## wpiea2022218-print-pdf

## Source details

**Canonical URL:** [wpiea2022218-print-pdf](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022218-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022218-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022218-print-pdf.pdf.json)

---

### Main objectives and context
- Investigate: How elastic are remittances to changes in transaction costs? What factors or policy interventions explain cross-country differences in the cost elasticity of remittances?
- Sample and period:
  - 71 countries over 2011Q1–2020Q4.
  - Quarterly remittance costs from “Remittance Prices Worldwide” and a newly constructed quarterly remittance flows dataset.
- Policy benchmark:
  - SDG 10.C target: “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”.

### Key empirical findings — aggregate and dynamic magnitudes
- Main short-run elasticity estimates:
  - Static fixed-effect estimates: transaction cost ($200, log) coefficients range from -0.201 [0.046]*** to -0.075 [0.021]*** across specifications (Table 1).
  - Dynamic Local Projection (LP) estimate: a 10 percent reduction in transaction costs results in a 0.87 percent rise in remittances in the first quarter after the shock.
  - Paper summary phrasing: “A 10 percent reduction in transaction costs leads to a 0.9 percent increase in remittances in the first quarter after the shock.”
  - LP-implied elasticity used in accounting exercise: 0.087 (i.e., 0.87 percent increase in remittances per 10 percent reduction in transaction costs).
- Aggregate counterfactual using 2020 baseline:
  - Remittances to low and middle-income countries in 2020: US$705.5 billion.
  - Average transaction cost in 2020: 6.3 percent.
  - Decline from 6.3 percent to UN target of 3 percent = 52.4 percent decline.
  - Applying elasticity 0.087 implies a 4.5 percent increase in remittances = US$32.2 billion.
  - Direct migrant cost-savings accounting: (6.3-3)*705.5 = US$23.28 billion.
  - Observation: US$32.2 billion is US$8.9 billion more than the saving in costs for migrants, indicating remittance increases exceed migrants’ direct cost savings.
- Robustness:
  - Results are robust to an instrumental variable (IV) approach; IV-LP confirms higher transaction costs undermine remittance flows.
  - IV first-stage: share of MTOs in remittance market is negative and highly significant for transaction costs.

### Stylized facts on remittance costs and providers
- Sample average: Average fee as a share of a US$200 remittance (2011Q1-2020Q4) = 7.5 percent.
- Geographic heterogeneity:
  - Very high costs in southern Africa (e.g., Angola, Botswana, Namibia) exceeding 2.5 times the sample average.
  - Lowest costs prevalent in eastern Europe.
  - Small islands exhibit high costs due to limited scale and financial integration.
- Time trends:
  - Average decline in remittance costs of about 5 percent during 2016Q1–2020Q1.
  - Larger decline in the first half of the decade than in the second half; overall 2011–2020 trend shows modest narrowing of gap between remittance cost and interest rate spread.
- Provider heterogeneity:
  - Banks charge higher remittance fees than Money Transfer Operators (MTOs) and post office providers.
- Correlation:
  - Negative correlation between quarterly remittance costs and remittance flows: lower costs tend to be associated with higher remittance flows.

### Heterogeneity analysis — cost-mitigation and cost-adaptation factors
- Method:
  - Interact transaction costs with policy/structural indicators and plot IRFs at 10th and 90th percentiles to show variation.
- Cost-mitigation findings:
  - Competition (average number of remittance service providers, RSPs):
    - 10th percentile (6 RSPs): first-quarter elasticity ≈ -0.1.
    - 90th percentile (19 RSPs): elasticity declines to around -0.3 and becomes insignificant as number of providers increases.
  - Financial sector development (private sector credit to GDP; branches per km^2; deposit accounts per 1,000 adults):
    - High financial depth (private sector credit = 69.7 percent of GDP at 90th percentile) yields elasticity not significant at conventional levels.
    - Banks often not the cheapest remittance channel—limits on the marginal effect of deeper banking alone.
  - Correspondent Banking Relationships (CBRs):
    - Remittances react more to costs where the number of correspondent banks is smaller.
    - Sub-regional CBR proxy used; interpretation with caution.
- Cost-adaptation findings:
  - Price transparency (share of RSPs classified as “transparent”):
    - 10th percentile (share transparent = 84 percent): first-quarter elasticity = -0.23.
    - 90th percentile (share transparent = 100 percent): elasticity = -0.09.
  - Financial literacy (gross secondary school enrollment rate; deposit accounts per adult):
    - 10th percentile (secondary enrollment = 41 percent): remittances more sensitive to transaction costs.
    - 90th percentile (secondary enrollment = 106 percent): remittances less sensitive to transaction costs.
  - ICT development and affordability:
    - Higher ICT development (IDI and sub-indices: access, capability) associated with lower cost-elasticity of remittances.
    - ICT use sub-index result less conclusive; complementarities across ICT dimensions matter.
    - ICT affordability index (fixed-broadband monthly charge; price of 3-minute mobile local call; price of 3-minute local call to fixed-line) shows stronger elasticity where ICT costs are high.
    - Augmented IDI combining access, use, capability and multiplicative inverse of affordability confirms results.

### Micro evidence — US-Mexico corridor
- Corridor facts (2011–2020 context):
  - Largest global corridor with volume ~US$25bn in 2017.
  - Number of RSPs increased from 20 in 2011 to 26 in 2020.
  - Number of banks providing remittances declined from 7 in 2011 to 4 in 2020; number of MTOs increased from 13 to 22.
  - Cost of a US$200 remittance: 5.5 percent in 2011, 4 percent in 2020.
- Survey data:
  - Bank of Mexico annual survey 2013–2017, sample per year varies from 6,800 to 13,000 individuals; pooled observations up to 37,389.
  - Cross-sectional identification; regressions include time dummies and fixed effects for US state of residence and Mexican state of family.
- Main micro results:
  - Transaction costs (percent of transferred amount, log) baseline coefficients (Table 2):
    - Column (1): -0.703 [0.035]***.
    - Column (2): -0.677 [0.038]***.
    - Column (3): -0.636 [0.037]***.
    - Column (4): -0.635 [0.031]***.
  - Financial literacy interactions (Table 3):
    - Education * Transaction costs (log): 0.012 [0.005]** (lower elasticity among more educated migrants).
    - Sender access to a bank account * Transaction costs (log): 0.133 [0.029]***.
    - Receiver access to a bank account * Transaction costs (log): 0.111 [0.016]***.
  - IV and robustness (Table 4):
    - IV (Remittances (log)): Transaction costs (log) = -0.185 [0.093]* (Column (1)).
    - Fixed effect (Remittances as share of migrant’s income (log)): Transaction costs (log) = -0.867 [0.013]***; Frequency of remittances = -0.418 [0.012]*** (Column (2)).
    - IV (Remittances as share of migrant’s income (log), instrumented): Transaction costs (log) = -0.443 [0.064]*** (Column (3)).
- Micro corroboration:
  - Migrants who face higher transaction costs remit less, robust to socio-economic controls and IV strategies.
  - Education and access to bank accounts mitigate cost-elasticity consistent with macro-level heterogeneity findings.

### Methodology highlights
- Dynamic Local Projections (Jordà, 2005) with lags n=4 and horizons h up to 5 months; LP augmented with Teulings and Zubanov (2014) correction.
- Static fixed-effect estimators with time dummies for cross-sectional specifications.
- Robust standard errors: Driscoll and Kraay (1998) for heteroscedasticity, autocorrelation and cross-sectional dependence.
- Instrumental variable approach: share of MTOs in the remittance service market as instrument for transaction costs in IV-IRF specification.

### Policy-relevant implications and recommendations
- Quantitative policy implications:
  - Reducing transaction costs has a measurable, short-term impact on remittance inflows (10 percent cost reduction → ~0.87–0.9 percent remittance increase in first quarter).
  - Reducing global average transaction costs from 6.3 percent to 3 percent could increase remittances by US$32.2 billion (4.5 percent), exceeding direct migrant savings of US$23.28 billion.
- Recommended policy actions:
  - Cost-mitigation policies:
    - Promote competition among banks and Money Transfer Operators (MTOs) through adapted regulation and market entry facilitation.
    - Deepen financial systems (branch and deposit account expansion, private sector credit development).
    - Preserve and expand correspondent banking relationships to avoid higher sensitivity to costs where CBRs are thin.
  - Cost-adaptation policies:
    - Enforce price transparency (require transfer companies to publish full fees and exchange-rate margins).
    - Improve financial literacy (education and targeted financial education) to reduce migrants’ and recipients’ sensitivity to fees.
    - Improve ICT access, capability and affordability to foster digital channels and ease cost-comparison across providers.
- Complementarities:
  - Simultaneous progress on multiple fronts (competition, transparency, financial inclusion, ICT) is likely to have multiplicative effects on reducing costs and lowering remittance cost-elasticity.
- Pandemic note:
  - Cost-elasticity of remittances was lower in 2020 than prior years; border closures and limited informal channels likely muted responsiveness.

### Concluding summary
- Transaction costs have a robust, negative and economically meaningful effect on remittance volumes, primarily in the short run.
- Heterogeneity across countries is substantial: competition, financial development, correspondent banking ties, price transparency, financial literacy and ICT development and affordability materially shape the sensitivity of remittances to transaction costs.
- Policy reforms that both mitigate costs and help migrants adapt (information, literacy, ICT) can increase remittance flows and amplify development financing to low- and middle-income countries without adding to public debt.

*IMF Working Paper — How Do Transaction Costs Influence Remittances? (wpiea2022218-print-pdf)*

### References .............................................................................................................

### wpiea2022218-print-pdf - References .............................................................................................................

### Figures and Tables (inventory)
- Figures listed include:
  - Average fee as a share of a $200 USD remittance (2011Q1-2020Q4) (percent)
  - Change in Average fee for a $200 USD remittance (2016Q1-2020Q1) (percent)
  - Trends in Average fee for a $200 USD remittance and Interest Rate Spread (2011-20) (percent)
  - Average fee for a $200 USD remittance by Type of Provider (2011Q1-2020Q4) (percent)
  - Remittances Flows and Average fee for a $200 USD remittance (2011Q1-2020Q4)
  - Cost-Elasticity of Remittances: Local Projections
  - Cost-Elasticity of Remittances with Respect to Competition in the Remittance Market
  - Cost-Elasticity of Remittances in Countries with Low and High Financial Development
  - Cost-Elasticity of Remittances and Correspondent banking relationships
  - Cost-Elasticity of Remittances and Price Transparency
  - Cost-Elasticity of Remittances and Education level
  - Cost-Elasticity of Remittances and ICT Development
  - Cost-Elasticity of Remittances and ICT Affordability

- Tables listed include:
  - Transaction Costs and Remittances: Fixed-Effect Estimates
  - Transaction Costs and Remittances in the US-Mexico Corridor
  - Transaction Costs and Remittances in the US-Mexico Corridor: The Role of Financial Literacy
  - Transaction Costs and Remittances in the US-Mexico Corridor: Instrumental Variable Approach

- Annexes and annex materials listed:
  - ANNEXES: Sample Composition; Summary Statistics and Correlation Matrix
  - ANNEX TABLES: Cost-Elasticity of Remittances: Local Projections; First-Stage Regression
  - ANNEX FIGURES: Change Cost-Elasticity of Remittances: Instrumental Variable Local Projections; Cost-Elasticity of Remittances: Low vs High Transaction Cost; Cost-Elasticity of Remittances with respect to the Geographical Coverage of Financial Institution; Cost-Elasticity of Remittances with respect to Access to Deposit Accounts; Cost-Elasticity of Remittances and ICT Access; Cost-Elasticity of Remittances and ICT Use; Cost-Elasticity of Remittances and ICT Capability

### Introduction: context and objectives
- Remittances are a key source of funding for developing countries; according to World Bank data, remittances to low- and middle-income countries more than doubled during the past 15 years to reach US$550 billion in 2021.
- Over half of remittances goes to people in rural areas; about 75 percent is used to cover basics such as food and medical or school expenses, while the remaining is invested in assets or saved (IFAD, 2021).
- The SDGs include the objective to “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” (SDG 10.C).
- High transaction costs (fees, exchange rate margin and other costs) imply between 5 and 15 percent of remittances are “lost” due to high transaction costs, depending on country and amounts (Ratha, 2021).
- Paper’s core questions:
  - How elastic are remittances to changes in transaction costs?
  - What factors or policy interventions explain cross-country differences in the cost elasticity of remittances?

### Literature and contribution
- Prior findings summarized:
  - Recorded remittances depend positively on migrant stocks and negatively on transaction costs and exchange rate restrictions (Freund and Spatafora, 2008).
  - Lower transaction costs associated with more developed financial systems and less volatile exchange rates.
  - Banking competition and lower barriers to access banking services reduce remittance costs (Beck and Martinez Peria, 2011).
  - Cost- and risk-based constraints and market structure hinder affordable remittance transaction costs (Beck, Janfils, and Kpodar, 2022).
  - Regulatory issues, lack of price transparency, restrictive licensing and thin remittance markets likely increase costs (da Silva Filho, 2021).
- Gap addressed:
  - Existing studies estimate cost elasticity but do not systematically analyze how elasticity varies across countries with different characteristics.
- Improvements introduced in this paper:
  - Use of a new quarterly database on remittances enabling high-frequency dynamic analysis via local projections (Jorda, 2005).
  - Systematic analysis of factors shaping cost elasticity, distinguishing cost-mitigation policies (competition, financial sector development, correspondent banking relationships) and cost-adaptation policies (price transparency, financial literacy, ICT development).
  - Microanalysis using USA-Mexico corridor data to assess heterogeneity with respect to financial literacy.

### Key empirical findings (sample and main estimates)
- Sample: 71 countries over the period 2011Q1-2020Q4.
- Main elasticity estimate:
  - A 10 percent reduction in transaction costs leads to a 0.9 percent increase in remittances in the first quarter after the shock.
  - The impact becomes statistically insignificant from zero in subsequent quarters, implying the response is essentially short-term in nature.
- Policy-target implication:
  - Moving from the 2020 level of transaction costs (6.3 percent) to the SDG target of 3 percent will generate an additional US$32 billion in remittances, much larger than the direct cost savings.
  - Interpreted implication: migrants would not only fully pass on the cost savings to families, but also send more, implying an absolute elasticity higher than one.
- Robustness:
  - Results are robust to an instrumental variable approach.

### Heterogeneity: cost-mitigation and cost-adaptation factors
- Cost-mitigation findings:
  - Where competition in the remittance market is high, the financial system is developed, and ties with correspondent banks hold up, the elasticity of remittances to transaction costs is much lower.
  - Interpretation: remittances are less sensitive to transaction costs where alternative informal channels to repatriation exist.
- Cost-adaptation findings:
  - Less opaque remittance transaction costs, improved financial literacy, and higher ICT development (dimensions: ICT use, access, capability and affordability) help explain why some countries may have a lower cost-elasticity of remittances.

### USA-Mexico corridor micro evidence
- Data: an annual survey of the Bank of Mexico during 2013-2017 covering over 37,000 individuals.
- Findings:
  - Migrants who face higher transaction costs tend to remit less, even after controlling for socio-economic characteristics.
  - Education level or access to a bank account, as proxies of financial literacy, mitigate the cost-elasticity of remittances, consistent with panel-level findings.

### Methodology and structure
- Methodology highlights:
  - Dynamic local projections (Jorda, 2005) used to estimate short- and medium-term impacts of shocks to transaction costs.
  - Instrumental variable approach employed for robustness.
- Paper structure:
  - Section II presents data, stylized facts, empirical model and methodology.

*IMF WORKING PAPERS How Do Transaction Costs Influence Remittances? INTERNATIONAL MONETARY FUND*

### Section III discusses the key results of the paper, with additional findings relegated to the

### Section III — Key Results: Data, Empirical Strategy, and Findings

### Data and Empirical Strategy
- Remittance cost data source: “Remittance Prices Worldwide” (World Bank), covering 365 corridors (48 sending and 105 receiving countries). Data available quarterly from 2011Q1 onwards. Main variable: remittance cost as a share of the amount transferred averaged at the country level.
- Annual remittance inflows database (World Bank) covers 215 countries and territories but is annual; to exploit quarterly variation the authors construct a new quarterly remittance flows dataset for 95 countries (18 high-income, 62 middle-income, 15 low-income). Data run from 1971Q1 for a handful of countries through 2020Q4 for most countries.
- Quarterly remittance compilation follows the international definition: remittances = personal transfers + compensation of employees. When compensation of employers missing, it is not included owing to marginal size relative to personal transfers.
- Combined quarterly remittance flows and costs yield a sample of 71 countries for 2011Q1–2020Q4.
- Control variables: income per capita (receiving and sending countries), US dollar/local currency exchange rate, number of migrants (stock), population as alternative proxy. Income per capita of sending countries computed as weighted average of host countries’ income per capita using host share in migrants as weights.
- Estimation approach:
  - Static model: fixed-effect estimator (time dummies included).
  - Dynamic model: Local Projection (LP) approach (Jordà, 2005) with lags n=4 and horizon h up to 5 months; LP augmented with Teulings and Zubanov (2014) correction factor to control for subsequent shocks.
  - Robust standard errors (Driscoll and Kraay, 1998) to address heteroscedasticity, autocorrelation and cross-sectional dependence.
  - Instrumental variable (IV) approach uses share of MTOs in remittance service market as instrument for transaction cost in IV-IRF specification.

### Stylized Facts on Remittance Costs
- Average fee as a share of a US$200 remittance (2011Q1-2020Q4) sample average: 7.5 percent.
- Geographic heterogeneity:
  - Very high costs in southern Africa (e.g., Angola, Botswana, Namibia) exceeding 2.5 times the sample average.
  - High costs in small islands (limited scale, limited financial integration).
  - Lowest costs prevalent in eastern Europe.
- Change over 2016Q1-2020Q1:
  - Average decline in remittance costs of about 5 percent during this five-year period.
  - Some countries experienced increases exceeding 40 percent (Gambia, The; Afghanistan; Kyrgyz Republic), driven mainly by higher exchange rate margins from volatile/depreciating currency.
  - Temporary declines in some countries (Lesotho, Eswatini) due to fee reductions by South African banks early in the pandemic.
- Trend 2011–2020:
  - Larger decline in remittance costs in the first half of the decade than in the second half; overall the 5 percent decline in the second half was modest relative to the earlier reduction.
  - Remittance cost consistently higher than interest rate spread over 2011–20; gap narrowed only marginally.
- Provider heterogeneity:
  - Banks charge higher remittance fees than Money Transfer Operators (MTOs) and post office providers.
- Correlation:
  - Clear negative correlation between quarterly remittance costs and remittance flows: countries with lower costs tend to have higher remittance flows (suggestive evidence that costs matter for remittances).

### Model Specification (as in paper)
- Dynamic LP specification (Eq(1)):
  ln(Rem)_{c,t+h} = Σ_{i=1}^n α_i ln(Rem)_{c,t−i} + Σ_{i=1}^n β_i ln(Cost)_{c,t−i} + Σ_{j=0}^h δ_j ln(Cost)_{c,t+j,h} + θ^h X_{c,t} + v_t + u_c + ε_{c,t+h}
  - where Rem = remittances (millions of US$), Cost = fee per US$200 remittance as share, X = controls, v = time dummies, u = country fixed effects, ε robust error term, n=4, h up to 5 months.
- LP favored for handling highly persistent data, non-linearities, and direct multi-step forecasting; Teulings and Zubanov (2014) correction included to control for subsequent shocks.

### Static Fixed-Effect Results — Key Coefficients (Table 1)
- Transaction cost ($200, log):
  - Column (1): -0.201 [0.046]***
  - Column (2): -0.198 [0.045]***
  - Column (3): -0.097 [0.029]***
  - Column (4): -0.091 [0.026]***
  - Column (5): -0.075 [0.021]***
- Transaction cost ($500, log):
  - Column (3) (reported): -0.115 [0.023]***
- GDP per capita (log), receiving country: coefficients across columns include 0.400 [0.064]***, 1.634 [0.744]**, 0.667 [0.048]***, 0.609 [0.056]***, 0.548 [0.052]***, 0.609 [0.057]***
- GDP per capita (log), sending country: coefficients around 0.470 to 0.506 with statistical significance in columns (where reported).
- GDP per capita (log) squared, receiving country: -0.076 [0.045] (marginally significant at 11 percent in text; implied threshold US$46,000).
- USD exchange rate (log): positive and significant (examples: 0.383 [0.030]***; 0.334 [0.035]***; 0.201 [0.068]***; 0.335 [0.035]***), implying remittances in US$ increase with local currency depreciation.
- Migrant population (log): positive (0.257 [0.096]**; 0.271 [0.099]***).
- Total population: 1.623 [0.355]*** (used in some specifications).
- Sample size and fit:
  - Observations: between 2,110 and 2,142 across specs.
  - Number of countries: 69–71.
  - R2: 0.17 to 0.33 (depending on specification).
- Notes: fixed effects estimations; time dummies included; robust standard errors in brackets; *,**,*** denote significance at 10 percent, 5 percent and 1 percent respectively. USD exchange rate denotes units of local currency per USD.

### Dynamic Results — Elasticities and Economic Magnitudes
- LP impulse response:
  - A 10 percent reduction in transaction costs results in a 0.87 percent rise in remittances in the first quarter (LP estimate), consistent with fixed-effect elasticity estimates.
  - The effect in subsequent quarters is not significantly different from zero (not persistent).
  - Similar results when using median cost and US$500 cost measures.
- Elasticity used in accounting exercise:
  - Elasticity implied: 0.087 (i.e., 0.87 percent increase in remittances per 10 percent reduction in transaction costs).
- Aggregate counterfactual (2020 figures):
  - Remittances to low and middle-income countries in 2020: US$705.5 billion.
  - Average transaction cost in 2020: 6.3 percent.
  - Bringing transaction costs to UN target of 3 percent (a decline from 6.3 percent to 3 percent = 52.4 percent decline) multiplied by elasticity 0.087 implies a 4.5 percent increase in remittances (equivalent to US$32.2 billion).
  - Accounting approach (assuming $1 saved by migrants converts $1 to households) gives US$23.28 billion ((6.3-3)*705.5).
  - The paper notes the US$32 billion figure is US$8.9 billion more than the saving in costs for migrants, indicating remittance increases exceed migrants’ direct cost savings.
- IV-LP robustness:
  - IV using share of MTOs as instrument (first-stage: share of MTOs negative and highly significant for transaction costs) confirms that higher transaction costs undermine remittance flows; IRF-IV supports main findings.
- Pandemic year interaction:
  - Interaction of transaction costs and 2020 dummy shows cost-elasticity of remittances was lower in 2020 than prior years (possible explanation: limited informal channels due to border closures).

### Heterogeneities in Cost Elasticity — Policy-Relevant Factors
- Methodology for heterogeneity:
  - Interaction of transaction cost with policy/structural indicators in Eq.(1); plot IRFs at 10th and 90th percentiles of the conditional variable to show variation in elasticity.
- Two policy categories:
  - Cost-mitigation policies: enhancing competition, financial development, addressing correspondent banking relationship (CBR) issues.
  - Cost-adaptation policies: price transparency, financial literacy, ICT and information lowering costs.
- Findings for cost-mitigation policies:
  - Competition in remittance market (proxy: average number of remittance service providers, RSPs):
    - At 10th percentile (6 RSPs): first-quarter elasticity ≈ -0.1.
    - At 90th percentile (19 RSPs): elasticity declines to around -0.3 and becomes insignificant as number of providers increases (implying less sensitivity to costs in more competitive markets).
  - Financial sector development (proxy: private sector credit to GDP; alternatives: branches per km^2, deposit accounts per 1,000 adults):
    - In high financial depth (private sector credit ratio=69.7 percent of GDP at 90th percentile), elasticity is not significant at conventional levels.
    - Difference between 10th and 90th percentiles marginal; banks often not cheapest remittance channel.
  - Correspondent Banking Relationships (CBRs):
    - Using weighted average number of active correspondents by sub-region (CPMI data) as proxy: remittances react more to costs where the number of correspondent banks is smaller.
    - Caution: sub-region series used; assumes regional CBR trends reflect country-level CBRs.
- Findings for cost-adaptation policies:
  - Price transparency (share of RSPs classified as “transparent” in Remittance Prices Worldwide):
    - 10th percentile (share transparent = 84 percent): first-quarter elasticity = -0.23.
    - 90th percentile (share transparent = 100 percent): elasticity = -0.09.
    - Interpretation: increased transparency reduces elasticity; gains from increasing transparency larger where transparency initially low.
  - Financial literacy (proxy: gross secondary school enrollment rate):
    - 10th percentile (secondary enrollment = 41 percent): remittances more sensitive to transaction costs.
    - 90th percentile (secondary enrollment = 106 percent): remittances less sensitive to transaction costs.
    - Alternative proxy: number of deposit accounts with commercial banks yields supportive results (financial inclusion linked to lower cost-sensitivity).

### Policy-Relevant Interpretations from Results
- Reducing transaction costs yields a measurable, short-term increase in remittance flows; a 10 percent cost reduction increases remittances by about 0.87 percent in the first quarter.
- Cost reductions can produce remittance gains that exceed direct migrant savings from lower fees (example: UN target to 3 percent yields US$32.2 billion additional remittances vs. US$23.28 billion migrants’ cost savings in 2020 baseline).
- Policies that increase competition (more remittance service providers, higher share of MTOs), deepen financial systems, preserve and expand correspondent banking relationships, increase price transparency, and boost financial literacy all lower either remittance costs or the sensitivity of remittances to costs.
- Heterogeneity matters: the elasticity of remittances to transaction costs is larger in high-cost, low-competition, low-transparency, low-financial-development settings; policy gains from cost-reduction and transparency improvements are therefore larger in these environments.

*IMF Working Paper — Section III summary based on the provided content.*

### Annex Figure 4).

### Annex Figure 4

### Information and communication technologies (ICT): Mechanisms and measurement
- ICT development has enabled Fintech entry, increasing competition and helping bring down transaction costs.
- ICT fosters digital remittances that increasingly bypass traditional payment systems, resulting in lower transaction costs for migrants.
- Mobile money demonstrates cost-effective financial services to unbanked and under-banked populations (example cited: M-PESA in Kenya).
- ICT reduces information asymmetries by increasing Internet access, allowing remitters to compare costs across remittance services and reduce information costs where price transparency is weak.

- ICT measurement used:
  - ICT Development Index (IDI) compiled by the International Telecommunication Union (ITU).
  - IDI comprises 11 indicators divided into 3 groups measuring ICT access, ICT use and ICT capability.
  - IDI in this study is the simple average of the 3 sub-indices, each computed as the simple average of the normalized value of their components.
  - Normalization: variables have zero mean and unit variance; a max-min transformation was also tested with similar econometric findings.
  - Note: ITU’s original IDI weights sub-indices 40 percent (access), 40 percent (use), 20 percent (capability); this paper gives equal weights to sub-indices to avoid subjective weighting.
  - Because ITU discontinued the IDI in 2017, the index is recalculated using underlying variables for this study.

### ICT development and remittance cost-elasticity: Findings
- Higher ICT development is associated with a lower cost-elasticity of remittances.
- Same result found for ICT access and ICT capability sub-indices.
- Result is not conclusive for the ICT use sub-index, suggesting complementarities across ICT dimensions matter.
- A missing dimension from IDI is ICT affordability; authors construct an ICT affordability index composed of:
  - fixed-broadband monthly subscription charge,
  - price of 3-minute mobile local call (off-peak rate),
  - price of a 3-minute local call to a fixed-telephone line (off-peak rate).
- Findings on affordability:
  - Elasticity of remittances to transaction costs is stronger in countries with high ICT costs than in those with low ICT costs (Figure 13).
  - An augmented IDI that combines access, use, capability and the multiplicative inverse of affordability confirms earlier results.

### Micro evidence: US-Mexico corridor (survey and methods)
- Corridor facts:
  - Largest corridor in the world with a volume of remittances of about US$25bn in 2017.
  - More than 70 percent higher than the second largest corridor: USA-China.
  - Number of remittance service providers increased from 20 in 2011 to 26 in 2020.
  - Number of banks providing remittance transfers declined from 7 in 2011 to 4 in 2020; number of MTOs increased from 13 to 22 over the same period.
  - Cost of a US$200 remittance: 5.5 percent in 2011, 4 percent in 2020.
- Micro-data source and scope:
  - Annual survey by the Bank of Mexico from 2013 through 2017, administered in December to Mexican citizens residing in the US who visited Mexico.
  - Sample size varies from 6,800 to 13,000 individuals.
  - Survey covers demographic characteristics, amount and frequency of remittances, fees paid and means of transfer.
  - Not a panel; identification of elasticity relies on cross-sectional variation.
- Empirical approach:
  - Regress annual remittances per individual on reported fees (transaction costs) and socio-economic controls.
  - Regressions include time dummies, fixed-effects for US state of residence, state of families in Mexico, and sector of employment.

### Main micro results: Fixed-effect and OLS estimates (Table 2 and Table 3)
- Transaction costs (percent of transferred amount, log): consistently negative and highly significant across specifications.
  - Table 2 estimates:
    - Column (1): -0.703 [0.035]***
    - Column (2): -0.677 [0.038]***
    - Column (3): -0.636 [0.037]***
    - Column (4): -0.635 [0.031]***
  - Number of observations: 37,389; 37,074; 37,064; 28,297 across columns (1) to (4).
  - R2: 0.26; 0.29; 0.36; 0.38 across columns (1) to (4).
- Control variable highlights (Table 2):
  - Income (log): positive and significant in some specifications (e.g., Column (3): 0.085 [0.031]***; Column (4): 0.146 [0.026]***).
  - Age (log): negative and significant in early columns (e.g., Column (1): -0.589 [0.079]***), becomes insignificant when number of years lived in the US is included.
  - Gender: men remit more than women (e.g., Column (1): -0.298 [0.034]*** where gender coding is 1 for men and 2 for women).
  - Level of education: more skilled migrants exhibit a lower propensity to remit (e.g., Column (1): -0.032 [0.007]***).

- Financial literacy interactions (Table 3):
  - Education * Transaction costs (log): positive and significant (Column (1) interaction coefficient 0.012 [0.005]**) indicating lower elasticity among skilled migrants.
  - Access to a bank account (sender) * Transaction costs (log): 0.133 [0.029]*** (columns with sender bank account interaction) indicating access to a bank account for sender associated with lower cost elasticity.
  - Access to a bank account (receiver) * Transaction costs (log): 0.111 [0.016]*** (receiver bank account interaction) indicating receiver bank access associated with lower cost elasticity.
  - Table 3 transaction cost baseline coefficients across columns:
    - Column (1): -0.703 [0.035]***
    - Column (2): -0.636 [0.037]***
    - Column (3): -0.697 [0.024]***
    - Column (4): -0.642 [0.033]***
    - Column (5): -0.729 [0.023]***
    - Column (6): -0.632 [0.030]***
    - Column (7): -0.667 [0.030]***
    - Column (8): -0.774 [0.029]***
  - Number of observations by column: 37,389; 37,064; 37,064; 20,862; 20,862; 17,648; 17,648; 17,636.
  - R2 ranges from 0.26 to 0.41 across columns.

### Addressing endogeneity: Instrumental variable and robustness (Table 4)
- Three approaches to address potential endogeneity of transaction costs:
  1. Instrument transaction costs by type of service provider/medium of transfer (exploiting cost differentials between banks and MTOs).
  2. Add frequency of remittances to control indirectly for remittance size; use remittances as a share of remitter’s income as dependent variable.
  3. Combine approach 1 and 2 (instrument cost while using remittances as share of income).
- Table 4 key estimates:
  - Column (1) IV (Dependent variable Remittances (log)):
    - Transaction costs (log): -0.185 [0.093]*.
    - Number of observations: 37,061.
    - R2: 0.15.
  - Column (2) Fixed effect (Dependent variable Remittances as a share of migrant’s income (log)):
    - Transaction costs (log): -0.867 [0.013]***.
    - Frequency of remittances: -0.418 [0.012]***.
    - Number of observations: 37,064.
    - R2: 0.88.
  - Column (3) IV (Remittances as a share of migrant’s income (log), instrumented):
    - Transaction costs (log): -0.443 [0.064]***.
    - Number of observations: 37,061.
    - R2: 0.77.
- Interpretation of IV and robustness results:
  - IV estimate in column (1) is smaller in magnitude than OLS but still negative and statistically significant at conventional levels.
  - Using remittances as share of income and controlling for frequency yields a strong negative association (column (2)).
  - Instrumenting in the share specification yields a negative and statistically significant elasticity (column (3]), though magnitude is lower than some fixed-effect estimates.

### Summary of substantive findings
- Transaction costs have a robust, negative, and economically significant effect on remittance volumes.
- ICT development—particularly access and capability—and ICT affordability materially shape the sensitivity of remittances to transaction costs.
- Financial literacy proxies (education level, access to bank accounts for sender and receiver) reduce the cost-elasticity of remittances.
- Micro evidence from the US-Mexico corridor corroborates macro-level findings and is robust to IV strategies and alternative dependent variables.

*Source: wpiea2022218-print-pdf - Annex Figure 4).*

### CONCLUSION

### CONCLUSION

### Key empirical findings
- The paper uses a novel quarterly data set for 71 countries over a 10-year period, from 2011Q1 to 2020Q4, to investigate the elasticity of remittances to transaction costs.
- A 10-percentage point decrease in transaction costs leads to a 0.9 percent increase in remittance in the short-run, but has no discernible impact in subsequent quarters, suggesting a short-run effect.
- Moving from the 2020 level of transaction costs (6.3 percent) to the SDG target of 3 percent will generate an additional US$32.2 billion in remittances, much larger than the direct cost savings.
- When competition in the remittance market is high, the financial system is more developed, and transactions with correspondent banks hold up, the elasticity of remittance to transaction costs is much lower, ceteris paribus.
- Greater transparency on remittance transaction costs, improved financial literacy, and higher ICT development inhibit the sensitivity of remittances to high transaction costs, all else equal.

### Microdata confirmation (USA–Mexico corridor)
- Micro data with over 37,000 individuals surveyed during 2013-17 confirm that migrants who face higher transaction costs tend to remit less, after controlling for socio-economic characteristics.
- Education levels and access to a bank account diminish some of the cost-elasticity of remittances, corroborating panel results.

### Policy implications and recommendations
- Governments can influence remittance decisions by promoting competition among banks and money transfer operators through adapted regulations.
- Forcing transfer companies to list transparently all their prices and providing information to migrants and their families could help them choose more cost-effective remittance services and ultimately drive costs down.
- Improving educational outcomes to facilitate the acquisition of financial literacy should be enhanced to reduce remittance sensitivity to transaction costs.
- Policies can be grouped into:
  - Cost-mitigation policies: directly tackle root causes of high remittance costs (e.g., competition, correspondent banking relationships, financial system development).
  - Cost-adaptation policies: reduce information asymmetries in the remittance market (e.g., price transparency, financial literacy, ICT access).

### Complementarities and broader development implications
- The factors identified individually modify the elasticity of remittances to costs and also exhibit interdependence that typically reinforces each other.
- Moving on several fronts simultaneously would have a multiplicative effect (for example, deepening the financial system and improving financial literacy simultaneously will support each other further and help remittance flows).
- Reducing the cost of remitting to the SDG target could significantly boost remittances, creating an important tool to enhance capital flows to developing countries to finance development without creating excessive government debt.

*IMF Working Paper — How Do Transaction Costs Influence Remittances?*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022218-print-pdf.pdf_