## _wp1017

## Source details

**Canonical URL:** [_wp1017](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1017.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1017.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1017.pdf.json)

---

### I. Introduction
- Tajikistan experienced very high inflation during its transition period starting in 1991.  
- Years of civil war (1992–97) complicated macroeconomic management and limited development of monetary policy framework and instruments.  
- Improvements in monetary policy management followed the end of civil strife, but significant gaps remain.  
- Contributing factors to difficulties in controlling inflation: a high level of dollarization, continuation of provision of directed credits to special sectors, a low level of financial intermediation, and underdeveloped financial markets.  
- Given elevated poverty levels, high inflation reduces purchasing power and hurts the most vulnerable groups.  
- Purpose: explain inflation dynamics and forecast inflation using the Vector Error Correction Model (VECM) and compare estimates with the Autoregressive Moving Average Model (ARMA).  
- Data limitations and potential structural instability motivate using monthly and quarterly data starting January 2001.  
- Main finding summary: the nominal exchange rate and, to a lesser extent, interest rate channels exert influence on prices.  
- Policy encouragements: strengthen the interest rate channel, make reserve requirements an effective instrument, resume operations with tradable securities, and foster financial deepening.

### II. Monetary Policy Transmission Channels in Tajikistan
- Channels examined: money supply, income, interest rates, and exchange rates.  
- Observations on channels:
  - Wealth and monetarist channels largely inoperative due to shallow financial markets, directed credits, low monetization, and high dollarization.
  - Monetization: broad money to GDP was 18 percent of GDP in 2009.
  - Money channel: decoupling between inflation and money growth observed from 2003 to 2004; relationship aligns as expected since 2005.
  - Interest rate channel: weak pass-through from NBT policy rate to market rates; disconnect between National Bank of Tajikistan (NBT) and commercial bank rates.
    - Banks set rates according to supply and demand for their own funds because of dollarization and scarce funding.
    - High deposit-lending spreads driven by limited credit risk assessment and mitigation mechanisms, weak competition, and deficiencies in liquidity management tools and infrastructure.
  - Since 2009 the NBT has set the interest rate on the basis of a small spread above inflation.
  - Exchange rate channel: important given imports and dollarization.
    - Imports-to-GDP ratio estimated at 62 percent of 2008 GDP.
    - Large import content of consumption basket, high dollarization, and remittance flows imply significant pass-through from exchange rate to prices.
    - A sudden large devaluation could have a significant inflationary effect given low domestic production capacity and energy shortages.
  - Dollarization:
    - Loan dollarization reported as 60 percent at end-2008 in one mention and 56 percent at end-2008 in another mention (both figures appear in source).
    - Dollarization reduces central bank influence on monetary aggregates, distorts domestic money–inflation link, and leads to asset–liability mismatches.
  - From 2006 until about the beginning of 2009 the NBT maintained a de facto exchange rate peg to the dollar, sustained by large inflows of remittances, foreign direct investment, and aid.
  - Narrow credit channel: supply of loanable funds remains limited despite rapid growth in deposits and lending.
    - At end-2008 commercial banks’ credit constituted 47 percent of their assets.
    - Lending primarily short term (between 1 and 3 years); majority directed towards agricultural lending.
    - Banks hold high excess reserves, reducing sensitivity to monetary policy decisions.

### III. Empirical Investigation — VECM Specification
- Model specification:
  - xt = (pt, mt, yt, et, p*t); where pt is CPI, mt is broad money, yt is real GDP, et is nominal effective exchange rate, p*t is foreign prices proxied by world GDP deflator.
  - Z includes seasonal dummies and (if necessary) time trend.
  - ∆ denotes first difference, ()Lφ denotes lag operator coefficients, δ is co-integration vector(s).
  - Lag length for the VECM selected to be 2 based on the Akaike Information Criterion (AIC).
- Unit root (Augmented Dickey-Fuller) results (Table 1):
  - Variables are I(1) except foreign prices, which is I(2).
  - ADF t-statistics (T-statistics, null hypothesis: unit root):
    - CPI: Level 2.69*; First Difference 5.95***
    - Broad money: Level 1.61; First Difference 6.52***
    - Real GDP: Level 2.44; First Difference 5.24***
    - Nominal exchange rate: Level 1.98; First Difference 4.20***
    - Interest rate: Level 1.69; First Difference 4.72***
    - Foreign prices: Level 1.73; First Difference 1.63
  - Significance notation: * indicates 10 percent, ** indicates 5 percent, and *** indicates 1 percent significance.
- Co-integration (Johansen) results:
  - Tests suggest one co-integration vector.
  - Unrestricted co-integration relation (equation 2): p = 0.09 + 1.11 m + 0.21 y + 3.66 e + p*∆ˆ
  - Restricted co-integration relation (equation 3, theoretical money demand normalization): p = −1.28 m + 5.33 y + 6.11 e + p*∆ˆ
  - Interpretations:
    - Income elasticity (transaction demand for money) is larger than unity and significant.
    - Nominal effective exchange rate depreciation leads to higher domestic prices.
    - International prices have a large and significant long-run impact on Tajik price level.
  - Johansen Trace test (Table 2):
    - Trace Statistic for "None" is 120.79 with 1% Critical Value 77.82 and P-Value 0.00*.
  - Max-Eigen test (Table 2):
    - Max-Eigen Statistic for "None" is 73.74 with 1% Critical Value 39.37 and P-Value 0.00*.
- Weak exogeneity and adjustment:
  - Weak exogeneity tests (Table 3) indicate broad money, real GDP, interest rate, and exchange rate are weakly exogenous to the co-integration relation, suggesting deviations from the long-run relation affect inflation.
  - Speed of adjustment to long-run equilibrium: convergence occurs at a speed of 33 percent per quarter (one-third of discrepancy corrected each period).
- Model restriction tests (Table 3):
  - B(1,2)=1 (unity of coefficient on money): Restricted Log-Likelihood 589.64; LR Statistic 15.76; Degrees of Freedom 1; Probability 0.00.
  - A(1,1)=0: Restricted Log-Likelihood 579.02; LR Statistic 37.00; Degrees of Freedom 1; Probability 0.00.
  - A(2,1)=0: Restricted Log-Likelihood 597.15; LR Statistic 0.75; Degrees of Freedom 1; Probability 0.39.
  - A(3,1)=0: Restricted Log-Likelihood 597.43; LR Statistic 0.18; Degrees of Freedom 1; Probability 0.67.
  - A(4,1)=0: Restricted Log-Likelihood 536.71; LR Statistic 0.07; Degrees of Freedom 1; Probability 0.79.
  - The test rejects the unity hypothesis on the coefficient on money.

### IV. Forecasts and Model Performance (VECM)
- Ex-post out-of-sample forecast:
  - Estimation period: 1999Q1–2007Q4; forecast for 2008Q1–2009Q2.
  - Observed: forecasted inflation decreases faster than actual values during the second half of 2008; actual values sometimes lie outside the 75 percent confidence bands.
  - Possible cause: small sample weakening forecasting ability.
- Dynamic forecast using IMF World Economic Outlook for exogenous international prices for 2009Q3–2010Q4:
  - Model projects inflation in Tajikistan is likely to increase to 6 percent by end-2009 (or two percentage points above the latest observation of end-September).
  - For 2010, CPI inflation is likely to follow a downward trend through the first quarter, driven by the lagged impact of international prices.
  - Average inflation in 2010 is likely to be around [text ends here in source].
- Additional projection fragment in source:
  - "7.5 percent. Toward the end of the year, CPI inflation is projected to pick up slightly to" (text fragment present in source).

### V. ARMA Model and Forecast
- Comparison measures plotted and rationale:
  - Four measures: overall year-on-year CPI inflation, 3-month average year-on-year CPI inflation, 12-month average year-on-year CPI inflation, and core CPI inflation.
  - Forecast focuses on the 3-month average year-on-year CPI inflation to reduce short-term noise while lagging actual inflation less than the 12-month average.
- ARMA specification and selection:
  - General form: ARMA(p, q) with parameters estimated by Maximum Likelihood Estimates (MLE), assuming error term (ε) is white noise and normally distributed.
  - Model selection steps: (i) test autocorrelation of residuals using Breusch-Godfrey Lagrange Multiplier test; (ii) test residual distribution using Jarque-Bera test; (iii) choose model with smallest SIC and AIC among models that satisfy (i) and (ii).
  - Table of 42 ARMA models indicates ARMA(2,2) and ARMA(5,5) are best candidates.
  - ARMA(2,2) has the lower SIC; ARMA(5,5) has the lower AIC.
  - Analysis is based on ARMA(2,2); ARMA(5,5) discussed in appendix.
- Fit and out-of-sample performance:
  - Adjusted R2 suggests that over 99 percent of the fluctuations in inflation can be explained by the estimated ARMA model.
  - Standard error of regression is less than one percentage point, implying the 95 percent confidence interval of the model during the sample period is at most plus/minus 1.5 percentage points from the inflation based on the estimated model.
  - ARMA(2,2) estimated using a smaller sample (through June 2008) produced a forecast for July 2008–June 2009 in which actual inflation stays within the 75 percent confidence interval and tracks forecasted inflation well.
  - The ARMA model ex-post, out-of-sample forecast outperforms the forecast of the VECM.
- ARMA forecast results:
  - 3-month average inflation will be at 6 percent by end-2009.
  - Inflation will gradually increase during 2010 to reach 12 percent in December 2010.
  - Earlier in text: CPI inflation projected to pick up slightly to 10.6 percent (toward the end of the year), linking trend to international prices and projected depreciation of the Tajik currency in 2010.

### VI. Empirical Findings (VECM and Comparative Results)
- VECM findings:
  - Broad money growth is highly significant in determining inflation in both the short and long terms.
  - Real GDP is an important determinant of inflation, especially in the long run.
  - Policy interest (refinancing) rate does not influence inflation because it is decoupled from the market rate and does not affect financial decisions of economic agents and the supply of credit.
  - Exchange rate has a strong impact on prices, supported by evidence from other dollarized economies.
  - Domestic price level is significantly affected by movement in international prices.
  - Estimated magnitude of the error correction term is 33 percent; any disequilibrium in the price level is likely to be eliminated within nine months.
- Comparative model forecasts:
  - Both VECM and ARMA show strong forecasting explanatory power during the sample period.
  - End-period inflation projections:
    - Projected to reach 6–9 percent in 2009.
    - Projected to increase to 10.5–12 percent in 2010.
  - These projections are consistent with observed exchange rate depreciation in 2009, expected to carry over to 2010.

### VII. Policy Recommendations
- Strengthen the interest rate channel:
  - Adjust policy interest rates to a level attractive to financial market participants to help manage credit provision (broad money) and influence aggregate demand and inflation over time.
  - Strengthen the NBT’s financial position and improve transparency to become a trustworthy counterpart to the private sector.
  - Note: The NBT’s plan to set up a deposit standing facility is cited as a step in this direction.
- Ensure reserve requirements become an effective instrument:
  - Reserve requirements can be a powerful prudential tool if set to reflect current monetary conditions and banking developments.
  - Recommendation: unify the treatment of both types of deposits and strictly enforce compliance.
- Resume regular operations with tradable securities (e.g., government bonds or central bank bills):
  - Use securities to mop up excess liquidity and serve as collateral for bank borrowing.
  - Gradually establish a reference yield curve to facilitate pricing of credit risk and develop a broader securities market.
  - Promote somoni-denominated medium- and long-term securities (private and government) to increase investment opportunities, hedge foreign exchange exposure, and remove structural liquidity (recognizing this will take time).
- Encourage financial deepening:
  - Promote deposit of remittances in the banking system; NBT could foster competition.
  - Introduce new products (e.g., payment cards) to reduce cash handling costs and enable remittance beneficiaries to hold deposit accounts.
  - Develop awareness campaigns, education mechanisms, and incentives to encourage remittance recipients to use these products.
  - NBT could foster competition among money transfer operators by increasing market transparency (e.g., publishing comparative data on pricing and service alternatives).
- Strengthen the bank lending channel:
  - Continue financial sector reforms and encourage foreign bank entry.
  - Strengthen corporate accounting standards and governance, bankruptcy procedures, use of collateral, and creditor rights.

### VIII. Key Statistics and Data Coverage
- Transition period start: 1991.  
- Civil war period: 1992–97.  
- Sample period used for modeling: monthly and quarterly data starting January 2001; VECM lag length 2.  
- Monetization (broad money to GDP): 18 percent of GDP in 2009.  
- Imports-to-GDP ratio: estimated at 62 percent of 2008 GDP.  
- Loan dollarization: 60 percent at end-2008 (one mention); 56 percent at end-2008 (another mention).  
- Commercial banks’ credit constituted 47 percent of their assets at end-2008.  
- VECM convergence speed: 33 percent per quarter.  
- Forecasted inflation: 6 percent by end-2009 (model projection); ARMA projects 12 percent in December 2010; projections for 2009 overall 6–9 percent and for 2010 10.5–12 percent.

*Source: Excerpts from the IMF working paper content unit _wp1017 - References.*

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

### _wp1017 - References

### I. Introduction
- Tajikistan experienced very high inflation during its transition period starting in 1991.  
- Years of civil war (1992–97) complicated macroeconomic management and limited development of monetary policy framework and instruments.  
- Improvements in monetary policy management followed the end of civil strife, but significant gaps remain.  
- Contributing factors to difficulties in controlling inflation: a high level of dollarization, continuation of provision of directed credits to special sectors, a low level of financial intermediation, and underdeveloped financial markets.  
- Given elevated poverty levels, high inflation reduces purchasing power and hurts the most vulnerable groups.  
- The paper attempts to explain inflation dynamics and forecast inflation using the Vector Error Correction Model (VECM) and compares estimates with the Autoregressive Moving Average Model (ARMA).  
- Data limitations and potential structural instability motivate using monthly and quarterly data starting January 2001.  
- Main finding summary: the nominal exchange rate and, to a lesser extent, interest rate channels exert influence on prices.  
- Policy encouragements: strengthen the interest rate channel, make reserve requirements an effective instrument, resume operations with tradable securities, and foster financial deepening.

### II. Monetary Policy Transmission Channels in Tajikistan
- The analysis examines four transmission channels: money supply, income, interest rates, and exchange rates (Exhibit 1).  
- Shallow financial markets, directed credits, low monetization, and high dollarization render wealth and monetarist channels largely inoperative.  
- Monetization: broad money to GDP was 18 percent of GDP in 2009.  
- Money channel: decoupling between inflation and money growth observed from 2003 to 2004; relationship aligns as expected since 2005 (Figure 1).  
- Expanded money definitions that include foreign cash holdings would better capture dollarization effects; such data are not available in Tajikistan.  
- Interest rate channel: weak pass-through from NBT policy rate to market rates; disconnect between National Bank of Tajikistan (NBT) and commercial bank rates (Figure 2).  
  - Banks set rates according to supply and demand for their own funds because of dollarization and scarce funding.  
  - High deposit-lending spreads driven by limited credit risk assessment and mitigation mechanisms, weak competition, and deficiencies in liquidity management tools and infrastructure (2008, IMF).  
- Since 2009 the NBT has set the interest rate on the basis of a small spread above inflation. Possible motivations: treating interest rates as a “growth” policy tool and NBT’s weak financial position limiting ability to offer market-close rates.  
- Exchange rate channel: important in open economies; Tajikistan’s high imports-to-GDP ratio (estimated at 62 percent of 2008 GDP), large import content of consumption basket, high dollarization, and remittance flows suggest significant pass-through from exchange rate to prices.  
  - A sudden large devaluation could have a significant inflationary effect given low domestic production capacity and energy shortages.  
- Dollarization: both payments and financial dollarization present; loan dollarization stood at 60 percent at end-2008 and 56 percent at end-2008 in different mentions (text notes loan dollarization stood at 60 percent at end-2008 and later states loan dollarization stood at 56 percent—both figures appear in source).  
  - Dollarization reduces central bank influence on monetary aggregates, distorts domestic money–inflation link, and leads to asset–liability mismatches.  
- From 2006 until about the beginning of 2009 the NBT maintained a de facto exchange rate peg to the dollar, sustainable due to large inflows of remittances, foreign direct investment, and aid.  
- Narrow credit channel: despite rapid growth in deposits and lending, supply of loanable funds remains limited; at end-2008 commercial banks’ credit constituted only 47 percent of their assets; lending primarily short term (between 1 and 3 years); majority of lending directed towards agricultural lending.  
- Banks hold high excess reserves, reducing sensitivity to monetary policy decisions.

### III. Empirical Investigation — VECM Specification
- VECM specification used: xt = (pt, mt, yt, et, p*t); pt is CPI, mt is broad money, yt is real GDP, et is nominal effective exchange rate, p*t is foreign prices proxied by world GDP deflator.  
- Z includes seasonal dummies and (if necessary) time trend.  
- ∆ denotes first difference, ()Lφ denotes lag operator coefficients, δ is co-integration vector(s).  
- Lag length for the VECM selected to be 2 based on the Akaike Information Criterion (AIC).  

### Unit Root Tests (Augmented Dickey-Fuller)
- ADF t-test results (Table 1) show variables are integrated of order one, I(1), except foreign prices, which is I(2).  
- The VECM uses levels for variables except foreign prices, which enters as a first difference.  
- Table 1 ADF t-statistics (T-statistics, null hypothesis: unit root):  
  - CPI: Level 2.69*; First Difference 5.95***  
  - Broad money: Level 1.61; First Difference 6.52***  
  - Real GDP: Level 2.44; First Difference 5.24***  
  - Nominal exchange rate: Level 1.98; First Difference 4.20***  
  - Interest rate: Level 1.69; First Difference 4.72***  
  - Foreign prices: Level 1.73; First Difference 1.63  
- Significance notation: * indicates 10 percent, ** indicates 5 percent, and *** indicates 1 percent significance.

### Co-Integration Tests (Johansen)
- Johansen co-integration tests (Table 2) suggest one co-integration vector.  
- Unrestricted co-integration relation (equation 2): p = 0.09 + 1.11 m + 0.21 y + 3.66 e + p*∆ˆ  
- Restricted co-integration relation (theoretical money demand normalization, equation 3): p = −1.28 m + 5.33 y + 6.11 e + p*∆ˆ  
- Interpretation: income elasticity (transaction demand for money) is larger than unity and significant; nominal effective exchange rate depreciation leads to higher domestic prices; international prices have a large and significant long-run impact on Tajik price level.  
- Johansen Trace test (Table 2): Trace Statistic for "None" is 120.79 with 1% Critical Value 77.82 and P-Value 0.00* (rejects null at 0.01 level).  
- Max-Eigen test (Table 2): Max-Eigen Statistic for "None" is 73.74 with 1% Critical Value 39.37 and P-Value 0.00* (rejects null at 0.01 level).  
- Weak exogeneity tests (Table 3) indicate broad money, real GDP, interest rate, and exchange rate are weakly exogenous to the co-integration relation, suggesting deviations from the long-run relation affect inflation.  
- Speed of adjustment to long-run equilibrium: convergence occurs at a speed of 33 percent per quarter (one-third of discrepancy corrected each period).

### Model Restriction and Weak Exogeneity Tests (Table 3)
- B(1,2)=1 (test of unity of coefficient on money): Restricted Log-Likelihood 589.64; LR Statistic 15.76; Degrees of Freedom 1; Probability 0.00.  
- A(1,1)=0: Restricted Log-Likelihood 579.02; LR Statistic 37.00; Degrees of Freedom 1; Probability 0.00.  
- A(2,1)=0: Restricted Log-Likelihood 597.15; LR Statistic 0.75; Degrees of Freedom 1; Probability 0.39.  
- A(3,1)=0: Restricted Log-Likelihood 597.43; LR Statistic 0.18; Degrees of Freedom 1; Probability 0.67.  
- A(4,1)=0: Restricted Log-Likelihood 536.71; LR Statistic 0.07; Degrees of Freedom 1; Probability 0.79.  
- The test rejects the unity hypothesis on the coefficient on money.

### Forecasts and Model Performance
- Ex-post out-of-sample forecast: estimation on 1999Q1–2007Q4, forecast for 2008Q1–2009Q2 (Figure 3).  
  - Observed pattern: forecasted inflation decreases faster than actual values during the second half of 2008; actual values sometimes lie outside the 75 percent confidence bands. Possible cause: small sample weakening forecasting ability.  
- Dynamic forecast using IMF World Economic Outlook for exogenous international prices for 2009Q3–2010Q4:  
  - Model projects inflation in Tajikistan is likely to increase to 6 percent by end-2009 (or two percentage points above the latest observation of end-September).  
  - For 2010, CPI inflation is likely to follow a downward trend through the first quarter, driven by the lagged impact of international prices. Average inflation in 2010 is likely to be around [text ends here in source].

### Key Statistics and Data Coverage
- Transition period start: 1991.  
- Civil war period: 1992–97.  
- Sample period used for modeling: monthly and quarterly data starting January 2001; VECM lag length 2.  
- Monetization (broad money to GDP): 18 percent of GDP in 2009.  
- Imports-to-GDP ratio: estimated at 62 percent of 2008 GDP.  
- Loan dollarization: 60 percent at end-2008 (one mention); 56 percent at end-2008 (another mention).  
- Commercial banks’ credit constituted 47 percent of their assets at end-2008.  
- VECM convergence speed: 33 percent per quarter.  
- Forecasted inflation: 6 percent by end-2009 (model projection).

### Policy Recommendations (from the source)
- Strengthen the interest rate channel.  
- Make reserve requirements an effective instrument.  
- Resume operations with tradable securities.  
- Foster financial deepening.  

*Source: Excerpts from the IMF working paper content unit _wp1017 - References.*

### 7.5 percent. Toward the end of the year, CPI inflation is projected to pick up slightly to

### _wp1017 - 7.5 percent. Toward the end of the year, CPI inflation is projected to pick up slightly to

### ARMA model and forecast
- Four measures plotted and compared: overall year-on-year CPI inflation, 3-month average year-on-year CPI inflation, 12-month average year-on-year CPI inflation, and core CPI inflation.
- Rationale for choice: 3-month average year-on-year CPI inflation reduces short-term noise while lagging actual inflation much less than the 12-month average; hence the forecast focuses on the 3-month average inflation.
- ARMA specification:
  - General form: ARMA(p, q) with parameters estimated by Maximum Likelihood Estimates (MLE), assuming error term (ε) is white noise and normally distributed.
  - Model selection steps: (i) test autocorrelation of residuals using Breusch-Godfrey Lagrange Multiplier test; (ii) test residual distribution using Jarque-Bera test; (iii) choose model with smallest SIC and AIC among models that satisfy (i) and (ii).
- Model selection outcome:
  - Table of 42 ARMA models indicates ARMA(2,2) and ARMA(5,5) are the best candidates.
  - ARMA(2,2) shows the lower SIC value; ARMA(5,5) shows the lower AIC value.
  - Analysis is based on ARMA(2,2); results for ARMA(5,5) discussed in appendix.
- Fit and out-of-sample performance:
  - The adjusted R2 suggests that over 99 percent of the fluctuations in inflation can be explained by the estimated ARMA model.
  - Standard error of regression is less than one percentage point, implying the 95 percent confidence interval of the model during the sample period is at most plus/minus 1.5 percentage points from the inflation based on the estimated model.
  - ARMA(2,2) estimated using a smaller sample (through June 2008) produced a forecast for July 2008–June 2009 in which actual inflation stays within the 75 percent confidence interval and tracks forecasted inflation well.
  - The ARMA model ex-post, out-of-sample forecast outperforms the forecast of the VECM.

- ARMA forecast results:
  - 3-month average inflation will be at 6 percent by end-2009.
  - Inflation will gradually increase during 2010 to reach 12 percent in December 2010.
  - Earlier in text: CPI inflation projected to pick up slightly to 10.6 percent (toward the end of the year), linking trend to international prices and projected depreciation of the Tajik currency in 2010.

### Empirical findings (VECM and comparative results)
- VECM results:
  - Broad money growth is highly significant in determining inflation in both the short and long terms.
  - Real GDP is an important determinant of inflation, especially in the long run.
  - Policy interest (refinancing) rate does not influence inflation because it is decoupled from the market rate and does not affect financial decisions of economic agents and the supply of credit.
  - Exchange rate has a strong impact on prices, supported by evidence from other dollarized economies.
  - Domestic price level is significantly affected by movement in international prices.
  - Estimated magnitude of the error correction term is 33 percent; any disequilibrium in the price level is likely to be eliminated within nine months.
- Comparative model forecasts:
  - Both VECM and ARMA show remarkable forecasting explanatory power during the sample period.
  - End-period inflation projections:
    - Projected to reach 6–9 percent in 2009.
    - Projected to increase to 10.5–12 percent in 2010.
  - These projections are consistent with observed exchange rate depreciation in 2009, expected to carry over to 2010.

### Current state of monetary policy management and policy recommendations
- Current constraints and observations:
  - NBT’s (National Bank of Tajikistan) ability to control inflation is quite limited.
  - Available instruments such as reserve requirements and a refinancing rate have proven ineffective.
  - Most banks comply with reserve requirements of 9 and 7 percent for non-resident and resident deposits, respectively.
  - Excess liquidity was observed in the banking system prior to the 2009 financial crisis.
  - The refinancing rate has not been an influential tool in directing money supply.
  - The only effective instrument currently employed by the NBT to control the money supply is transactions with foreign currencies.
- Steps to improve the monetary policy toolkit:
  - Strengthen the interest rate channel:
    - Adjust policy interest rates to a level attractive to financial market participants to help manage credit provision (broad money) and influence aggregate demand and inflation over time.
    - Strengthen the NBT’s financial position and improve transparency to become a trustworthy counterpart to the private sector.
    - Note: The NBT’s plan to set up a deposit standing facility is a step in this direction.
  - Ensure reserve requirements become an effective instrument:
    - Reserve requirements can be a powerful prudential tool if set to reflect current monetary conditions and banking developments.
    - Past practice: NBT changed reserve requirement regulations multiple times and they remain differentiated.
    - Recommendation: unify the treatment of both types of deposits and strictly enforce compliance.
  - Resume regular operations with tradable securities (e.g., government bonds or central bank bills):
    - Use securities to mop up excess liquidity and serve as collateral for bank borrowing.
    - Gradually establish a reference yield curve to facilitate pricing of credit risk and develop a broader securities market.
    - Promote somoni-denominated medium- and long-term securities (private and government) to increase investment opportunities, hedge foreign exchange exposure, and remove structural liquidity (recognizing this will take time).
  - Encourage financial deepening:
    - Despite rapid growth in credit and deposits, the banking sector remains rudimentary and cash transactions are still preferred.
    - Banks could promote deposit of remittances in the banking system; NBT could foster competition.
    - Introduce new products (e.g., payment cards) to reduce cash handling costs and enable remittance beneficiaries to hold deposit accounts.
    - These initiatives would help authorities better gauge the “true” state of the monetary base (including foreign holdings) and improve monetary policy management.
    - Note: It is indispensable to develop awareness campaigns, education mechanisms, and incentives to encourage remittance recipients to use these products. The NBT could foster competition among money transfer operators by increasing market transparency (e.g., publishing comparative data on pricing and service alternatives).
  - Strengthen the bank lending channel:
    - Continue financial sector reforms and encourage foreign bank entry.
    - Further efforts needed to strengthen corporate accounting standards and governance, bankruptcy procedures, use of collateral, and creditor rights.

*Sources: Tajik authorities; and authors' estimates and forecast.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1017.pdf_