## _wp05234

## Source details

**Canonical URL:** [_wp05234](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05234.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05234.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05234.pdf.json)

---

### Introduction and motivation
- Remittances are "currently the second largest source of foreign exchange, both in absolute terms and as a percentage of GDP."
- For some countries, remittances "represent more than 10 percent of GDP." Examples: small Caribbean and Pacific Islands, Albania, El Salvador, the Philippines.
- Paper’s core questions:
  - What is the macroeconomic effect of remittances?
  - How does financial development influence the growth effect of remittances?
  - Are remittances important for productive investment?
- Main contributions:
  - Newly constructed remittances measure covering "about 100 developing countries" for descriptive work and "about 70" for econometric analysis.
  - Analysis of interaction between remittances and financial development (substitution vs complementarity) and investigation of investment channels.

### Data and measurement
- Sample and period:
  - Sample: 73 developing countries (descriptive/econometrics vary between "about 70" and 73).
  - Period: 1975-2002.
  - Panel: six nonoverlapping five-year periods (last period averaged over three years).
  - Number of five-year-average observations reported: 306.
- Remittances definition:
  - Rem/GDP = sum of "Workers’ Remittances," "Compensation of Employees," and "Migrant Transfers" from IMF BOPSY, with country-specific exclusions/adjustments (exclusions listed for 20 named countries; additional desk-provided adjustments for 14 named country cases).
- Financial development indicators (banking-sector focused; sources IFS and WDI):
  - M2/GDP: liquid liabilities of the financial system divided by GDP.
  - Dep/GDP: sum of demand, time, saving, and foreign currency deposits divided by GDP.
  - Loan/GDP: claims on the private sector divided by GDP.
  - Credit/GDP: domestic credit provided by banking sector divided by GDP.
- Control variables:
  - Dependent variable: growth of real per capita GDP (WDI).
  - Controls: inflation (annual % change CPI), openness (exports + imports as % of GDP), human capital (average years of secondary schooling, Barro and Lee), central government fiscal balance/GDP, investment ratio (gross fixed capital formation/GDP), population growth (log difference).
  - Investment regressions use lending interest rate or interest rate spread as user cost proxies.

### Summary statistics (five-year averages, 1975-2002; Table 2)
- GDP growth: Mean 1.2; Median 1.3; Standard Deviation 3.4; Minimum -14.2; Maximum 11.0; Observations 306.
- Rem/GDP: Mean 2.9; Median 1.5; Standard Deviation 4.0; Minimum 0; Maximum 22.6; Observations 306.
- Dep/GDP: Mean 32.2; Median 27.7; Standard Deviation 20.6; Minimum 5.8; Maximum 142.5; Observations 306.
- M2/GDP: Mean 37.8; Median 30.9; Standard Deviation 25.1; Minimum 8.1; Maximum 164.5; Observations 306.
- Loan/GDP: Mean 27.7; Median 22.6; Standard Deviation 20.4; Minimum 2.4; Maximum 133.3; Observations 305.
- Credit/GDP: Mean 47.1; Median 31.8; Standard Deviation 31.8; Minimum 0.9; Maximum 193.8; Observations 306.
- Country extremes examples: Remittances <1% (Chile) to 23% (Jordan in 2002); deposits 6% (Niger) to 142% (China); liquidity 8% (Brazil) to 164% (Malta); claims to private sector 2% (Sudan) to 133% (China); bank credit 1% (Botswana) to 194% (Nicaragua).
- Outliers: some observations excluded (e.g., Lesotho remittances > 50% of GDP dropped).

### Empirical methodology
- Baseline growth equation (eq. (1)): growth regressed on initial log GDP per capita, Rem/GDP, controls X, time effects μt, country fixed effects ηi.
- Interaction specification (eq. (2)): includes Rem, FinDev, and (Rem × FinDev) to test substitutability/complementarity.
  - Interpretation: negative coefficient on interaction → remittances substitute for financial development; positive → remittances complement financial development.
- Estimators: OLS, fixed effects (FE), and system GMM (SGMM) using two lags as instruments; five-year averages to smooth cycles.
- Endogeneity addressed with SGMM (Arellano and Bover (1995) / Blundell and Bond (1997)), testing AR(1), AR(2), and Hansen overidentification.

### Baseline growth regression results (Table 4)
- Representative coefficients (point estimate (robust std. error)):
  - LogInGDP (convergence):
    - OLS: -0.698*** (0.244)
    - FE: -5.896*** (0.992)
    - SGMM: -1.059 (1.038)
  - LogInvGDP:
    - OLS: 4.698*** (0.571)
    - FE: 5.232*** (0.795)
    - SGMM: 5.039*** (1.138)
  - GovFiscalBal:
    - OLS: 0.119** (0.050)
    - FE: 0.156** (0.064)
    - SGMM: 0.209 (0.180)
  - Inflation:
    - OLS: -0.022** (0.010)
    - FE: -0.007 (0.007)
    - SGMM: -0.035** (0.015)
  - Rem/GDP:
    - OLS: 0.043 (0.051)
    - FE: 0.022 (0.087)
    - SGMM: 0.010 (0.096)
- Observations: OLS/FE/SGMM reported 315; Number of countries: 73.
- Diagnostics: AR(1) test 0.00; AR(2) test 0.52; Hansen test p-value 0.55.
- Interpretation: Rem/GDP alone is small and statistically insignificant; standard controls behave as expected.

### Interaction results: remittances × financial depth (Tables 5–7; OLS and SGMM)
- Main empirical pattern (consistent across four FD measures and across OLS and SGMM):
  - Strong evidence of a positive and significant coefficient on Rem/GDP (when interaction included).
  - Robust evidence of a negative and significant interaction between Rem/GDP and financial depth (Dep/GDP, Loan/GDP, Credit/GDP, M2/GDP).
  - Interpretation: marginal impact of remittances on growth declines with higher financial development — remittances substitute for financial intermediation in shallower systems.
  - At very high financial development (above the 75th percentile), marginal effect of remittances can be zero or negative.

### Key estimated magnitudes (SGMM and OLS selected results)
- SGMM RemGDP coefficients: 0.406**, 0.397**, 0.251*, 0.389** (columns corresponding to DEP/GDP, LOAN/GDP, CREDIT/GDP, M2/GDP).
- SGMM interaction coefficients (RemGDP × FD):
  - RemGDP*DepGDP: -0.008*** (0.003)
  - RemGDP*LoanGDP: -0.009*** (0.003)
  - RemGDP*CreditGDP: -0.005*** (0.002)
  - RemGDP*M2GDP: -0.006*** (0.002)
- OLS RemGDP coefficients: 0.253**, 0.228**, 0.213*, 0.197*.
- OLS interaction terms:
  - RemGDP*DepGDP: -0.004** (0.002)
  - RemGDP*LoanGDP: -0.005** (0.002)
  - RemGDP*CreditGDP: -0.003* (0.002)
  - RemGDP*M2GDP: -0.003** (0.001)
- SGMM diagnostics (by FD column): Number of countries 72, 71, 73, 73; AR(1) test 0.00; AR(2) test 1.00/0.78/0.91/0.75; Hansen p-values 0.86/0.77/0.77/0.57; R-squared 0.23/0.10/0.24/0.21.

### Marginal effects and thresholds (Table 7 and threshold estimation)
- Marginal effect of remittances at the median level of financial depth (with investment included):
  - DEP/GDP 0.18; LOAN/GDP 0.19; CREDIT/GDP 0.09; M2/GDP 0.20.
- Marginal effect at the mean level (with investment):
  - DEP/GDP 0.15; LOAN/GDP 0.15; CREDIT/GDP 0.02; M2/GDP 0.16.
- Marginal effect increases when investment is omitted (interpretation: part of remittances’ effect operates through investment). With investment omitted at median:
  - DEP/GDP 0.27; LOAN/GDP 0.23; CREDIT/GDP 0.21; M2/GDP 0.29.
- Financial depth level at which marginal effect of remittances is zero (percent; Table 7):
  - DEP/GDP: with investment 50.8; without investment 57.2.
  - LOAN/GDP: with investment 44.1; without investment 45.1.
  - CREDIT/GDP: with investment 50.2; without investment 74.4.
  - M2/GDP: with investment 64.8; without investment 72.1.
- Median marginal impacts reported elsewhere in text:
  - 0.19 at the median loans to GDP ratio.
  - 0.09 at the median banking sector credit ratio.
  - 0.20 at the median M2 to GDP level.
- Note: "These statistics are based on SGMM estimates and are statistically significant at 5 percent significance level."

### Investment channel: remittances → investment → growth
- Evidence that remittances raise investment:
  - Dropping investment from growth regressions increases the marginal impact of remittances (indirect evidence of investment channel).
  - Investment regressions (SGMM, dependent variable INV/GDP) — selected coefficients:
    - Lagged InvGDP: 0.874*** (0.110); 0.837*** (0.095); 0.865*** (0.104); 0.854*** (0.110).
    - Real GDP Growth: 0.534** (0.214); 0.518*** (0.181); 0.555*** (0.189); 0.528** (0.208).
    - Lending rate: -0.014 (0.019); -0.015 (0.020); -0.021 (0.019); -0.005 (0.014).
    - RemGDP: 0.398* (0.231); 0.710** (0.341); 0.507** (0.242); 0.690** (0.295).
    - RemGDP × FD interactions negative and significant (e.g., RemGDP*DepGDP: -0.006** (0.003); RemGDP*LoanGDP: -0.012** (0.006)).
  - SGMM diagnostics for investment regressions: AR(1) test 0.000; AR(2) tests 0.83/0.78/0.89/0.87; Hansen p-values 0.81/0.76/0.80/0.81; R-squared 0.58/0.54/0.56/0.58.
- Quantitative investment magnitudes:
  - Marginal impact of remittances on investment is positive across most financial development levels.
  - Median marginal impact on investment reported as 0.4 at the median level of financial development; at the lowest quartile it can surpass 0.5.
  - Investment regressions do not show an independent, statistically positive effect of financial development.

### Robustness: median split and endogenous threshold tests
- Median split SGMM growth estimates:
  - Impact of remittances positive for countries with low financial development (below median).
  - Impact nil or negative for countries above median.
  - Standard t-test rejects equality across subsamples in only one reported case (T-stat entries: 1.3, 1.3, 1, 2.9).
- Endogenous threshold estimation (Hansen 1996/2000; Table 10):
  - Estimated thresholds (percent of GDP) and 95 percent confidence intervals:
    - Dep/GDP threshold: 22.6; 95% CI [11, 73].
    - Loan/GDP threshold (claims on private sector): 20.8; 95% CI [16, 22].
    - Credit/GDP threshold: 30; 95% CI [29, 33].
    - M2/GDP threshold: 20.8; 95% CI [16, 22].
  - Bootstrap evidence (F-test / Bootstrap P-value):
    - DEP/GDP 32.16 0.017; LOAN/GDP 38.94 0.004; CREDIT/GDP 56.25 0.000; M2/GDP 42.14 0.000.
  - Subsample OLS patterns: remittances coefficients often larger and statistically significant in low-FD regimes (examples provided).

### Cyclical behavior of remittances
- Conceptual test: compensatory vs profit-driven remittances based on cyclical correlation with home-country GDP.
  - Compensation/altruism → countercyclical (negative correlation).
  - Profit/investment → procyclical (positive correlation).
- Empirical findings:
  - Using Hodrick-Prescott decomposition for about 100 developing countries, remittances are procyclical for two thirds of countries.
  - Aggregate average correlation (equally weighted) ≈ 0.1.
  - Countries where remittances are more procyclical tend to have less developed financial systems.
- Quantitative association (Table 11):
  - DEP/GDP: Correlations -0.18*; Reg Estimates -0.29*.
  - LOAN/GDP: Correlations -0.16*; Reg Estimates -0.35*.
  - CREDIT/GDP: Correlations -0.13; Reg Estimates -0.18.
  - M2/GDP: Correlations -0.17*; Reg Estimates -0.23*.
  - Note: star denotes significance at the 10 percent level or better.

### Main conclusions and interpretation
- Core finding: Remittances promote growth in less financially developed countries; evidence robust to SGMM endogeneity controls, alternative FD measures, and robustness tests.
- Mechanism: Remittances alleviate liquidity constraints, substitute for scarce credit/insurance, finance investment, and improve allocation of capital in shallow financial systems.
- At high financial development levels, remittances’ marginal growth effect is nil or possibly negative (moral hazard or labor-supply effects).
- Cyclicality: Remittances tend to be more procyclical in shallower financial systems, consistent with an investment/profit-driven channel; remittances are more countercyclical in countries with deeper financial systems.
- Policy implications:
  - In economies with shallow financial systems, facilitating remittance flows and policies to channel remittances to productive investment can help alleviate credit constraints and raise growth.
  - Strengthening financial sector depth may reduce remittances’ direct growth role and make remittances more countercyclical; policymakers should consider the interaction between remittances and financial development when designing policies to channel remittances toward productive investment and when assessing labor-supply and social effects.

*Source: _wp05234 (extracted sections II–VI, references, tables and appendices as provided in the source content).*

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

### _wp05234 - References

### Introduction and motivation
- Remittances have become increasingly important in international capital flows over the past decade.
- In the aggregate, remittances are "currently the second largest source of foreign exchange, both in absolute terms and as a percentage of GDP" (Figures 1 and 2).
- For some countries, remittances "represent more than 10 percent of GDP." Examples cited include small Caribbean and Pacific Islands, and labor-exporting countries such as Albania, El Salvador, and the Philippines (Figure 3).
- The paper addresses a gap: the lack of comprehensive cross-country macroeconomic evidence on the impact of remittances on growth.

### Research questions and contribution
- The paper asks three specific questions:
  - What is the macroeconomic effect of remittances?
  - How does financial development influence the growth effect of remittances?
  - Are remittances important for productive investment?
- Two main contributions:
  - Substantially expand data on remittance flows using a "newly constructed measure for remittances," covering "about 100 developing countries for the descriptive part and about 70 for the econometric analysis."
  - Analyze the interaction between remittances and the financial sector, specifically whether remittances substitute for or complement financial development in promoting growth, and whether remittances promote productive investment.

### Literature context
- Prior cross-country evidence is limited. Chami, Fullenkamp, and Jahjah (2003) is noted as the "only cross-country study on remittances and growth" and finds a negative impact; the present paper investigates investment channels that that study disregarded.
- Two strands of literature the paper contributes to:
  - Development impact of remittances (macro and micro evidence).
  - Link between remittances and financial development, including transaction costs, competition among money transfer firms, innovative financial products, and microfinance institutions.
- The paper distinguishes itself by analyzing "how a country’s capacity to use remittances and its effectiveness in doing so might be influenced by local financial sector conditions," and by examining complementarity/substitutability between remittances and financial development for growth.

### Theoretical mechanisms and hypotheses
- Two ambiguous theoretical forces:
  - Financial development may enhance the growth benefits of remittances by lowering transaction costs and directing remittances to high-return projects.
  - Remittances may substitute for poor financial systems by loosening liquidity constraints and providing start-up capital where credit markets are imperfect.
- Hypothesis: "Voluminous migrant remittances can substitute for a lack of financial development and hence promote economic growth via investment."
- Rationale: Entrepreneurs in developing countries face inefficient credit markets, lack of collateral, and high lending costs; remittances can alleviate these constraints and improve allocation of capital.

### Micro-level evidence cited
- Evidence that remittances support enterprise formation and start-up capital:
  - Dustmann and Kirchkamp (2001): "50 percent of a sample of Turkish emigrants returning from Germany started a micro enterprise within four yeas of resettling in Turkey using money saved while working abroad."
  - Massey and Parrado (1998): earnings from U.S. work provided start-up capital in "21 percent of the new business formations" in their sample.
  - Woodruff and Zenteno (2001): remittances are responsible for "almost 20 percent of the capital invested in micro enterprises throughout urban Mexico."

### Empirical approach and data
- Use of standard financial market indicators in growth regressions to study the interaction between remittances and financial development.
- Employs methods that "deal with endogeneity."
- Sample scope: descriptive analysis covers "about 100 developing countries"; econometric analysis covers "about 70" countries.
- Paper structure: Section II describes the data; Section III presents empirics of remittances and growth; Section IV contains robustness tests.

### Key empirical finding (as stated)
- "Remittances may play a significant role in promoting growth in countries with shallower financial systems."  
- This result "holds true after addressing concerns regarding endogeneity."

### Policy-relevant implications (inferred from paper's analysis and findings)
- In contexts of limited financial development, facilitating remittance flows and policies that encourage use of remittances for productive investment could help alleviate credit constraints and raise growth.
- Strengthening financial sector infrastructure may alter how remittances are channeled—either enhancing their productivity or reducing the need for remittances as a substitute for formal finance.

*Source: _wp05234 - References (excerpted content provided from the PDF)_*

### Section V analyzes the cyclical behavior of remittances, and Section VI concludes.

### _wp05234 — Sections II–III summary (data, methodology, and empirical results)

### Data: remittances, financial development, and control variables
- Sample and period
  - Sample: 73 developing countries.
  - Period: 1975-2002.
  - Panel construction: six nonoverlapping five-year periods (last period averaged over three years).
  - Number of five-year-average observations reported in tables: 306.

- Remittances definition and data construction
  - Remittances (REM/GDP) = sum of three items in the IMF’s Balance of Payment Statistics Yearbook: workers’ remittances, compensation of employees, and migrant transfers (details in Appendix 1).
  - Country-by-country inspection of BOPSY country notes; exclusion of compensation of employees in about 20 countries where it did not qualify as remittances.
  - Additional data and clarifications obtained from IMF desk economists and country authorities for more than 29 countries (list in Appendix I).
  - All regressions employ the ratio of remittances to GDP (Rem/GDP).

- Financial development indicators (banking-sector focused — data sources: IFS and WDI)
  - M2/GDP: liquid liabilities of the financial system (currency + demand + interest-bearing liabilities of banks and nonfinancial intermediaries) divided by GDP.
  - Dep/GDP: sum of demand, time, saving, and foreign currency deposits divided by GDP.
  - Loan/GDP: claims on the private sector divided by GDP.
  - Credit/GDP: domestic credit provided by banking sector divided by GDP.
  - Broader coverage than previous datasets (e.g., claims on private sector available for 73 countries here vs. 44 in other datasets).

- Other variables and measurement
  - Dependent variable for growth regressions: growth of real per capita GDP in constant dollars (WDI).
  - Controls: inflation (annual percentage change in CPI), openness (exports + imports as share of GDP), human capital (average years of secondary schooling, Barro and Lee), central government fiscal balance/GDP, investment ratio (gross fixed capital formation/GDP), population growth (log difference).
  - Investment regressions use user cost of capital proxied by either lending interest rate or interest rate spread (lending minus deposit rate), from WDI.

- Selected summary statistics (five-year averages, period 1975-2002; from Table 2)
  - GDP growth: Mean 1.2; Median 1.3; Standard Deviation 3.4; Minimum -14.2; Maximum 11.0; Observations 306.
  - Rem/GDP: Mean 2.9; Median 1.5; Standard Deviation 4.0; Minimum 0; Maximum 22.6; Observations 306.
  - Dep/GDP: Mean 32.2; Median 27.7; Standard Deviation 20.6; Minimum 5.8; Maximum 142.5; Observations 306.
  - M2/GDP: Mean 37.8; Median 30.9; Standard Deviation 25.1; Minimum 8.1; Maximum 164.5; Observations 306.
  - Loan/GDP: Mean 27.7; Median 22.6; Standard Deviation 20.4; Minimum 2.4; Maximum 133.3; Observations 305.
  - Credit/GDP: Mean 47.1; Median 31.8; Standard Deviation 31.8; Minimum 0.9; Maximum 193.8; Observations 306.
  - Example country extremes: remittances less than 1 percent of GDP (Chile) to 23 percent (Jordan in 2002); deposits as low as 6 percent (Niger) and high as 142 percent (China); liquidity low 8 percent (Brazil) high 164 percent (Malta); claims to private sector low 2 percent (Sudan) high 133 percent (China); bank credit low 1 percent (Botswana) high 194 percent (Nicaragua).
  - Notes on outliers: some observations excluded (e.g., Lesotho observations where remittances > 50 percent of GDP were dropped).

- Bivariate correlations (Table 3, stars denote significance at 10 percent or better)
  - Growth positively correlated with investment, government fiscal balance, years of education, and openness; negatively correlated with inflation.
  - Rem/GDP positively correlated with growth, investment, and openness; negatively correlated with inflation.
  - Correlations between remittances and financial development indicators are positive across the four indicators (very small for loans and bank credit).

### Empirical methodology
- Regression frameworks
  - Baseline growth equation (equation (1)):
    - Growth (or change in log GDP per capita) regressed on initial log GDP per capita, Rem/GDP, matrix of controls X, time effects μt, country fixed effects ηi, and error εit.
  - Interaction specification (equation (2)):
    - Allows remittance impact to vary with financial development: includes Rem, FinDev, and (Rem × FinDev) interaction as regressors (both Rem and FinDev also included separately).
    - Interpretation: negative coefficient on interaction → remittances more effective in countries with shallower financial systems (substitutability); positive coefficient → remittances complement deep financial systems.
  - Econometric approaches: OLS, fixed effects (FE), and system GMM (SGMM) following Arellano and Bover (1995) / Blundell and Bond (1997).
    - Panel constructed from five-year averages to smooth business cycle fluctuations.
    - SGMM used to address endogeneity: two lags of endogenous variables used as instruments; autocorrelation and Hansen tests reported.

- Treatment of endogeneity
  - Potential reverse causality between growth, remittances, and financial development acknowledged.
  - SGMM (system GMM) employed to mitigate endogeneity, using internal instruments (lagged levels and differences) under assumptions of no serial correlation in ε and stationarity conditions described in the text.

### Estimation results and key findings
- Baseline (Table 4 — linear growth effects of remittances; dependent variable GDP per capita growth)
  - Selected coefficient estimates (point estimate followed by robust standard error in parentheses):
    - LogInGDP (convergence term)
      - OLS: -0.698*** (0.244)
      - FE: -5.896*** (0.992)
      - SGMM: -1.059 (1.038)
    - LogInvGDP (investment)
      - OLS: 4.698*** (0.571)
      - FE: 5.232*** (0.795)
      - SGMM: 5.039*** (1.138)
    - GovFiscalBal
      - OLS: 0.119** (0.050)
      - FE: 0.156** (0.064)
      - SGMM: 0.209 (0.180)
    - Inflation
      - OLS: -0.022** (0.010)
      - FE: -0.007 (0.007)
      - SGMM: -0.035** (0.015)
    - Rem/GDP
      - OLS: 0.043 (0.051)
      - FE: 0.022 (0.087)
      - SGMM: 0.010 (0.096)
  - Sample / diagnostics:
    - Observations: OLS and FE reported 315; SGMM 315.
    - Number of countries: 73.
    - AR(1) test: 0.00
    - AR(2) test: 0.52
    - Hansen test p-value: 0.55
  - Summary interpretation:
    - When Rem/GDP is simply added to a standard growth regression, its coefficient is small and statistically insignificant across OLS, FE, and SGMM specifications.
    - Coefficient estimates on other controls broadly align with expectations (positive effect of investment, negative effect of inflation).

- Interaction results (Tables 5–7; discussion summarized in text)
  - Main pattern (consistent across four measures of financial depth and across OLS and SGMM):
    - Strong evidence of a positive and significant coefficient on remittance flows.
    - Robust evidence of a negative and significant interaction between remittances and financial depth.
  - Interpretation:
    - The marginal impact of remittances on growth decreases with higher levels of financial development.
    - Remittances act as a substitute for financial intermediation in countries with shallower financial systems by relaxing liquidity constraints and providing credit and insurance where markets are incomplete or inefficient.
    - In countries with well-developed financial systems, remittances-driven growth is less important, and the marginal effect of remittances on growth can become zero or negative at high levels of financial development (above the 75th percentile).
  - Magnitude example reported:
    - An increase by one percentage point in the deposits to GDP ratio from the median level of 29 percent would enhance growth by 0.18 percentage points (marginal effect of remittances at the median level of financial development reported in Table 7).

### Synthesis of empirical implications
- Heterogeneous effects: The growth effect of remittances is not uniform across countries; it is conditioned by the recipient country’s financial development.
- Role of remittances in financially shallow economies:
  - Remittances can substitute for missing or inefficient financial services — providing liquidity, credit, and insurance that facilitate investment and growth.
- Policy-relevant inference:
  - In economies with underdeveloped financial systems, remittances may play an important growth-enhancing role by easing liquidity constraints.
  - In economies with deeper financial systems, the marginal growth contribution of additional remittance flows diminishes and may be nonpositive at high financial depth.

*Source: _wp05234 — Sections II–III (extracted from the provided PDF content)._*

### 0.19 at the median loans to GDP ratio, 0.09 at the median banking sector credit ratio, and

### _wp05234 - 0.19 at the median loans to GDP ratio, 0.09 at the median banking sector credit ratio, and

### Key empirical magnitudes and interpretation
- Median marginal impacts reported in the text:
  - 0.19 at the median loans to GDP ratio.
  - 0.09 at the median banking sector credit ratio.
  - 0.20 at the median M2 to GDP level.
- The text notes: "These effects can be twice as large in the presence of stringent lending and borrowing restrictions."
- With limited capital market imperfections, remittances may be less essential for financing investment and instead be devoted to nongrowth-generating activities (conspicuous consumption) or reduce labor supply; this helps explain why the impact of remittances declines with financial depth. A potential marginally negative impact at very high levels of financial development is discussed on moral hazard grounds (Chami, Fullenkamp, and Jahjah (2003)).

### SGMM estimates (selected coefficients and diagnostics)
- Remittances and interaction terms (SGMM estimates, dependent variable = GDP per capita growth):
  - RemGDP coefficients: 0.406**, 0.397**, 0.251*, 0.389** (corresponding to DEP/GDP, LOAN/GDP, CREDIT/GDP, M2/GDP columns).
  - RemGDP*DepGDP: -0.008*** (standard error (0.003)).
  - LoanGDP: 0.084*** (standard error (0.026)).
  - RemGDP*LoanGDP: -0.009*** (standard error (0.003)).
  - CreditGDP: 0.034*** (standard error (0.012)).
  - RemGDP*CreditGDP: -0.005*** (standard error (0.002)).
  - M2GDP: 0.047*** (standard error (0.015)).
  - RemGDP*M2GDP: -0.006*** (standard error (0.002)).
- Other notable SGMM coefficient signs and significance:
  - LogInvGDP: 3.200***, 2.626**, 4.041***, 3.629***.
  - Inflation: -0.029**, -0.024**, -0.034**, -0.027**.
  - LogInGDP: -1.394*, -2.462***, -1.482*, -1.974**.
- SGMM diagnostics:
  - Number of countries: 72, 71, 73, 73 (by column).
  - AR(1) test: 0.00, 0.00, 0.00, 0.00.
  - AR(2) test: 1.00, 0.78, 0.91, 0.75.
  - P-value Hansen test: 0.86, 0.77, 0.77, 0.57.
  - R-squared: 0.23, 0.10, 0.24, 0.21.

### OLS estimates (selected coefficients)
- RemGDP coefficients (OLS, dependent variable = GDP per capita growth): 0.253**, 0.228**, 0.213*, 0.197*.
- Interaction terms (OLS):
  - RemGDP*DepGDP: -0.004** (standard error (0.002)).
  - RemGDP*LoanGDP: -0.005** (standard error (0.002)).
  - RemGDP*CreditGDP: -0.003* (standard error (0.002)).
  - RemGDP*M2GDP: -0.003** (standard error (0.001)).
- Other OLS coefficients consistent with standard growth regressions (examples):
  - LogInvGDP: 4.255***, 4.312***, 4.580***, 4.091***.
  - Inflation: -0.019**, -0.019**, -0.023**, -0.018**.
  - LogInGDP: -0.654**, -0.661***, -0.661***, -0.734***.
- OLS R-squared reported: 0.36, 0.35, 0.34, 0.35 (by column).
- Note: Robust standard errors in parentheses; significance: * 10 percent, ** 5 percent, *** 1 percent. All regressions include time dummies.

### Marginal effect of remittances on growth by financial depth (Table 7, based on SGMM)
- Financial depth level at which marginal effect of remittances is zero (percent):
  - DEP/GDP: with investment 50.8; without investment 57.2.
  - LOAN/GDP: with investment 44.1; without investment 45.1.
  - CREDIT/GDP: with investment 50.2; without investment 74.4.
  - M2/GDP: with investment 64.8; without investment 72.1.
- Marginal effect of remittances at the median level of financial depth:
  - With investment: DEP/GDP 0.18; LOAN/GDP 0.19; CREDIT/GDP 0.09; M2/GDP 0.20.
  - Without investment: DEP/GDP 0.27; LOAN/GDP 0.23; CREDIT/GDP 0.21; M2/GDP 0.29.
- Marginal effect of remittances at the mean level of financial depth:
  - With investment: DEP/GDP 0.15; LOAN/GDP 0.15; CREDIT/GDP 0.02; M2/GDP 0.16.
  - Without investment: DEP/GDP 0.23; LOAN/GDP 0.17; CREDIT/GDP 0.14; M2/GDP 0.24.
- Note: "These statistics are based on SGMM estimates and are statistically significant at 5 percent significance level."

### Channels: remittances → investment → growth
- Evidence on channels:
  - Dropping investment from growth regressions increases the marginal impact of remittances, providing indirect evidence that a channel operates through productive investment.
  - Table 7: marginal impact of remittances at median and mean financial development increases by about 50 percent in the case of deposits and M2 to GDP; increase is between two and six times larger for total credit from the banking sector.
  - The authors interpret these results as suggesting an important channel through which remittances influence growth is the volume of investments; other channels may include efficiency of investments, investment in human capital, and multiplicative effects from higher savings and internal demand.
- Direct investment equation estimated (equation (4) specification summary):
  - Dependent variable: INV/GDP (total investment to GDP).
  - Controls (Z) include per capita real GDP growth and lending interest rate (user cost of capital proxy); other variants tested (inflation, openness) do not alter main results.
- Investment equation results (text summary):
  - Lagged investment coefficient: large and positive.
  - Output growth elasticity of investment: positive and significant.
  - Lending interest rate: negative sign but not statistically significant (similar outcome when using interest rate spread).
  - Remittances: positive and significant across all specifications.
  - Interaction remittances × financial depth: negative and significant.
  - Interpretation: marginal impact of remittances on investment is positive across largely all levels of financial development; largest remittances-driven increases in investment have taken place in less financially developed countries. The marginal impact of remittances on investment ranges between 0.2 and

*Source: _wp05234 - 0.19 at the median loans to GDP ratio, 0.09 at the median banking sector credit ratio, and*

### 0.4 at the median level of financial development, the impact can surpass 0.5 at the lowest

### _wp05234 - 0.4 at the median level of financial development, the impact can surpass 0.5 at the lowest

### Investment regressions and main empirical findings
- Investment regressions do not show an independent, statistically positive effect of financial development.
- The marginal effect of remittances:
  - Only becomes zero at very high levels of financial depth, beyond the 90-95 percentile of the distribution.
  - At the median level of financial development is reported as 0.4; at the lowest quartile of the distribution the impact can surpass 0.5.
- SGMM estimates (Table 8) — dependent variable is investment to GDP — selected coefficients (robust standard errors in parentheses; significance: * 10 percent; ** 5 percent; *** 1 percent):
  - Lagged InvGDP: 0.874*** (0.110); 0.837*** (0.095); 0.865*** (0.104); 0.854*** (0.110)
  - Real GDP Growth: 0.534** (0.214); 0.518*** (0.181); 0.555*** (0.189); 0.528** (0.208)
  - Lending rate: -0.014 (0.019); -0.015 (0.020); -0.021 (0.019); -0.005 (0.014)
  - RemGDP: 0.398* (0.231); 0.710** (0.341); 0.507** (0.242); 0.690** (0.295)
  - Interaction terms (RemGDP * financial depth measures):
    - RemGDP*DepGDP: -0.006** (0.003)
    - RemGDP*LoanGDP: -0.012** (0.006)
    - RemGDP*CreditGDP: -0.005* (0.003)
    - RemGDP*M2GDP: -0.008** (0.003)
  - Sample and diagnostics (four specifications): Observations 3433 4433 3433 350; Number of countries 109110112112; AR(1) test 0.000.000.000.00; AR(2) test 0.830.780.890.87; P-value Hansen test 0.810.760.800.81; R-squared 0.580.540.560.58
- Diagnostic tests: Hansen and first- and second-order autocorrelation reveal no evidence against the validity of instruments used by the SGMM estimator.
- Interpretation:
  - Remittances have a significant positive impact on growth and investment once interaction with financial development is accounted for.
  - The interaction term is statistically negative, indicating remittances and financial development act as substitutes in promoting economic growth.
  - In economies with limited access to credit and insurance, remittances alleviate liquidity constraints, finance investment, and support consumption smoothing/insurance.
  - Where the financial system is sufficiently developed, remittances have a much lower impact on growth, and can be nil or possibly negative.

### Robustness: median split and threshold estimation
- Exogenous split by median of financial development:
  - SGMM growth estimates (Table 9) indicate:
    - Impact of remittances is positive for countries with low financial development (below median).
    - Impact is nil or negative for countries with deeper financial systems (above median).
    - Using a standard t-test, the null that marginal impact of remittances is equal across subsamples is rejected in only one case reported (T-stat Ho: Above Med = Below Med entries include 1.3, 1.3, 1, 2.9).
- Endogenous threshold estimation (Hansen 1996/2000 approach; Table 10):
  - Estimated thresholds (percent of GDP) and confidence intervals (where reported in text):
    - Deposits (Dep/GDP) threshold: 22.6 percent of GDP, with a 95 percent confidence interval [11, 73].
    - Claims to the private sector threshold: 20.8 percent, with a 95 percent confidence interval [16, 22].
    - Banking credit (Credit/GDP) threshold: 30 percent, with a 95 percent confidence interval [29, 33].
    - M2/GDP threshold: 20.8 percent, with a 95 percent confidence interval [16, 22].
  - Bootstrap evidence:
    - F-test for no threshold / Bootstrap P-value (Table 10): DEP/GDP 32.16 0.017; LOAN/GDP 38.94 0.004; CREDIT/GDP 56.25 0.000; M2/GDP 42.14 0.000
  - Subsample OLS coefficient patterns (selected illustrative values from Table 10; robust standard errors in parentheses; significance as in the table):
    - LogInvGDP in low financial development regimes is positive and often highly significant (e.g., 3.264** (1.244) for Dep/GDP ≤22.6; 3.648*** (0.696) for Credit/GDP ≤30).
    - RemGDP coefficients by regime:
      - Dep/GDP >22.6: 0.027 (0.053)
      - Dep/GDP ≤22.6: 0.212 (0.161)
      - LOAN/GDP >20.8: 0.052 (0.058)
      - LOAN/GDP ≤20.8: 0.178 (0.119)
      - CREDIT/GDP >30: -0.004 (0.052)
      - CREDIT/GDP ≤30: 0.216** (0.104)
      - M2/GDP >20.8: 0.018 (0.052)
      - M2/GDP ≤20.8: 0.467** (0.195)
  - Findings from threshold estimation:
    - Evidence for a regime change at determined levels of financial development.
    - Marginal impact of remittances is not statistically different from zero in the high financial development regime.
    - Remittances have a larger positive impact, often statistically significant, in the low-financial development subsample.
    - Many control variables in the growth regressions display different behavior across subsamples.

### Cyclical behavior of remittances: compensatory transfers vs profit-driven flows
- Conceptual test:
  - If remittances are compensatory (altruistic/insurance), they should be negatively correlated with home country GDP (countercyclical).
  - If remittances are profit-driven (investment-seeking), they should be positively correlated with home country GDP (procyclical).
  - Methodology: Hodrick-Prescott filter to decompose series; classify remittances as countercyclical/procyclical/acyclical based on sign and significance of correlation between cyclical components (following Kaminsky, Reinhart, and Végh (2004)).
- Empirical evidence (Figure 4 and summary):
  - For about a hundred developing countries, remittances are procyclical—to different degrees—for two thirds of the countries (note 19).
  - Aggregate average correlation across all countries, weighing them equally, is about 0.1.
  - Interpretation: migrants tend to send remittances when the economic situation in the country of origin is favorable, possibly to take advantage of investment opportunities — implying an investment/profit-driven channel that helps explain positive link between remittances and growth (but raises endogeneity/reverse causality concerns).
- Association between financial development and remittance cyclicality:
  - Hypothesis: remittances should be more procyclical in countries with shallower financial systems if procyclical remittances reflect investment-seeking behavior and remittances are more effective where financial systems are shallow.
  - Evidence (summary of Table 11 results):
    - Correlations and bivariate regression coefficients between the cyclical indicator of remittances and measures of financial development are negative across all measures of financial development and range from −0.13 to −0.35.
    - All coefficients are significant at the 10 percent level, except those associated with the banking credit variable.
    - Interpretation: countries where remittances are more procyclical are associated with less developed financial systems, consistent with remittances being more effective in shallower financial systems.

### Policy-relevant implications and interpretation
- Remittances function as a partial substitute for financial development in alleviating liquidity constraints and financing investment in economies with underdeveloped financial systems.
- In countries with deeper financial systems, formal financial intermediation may better meet financing needs for investment; remittances in such contexts are less growth-enhancing and may be allocated to non-growth-promoting uses.
- The procyclical nature of remittances in many countries suggests a substantial profit-driven/investment channel; this is particularly relevant in shallower financial systems where remittances have larger growth effects.
- Robustness checks (median split and endogenous threshold estimation) consistently support the conclusion that remittances have a larger positive impact on growth in shallower financial systems and negligible impact in more financially developed countries.

*Source: _wp05234 - 0.4 at the median level of financial development, the impact can surpass 0.5 at the lowest*

### conclusion of the paper, that there is substitutability between remittances and the financial

### Conclusion — substitutability between remittances and the financial sector

### Main empirical findings
- Remittances have promoted growth in less financially developed countries.
- Evidence is robust to:
  - Controlling for endogeneity of remittances and financial development using a SGMM approach.
  - Using alternative measures of financial sector development.
  - A number of robustness tests.
- Interpretation: remittances help alleviate credit constraints on the poor, substituting for the lack of financial development, improving the allocation of capital, and accelerating economic growth. There is an investment channel through which remittances can promote growth where the financial sector does not meet the credit needs of the population.
- At high levels of financial development, remittances show nil or even negative impact on growth, interpreted as suggestive evidence that remittances are more likely to discourage labor supply in more financially developed countries.
- Cyclical properties:
  - Large variation in the cyclical behavior of remittance flows across countries.
  - Remittances tend to be more procyclical where the financial system is less developed.
  - Procyclical remittances are more likely motivated by investment opportunities rather than altruism, consistent with larger impacts of remittances in shallower financial systems.
- Paradoxical (a priori) result: remittances tend to be more countercyclical (more compensatory) in countries with deeper financial systems, where agents could, in principle, obtain insurance more easily; conversely, remittances are more likely to seek investment opportunities by responding to home-country economic conditions in countries with less developed financial sectors. Financial depth smooths, rather than amplifies, the cyclicality of remittances.

### Quantitative evidence on cyclicality and financial depth (Table 11)
- DEP/GDP
  - Correlations 1/: -0.18*
  - Reg Estimates 2/: -0.29*
- LOAN/GDP
  - Correlations 1/: -0.16*
  - Reg Estimates 2/: -0.35*
- CREDIT/GDP
  - Correlations 1/: -0.13
  - Reg Estimates 2/: -0.18
- M2/GDP
  - Correlations 1/: -0.17*
  - Reg Estimates 2/: -0.23*

Notes from table:
- 1/ Displays pair wise correlation coefficients between the cyclical indicator of remittances in Figure 4 and the median over the period 1975-2002 of each of the financial depth (FD) variables for 116 developing countries.
- 2/ Displays regression estimates from OLS regressions of the cyclical indicator of remittances and each of the FD variables. A star denotes significance at the 10 percent level or better.

### Data and variable construction
- Panel used: 70 developing countries, period 1975-2002.
- Total remittances = sum of three items in the IMF’s Balance of Payment Statistics Yearbook (BOPSY):
  - “Workers’ Remittances,”
  - “Compensation of Employees,” and
  - “Migrant Transfers.”
- Definitions:
  - Workers’ Remittances: current transfers made by migrants who are employed and resident in another economy (typically those who move and stay, or are expected to stay, a year or longer).
  - Compensation of Employees: wages, salaries, and other benefits earned by nonresident workers for work performed for residents of other countries (includes border and seasonal workers and some other categories).
  - Migrant Transfer: financial items that arise from the migration (change of residence) of individuals from one economy to another.
- Exclusions and recording specifics (per BOPSY):
  - Compensation of Employees is excluded from total remittances for: Argentina, Azerbaijan, Barbados, Belize, Benin, Brazil, Cambodia, Cape Verde, China, Côte d’Ivoire, Dominican Republic, Ecuador, El Salvador, Guyana, Panama, Rwanda, Senegal, Seychelles, Turkey, and Venezuela.
  - BOPSY specifies migrants transfers are recorded under “Other Current Transfers” for Kenya, Malaysia, and the Syrian Arab Republic.
- Additional country-specific adjustments based on IMF desk and national authority information:
  1. Bosnia and Herzegovina: Desk provided data from 1998-2003.
  2. Bulgaria: Other current transfers are included in remittances.
  3. Caribbean: Desk provided data for 1991-2002. (Caribbean region defined in source.)
  4. I.R. of Iran: Other current transfers are used as figure for remittances.
  5. Lebanon: Desk provided data for 1997-2003.
  6. Lesotho: Desk provided data for 1982-2003.
  7. Macedonia, FYR: Desk provided data for 1993-1997.
  8. Moldova: Desk provided data for 2000.
  9. Niger: Desk provided data for 1995-2003.
  10. Romania: Desk provided data for 2000-2003.
  11. Slovak Republic: Desk provided data for 1999-2003.
  12. Tajikistan: Desk provided data for 1997-2001.
  13. Ukraine: Desk provided data for 2000.
  14. Venezuela: Desk provided data for 1997-2003.

### Limitations and open questions
- The analysis does not explore all channels through which remittances may affect growth; other country characteristics (including institutional aspects) may help explain the observed effects and could be omitted variables.
- Potential moral hazard implications of remittances were not explored in detail.
- The findings are conditional on the specification used; omitted variables cannot be fully ruled out as drivers of some results.

### Policy implications
- In economies with shallow financial systems, remittance flows can play an important role in alleviating credit constraints and promoting investment and growth.
- Strengthening financial sector depth may change the cyclical role of remittances (making them more countercyclical) and could reduce remittances’ direct growth-enhancing investment role.
- Policymakers in remittance-receiving countries should consider the interaction between remittances and financial development when designing policies to channel remittances toward productive investment and when assessing labor-supply and social effects.

*Source: Conclusion and appendices (remittance constructions and data adjustments) from the paper summarized in the provided IMF PDF content.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05234.pdf_