Strong evidence that higher trade volume in the corridor is associated with lower remittance costs.
  - Distance: no clear relationship for MTOs.
  - Larger migrant population associated with lower remittance costs.
  - Number of market participants: significant and negative association with remittance costs (competition effect) across specifications (e.g., -0.175*** in pooled regressions).
  - Share of banks: positive association with MTO remittance costs (e.g., 5.018*** in column 6).
  - Pegged exchange rate source country associated with lower MTO fees (e.g., -2.423*** and -13.67*** in some specs).
  - AML index source country sometimes associated with lower MTO fees (e.g., -1.345*** in column 6).

### Provider-specific regressions: Western Union and MoneyGram (Tables 7–8)
- Western Union (Table 7):
  - Higher GDP per capita in sending or receiving country often associated with higher remittance costs (coefficients frequently positive; significance varies).
  - Rural population share in source country mostly positively associated with remittance costs where significant.
  - Bilateral trade tends to predict lower remittance costs (negative coefficients: e.g., -0.276* column 1; larger negative coefficients in pooled/fixed specs).
  - Number of market participants generally reduces fees for Western Union (negative and often significant pooled/fixed estimates).
  - Share of banks where significant tends to be associated with higher fees for Western Union (e.g., 3.067** column 5; 3.318*** column 6).
  - Pegged exchange rate source country associated with lower fees (e.g., -3.793*** column 11; -4.641*** column 12).
- MoneyGram (Table 8):
  - Similar patterns to Western Union: GDP per capita often positive, bilateral trade negative, number of market participants negative and significant across specifications (e.g., -0.156*** column 1; -0.145*** column 11).
  - Pegged exchange rate source country strongly associated with lower fees (e.g., -4.921*** column 11; -5.590*** column 12; -12.08*** column 15).
  - AML index source country shows large negative coefficients in some specifications (e.g., -2.365*** column 11).

### Firm-level regressions and product characteristics (Table 9 and firm-level analysis)
- Regression designs:
  - Five specifications in Table 9 including baseline clustered at corridor level (column 1), cross-sectional 2018 with and without corridor fixed effects (columns 2–3), and panel regressions with and without corridor-year fixed effects (columns 6–8).
- Provider-type effects (exact reported ranges):
  - Banks charge higher fees: in 2018 cross-sectional regressions banks have 1.3 to 1.9 percentage points higher fees than other providers.
  - MTOs charge lower fees: in 2018 cross-sectional regressions MTOs have 0.9 to 1.3 percentage points lower fees than other providers.
  - In panel regressions over time, bank coefficient sizes are larger; MTO coefficients turn insignificant in panel specifications.
- Delivery and channel effects (exact reported ranges):
  - Internet transactions: 1.7 to 2.7 percentage points lower fees (Internet dummy: -2.376***, -2.690***, -2.391*** in columns 1–3).
  - Paying at bank branches: more expensive than paying at agents or via Internet, but cheaper than paying at post offices or other outlets (Branch and Agent coefficients: Agent around -2.157*** to -2.762***; Branch around -1.213* to -1.340* depending on spec).
  - Cash payments: associated with higher fees in 2018 cross-sectional regressions—0.7 to 0.9 percentage points higher (Cash payment coefficients: 0.742**, 0.892**, 0.770** in columns 1–3).
  - Fast transfer (same day) dummy: not significant at 5 percent in any regression (e.g., -0.183 to 0.265 across specs).
- Corridor- and country-level firm-regression findings:
  - Share of banks in corridor associated with higher fees in firm-level regressions (e.g., Share of banks 5.875*** column 1; 2.584** column 2).
  - Number of market participants generally associated with lower fees in firm-level cross-sections (e.g., -0.0811** column 1; -0.0822** column 2; -0.0859*** column 3).
  - Ln(distance) positive and significant in firm-level cross-sections (e.g., 0.558**, 0.592**, 0.545** in columns 1–3); very large positive coefficients in some fixed-effect specs (19.12***, 19.51*** in columns 6–7 suggesting specification differences when fixed effects included).

### Summary conclusions (V. CONCLUSIONS)
- Five key results highlighted:
  1. Higher GDP per capita in sending countries and easier geographic access to financial institutions (especially for banks) are associated with lower fees.
  2. Scale economies matter: closer economic ties and larger migrant stocks are associated with lower costs; shorter distance also lowers costs.
  3. Market structure is important:
     - Banks charge higher fees than MTOs.
     - A larger share of banks among providers is associated with higher fees charged by MTOs.
     - MTO fees respond to competition (more market players → lower MTO fees); bank fees do not respond similarly to increased competition.
  4. Pegged exchange rates in the sending country are associated with lower remittance costs for both banks and MTOs.
  5. Product/channel effects: cash payments attract higher fees; internet payments lower fees.
- Regulatory framework impact remains inconclusive due to mixed coefficient signs and potential endogeneity.
- Policy-relevant implication: structural country factors may limit further fee reductions, but stronger competition—particularly from non-bank providers—and digitalization can help reduce remittance costs.

*Italic: Content extracted from the provided IMF working paper content unit.*

### Appendix Table. Variables, Definitions and Sources

### Appendix Table. Variables, Definitions and Sources

### Variables, Definitions and Sources

- Remi ttance fee  
  - Definition: Remi ttance fee per 200 (500) US dol l ars  
  - Source: Remi ttance Pri ces  Worl dwi de (Worl d Bank)

- Ln (GDP per capi ta) s ource country  
  - Definition: Log of GDP per capi ta i n s endi ng country  
  - Source: Worl d Devel opment Indi cators

- Ln (GDP per capi ta) des ti nati on country  
  - Definition: Log of GDP per capi ta i n s endi ng country  
  - Source: Worl d Devel opment Indi cators

- Rural  popul ati on s hare s ource country  
  - Definition: Percentage of popul ati on l i vi ng i n rural  areas  i n s endi ng country  
  - Source: Worl d Devel opment Indi cators

- Rural  popul ati on s hare des ti nati on country  
  - Definition: Percentage of popul ati on l i vi ng i n rural  areas  i n recei vi ng country  
  - Source: Worl d Devel opment Indi cators

- Ln (Bi l ateral  trade)  
  - Definition: Log of bi l ateral  trade  
  - Source: Worl d Integrated Trade Sol uti on (WITS) - Worl d

- Ln(di s tance)  
  - Definition: Log of di s tance between s endi ng and recei vi ng c ountri es  (i n km, between the capi tal  ci ti es )  
  - Source: Di s tances  Databas e (CEPII)

- Branch penetrati on s ource country  
  - Definition: Branches  of commerci al  banks  per capi ta (100,000 adults ) i n s endi ng country  
  - Source: Databas e on Acces s  to Fi nanci al  Servi ces  (IMF)

- Branch penetrati on des ti nati on country  
  - Definition: Branches  of commerci al  banks  per capi ta (100,000 adults ) i n recei vi ng country  
  - Source: Databas e on Acces s  to Fi nanci al  Servi ces  (IMF)

- Share of banks  
  - Definition: percentage of banks  among res pondents  i n a corri dor and year  
  - Source: authors ' cal cul ati on, bas ed on Remi ttance Pri ces  Worl dwi de (Worl d Bank)

- Number market parti ci pants  
  - Definition: number of res pondents  i n a corri dor and year  
  - Source: authors ' cal cul ati on, bas ed on Remi ttance Pri ces  Worl dwi de (Worl d Bank)

- Ln(mi grants ) 2017  
  - Definition: Log of the number of mi grants  i n each corri dor (by ori gi n and des ti nati on), "mi grant" bei ng defi ned as  forei gn-born popul ati on when the data i s  avai l abl e, otherwi s e by country of ci ti zens hi p  
  - Source: Internati onal  Mi grant Stock Databas e. Trends  i n Internati onal  Mi grant Stock: The 2017 revi s i on. (Uni ted Nati ons , Department of Economi c and Soci al  Affai rs , Popul ati on Di vi s i on)

- Regul ati on Source country 2008  
  - Definition: Index of i mportance of  banks  i n the provi s i on of remi ttances  i n the s endi ng country, degree to whi ch central  banks  cons i der commerci al  banks  to be s i gni fi cant remi ttance s ervi ce provi ders  (hi gher i s  more rel evant)  
  - Source: Worl d Bank Gl obal  Payment Sys tems  Survey (Worl d Bank 2008)

- Regul ati on des ti nati on country 2008  
  - Definition: Index of i mportance of  banks  i n the provi s i on of remi ttances  i n the recei vi ng country, degree to whi ch central  banks  cons i der commerci al  banks  to be s i gni fi cant remi ttance s ervi ce provi ders  (hi gher i s  more rel evant)  
  - Source: Worl d Bank Gl obal  Payment Sys tems  Survey (Worl d Bank 2008)

- Pegged exchange rate s ource country  
  - Definition: Dummy for pegged exchange rate  
  - Source: Exchange Rate Regi me Cl as s i fi cati on (Il zetzki , Rei nhart, and Rogoff)

- Pegged exchange rate des ti nati on country  
  - Definition: Dummy for pegged exchange rate  
  - Source: Exchange Rate Regi me Cl as s i fi cati on (Il zetzki , Rei nhart, and Rogoff)

- AML i ndex s ource country  
  - Definition: Index of AML ri s k i n s endi ng country  
  - Source: Bas el  Ins ti tute on Governance

- AML i ndex des ti nati on country  
  - Definition: Index of AML ri s k i n recei vi ng country  
  - Source: Bas el  Ins ti tute on Governance

- Capi tal  control s  des ti nati on country  
  - Definition: Takes  the val ue of 1 i f control s  on pers onal  capi tal  trans acti ons  i n the recei vi ng country  
  - Source: Annual  Report on Exchange Arrangements  and Res tri cti ons  (IMF)

- Capi tal  control s  s ource country  
  - Definition: Takes  the val ue of 1 i f control s  on pers onal  capi tal  trans acti ons  i n the s endi ng country  
  - Source: Annual  Report on Exchange Arrangements  and Res tri cti ons  (IMF)

- Bank  
  - Definition: Dummy i f remi ttance s ervi ce provi der i s  a bank  
  - Source: Remi ttance Pri ces  Worl dwi de (Worl d Bank)

- MTO  
  - Definition: Dummy i f remi ttance s ervi ce provi der i s  a money trans fer operator  
  - Source: Remi ttance Pri ces  Worl dwi de (Worl d Bank)

- Branch  
  - Definition: Dummy i f remi ttance has  to be s ent through branch  
  - Source: Remi ttance Pri ces  Worl dwi de (Worl d Bank)

- Agent  
  - Definition: Dummy i f remi ttance can be s ent through agent  
  - Source: Remi ttance Pri ces  Worl dwi de (Worl d Bank)

- Internet  
  - Definition: Dummy i f remi ttance can be s ent through Internet  
  - Source: Remi ttance Pri ces  Worl dwi de (Worl d Bank)

- Cas h payment  
  - Definition: Dummy i f remi ttance payment i s  done i n cas h  
  - Source: Remi ttance Pri ces  Worl dwi de (Worl d Bank)

- Fas t trans fer  
  - Definition: Dummy i f payment arri ves  the s ame day  
  - Source: Remi ttance Pri ces  Worl dwi de (Worl d Bank)

*Appendix Table. Variables, Definitions and Sources — What Explains Remittance Fees? Panel Evidence Working Paper No. WP/2022/063*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022063-print-pdf.pdf_