## 6. Out-of-sample Forecast of Select Indicators, June 2003-December 2003

## Source details

**Canonical URL:** [6. Out-of-sample Forecast of Select Indicators, June 2003-December 2003](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp04229.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp04229.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp04229.pdf.json)

---

### Introduction and data
- Scope: builds a three-market macroeconomic model of Armenia (foreign exchange, money, labor) to analyze determinants of inflation using cointegration and a vector error correction model (VECM).
- Sample period used for estimation: January 1996 to June 2003.
- Data sources: Central Bank of Armenia; International Monetary Fund’s International Financial Statistics database.
- Data limitations noted: issues in external and real sector coverage, timeliness, internal consistency, and periodicity.

### Monetary policy background and stylized facts
- Historical inflation and stabilization:
  - Hyperinflation of 1993–94 (about 5,000 percent in 1994) followed by stabilization.
  - Since 1998 the economy operated in a low single-digit inflation environment.
- Stages of monetary policy development:
  - Stage 1: stabilization with exchange rate as anchor, early 1990s.
  - Stage 2: 1995–2002 — focus on price stability, development of market instruments, liberalization.
  - Stage 3 (beginning): 2003 — research on possible switch to inflation targeting.
- Key structural constraints on monetary transmission:
  - High dollarization (figures and charts reported in paper).
  - Narrow credit base and low financial intermediation.
  - Large inflows of external financing (budgetary and private), strong role of foreign reserves and FX interventions.
  - Large share of imported goods → strong exchange rate pass-through.
  - Monopolistic and administered prices important.
- Treasury bill yields and market narrowness:
  - Annualized treasury bill yields declined from an average of 50 percent to around 20 percent in 2000 and to approximately 12 percent in 2003.
  - Government reliance on external concessional financing persisted; treasury bill market remained narrow and non-liquid.

### Model setup and identification
- Model form: VAR with k lags transformed to an error correction form.
- Variables (log or levels as indicated):
  - e: (log) nominal exchange rate (drams per U.S. dollar)
  - p: (log) domestic CPI
  - p*: (log) foreign (U.S.) price level
  - m: (log) narrow money (currency + demand deposits)
  - i: average nominal rate on time deposits (measured in fractions)
  - tot: terms of trade index
  - y: (log) real GDP (1995 drams)
  - ydev: deviation of real GDP from potential (HP filter)
  - w: (log) average nominal wage
- Deterministic components: constants, trends, 11 monthly dummies; a dummy for bank liquidation beginning November 2000 was tested and dropped as insignificant.
- Lag selection: shortest lag length producing autocorrelation-free residuals; two lags used.

### Data properties and unit roots
- Augmented Dickey-Fuller test results (levels p-values reported):
  - e: 0.45 (lags included 0)
  - p: 0.13 (lags included 1)
  - m: 0.05 (lags included 0)
  - p*: 0.62 (lags included 2)
  - tot: 0.04 (lags included 1)
  - i: 0.76 (lags included 4)
  - y: 0.99 (lags included 11)
  - ydev: 0.85 (lags included 9; test with no intercept or trend)
- First differences: p-values all reported as 0.00 for the listed series (lags indicated).

### Long-run cointegrating relationships
- Foreign exchange market (cointegrating relation, Equation (6)):
  - e + 7.07 p* − 0.007 tot − 0.0001 ydev + 0.004 trend + ... = p
  - Reported t-statistics where shown: (15.2) for p*, (-1.51) for tot, (1.50) for trend.
  - Interpretation: real exchange rate in long run equals 1 when adjusted for terms-of-trade and productivity shocks.
- Money market (Equation (7)):
  - m − p = −6.25 + 9.26 y − 1.269 i − 0.01 trend
  - Reported t-statistics: (10.3) on y, (-10.4) on i, (-7.6) on intercept terms as reported.
  - Interpretation: real money balances positively related to income, negatively to nominal interest rate; trend negative interpreted as financial innovations reducing money demand.
  - Note: income coefficient 9.26 much greater than monetarist/Baumol-Tobin benchmarks; possible explanations include remittances/transfers or cash-in-advance features.
- Labor market (Equation (8)):
  - w = 14.586 − 1.001 p + 3.003 ydev − 0.004 trend
  - Reported t-statistics: (6.02) for p, (7.04) for ydev, (-0.92) for intercept consistency note.
  - Interpretation: wages highly sensitive to price movements; measurement issues or nontradable-sector wage volatility may explain large coefficient.

### Short-run dynamics, weak exogeneity, and estimation results
- Error correction terms:
  - ecm(ppp), ecm(m), ecm(w) constructed from long-run relations; plotted and described as well behaved.
- Weak exogeneity tests (null: weak exogeneity; chi-square, df=3, p-values):
  - e: chi-square 14.3, p-value 0.00 → rejects weak exogeneity
  - m: chi-square 6.3, p-value 0.10 → cannot reject weak exogeneity at 5 percent (money appears weakly exogenous)
  - w: chi-square 9.8, p-value 0.02 → rejects weak exogeneity
  - p: chi-square 41.4, p-value 0.00 → rejects weak exogeneity
  - p*: chi-square 5.5, p-value 0.14 → p* weakly exogenous
  - tot: chi-square 2.6, p-value 0.46 → tot weakly exogenous
  - y: chi-square 21.3, p-value 0.00 → rejects weak exogeneity
  - ydev: chi-square 50.0, p-value 0.00 → rejects weak exogeneity
  - i: chi-square 29.1, p-value 0.00 → rejects weak exogeneity
- Short-run estimation highlights:
  - Prices adjust strongly to foreign exchange disequilibria: 0.18 (18 percent) of the deviation adjusted in one month (implying about six months for full adjustment).
  - Adjustment to money disequilibrium in prices is much smaller and statistically insignificant (coefficient on mecm small).
  - No evidence of above-productivity wage pressures being passed to consumers in long run: wage error correction coefficient small and highly insignificant.
  - Strong price inertia: coefficient on once-lagged change in price equals 0.40 and is highly significant.
  - Short-run responsiveness of prices to changes in money: once-lagged change in money positive and significant (coefficient 0.06** reported).
  - Money appears to have no effect on the real exchange rate: coefficients on money in price and exchange rate equations have equal magnitudes (neutrality of money in trade account movements).
  - Short-run coefficient on lagged exchange rate in price equation negative (discussed as possible overshooting explanation).
  - Interest rate and terms of trade: no short-term impact on prices, exchange rate, and wages (insignificant).
- Model diagnostics:
  - Jarque-Bera normality test p-values for residuals: e 0.188, m 0.911, w 0.084, p 0.753 → residuals appear normally distributed at 5 percent confidence level.
  - Stability: recursive coefficient estimates reported and discussed; in-sample fit illustrated.

### Out-of-sample forecast performance (June 2003–December 2003)
- Forecast horizon: June 2003–December 2003.
- Forecast performance summary:
  - Exchange rate and price level: model captured non-monotonic patterns with reasonable accuracy (visual fit in figures).
  - Narrow money: relatively poor predictions — may be due to money’s weak exogeneity.
  - Wages: forecasts presented alongside actuals (figures in source).

### Policy implications and recommendations
- Constraints on monetary policy:
  - Exchange rate pass-through to prices is strong.
  - Heavy dependence of the monetary base on external inflows (remittances and capital inflows).
  - Limited ability to sterilize due to small portfolio of domestic money market instruments.
  - De facto preference for exchange rate stability via reserve accumulation, which conflicts with formal policy emphasis on monetary aggregates and price stability.
- Implications for potential switch to inflation targeting:
  - With virtually no pass-through from money to prices and a strong pass-through from exchange rate to prices, the ability to target prices via monetary aggregates is limited.
  - Recommendation: in coordination with the government, develop and accumulate more money market instruments to mitigate policy dependence on external flows and enhance control over monetary aggregates and price movements.
- Suggested areas for subsequent research:
  - Employ alternative income measures incorporating transfers and remittances.
  - Model elements of high dollarization and their impact on transmission channels.
  - Use trade-weighted foreign price index and nominal exchange rate for foreign exchange market dynamics.

*Source: _wp04229 - References; https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp04229.pdf*

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

### _wp04229 - References..............................................................................................................

### Tables
- 1. Stationarity Tests .................................................................................................................11
- 2. Cointegration Analysis of the Full Model ...........................................................................14
- 3. Weak Exogeneity Tests........................................................................................................17
- 4. Short-Run Specifications Based on a Full Information Model............................................21

### Figures
- 1. Armenia: Money and Credit Indicators, 1995-2003..............................................................8
- 2. The Structure of Error Correction Terms.............................................................................20
- 3. Recursive Estimates of Coefficients in the Price Equation  1/ ............................................22
- 4. Actual versus Fitted Values of Selected Indicators..............................................................23
- 5. Overshooting of Prices as a Result of a One-Step Increase in the Exchange Rate..............24

*Source: _wp04229 - References; https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp04229.pdf*

### 6. Out-of-sample Forecast of Select Indicators, June 2003-December 2003 ..........................25

### 6. Out-of-sample Forecast of Select Indicators, June 2003-December 2003

### Introduction
- Scope: builds a three-market macroeconomic model of Armenia (foreign exchange, money, labor) to analyze determinants of inflation using cointegration and a vector error correction model (VECM).
- Sample period used for estimation: January 1996 to June 2003.
- Data sources: Central Bank of Armenia; International Monetary Fund’s International Financial Statistics database.
- Data limitations noted: issues in external and real sector coverage, timeliness, internal consistency, and periodicity.

### Monetary policy and price developments — background and stylized facts
- Armenia experienced hyperinflation of 1993–94 (about 5,000 percent in 1994) and subsequent stabilization; since 1998 the economy operated in a low single-digit inflation environment.
- Stages of monetary policy development:
  - Stage 1: stabilization with exchange rate as anchor, early 1990s.
  - Stage 2: 1995–2002 — focus on price stability, development of market instruments, liberalization.
  - Stage 3 (beginning): 2003 — research on possible switch to inflation targeting.
- Key structural constraints on monetary transmission:
  - High dollarization (figures and charts reported in paper).
  - Narrow credit base and low financial intermediation.
  - Large inflows of external financing (budgetary and private), strong role of foreign reserves and FX interventions.
  - Large share of imported goods → strong exchange rate pass-through.
  - Monopolistic and administered prices important.
- Treasury bill yields and market narrowness:
  - Annualized treasury bill yields declined from an average of 50 percent to around 20 percent in 2000 and to approximately 12 percent in 2003.
  - Government reliance on external concessional financing persisted; treasury bill market remained narrow and non-liquid.

### Model setup and identification
- Overall model: VAR with k lags transformed to an error correction form (Equations (1) and (2) in source).
- Variables (log or levels as indicated):
  - e: (log) nominal exchange rate (drams per U.S. dollar)
  - p: (log) domestic CPI
  - p*: (log) foreign (U.S.) price level
  - m: (log) narrow money (currency + demand deposits)
  - i: average nominal rate on time deposits (measured in fractions)
  - tot: terms of trade index
  - y: (log) real GDP (1995 drams)
  - ydev: deviation of real GDP from potential (HP filter)
  - w: (log) average nominal wage
- Deterministic components: constants, trends, 11 monthly dummies; a dummy for bank liquidation beginning November 2000 was tested and dropped as insignificant.
- Lag selection: shortest lag length producing autocorrelation-free residuals; two lags used.

### Data properties and unit roots
- Augmented Dickey-Fuller test results (levels p-values reported in source):
  - e: 0.45 (lags included 0)
  - p: 0.13 (lags included 1)
  - m: 0.05 (lags included 0)
  - p*: 0.62 (lags included 2)
  - tot: 0.04 (lags included 1)
  - i: 0.76 (lags included 4)
  - y: 0.99 (lags included 11)
  - ydev: 0.85 (lags included 9; test with no intercept or trend)
- First differences: p-values all reported as 0.00 for the listed series (lags indicated).

### Long-run cointegrating relationships (results)
- Foreign exchange market long-run relationship (restricted coefficients on p and p* to -1 and 1):
  - Equation (6) (cointegrating vector reported):
    - e + 7.07 p* − 0.007 tot − 0.0001 ydev + 0.004 trend + ... = p
    - Coefficient t-statistics (where reported): (15.2) for p*, (-1.51) for tot, (1.50) for trend.
  - Interpretation: real exchange rate in long run equals 1 when adjusted for terms-of-trade and productivity shocks.
- Money market long-run relationship (restricted homogeneity in prices):
  - Equation (7):
    - m − p = −6.25 + 9.26 y − 1.269 i − 0.01 trend
    - t-statistics: (10.3) on y, (-10.4) on i, (-7.6) on intercept terms as reported.
  - Interpretation: well-behaved money demand — real money balances positively related to income, negatively to nominal interest rate; trend negative interpreted as financial innovations reducing money demand.
  - Note: income coefficient (9.26) much greater than monetarist/Baumol-Tobin benchmarks; possible explanations include remittances/transfers or cash-in-advance features.
- Labor market long-run relationship:
  - Equation (8):
    - w = 14.586 − 1.001 p + 3.003 ydev − 0.004 trend (coefficients and t-statistics reported)
    - Reported t-statistics: (6.02) for p, (7.04) for ydev, (-0.92) for intercept consistency note.
  - Interpretation: wages highly sensitive to price movements (one percent increase in prices leads to a five percent increase in nominal wage in specification); measurement issues or nontradable-sector wage volatility may explain large coefficient.

### Short-run structure, weak exogeneity, and estimation results
- Error correction terms constructed from long-run relations; ecm(ppp), ecm(m), ecm(w) plotted and described as well behaved.
- Weak exogeneity tests (null: presence of weak exogeneity); chi-square, df=3, p-values:
  - e: chi-square 14.3, p-value 0.00 → rejects weak exogeneity
  - m: chi-square 6.3, p-value 0.10 → cannot reject weak exogeneity at 5% (money appears weakly exogenous)
  - w: chi-square 9.8, p-value 0.02 → rejects weak exogeneity
  - p: chi-square 41.4, p-value 0.00 → rejects weak exogeneity
  - p*: chi-square 5.5, p-value 0.14 → p* weakly exogenous
  - tot: chi-square 2.6, p-value 0.46 → tot weakly exogenous
  - y: chi-square 21.3, p-value 0.00 → rejects weak exogeneity
  - ydev: chi-square 50.0, p-value 0.00 → rejects weak exogeneity
  - i: chi-square 29.1, p-value 0.00 → rejects weak exogeneity
- Short-run estimation highlights (from Table 4 and narrative):
  - Prices adjust strongly to foreign exchange disequilibria: 0.18 (18 percent) of the deviation adjusted in one month (implying about six months for full adjustment).
  - Adjustment to money disequilibrium in prices is much smaller and statistically insignificant (coefficient on mecm small).
  - No evidence of above-productivity wage pressures being passed to consumers in long run: wage error correction coefficient small and highly insignificant.
  - Strong price inertia: coefficient on once-lagged change in price equals 0.40 and is highly significant.
  - Short-run responsiveness of prices to changes in money: once-lagged change in money positive and significant (coefficient 0.06** reported in Table 4).
  - Money appears to have no effect on the real exchange rate: coefficients on money in price and exchange rate equations have equal magnitudes (neutrality of money in trade account movements).
  - Short-run coefficient on lagged exchange rate in price equation negative (discussed as possible overshooting explanation).
  - Interest rate and terms of trade: no short-term impact on prices, exchange rate, and wages (insignificant).
- Model diagnostics:
  - Jarque-Bera normality test for residuals (equations e, m, w, p) — p-values: e 0.188, m 0.911, w 0.084, p 0.753 → residuals appear normally distributed at 5 percent confidence level.
  - Stability: recursive coefficient estimates reported and discussed; in-sample fit illustrated.

### Out-of-sample forecast performance (June 2003–December 2003)
- The model produced out-of-sample forecasts that reflected non-monotonic dynamics of exchange rate and prices in the second half of 2003 (Figure 6).
- Forecast accuracy:
  - Exchange rate and price level: model captured non-monotonic patterns with reasonable accuracy (visual fit in figures).
  - Narrow money: relatively poor predictions — may be due to money’s weak exogeneity.
  - Wages: forecasts presented alongside actuals (figures in source).

### Policy implications and recommendations
- Exchange rate pass-through to prices is strong; monetary policy in Armenia is constrained by:
  - Heavy dependence of the monetary base on external inflows (remittances and capital inflows).
  - Limited ability to sterilize due to small portfolio of domestic money market instruments.
  - De facto preference for exchange rate stability due to reserve accumulation to avoid nominal appreciation, which conflicts with formal policy emphasis on monetary aggregates and price stability.
- Implications for potential switch to inflation targeting:
  - With virtually no pass-through from money to prices and a strong pass-through from exchange rate to prices, the ability to target prices via monetary aggregates is limited.
  - Recommendation: in coordination with the government, develop and accumulate more money market instruments to mitigate policy dependence on external flows and enhance control over monetary aggregates and price movements.
- Suggested areas for subsequent research:
  - Employ alternative income measures incorporating transfers and remittances.
  - Model elements of high dollarization and their impact on transmission channels.
  - Use trade-weighted foreign price index and nominal exchange rate for foreign exchange market dynamics.

*Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp04229.pdf*

### References

### _wp04229 - References

### References
- Billmeier, A. and L. Bonato, 2002, “Exchange Rate Pass-Through and Monetary Policy in Croatia,” IMF Working Paper 02/109, Washington: International Monetary Fund.
- Calvo, Guillermo, and Carmen Rainhart, 2000, “Fixing for Your Life,” NBER Working Paper No. 8006, Cambridge, Massachusetts: National Bureau of Economic Research.
- de Melo, Martha, and Cevdet Denizer, 1997, “Monetary Policy During Transition: An Overview,” Policy Research Working Paper No. 1706, Washington: World Bank.
- Duguay P., 1994, “Empirical Evidence on the Strength of the Monetary Transmission Mechanism in Canada: An Aggregate Approach,” Journal of Monetary Economics, Vol. 33, No. 1.
- Grigorian, David, 2003, “Banking Sector in Armenia: How to Turn a Basket Case Into a Beauty Case?” Armenian Journal of Public Policy, Vol. 1, No. 1.
- Hunt, Benjamin, and Peter Isard, 2003, “Some Implications for Monetary Policy of Uncertain Exchange Rate Pass-Through,” IMF Working Paper 03/25, Washington: International Monetary Fund.
- Johansen, Soren, 1992, “Weak Exogeneity and Cointegration in the U.K. Money,” Journal of Policy Modeling, Vol. 14, No. 3.
- Kuijs, Louis, 2002. “Monetary Policy Transmission and Inflation in Slovakia,” IMF Working Paper 02/80, Washington: International Monetary Fund.
- Meltzer, A., 1995, “Monetary Credit and (Other) Transmission Processes: A Monetarist Perspective,” Journal of Economic Perspectives, Vol.9, No. 4.
- McCallum, Bennett, 1994, “A Monetary Impulse Measure For Medium-Term Policy Analysis,” IMF Working Paper 94/146, Washington: International Monetary Fund.
- Poghosyan, T., and A. Khachatryan, 2001, “Factor Analysis of Interest Rate Spread in the Armenian Banking System,” Central Bank of Armenia Journal of Quarterly Publications, No. 01(3).
- Poghosyan, T., A. Mkrtchyan, and A. Khachatryan, 2001, “Implication of MCI for Monetary Policy Implementation in Armenia,” Central Bank of Armenia Journal of Quarterly Publications, No. 01(1).
- __________, 2002, “Implication of Monetary Policy Rules in Armenia: Efficiency of the Base Money Rule,” Central Bank of Armenia Journal of Quarterly Publications, No. 02(3).
- Thiessen, Gordon, 1995, “Uncertainty and the Transmission of Monetary Policy In Canada,” unpublished; Toronto: York University.
- Vardanyan V., A. Mkrtchyan, and A. Khachatryan, 1998, “Practical Implication of Monetary Impulses In Armenia,” Central Bank of Armenia Journal of Quarterly Publications, No. 98(2).

### Appendix — Chronology of Key Events in the Monetary and Financial Sector

1994:
- first monetary policy program was developed
- the CBA established an interest rates floor for non-auction loans
- all loan rates were determined by the CBA`s refinance rate
- the CBA began guaranteeing interbank credits, a practice that was quickly abolished as a result of several bank closures
- the concept of reserve requirements was introduced and the banks were asked to deposit 15 percent of their dram and foreign exchange liabilities with the CBA
- first program with the IMF was launched in December

1995:
- introduction of additional requirements for banks wishing to participate in credit auctions (the intention was to restrict participation of problem banks)
- preliminary work on the introduction of the treasury bills market and issuance of the “Procedures for Allocation, Circulation, and Maturity of Treasury Bills”
- the CBA started to offer liquidity, using treasury bills as collateral
- surrender requirements for export proceeds were abolished and the interbank foreign exchange transactions were permitted
- the official exchange rate was determined on a daily basis instead of tri-weekly
- the Monetary Policy Committee was established, which meets weekly to discuss developments in markets and prices

1996:
- the CBA introduced deposit auctions, which could serve as an alternative instrument to absorb excess liquidity in the banking system
- short-term liquidity projection scheme was developed
- first steps toward monetary and fiscal policies coordination were made. The regular meetings of the Minister of Finance and the Chairman of the CBA focused on policy coordination and implementation of monetary policy
- all restrictions on current and capital account operations were abolished

1997:
- modifications to the “Procedures for Allocation, Circulation, and Maturity of Treasury Bills” were made, creating an environment for institutional coordination between monetary and fiscal policies
- Armenia accepted the IMF’s Article VIII obligations

1998:
- the stock of the CBA's credit to the government was securitized and steps to cease the practice of direct credit were made
- the CBA and the Ministry of Finance signed an agreement about the coordination of monetary and fiscal policies and the operation of treasury accounts in the banking system
- the CBA announced its intentions to develop monetary policy instruments and place special attention on the path of the interest rates
- first repurchase agreements with commercial banks were signed

1999:
- further revisions of the regulation on conduct of monetary policy and instruments were made
- the CBA began disclosing the minutes of the Board discussions on the short-term monetary policy program, which were discussed on a bi-weekly basis
- new regime for reserve requirements came into effect in April, according to which the reserve requirement against both dram and foreign currency liabilities should be kept only in Armenian dram
- the monetary policy program introduced the notion of the band (± 2 percent)
- resolution on new principles and approaches for use of the CBA monetary policy instruments was adopted. It underlined that the CBA announces the interest rate only for repurchase agreements for 14 days (as a main signal of the monetary policy stance), and for stand-by facilities as the principal instrument for short-term liquidity management with the Central Bank

2000:
- Financial Sector Assessment Program and Review of Observance of Standards
- Regulation on the “Interrelations Between the Central Bank and the Ministry of Finance and Economy” was adopted by the CBA and the MOFE
- the CBA announced its compliance with the International Code of Transparency in Monetary and Financial Policies

2001:
- Armenia joined General Data Disseminated System (GDDS)

2002:
- the legal framework of the banking system was modified
- the Law on Credit Organizations was adopted

2003:
- commercial bank reserve requirements were reduced to 6 percent
- Armenia joined IMF’s Special Data Dissemination Standard (SDDS)

*Source: _wp04229 - References*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp04229.pdf_