## GROWTH-REMITTANCE SPILLOVER ANALYSIS

## Source details

**Canonical URL:** [GROWTH-REMITTANCE SPILLOVER ANALYSIS](https://www.imf.org/-/media/files/publications/cr/2021/english/1tjkea2021003.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2021/english/1tjkea2021003.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2021/english/1tjkea2021003.pdf.json)

---

### A. Context and Summary
- Tajikistan relies primarily on earnings of migrants working in Russia; remittances are “almost entirely originating in Russia and equivalent to nearly half of Tajik GDP.”
- From end-2013, Russia experienced a marked slowdown in national income and real economic growth, causing remittances to shrink.
- Two modeling approaches are used to project remittance inflows and their impact on Tajikistan’s growth:
  - A small model of Tajik remittance determination based on a construction index in Russia → projects remittance inflows likely to contract by 18.5 percent in 2015.
  - An alternative model linking Tajik remittances explicitly to income in Russia, the ruble/U.S. dollar exchange rate, and the oil price → projects remittance inflows will fall by 37 percent in 2015.
- Data through April (remittances contracting by 34 percent) favors the projection of the alternative model.
- Two approaches to estimate growth impacts (heuristic exploiting 2008 crisis turnaround; VAR model for 2002–14) imply the projected fall in remittances would lower GDP growth by an estimated 1.8 to 5.3 percentage points in 2015.

### Tajik remittances: Model with construction index in Russia
- Rationale:
  - The construction sector (official and unofficial) is the biggest employer of Tajik migrants in Russia; many migrants also work in retail trade, maintenance, and transportation but construction dominates.
- Model specification:
  - Relates log(Rem$ t ) to log(C_rus t ) (real construction index by Russian State Statistics Committee) with error term assumed AR(4) due to quarterly data (2004:Q1–2014:Q2) and strong seasonality.
  - Functional form: log(Rem$) = a1 + a2*log(Construction_RUS) + eps; eps = rho*eps(-4) + u.
- Estimation results (selected):
  - LOG(ICONST) coefficient = 2.385017 (Std. Error = 0.225052, t-Statistic = 10.59763, Prob. = 0.0000)
  - AR(4) = 0.978082 (Std. Error = 0.036660, t-Statistic = 26.68004, Prob. = 0.0000)
  - R-squared = 0.967756; Adjusted R-square = 0.965859; S.E. of regression = 0.123490
  - Durbin-Watson stat = 1.578227
- Forecasting assumptions and method:
  - Assume Russian construction sector contracts by 5.3 percent on average in 2014 and by 9 percent on average in 2015.
  - Construct dynamic forecast (resolving autocorrelated errors) and static forecast (drop error term), then aggregate using weights 0.2 (dynamic) and 0.8 (static) to form a single-point estimate (tilted toward static because of unusual 2015 uncertainty).
- Projection:
  - Model suggests remittance inflows likely to contract by 18.5 percent in 2015.

### Tajik remittances: Model with income, exchange rate, and oil price
- Key facts and exposure:
  - About 75 percent of remittances are denominated in rubles.
  - 2013 data: 97 percent of remittance inflows came from Russia; 74.2 percent of these transfers were wired in Russian rubles.
  - In 2014, the ruble depreciated by 42 percent against the U.S. dollar.
- Modeling approach:
  - Decompose US$ remittances (Rem$) = Rem_RUB / ER (RUB/USD exchange rate).
  - Project growth rates of Rem_RUB and RUB/USD through end-2014 and in 2015.
  - Growth rate of Rem_RUB: assumes slowdown from 8.3 percent in 2014 to 2.5 percent in 2015 (based on historical trends).
  - Project RUB/USD from likely path of Urals crude oil price and Russian budget parameters (budgeted oil price = 96 USD/barrel; budgeted exchange rate = 37.7 RUB/USD).
  - Assumption: the ruble value of 1 barrel of Urals crude oil will not change in 2015; fluctuations in exchange rate compensate for oil price changes.
  - Two oil scenarios used; both have the same average price of 58 USD/barrel but differ in monthly pattern (Scenario 2 uses 2013 historic monthly pattern).
- Projections:
  - Annual US$ remittance inflows projected to decline by 8.4 percent in 2014 and by 37 percent in 2015.
  - “2014 monthly pattern” distributes growth rates of ruble-denominated amounts according to monthly dynamics observed in 2014 and generates a given annual average growth rate for 2015.
- Empirical note:
  - Data through April with remittances contracting by 34 percent supports the alternative (income/exchange/oil) model projection.

### Remittance-growth spillovers: Two approaches
- Stylized facts:
  - Over the last two years, remittances and economic growth in Tajikistan appear tightly connected with a lag of one quarter; correlation coefficient between remittances (lagged three months) and real GDP growth is 85 percent.
- Two approaches used to estimate impact on growth:
  - Approach 1 (Heuristic exploiting 2008 turnaround):
    - Decomposition: GDP by production into agriculture, industry, construction, and services.
    - Assumption: a sizable part of services value-added is proportional to remittance inflows.
    - Derived relation: % chg real_GDP = (% chg REM$) * (REM/GDP) * α  (where α captures remittance-to-service value-added linkage).
    - Elasticity estimates:
      - 2009 calibration: fall in real GDP from 7.9 in 2008 to 3.4 percent in 2009 attributed to remittance decline (-31.3 percent) → α = 0.39; share of non-remittance related value added in services = 74 percent → 2009 real GDP-remittance elasticity = 0.1436.
      - 2012 calibration: using same share yields α = 0.29 and real GDP-remittance elasticity = 0.29 * 49.6% = 0.1262.
    - Implication: If remittances contract by 10 percent, real GDP is estimated to decline by 1.26–1.44 percentage points.
  - Approach 2 (VAR model for 2002–14):
    - Variables: Real GDP growth; CPI inflation; $ remittance inflow growth (growth rates are quarter over corresponding quarter of preceding year).
    - VAR specification: unrestricted VAR with 2 lags chosen (lag selection criteria ranged from 1 (Schwarz) to 4 (AIC); two criteria out of five favored 2 lags).
    - Fit and results (selected):
      - Remittance growth equation: adjusted R-squared = 92 percent (best fit).
      - Inflation equation: medium fit; GDP growth equation: lowest (but still good) fit.
      - Selected coefficients:
        - REMIT_ACHG(-1) coefficient in REMIT_ACHG equation = 1.165077 (Std. Error = 0.13644, t-stat = 8.53896)
        - REMIT_ACHG(-2) in REMIT_ACHG equation = -0.277960 (Std. Error = 0.12343, t-stat = -2.25189)
        - CPI_INFL_ACHG(-1) in NGDP_R_ACHG equation = -0.378308 (Std. Error = 0.14702, t-stat = -2.57323)
        - CPI_INFL_ACHG(-2) in NGDP_R_ACHG equation = 0.339708 (Std. Error = 0.15695, t-stat = 2.16438)
      - Overall R-squared: NGDP_R_ACHG = 0.413131; CPI_INFL_ACHG = 0.794215; REMIT_ACHG = 0.935027.
    - Use of VAR:
      - Helps with remittance-growth causality in Granger sense and captures joint dynamics.
- Combined implication:
  - Both approaches suggest strong real effects: the projected fall in remittances would lower GDP growth by an estimated 1.8 to 5.3 percentage points in 2015.

---

### Section 2

### Estimation results and model diagnostics
- R-squared 0.312525 0.758937 0.923889
- Sum sq. resids 0.013180 0.017453 0.267459
- S.E. equation 0.019406 0.022331 0.087417
- F-statistic 4.106418 22.51339 83.94775
- Log likelihood 109.8052 103.9084 46.59022
- Akaike AIC -4.895484 -4.614685 -1.885249
- Schwarz SC -4.605872 -4.325073 -1.595637
- Mean dependent 0.070865 0.091485 0.326321
- S.D. dependent 0.023405 0.045482 0.316862
- Determinant resid covariance (dof adj.) 1.34E-09
- Determinant resid covariance 7.75E-10
- Log likelihood 261.7637
- Akaike information criterion -11.46494
- Schwarz criterion -10.59610

- Breusch-Godfrey Serial Correlation LM Test:
  - F-statistic 1.035567     Prob. F(4,30) 0.4052
  - Obs*R-squared 4.504628     Prob. Chi-Square(4) 0.3420
  - Test sample: 2005Q1 2014Q1; Included observations: 37
  - R-squared 0.121747     Mean dependent var 8.49E-06
  - Adjusted R-squared -0.053904     S.D. dependent var 0.120039
  - S.E. of regression 0.123232     Akaike info criterion -1.180839
  - Sum squared resid 0.455583     Schwarz criterion -0.876071
  - Log likelihood 28.84552     Hannan-Quinn criter. -1.073394
  - F-statistic 0.693118     Durbin-Watson stat 1.894828
  - Prob(F-statistic) 0.656923

### Key empirical findings and interpretation
- The real GDP growth equation (adj. R2 = 31 percent) shows that the coefficients of remittance inflows lags are both significant and their sum is positive, implying a strong positive relationship between remittance inflows growth and real GDP growth.
- Granger causality test results:
  - Remittance inflows growth helps explain the dynamics of real GDP growth (dropping remittance growth from the model materially deteriorates fit).
  - There is no Granger-causality from real GDP growth to remittance inflows.
  - There is a two-way endogenous relationship between price inflation and remittance inflows growth: higher remittances drive up domestic prices, and higher prices help attract greater remittances.
- Granger Causality/Block Exogeneity Wald Tests (Date: 09/10/14; Sample: 2002Q1 2014Q2; Included observations: 42):
  - Dependent variable: NGDP_R_ACHG
    - CPI_INFL_ACHG  6.621944 2  0.0365
    - REMIT_ACHG 18.212952 2  0.0001
    - All  18.84195 4  0.0008
  - Dependent variable: CPI_INFL_ACHG
    - NGDP_R_ACHG  2.530938 2  0.2821
    - REMIT_ACHG  6.002358 2  0.0497
    - All  7.288732 4  0.1214
  - Dependent variable: REMIT_ACHG
    - NGDP_R_ACHG  0.621892 2  0.7328
    - CPI_INFL_ACHG 13.965342 2  0.0009
    - All 13.965434 4  0.0074
- The real GDP–remittance growth elasticity is estimated to range from 0.095 to 0.097 based on 2014 data.
  - Method: use remittance inflows growth in Q1, Q2 and Q3 of 2014 (actual) and Q4 (predicted by the dynamic and static forecast VAR with a Russian construction index), compute contributions to real GDP growth from observed and predicted remittance growth rates, average predicted contributions across quarters in 2014, and divide by the average of remittance inflows growth rates lagged one quarter to obtain the elasticity estimate.

### VAR model selection, residual tests, and fit
- VAR Lag Order Selection Criteria (Endogenous variables: NGDP_R_ACHG CPI_INFL_ACHG REMIT_ACHG; Exogenous variables: C; Sample: 2002Q1 2014Q2; Included observations: 40)
  - Lag 0: LogL 171.1087; FPE 4.49e-08; AIC -8.405435; SC -8.278769; HQ -8.359637
  - Lag 1: LogL 237.6508; LR 119.7758; FPE 2.53e-09; AIC -11.28254; SC -10.77588*; HQ -11.09935
  - Lag 2: LogL 253.9536; LR 26.89968*; FPE 1.77e-09; AIC -11.64768; SC -10.76102; HQ -11.32709*
  - Lag 3: LogL 264.0201; LR 15.09964; FPE 1.72e-09*; AIC -11.70100; SC -10.43434; HQ -11.24302
  - Lag 4: LogL 273.5157; LR 12.81905; FPE 1.75e-09; AIC -11.72578*; SC -10.07913; HQ -11.13040
  - Note: * indicates lag order selected by the criterion.
- VAR Residual Serial Correlation LM Test (Null Hypothesis: no serial correlation at lag order h; Date: 09/10/14; Sample: 2002Q1 2014Q2; Included observations: 42)
  - Lag 1: LM-Stat 19.22657  Prob 0.0233
  - Lag 2: LM-Stat 10.86233  Prob 0.2853
  - Lag 3: LM-Stat 8.908483  Prob 0.4458
  - Probs from chi-square with 9 df.

### Auxiliary model: Tajik remittances and Russian construction index
- Model and figure: Tajik Remittance Inflows and Construction Index in Russia, 2004:Q1–2014:Q1; plotted series LOG(REMITTANCES_TJK) and LOG(ICONSTRUCTION_RUS).
- Serial correlation test for remittance model: no evidence of significant serial correlation given Prob. F(4,30) 0.4052 and Prob. Chi-Square(4) 0.3420.

*Prepared by Maxym Kryshko; IMF Country Report No. 21/197, Selected Issues (May 18, 2015).*

*REPUBLIC OF TAJIKISTAN — INTERNATIONAL MONETARY FUND*

### Section 1

### GROWTH-REMITTANCE SPILLOVER ANALYSIS

### A. Context and Summary
- Tajikistan relies primarily on earnings of migrants working in Russia; remittances are “almost entirely originating in Russia and equivalent to nearly half of Tajik GDP.”
- From end-2013, Russia experienced a marked slowdown in national income and real economic growth, causing remittances to shrink.
- Two modeling approaches are used to project remittance inflows and their impact on Tajikistan’s growth:
  - A small model of Tajik remittance determination based on a construction index in Russia → projects remittance inflows likely to contract by 18.5 percent in 2015.
  - An alternative model linking Tajik remittances explicitly to income in Russia, the ruble/U.S. dollar exchange rate, and the oil price → projects remittance inflows will fall by 37 percent in 2015.
- Data through April (remittances contracting by 34 percent) favors the projection of the alternative model.
- Two approaches to estimate growth impacts (heuristic exploiting 2008 crisis turnaround; VAR model for 2002–14) imply the projected fall in remittances would lower GDP growth by an estimated 1.8 to 5.3 percentage points in 2015.

### Tajik remittances: Model with construction index in Russia
- Rationale:
  - The construction sector (official and unofficial) is the biggest employer of Tajik migrants in Russia; many migrants also work in retail trade, maintenance, and transportation but construction dominates.
- Model specification:
  - Relates log(Rem$ t ) to log(C_rus t ) (real construction index by Russian State Statistics Committee) with error term assumed AR(4) due to quarterly data (2004:Q1–2014:Q2) and strong seasonality.
  - Functional form shown: log(Rem$) = a1 + a2*log(Construction_RUS) + eps; eps = rho*eps(-4) + u.
- Estimation results (selected):
  - LOG(ICONST) coefficient = 2.385017 (Std. Error = 0.225052, t-Statistic = 10.59763, Prob. = 0.0000)
  - AR(4) = 0.978082 (Std. Error = 0.036660, t-Statistic = 26.68004, Prob. = 0.0000)
  - R-squared = 0.967756; Adjusted R-square = 0.965859; S.E. of regression = 0.123490
  - Durbin-Watson stat = 1.578227
- Forecasting assumptions and method:
  - Assume Russian construction sector contracts by 5.3 percent on average in 2014 and by 9 percent on average in 2015.
  - Construct dynamic forecast (resolving autocorrelated errors) and static forecast (drop error term), then aggregate using weights 0.2 (dynamic) and 0.8 (static) to form a single-point estimate (tilted toward static because of unusual 2015 uncertainty).
- Projection:
  - Model suggests remittance inflows likely to contract by 18.5 percent in 2015.

### Tajik remittances: Model with income, exchange rate, and oil price
- Key facts and exposure:
  - About 75 percent of remittances are denominated in rubles.
  - 2013 data: 97 percent of remittance inflows came from Russia; 74.2 percent of these transfers were wired in Russian rubles.
  - In 2014, the ruble depreciated by 42 percent against the U.S. dollar.
- Modeling approach:
  - Decompose US$ remittances (Rem$) = Rem_RUB / ER (RUB/USD exchange rate).
  - Project growth rates of Rem_RUB and RUB/USD through end-2014 and in 2015.
  - Growth rate of Rem_RUB: assumes slowdown from 8.3 percent in 2014 to 2.5 percent in 2015 (based on historical trends).
  - Project RUB/USD from likely path of Urals crude oil price and Russian budget parameters (budgeted oil price = 96 USD/barrel; budgeted exchange rate = 37.7 RUB/USD).
  - Assumption: the ruble value of 1 barrel of Urals crude oil will not change in 2015; fluctuations in exchange rate compensate for oil price changes.
  - Two oil scenarios used; both have the same average price of 58 USD/barrel but differ in monthly pattern (Scenario 2 uses 2013 historic monthly pattern).
- Projections:
  - Annual US$ remittance inflows projected to decline by 8.4 percent in 2014 and by 37 percent in 2015.
  - “2014 monthly pattern” distributes growth rates of ruble-denominated amounts according to monthly dynamics observed in 2014 and generates a given annual average growth rate for 2015.
- Empirical note:
  - Data through April with remittances contracting by 34 percent supports the alternative (income/exchange/oil) model projection.

### Remittance-growth spillovers: Two approaches
A. Stylized Facts
- Over the last two years, remittances and economic growth in Tajikistan appear tightly connected with a lag of one quarter; correlation coefficient between remittances (lagged three months) and real GDP growth is 85 percent.

B. Two approaches used to estimate impact on growth
- Approach 1: Heuristic approach exploiting sharp remittance turnaround after 2008 global financial crisis to gauge remittance-growth elasticity and establish an upper bound on GDP growth decline given remittance contraction.
  - Decomposition: GDP by production into agriculture, industry, construction, and services.
  - Assumption: a sizable part of services value-added is proportional to remittance inflows.
  - Derived relation: % chg real_GDP = (% chg REM$) * (REM/GDP) * α  (where α captures remittance-to-service value-added linkage).
  - Elasticity estimates:
    - 2009 calibration: fall in real GDP from 7.9 in 2008 to 3.4 percent in 2009 attributed to remittance decline (-31.3 percent) → α = 0.39; share of non-remittance related value added in services = 74 percent → 2009 real GDP-remittance elasticity = 0.1436.
    - 2012 calibration: using same share yields α = 0.29 and real GDP-remittance elasticity = 0.29 * 49.6% = 0.1262.
  - Implication: If remittances contract by 10 percent, real GDP is estimated to decline by 1.26–1.44 percentage points.
- Approach 2: VAR model (2002–14) with three variables evolving jointly and endogenously:
  - Variables: Real GDP growth; CPI inflation; $ remittance inflow growth (growth rates are quarter over corresponding quarter of preceding year).
  - VAR specification: unrestricted VAR with 2 lags chosen (lag selection criteria ranged from 1 (Schwarz) to 4 (AIC); two criteria out of five favored 2 lags).
  - Fit and results (selected):
    - Remittance growth equation: adjusted R-squared = 92 percent (best fit).
    - Inflation equation: medium fit; GDP growth equation: lowest (but still good) fit.
    - Selected coefficients in VAR (standard errors and t-stats reported in estimation):
      - REMIT_ACHG(-1) coefficient in REMIT_ACHG equation = 1.165077 (Std. Error = 0.13644, t-stat = 8.53896)
      - REMIT_ACHG(-2) in REMIT_ACHG equation = -0.277960 (Std. Error = 0.12343, t-stat = -2.25189)
      - CPI_INFL_ACHG(-1) in NGDP_R_ACHG equation = -0.378308 (Std. Error = 0.14702, t-stat = -2.57323)
      - CPI_INFL_ACHG(-2) in NGDP_R_ACHG equation = 0.339708 (Std. Error = 0.15695, t-stat = 2.16438)
    - Overall R-squared: NGDP_R_ACHG = 0.413131; CPI_INFL_ACHG = 0.794215; REMIT_ACHG = 0.935027.
  - Use of VAR:
    - Helps with remittance-growth causality in Granger sense and captures joint dynamics.
- Combined implication from heuristic and VAR approaches:
  - Both approaches suggest strong real effects: the projected fall in remittances would lower GDP growth by an estimated 1.8 to 5.3 percentage points in 2015.

*Prepared by Maxym Kryshko; IMF Country Report No. 21/197, Selected Issues (May 18, 2015).*

### Section 2

### Section 2

### Estimation results and model diagnostics
- R-squared0.312525 0.758937 0.923889
- Sum sq. resids  0.013180  0.017453  0.267459
- S.E. equation0.019406 0.022331 0.087417
- F-statistic4.106418 22.51339 83.94775
- Log likelihood109.8052 103.9084 46.59022
- Akaike AIC -4.895484 -4.614685 -1.885249
- Schwarz SC-4.605872-4.325073 -1.595637
- Mean dependent0.070865 0.091485 0.326321
- S.D. dependent0.023405 0.045482 0.316862
- Determinant resid covariance (dof adj.) 1.34E-09
- Determinant resid covariance 7.75E-10
- Log likelihood 261.7637
- Akaike information criterion -11.46494
- Schwarz criterion-10.59610

- Breusch-Godfrey Serial Correlation LM Test:
  - F-statistic1.035567     Prob. F(4,30)0.4052
  - Obs*R-squared4.504628     Prob. Chi-Square(4)0.3420
  - Test sample: 2005Q1 2014Q1; Included observations: 37
  - R-squared0.121747     Mean dependent var8.49E-06
  - Adjusted R-squared-0.053904     S.D. dependent var0.120039
  - S.E. of regression0.123232     Akaike info criterion-1.180839
  - Sum squared resid0.455583     Schwarz criterion-0.876071
  - Log likelihood28.84552     Hannan-Quinn criter.-1.073394
  - F-statistic0.693118     Durbin-Watson stat1.894828
  - Prob(F-statistic)0.656923

### Key empirical findings and interpretation
- The real GDP growth equation (adj. R2 = 31 percent) shows that the coefficients of remittance inflows lags are both significant and their sum is positive, implying a strong positive relationship between remittance inflows growth and real GDP growth.
- Granger causality test results:
  - Remittance inflows growth helps explain the dynamics of real GDP growth (dropping remittance growth from the model materially deteriorates fit).
  - There is no Granger-causality from real GDP growth to remittance inflows.
  - There is a two-way endogenous relationship between price inflation and remittance inflows growth: higher remittances drive up domestic prices, and higher prices help attract greater remittances.

- Granger Causality/Block Exogeneity Wald Tests (Date: 09/10/14; Sample: 2002Q1 2014Q2; Included observations: 42)
  - Dependent variable: NGDP_R_ACHG
    - CPI_INFL_ACHG  6.621944 2  0.0365
    - REMIT_ACHG18.2129520.0001
    - All  18.84195 4  0.0008
  - Dependent variable: CPI_INFL_ACHG
    - NGDP_R_ACHG  2.530938 2  0.2821
    - REMIT_ACHG  6.002358 2  0.0497
    - All  7.288732 4  0.1214
  - Dependent variable: REMIT_ACHG
    - NGDP_R_ACHG  0.621892 2  0.7328
    - CPI_INFL_ACHG13.9653420.0009
    - All13.9654340.0074

- The real GDP–remittance growth elasticity is estimated to range from 0.095 to 0.097 based on 2014 data.
  - Method: use remittance inflows growth in Q1, Q2 and Q3 of 2014 (actual) and Q4 (predicted by the dynamic and static forecast VAR with a Russian construction index), compute contributions to real GDP growth from observed and predicted remittance growth rates, average predicted contributions across quarters in 2014, and divide by the average of remittance inflows growth rates lagged one quarter to obtain the elasticity estimate.

### VAR model selection, residual tests, and fit
- VAR Lag Order Selection Criteria (Endogenous variables: NGDP_R_ACHG CPI_INFL_ACHG REMIT_ACHG; Exogenous variables: C; Sample: 2002Q1 2014Q2; Included observations: 40)
  - Lag 0: LogL 171.1087; FPE 4.49e-08; AIC -8.405435; SC -8.278769; HQ -8.359637
  - Lag 1: LogL 237.6508; LR 119.7758; FPE2.53e-09; AIC -11.28254; SC -10.77588*; HQ -11.09935
  - Lag 2: LogL 253.9536; LR 26.89968*; FPE1.77e-09; AIC -11.64768; SC -10.76102; HQ -11.32709*
  - Lag 3: LogL 264.0201; LR 15.09964; FPE1.72e-09*; AIC -11.70100; SC -10.43434; HQ -11.24302
  - Lag 4: LogL 273.5157; LR 12.81905; FPE1.75e-09; AIC -11.72578*; SC -10.07913; HQ -11.13040
  - Note: * indicates lag order selected by the criterion.

- VAR Residual Serial Correlation LM Test (Null Hypothesis: no serial correlation at lag order h; Date: 09/10/14; Sample: 2002Q1 2014Q2; Included observations: 42)
  - Lag 1: LM-Stat 19.22657  Prob 0.0233
  - Lag 2: LM-Stat 10.86233  Prob 0.2853
  - Lag 3: LM-Stat 8.908483  Prob 0.4458
  - Probs from chi-square with 9 df.

- Figure 2 indicates Actual and VAR Model fitted Real GDP Growth (quarter-over-quarter in previous year) with plotted series for Real GDP growth and Real GDP growth fitted (visual fit information provided in figure).

### Auxiliary model: Tajik remittances and Russian construction index
- Model and figure: Tajik Remittance Inflows and Construction Index in Russia, 2004:Q1–2014:Q1; plotted series LOG(REMITTANCES_TJK) and LOG(ICONSTRUCTION_RUS).
- Serial correlation test for remittance model: no evidence of significant serial correlation given Prob. F(4,30)0.4052 and Prob. Chi-Square(4)0.3420.

*REPUBLIC OF TAJIKISTAN — INTERNATIONAL MONETARY FUND*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2021/english/1tjkea2021003.pdf_