## 7. The Gravity Model including Belarus, Moldova, and Ukraine

## Source details

**Canonical URL:** [7. The Gravity Model including Belarus, Moldova, and Ukraine](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020128-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020128-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020128-print-pdf.pdf.json)

---

### I. Introduction
- Remittances are an important source of external financing in Caucasus and Central Asia (CCA), reaching more than 30 percent of GDP in some countries.
- Remittance inflows exceeded 5 percent of GDP in 57 countries in 2019.
- Remittance inflows to low- and middle-income countries reached USD 554 bln in 2019.
- CCA remittances largely originate from Russia and are in part motivated by altruism.
- Remittances can stabilize output fluctuations (countercyclical behavior) but can also destabilize recipient economies when sending-country shocks reduce outward remittances.
- Analysis period and scope: gravity model for 2010-17 estimating determinants of bilateral remittances involving Russia and CCA countries; expanded-sample results include Belarus, Moldova, and Ukraine.
- Aggregated elasticity findings (summary): sending-country GDP elasticity ranges between 0.4 – 1.4 depending on specification; receiving-country GDP elasticity reported as 2.5 in summary (model-specific estimates vary).

### II. Literature summary (key points)
- Gravity determinants (sending and receiving GDP, distance) explain more than 60 percent of variation in bilateral remittances in prior studies.
- Reported elasticities of remittances to sending-country GDP vary widely across studies: typically between 0.25 – 3.9, with some studies reporting 5 – 10.
- Institutional and geographic predictors identified:
  - Common official language facilitates migration and remittances.
  - Common colonial links can matter.
  - Sharing a border has mixed evidence (may increase formal remittances in some studies and decrease them in others due to informal cash flows).
  - Number of migrants is positively associated with remittances.
  - Inflation differentials, exchange rate behavior, and other macro variables have mixed findings.
- Remittances have been found to be countercyclical in several studies but high remittance dependence can increase synchronization and vulnerability to sending-country shocks.

### III. Data and stylized facts
- Sample and period: Russia and eight CCA countries (Armenia, Azerbaijan, Georgia, Kazakhstan, Kyrgyzstan, Tajikistan, Turkmenistan, Uzbekistan), 2010-17 (bilateral remittance data availability).
- Main data sources:
  - Bilateral remittances (2010-17): World Bank.
  - Geographical distances and language dummy: CEPII.
  - Nominal GDP (USD): World Bank.
  - Nominal exchange rate vis-à-vis USD: Penn World Table.
  - Inflation: WEO (IMF).
  - Number of migrants (2013 and 2017) and share of female migrants (2000): World Bank.
  - Oil prices: Datastream.
- Stylized facts (selected):
  - Remittance-to-GDP ratio varies widely: highest in Tajikistan and Kyrgyzstan (ranging from 20 to 42 percent in different years), lowest in Turkmenistan and Russia (less than 1 percent).
  - Russia is the main remittance-sending country to CCA; largest recipients from Russia are Kyrgyzstan, Tajikistan, and Uzbekistan.
  - For some pairs, remittances are very small or zero (e.g., Kazakhstan, Turkmenistan, Uzbekistan receive few remittances from countries other than Russia).
  - Empirical elasticity between bilateral remittances and number of migrants ≈ 0.5 (a 1 percent increase in migrants associated with a 0.5 percent increase in remittances).

### IV. Gravity model specification and estimation
- Baseline gravity equation (REMijt in USD):
  REMijt = β0 + β1 ln(GDPit) + β2 ln(GDPjt) + β3 ln(DISTij) + β4' Xijt + μi + uj + ηt + εijt
  - REMijt: remittances from sending country i to receiving country j in year t (USD).
  - GDPit, GDPjt: nominal GDPs (USD) of sending and receiving countries.
  - DISTij: physical distance (kilometers) between capitals.
  - Xijt: contiguity, common language, migrants, inflation, exchange rate, age dependency, oil price.
  - μi, uj, ηt: country and time fixed effects.
- Countercyclicality test:
  REMijt = β0 + β1 (GAPit − GAPjt) + β2 ln(DISTij) + β3' Xijt + μi + uj + ηt + εijt
  - (GAPit − GAPjt): difference in output gaps (HP filter, smoothing parameter 6.5).
  - Positive β1 is evidence of countercyclicality.
- Estimation method: Pseudo Poisson Maximum Likelihood (PPML) to handle many zeros in remittance flows.
- Coefficient interpretation: β implies 100*(e^β − 1) percent change in bilateral remittances for a unit change in the independent variable.

### V. Estimation results — key findings
- Model fit:
  - Pseudo R-squared ranges between 0.65 and 0.93 across specifications.
- Sending-country GDP:
  - Elasticity ranges between 0.4 – 1.4 depending on specification.
  - Example point estimates in Table 2: 0.89*** (col 1 pooled), 0.31** (col 2 with country FE), 0.33** (col 4 country-pair FE).
- Receiving-country GDP:
  - Positive association; summary text reports elasticity = 2.5 (model-specific estimates vary; Table 2 shows coefficients such as 1.25** in some specifications).
- Geographical and institutional variables:
  - log(Distance) elasticity ≈ -0.7 (negative association); example estimates include -1.31*** and -1.30***.
  - Contiguity (share a border) elasticity ≈ -0.6 (negative association); example estimates -0.85***, -0.84**.
  - Common official language elasticity ≈ 5.8 (positive association); Table 2 shows 1.91*** (interpretation via 100*(e^β −1) yields larger percent changes).
- Baseline model with additional controls (Table 3):
  - Sending-country GDP elasticity narrows to 0.6 – 0.8 in some specifications; example coefficients: 0.45**, 0.57***, 0.59***.
  - Number of migrants (log) positively associated with remittances (Table 3 example: 0.57***).
  - Share of female migrants negatively associated with bilateral remittances (examples: -0.47*, -0.46***).
  - Receiving-country inflation positively associated with remittances in some specs (examples: 4.05*, 4.08**).
  - Exchange rate depreciation (vis-à-vis USD) in receiving country associated with lower USD remittances (examples: -0.90**, -0.89*).
  - Age dependency ratio positive when significant (supports altruism hypothesis).
- Countercyclicality (Table 4):
  - Difference in output gaps (sending versus receiving) coefficient is positive and significant across specifications: example estimates 0.13*** (col 1), 0.05** (col 2), 0.10* (col 3).
  - Average elasticity of the difference in output gaps variable ≈ 0.1:
    - Interpretation: a one percentage point decline in output below potential in the recipient country is associated with a 0.1 percent increase in remittance inflows.
  - Policy implication: remittances act as a stabilizer in normal times but global shocks that reduce sending-country output can cause downward shifts in remittances and increase volatility in recipient countries.
- Robustness checks (Tables 5–7):
  - Direct effect of oil prices:
    - Adding log(oil prices) (excluding time fixed effects because oil prices are common) yields insignificant coefficients on oil prices in the full Russia+CCA sample (Table 5: log(oil prices) estimates -0.13, -0.42, -0.42 are insignificant).
    - Indirect effect possible: when Russia is excluded from the sample, oil price coefficient becomes positive and significant (results available upon request in source).
  - Endogeneity of GDP:
    - Using lagged GDP (Table 6) produces comparable coefficients to baseline, suggesting limited bias from reverse causality in this sample.
  - Expanded sample including Belarus, Moldova, and Ukraine (Table 7):
    - Results remain qualitatively unchanged.
    - Sending-country GDP elasticity ranges between 0.5 – 0.8 in expanded-sample specifications (Table 7 example coefficients: 0.57***, 0.40***, 0.41***).

### VI. Conclusions and policy-relevant points
- Remittances are a major external financing source in CCA and have grown over time, especially in Kyrgyzstan and Tajikistan.
- Russia is the dominant source of remittances to the region due to size and historical links.
- Gravity model implications for projections:
  - Larger sending-country GDP → more remittances outflows (elasticity generally positive).
  - Larger receiving-country GDP → more remittances inflows (positive association).
  - Greater physical distance → lower remittance volumes (negative elasticity).
  - Shared border → lower recorded formal remittances (suggesting informal cash channels).
  - Common official language → higher remittance flows.
- Additional controls and mechanisms:
  - More compatriots abroad → more remittances; effect attenuated with higher share of female migrants (gender bias).
  - Higher receiving-country inflation → higher remittances (to preserve purchasing power).
  - Exchange rate depreciation → lower USD remittances.
  - Higher age dependency ratio → higher remittances (altruism motive).
- Oil prices:
  - No direct effect on remittances in the full Russia+CCA sample, but oil price effects can operate indirectly via sending-country GDP (notably Russia).
- Countercyclicality and shock transmission:
  - Remittances are countercyclical: they increase when sending-country output is above potential or when receiving-country output is below potential, stabilizing recipients in normal times but exposing them to downward shifts when sending countries suffer global shocks.
- Relevance for COVID-19 and oil price shocks:
  - Projected shrinking of the Russian economy implies suppressed outward remittances from Russia.
  - Elasticity estimates from the gravity model can be used to project remittance declines into CCA and to assess transmission to tax revenues using approaches such as Abdih and others (2012).

*Source: IMF staff analysis from “The Gravity Model including Belarus, Moldova, and Ukraine” (chapter 7, wpiea2020128-print-pdf).*

### References ___________________________________________________________ 22

### References ___________________________________________________________ 22

### FIGURES
- 1. Remittance-to-GDP Ratio in Russia and CCA (2010-2017) _____________________ 12
- 2. Heat Map of Bilateral Remittances (in logs, 2017) ____________________________ 13
- 3. Bilateral Remittances and Migration (2013 and 2017) _________________________ 14

### TABLES
- 1. Variables and Their Sources ____________________________________________ 15
- 2. The Baseline Gravity Model ____________________________________________ 16
- 3. The Baseline Gravity Model with Additional Controls _________________________ 17
- 4. Are Remittances Countercyclical? ________________________________________ 18
- 5. Do Oil Prices Have a Direct Effect on Remittances?___________________________ 19
- 6. The Baseline Gravity Model with Lagged GDP Variables ______________________ 20

*https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020128-print-pdf.pdf*

### 7. The Gravity Model including Belarus, Moldova, and Ukraine ___________________ 21

### 7. The Gravity Model including Belarus, Moldova, and Ukraine

### I. Introduction
- Remittances are an important source of external financing in Caucasus and Central Asia (CCA), reaching more than 30 percent of GDP in some countries.
- Remittance inflows exceeded 5 percent of GDP in 57 countries in 2019.
- Remittance inflows to low- and middle-income countries reached USD 554 bln in 2019.
- CCA remittances largely originate from Russia and are in part motivated by altruism.
- Remittances can stabilize output fluctuations (countercyclical behavior) but can also destabilize recipient economies when sending-country shocks reduce outward remittances.
- This analysis uses a gravity model for the 2010-17 period to estimate determinants of bilateral remittances involving Russia and CCA countries and to assess remittances’ cyclical properties and sensitivity to oil prices.
- Key aggregated elasticity findings from the gravity model: sending-country GDP elasticity ranges between 0.4 – 1.4 depending on specification; receiving-country GDP elasticity reported as 2.5 in summary (model-specific estimates vary).

### II. Literature summary
- Gravity determinants (sending and receiving GDP, distance) explain more than 60 percent of variation in bilateral remittances in prior studies.
- Reported elasticities of remittances to sending-country GDP in the literature vary widely, typically between 0.25 – 3.9, with some studies reporting 5 – 10.
- Institutional and geographic predictors identified in the literature:
  - Common official language facilitates migration and remittances.
  - Common colonial links can matter.
  - Sharing a border has mixed evidence; it may increase formal remittances in some studies and decrease them in others (informal cash flows).
  - Number of migrants is positively associated with remittances.
  - Inflation differentials, exchange rate behavior, and other macro variables have been explored with mixed findings.
- Remittances have been found to be countercyclical in several studies, helping smooth recipient-country output; however, large remittance dependence can increase synchronization and vulnerability to sending-country shocks.

### III. Data and stylized facts
- Sample: Russia and eight CCA countries: Armenia, Azerbaijan, Georgia, Kazakhstan, Kyrgyzstan, Tajikistan, Turkmenistan, Uzbekistan. Time period: 2010-17 (bilateral remittance data availability).
- Main datasets and sources:
  - Bilateral remittances (2010-17): World Bank.
  - Geographical distances and language dummy: CEPII.
  - Nominal GDP (USD): World Bank.
  - Nominal exchange rate vis-à-vis USD: Penn World Table.
  - Inflation: WEO (IMF).
  - Number of migrants (2013 and 2017) and share of female migrants (2000): World Bank.
  - Oil prices: Datastream.
- Stylized facts:
  - Remittance-to-GDP ratio varies widely: highest in Tajikistan and Kyrgyzstan (ranging from 20 to 42 percent in different years), lowest in Turkmenistan and Russia (less than 1 percent).
  - Russia is the main remittance-sending country to CCA; largest recipients from Russia are Kyrgyzstan, Tajikistan, and Uzbekistan.
  - For some pairs, remittances are very small or zero (e.g., Kazakhstan, Turkmenistan, Uzbekistan receive few remittances from countries other than Russia).
  - Empirical elasticity between bilateral remittances and number of migrants ~0.5 (a 1 percent increase in migrants associated with a 0.5 percent increase in remittances).

### IV. Gravity model specification and estimation
- Baseline gravity specification (remittances REMijt in USD):
  REMijt = β0 + β1 ln(GDPit) + β2 ln(GDPjt) + β3 ln(DISTij) + β4' Xijt + μi + uj + ηt + εijt
  - REMijt: remittances from sending country i to receiving country j in year t (USD).
  - GDPit, GDPjt: nominal GDPs (USD) of sending and receiving countries.
  - DISTij: physical distance (kilometers) between capitals.
  - Xijt: vector of other controls (contiguity, common language, migrants, inflation, exchange rate, age dependency, oil price).
  - μ, u, η: country and time fixed effects.
- Countercyclicality test specification:
  REMijt = β0 + β1 (GAPit − GAPjt) + β2 ln(DISTij) + β3' Xijt + μi + uj + ηt + εijt
  - (GAPit − GAPjt): difference in output gaps between sending and receiving countries (output gaps estimated via HP filter with smoothing parameter 6.5).
  - Positive β1 supports countercyclicality of remittances.
- Estimation method: Pseudo Poisson Maximum Likelihood (PPML) to handle many zeros in remittance flows.
- Interpretation: coefficient β implies 100*(e^β − 1) percent change in bilateral remittances for a unit change in the independent variable.

### V. Estimation results — key findings
A. Baseline gravity model (Table 2: several specifications including country, country-pair, and year fixed effects)
- Model fit: pseudo R-squared ranging between 0.65 and 0.93 across specifications.
- Sending-country GDP:
  - Elasticity in baseline summaries ranges between 0.4 – 1.4 depending on specification.
  - Example point estimates in Table 2: 0.89*** (col 1 pooled), 0.31** (col 2 with country FE), 0.33** (col 4 country-pair FE).
- Receiving-country GDP:
  - Positive association; example elasticity = 2.5 in summary text (model-specific estimates vary; Table 2 shows coefficients such as 1.25** in some specifications).
- Geographical and institutional variables:
  - log(Distance) elasticity ≈ -0.7 (negative association; Table 2 column estimates include -1.31*** and -1.30***).
  - Contiguity (share a border) elasticity ≈ -0.6 (negative association; Table 2 shows -0.85***, -0.84** in country FE/pair FE specifications).
  - Common official language elasticity ≈ 5.8 (positive association; Table 2 shows 1.91*** in some specifications—interpretation via 100*(e^β −1) yields larger percent changes).

B. Baseline model with additional controls (Table 3)
- Sending-country GDP elasticity narrows to 0.6 – 0.8 in some specifications; sample coefficient examples: 0.45**, 0.57***, 0.59***.
- Number of migrants (log) positively associated with remittances (significant in some columns; Table 3 shows 0.57*** in column 1).
- Share of female migrants negatively associated with bilateral remittances (examples: -0.47*, -0.46***).
- Inflation in receiving country positively associated with remittances in some specifications (examples: 4.05* and 4.08** in certain columns).
- Exchange rate depreciation (vis-à-vis USD) in receiving country associated with lower USD remittances (negative coefficients: -0.90**, -0.89* in some columns).
- Age dependency ratio positive when significant (supports altruism hypothesis).

C. Countercyclicality (Table 4)
- Difference in output gaps (sending versus receiving) coefficient is positive and significant across specifications:
  - Example estimates: 0.13*** (col 1), 0.05** (col 2), 0.10* (col 3).
- Average elasticity of the difference in output gaps variable ~0.1:
  - Interpretation: a one percentage point decline in output below potential in the recipient country is associated with a 0.1 percent increase in remittance inflows.
- Policy implication: remittances act as a stabilizer in normal times but global shocks that reduce sending-country output can cause downward shifts in remittances and increase volatility in recipient countries.

D. Robustness checks (Tables 5–7)
- Direct effect of oil prices:
  - Adding log(oil prices) (excluding time fixed effects because oil prices are common) yields insignificant coefficients on oil prices in the full Russia+CCA sample (Table 5: log(oil prices) estimates -0.13, -0.42, -0.42 are insignificant).
  - Indirect effect possible: when Russia is excluded from the sample, oil price coefficient becomes positive and significant (results available upon request in source).
- Endogeneity of GDP:
  - Using lagged GDP for sending and receiving countries (Table 6) produces comparable coefficients to baseline, suggesting limited bias from reverse causality in this sample.
- Expanded sample including Belarus, Moldova, and Ukraine (Table 7):
  - Results remain qualitatively unchanged.
  - Sending-country GDP elasticity ranges between 0.5 – 0.8 in expanded-sample specifications (Table 7 example coefficients: 0.57***, 0.40***, 0.41***).

### VI. Conclusions and policy-relevant points
- Remittances are a major external financing source in CCA and have grown over time, especially in Kyrgyzstan and Tajikistan.
- Russia is the dominant source of remittances to the region due to size and historical links.
- The gravity model provides a good fit and can be used for remittance projections:
  - Larger sending-country GDP → more remittances outflows (elasticity generally positive).
  - Larger receiving-country GDP → more remittances inflows (positive association).
  - Greater physical distance → lower remittance volumes (negative elasticity).
  - Shared border → lower recorded formal remittances (suggesting informal cash channels).
  - Common official language → higher remittance flows.
- Additional controls:
  - More compatriots abroad → more remittances; effect attenuated with higher share of female migrants (gender bias).
  - Higher receiving-country inflation → higher remittances (to preserve purchasing power).
  - Exchange rate depreciation → lower USD remittances.
  - Higher age dependency ratio → higher remittances (altruism motive).
- Oil prices do not show a direct effect on remittances in the full Russia+CCA sample, but oil price effects can operate indirectly via sending-country GDP (notably Russia).
- Remittances are countercyclical: they increase when sending-country output is above potential or when receiving-country output is below potential, stabilizing recipients in normal times but exposing them to downward shifts when sending countries suffer global shocks.
- Relevance for COVID-19 and oil price shocks:
  - Projected shrinking of the Russian economy implies suppressed outward remittances from Russia.
  - Elasticity estimates from the gravity model can be used to project remittance declines into CCA and to assess transmission to tax revenues using approaches such as Abdih and others (2012).

*Source: IMF staff analysis from “The Gravity Model including Belarus, Moldova, and Ukraine” (chapter 7, wpiea2020128-print-pdf).*

### References

### References

### Empirical studies on remittances and migration
- Abdih, Y., Barajas, A., Chami, R. and C. Ebeke. 2012. “Remittances Channel and Fiscal Impact in the Middle East, North Africa, and Central Asia,” IMF Working Paper WP/12/104 (Washington, D.C.).  
- Barajas, A., Chami, R., Ebeke., C. and S. Tapsoba. 2012. “Workers’ Remittances: An Overlooked Channel of International Business Cycle Transmission?” IMF Working Paper WP/12/251 (Washington, D.C.).  
- Bettin, G., Presbitero, A., and N. Spatafora. 2017. “Remittances and Vulnerability in Developing Countries,” The World Bank Economic Review, 31 (1): pp. 1-23.  
- Chami, R., Fullenkamp, C. and S. Janjah. 2005. “Are Immigrant Remittance Flows a Source of Capital for Development?” IMF Staff Papers, 52(1): pp. 55–81.  
- Chami, R., Hakura, D. and P. Montiel. 2012. “Do Worker Remittances Reduce Output Volatility in Developing Countries?” Journal of Globalization and Development, 3 (1): pp. 1–25.  
- Docquier, F., and H. Rapoport. 2006. “The Economics of Migrants’ Remittances,” Handbook of the Economics of Giving, Altruism and Reciprocity, vol 1: pp. 1135–98.  
- Docquier, F., Rapoport, H. and S. Salomone. 2012. “Remittances, Migrants’ Education and Immigration Policy: Theory and Evidence from Bilateral Data,” Regional Science and Urban Economics, 42: pp. 817–28.  
- Frankel, J. 2011. “Are Bilateral Remittances Countercyclical?” Open Economies Review, 22 (1): pp. 1–16.  
- Lueth, E. and M. Ruiz-Arranz. 2008. “Determinants of Bilateral Remittance Flows,” The B.E. Journal of Macroeconomics, 8 (1): pp. 1–21.  
- Le Golf, M. and S. Salomone. 2015. “Changes in Migration Patterns and Remittances: Do Females and Skilled Migrants Remit More?” CEPII Working Paper 2015-15 (Paris).  
- Yang, D. 2011. “Migrant Remittances,” Journal of Economic Perspectives, 25(3): pp. 129-52.  
- Sayeh, A. and R. Chami. 2020. “Lifelines in Danger,” Finance and Development, 57(2): pp. 16-19. (link: https://www.imf.org/external/pubs/ft/fandd/2020/06/pdf/COVID19-pandemic-impact-on-remittance-flows-sayeh.pdf)

### Methodological and theoretical references
- Santos Silva, J. and S. Tenreyro. 2006. “The Log of Gravity,” Review of Economics and Statistics, 88(4): pp. 641-658.  
- Yotov, Y., Piermartini, R., Monteiro, J.-A. and M. Larch. 2016. “An Advanced Guide to Trade Policy Analysis: The Structural Gravity Model,” World Trade Organization (Geneva).

### Regional analyses and COVID-19 impact
- IMF. 2012. “Regional Economic Outlook: Middle East and Central Asia,” November (Washington: International Monetary Fund).  
- World Bank. 2020. “COVID-19 Crisis Through a Migration Lens,” Migration and Development Brief 32 (Washington, D.C.).  

*Source: wpiea2020128-print-pdf - References*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020128-print-pdf.pdf_