## _wp03212

## Source details

**Canonical URL:** [_wp03212](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2003/_wp03212.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2003/_wp03212.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2003/_wp03212.pdf.json)

---

### I. Introduction
- Georgia stabilized after hyperinflation in 1993–1994: inflation fell from 15,607 percent in 1994 to 163 percent in 1995 and reached single-digit level in 1997.
- Devaluation of the lari after the Russian financial crises raised the price level but did not have a lasting effect on inflation.
- Monetary policy challenges: high dollarization; low monetization; volatile money demand.
- Purpose of the paper:
  - Test for economically interpretable long-run relationships between prices, money, and the exchange rate.
  - Estimate a short-run inflation equation.
  - Test robustness and stability of the inflation specification.
- Key qualitative conclusions:
  - Exchange rate plays a dominant role in inflation dynamics.
  - Inflation persistence in Georgia is low.
  - After the Russian crisis depreciation, inflation rose sharply but was quickly subdued once the authorities maintained the exchange rate at a higher, stable level.

### II. Macroeconomic developments in Georgia (selected indicators and narrative)
- Policy history:
  - Coupon introduced April 1993; stabilization program mid-1994 ended hyperinflation.
  - New currency (lari) introduced October 1995; lari effectively pegged to the U.S. dollar October 1995–December 1998.
  - NBG interventions and sterilization operations used to maintain exchange rate stability; interbank liquidity provision in late-1996 abandoned in second half of 1997.
- Russian crisis and aftermath (late 1998–2002):
  - Broad money declined by 25 percent in nominal terms from August to November 1998.
  - Lari floated December 7, 1998; exchange rate fell by 20 percent and monthly inflation soared to 12 percent.
  - Dollarization rose from 73 percent at end-1998 to 86 percent at end-1999 (ratio of foreign to total deposits).
- Selected macroeconomic indicators (1995–2002) — preserve exact series:
  - Real sector — GDP volume: 1995: 2.6; 1996: 10.5; 1997: 10.6; 1998: 2.9; 1999: 3.0; 2000: 1.9; 2001: 4.7; 2002: 5.6.
  - Consumer prices, period average: 1995: 162.7; 1996: 39.4; 1997: 7.0; 1998: 3.6; 1999: 19.1; 2000: 4.0; 2001: 4.7; 2002: 5.6.
  - Consumer prices, end-of-period: 1995: 57.4; 1996: 13.7; 1997: 7.2; 1998: 10.7; 1999: 10.9; 2000: 4.6; 2001: 3.4; 2002: 5.4.
  - Revenue and grants (percent of GDP): 1995: 10.6; 1996: 14.1; 1997: 14.4; 1998: 15.6; 1999: 15.4; 2000: 15.2; 2001: 16.3; 2002: 15.8.
  - Of which: revenue (percent of GDP): 1995: 7.7 12.2 14.1 14.7 14.6 14.9 15.6 15.6 (as presented).
  - Expenditure and net lending (percent of GDP): 1995: 18.7; 1996: 21.5; 1997: 21.2; 1998: 21.8; 1999: 22.1; 2000: 19.2; 2001: 18.3; 2002: 17.8.
  - Balance (commitments, percent of GDP): 1995: -8.1; 1996: -7.4; 1997: -6.8; 1998: -6.1; 1999: -6.7; 2000: -4.0; 2001: -2.0; 2002: -2.0.
  - Balance (cash, percent of GDP): 1995: -7.0; 1996: -7.2; 1997: -6.1; 1998: -4.9; 1999: -5.0; 2000: -2.6; 2001: -1.6; 2002: -1.9.
  - Net domestic assets of the NBG (end-period, percent change / levels): 1995: 156.5; 1996: 158.5; 1997: 82.2; 1998: 69.9; 1999: 15.3; 2000: 6.9; 2001: -3.9; 2002: 8.8.
  - Credit to the rest of the economy: 1995: 40.0; 1996: -14.0; 1997: 63.4; 1998: 40.3; 1999: 34.3; 2000: 30.9; 2001: 10.8; 2002: 24.7.
  - Broad money including foreign currency deposits: 1995: 135.2; 1996: 41.9; 1997: 45.5; 1998: -1.2; 1999: 20.7; 2000: 39.0; 2001: 18.5; 2002: 17.9.
  - Broad money excluding foreign currency deposits: 1995: 319.7; 1996: 38.2; 1997: 35.4; 1998: -11.5; 1999: 9.6; 2000: 33.5; 2001: 5.7; 2002: 14.5.
  - Exports (goods): 1995: 362.6; 1996: 417.0; 1997: 493.5; 1998: 478.3; 1999: 477.0; 2000: 527.7; 2001: 473.0; 2002: 468.6.
  - Imports (goods): 1995: -700.1; 1996: -767.9; 1997: -1,052.4; 1998: -1,163.7; 1999: -1,013.0; 2000: -937.0; 2001: -934.8; 2002: -944.1.
  - Trade balance: 1995: -337.4; 1996: -350.9; 1997: -558.9; 1998: -685.4; 1999: -536.0; 2000: -409.2; 2001: -461.8; 2002: -475.6.
  - Current account (including transfers): 1995: -404.8; 1996: -415.4; 1997: -364.1; 1998: -370.3; 1999: -217.5; 2000: -175.6; 2001: -179.2; 2002: -202.5.
  - Official reserves, end-of-period (In millions of U.S. dollars): 1995: 156.7; 1996: 158.0; 1997: 173.3; 1998: 118.4; 1999: 132.4; 2000: 109.4; 2001: 161.1; 2002: 197.7.
  - Official reserves (In months of imports): 1995: 2.7; 1996: 2.5; 1997: 1.5; 1998: 1.0; 1999: 1.3; 2000: 1.0; 2001: 1.4; 2002: 1.8.
  - Memorandum items: GDP, current prices (in millions of lari): 1995: 2,432; 1996: 3,768; 1997: 4,638; 1998: 5,040; 1999: 5,665; 2000: 6,013; 2001: 6,638; 2002: 7,457.
  - Exchange rate, lari/U.S. dollar — Period average: 1995: 1.29; 1996: 1.26; 1997: 1.30; 1998: 1.39; 1999: 2.02; 2000: 1.98; 2001: 2.07; 2002: 2.20.
  - Exchange rate, lari/U.S. dollar — End-of-period: 1995: 1.23; 1996: 1.27; 1997: 1.30; 1998: 1.79; 1999: 1.96; 2000: 1.98; 2001: 2.06; 2002: 2.09.

### III. Model of inflation — theoretical framework and data
- Theoretical setup:
  - Aggregate demand in logs: yD = α1 (m – p) + α2 (e – p)    (equation (1))
  - Equilibrium: y = yS = yD    (equation (2)
  - Long-run price equation under goods-market equilibrium:
    - p = α1/(α1 + α2) m + α2/(α1 + α2) e – 1/(α1 + α2) y    (equation (3))
    - Testable restriction: coefficients of money and exchange rate sum to one (homogeneity).
- Data and transformations:
  - Monthly data January 1996–February 2003.
  - Domestic CPI and GDP from Georgian State Department of Statistics; quarterly GDP interpolated assuming unit root.
  - Exchange rate: average lari/U.S. dollar from NBG.
  - Money measured by M2; M3 also attempted.
  - Fruit and vegetable prices from disaggregated CPI; average oil prices from IMF’s WEO.
  - All series in logs and seasonally adjusted (X-12 variant).
- Time-series properties:
  - ADF tests indicate logs of CPI, M2, and the lari/U.S. dollar exchange rate are integrated of order one in the Jan 1996–Feb 2003 sample.
  - Visual/test evidence: after introduction of lari, inflation and exchange rate depreciation became integrated of order zero, while real money balances remained I(1) — implications for cointegration and inclusion of expected inflation/depreciation.

### IV. Testing and estimation of cointegrating vectors
- Johansen (1988) procedure:
  - Endogenous variables: prices (p), exchange rate (e), money (m).
  - Real GDP treated as exogenous (restricted to cointegration space).
  - Two exogenous short-run variables: changes in relative prices of fruits and vegetables, and changes in oil prices.
  - Two impulse dummies: September 1998 (D0998) and December 1998 (D1298).
  - VAR with six lags; effective estimation period June 1996–February 2003.
- Tests for number of cointegrating vectors:
  - Results indicate only one cointegrating vector.
  - Reported test statistics (selected):
    - Rank 0: λmax 39.25 [0.003]**, λmax using T-nk 28.38 [0.003]**, λtrace 30.53 [0.041]*, λtrace using T-nk 22.07 [0.035]*.
    - Rank 1: λmax 10.87 [0.223], λmax using T-nk 9.09 [0.285], λtrace 8.46 [0.425], λtrace using T-nk 7.07 [0.489].
    - Rank 2: λmax 1.78 [0.182], λmax using T-nk 1.78 [0.182], λtrace 1.38 [0.239], λtrace using T-nk 1.38 [0.239].
- Homogeneity and weak exogeneity tests:
  - Homogeneity restriction (sum of coefficients of money and exchange rate = 1): χ2 (1) = 2.8404 [0.0919].
  - Weak exogeneity tests (no other restrictions imposed):
    - Prices: χ2 (2) = 17.868 [0.0001]**, χ2 (1) = 17.606 [0.0000]**.
    - Money: χ2 (2) = 3.1609 [0.2059], χ2 (1) = 3.0294 [0.0818].
    - Exchange rate (eUSD): χ2 (1) = 0.2559 [0.6130].
  - Weak exogeneity tests (with homogeneity restriction imposed):
    - Prices: χ2 (3) = 19.273 [0.0002]**, χ2 (2) = 22.616 [0.0000]**.
    - Money: χ2 (3) = 6.8798 [0.0758], χ2 (2) = 6.7213 [0.0347]*.
    - Exchange rate (eUSD): χ2 (2) = 4.3538 [0.1134].
  - Inference:
    - Homogeneity restriction, weak exogeneity of the exchange rate, and weak exogeneity of money are not rejected at the 5 percent level.
    - Weak exogeneity of prices is strongly rejected.
    - Exogeneity of the exchange rate implies prices adjust to restore goods-market equilibrium; weak exogeneity of money implies prices adjust to generate necessary changes in real money balances.
- Monetary aggregate robustness:
  - Estimation with M3 yields similar results; weak exogeneity of money harder to reject with M3.
  - Correlation between log changes in M2 and M3 is 0.88.
- Chosen long-run specification:
  - Restrictions imposed: homogeneity; weak exogeneity of money (M2) and exchange rate.
  - Key estimated long-run coefficients (specification A, Table 4):
    - β: p_t = 1.0000; m_t = -0.2530 (0.0457); eUSD_t = -0.5213 (0.0314); y_t = 0.4418 (0.1478).
    - α (adjustment coefficients): p_t = -0.1561 (0.0394); m_t = 0.3271 (0.2015); eUSD_t = -0.0451 (0.0888).
  - Notes:
    - Exchange rate coefficient generally higher than money coefficient; real income coefficient negative as theory predicts.
    - Recursive parameter estimates show stability over time.
    - Largest deviation from equilibrium appears after post-Russian-crisis devaluation—prices did not fully adjust, producing upward pressure on inflation.
- Econometric implication:
  - Given weak exogeneity of exchange rate and money, the subsequent error-correction model for inflation conditions on these two variables.
  - Robustness: instrumental variables estimation using past changes in money and exchange rate as instruments.

### V. Error-correction model for inflation (short-run dynamics and diagnostics)
- Specification search and final parsimonious equation (Table 5):
  - Unrestricted candidate included: five lags of inflation; five lagged and current values of ∆ log money and ∆ log exchange rate; lagged ECM; changes in oil prices and relative food prices; dummies D1298 and D0998.
  - Parsimonious final inflation equation:
    - ∆p_t = 0.4633 + 0.0935 ∆eUSD_t + 0.0973 1/3 (∆m_t + ∆m_{t-1} + ∆m_{t-2}) + 0.0589 ECM_{t-1}
      + 0.1322 ∆(p_t^{food} - p_t) + 0.0287 ∆p_t^{oil} + 0.0872 D1298
    - Standard errors (in parentheses): (0.1078) (0.0274) (0.0253) (0.0137) (0.0225) (0.0076) (0.0081).
    - Goodness of fit and sample:
      - R^2 = 0.8583.
      - Equation standard error: 0.0054.
      - DW = 2.18.
      - Sample: 1996.6 – 2003.2; Number of observations: 81.
    - Note: ∆ indicates first difference. D0998 = 1 for September 1998, zero otherwise. D1298 = 1 for December 1998, zero otherwise.
- Key empirical findings from the short-run model:
  - Inflation is strongly affected by exchange rate changes and pass-through is fast: coefficient on ∆eUSD_t = 0.0935 (0.0274).
  - Money changes significantly impact inflation but with slower transmission: 1/3 MA(∆m) coefficient = 0.0973 (0.0253).
  - Error-correction term is highly significant: ECM_{t-1} coefficient = 0.0589 (0.0137), indicating prices adjust toward long-run equilibrium defined by money, exchange rate, and output.
  - Lagged inflation terms drop out of final specification — conditional on exchange rate, money, and relative prices, inflation persistence is very low.
  - Supply shocks from agriculture (∆(p^{food}-p)) have a high and significant short-term impact: coefficient = 0.1322 (0.0225).
  - Changes in oil import prices have a smaller but significant impact: coefficient = 0.0287 (0.0076).
- Diagnostics and stability:
  - Specification tests (Table 6) — do not reject adequacy:
    - AR 1-5 test F(5,69) = 1.5659 [0.1813].
    - ARCH 1-5 test F(5,64) = 0.17575 [0.9707].
    - Normality test χ2 (2) = 1.6839 [0.4309].
    - hetero test F(11,62) = 0.77122 [0.6668].
    - hetero-X test F(21,52) = 0.57618 [0.9165].
    - RESET test F(1,73) = 0.93119 [0.3377].
  - Residuals:
    - One-step ahead residuals lie within a 2 standard deviations band.
    - One-step ahead and break-point Chow tests do not reject parameter stability at the 1 percent level.
  - Robustness:
    - Re-estimation on shorter sample (June 1999–Feb 2003) yields results close to full-sample estimates.
    - IV estimation using lagged changes of exchange rate and lagged MA(∆m) as instruments yields estimates close to OLS.

### VI. Conclusions and policy-relevant observations
- Main conclusions:
  - It is feasible to estimate robust price and inflation equations for Georgia.
  - Long-run price equation: prices are a function of money, the exchange rate, and real income — interpreted as goods-market equilibrium.
  - Short-run inflation dynamics are strongly affected by current exchange rate changes, money growth, and changes in relative food prices and oil prices.
  - Estimated long- and short-run relationships are stable and useful for policy formulation and evaluation.
  - Inflation in Georgia exhibits very low persistence, possibly in part due to relatively short-term nominal wage contracts.
- Policy-relevant observations and recommendations:
  - Constraints facing the National Bank of Georgia (NBG):
    - Public memory of hyperinflation remains; shocks exert strong pressure on the exchange rate.
    - NBG’s stock of foreign exchange reserves remains small, limiting scope to lean against downward pressure on the lari, especially when fiscal financing needs complicate monetary tightening.
  - Policy recommendations to strengthen NBG capacity:
    - Further accumulation of foreign reserves.
    - Development of indirect monetary control instruments, facilitated by a deeper treasury bill market.
    - These measures would increase the NBG’s capacity to respond to shocks.
  - Despite constraints, the NBG has substantially succeeded in keeping inflation low and relatively stable.

### Appendix I — Tests for stationarity (ADF statistics, January 1996–February 2003)
- Seasonally Adjusted Series:
  - p: Level lag 2 test statistic -2.951; First difference lag 1 test statistic -3.873**.
  - m: Level lag 1 test statistic -3.180; First difference lag 1 test statistic -6.250**.
  - eUSD: Level lag 3 test statistic -1.628; First difference lag 2 test statistic -5.172**.
- Unadjusted Series:
  - p: Level lag 2 test statistic -2.732; First difference lag 4 test statistic -4.344**.
  - m: Level lag 1 test statistic -3.181; First difference lag 1 test statistic -6.596**.
  - eUSD: Level lag 3 test statistic -1.750; First difference lag 2 test statistic -4.854**.
- Notes on tests:
  - Null hypothesis: series contains a unit root against a stationary alternative.
  - Number of lags chosen based on F-tests and Akaike information criterion, starting from six lags.
  - Level equations for unadjusted series include constant, seasonals and linear trend; levels of seasonally adjusted series include constant and linear trend.
  - First differences not trended; inclusion of trend into first-difference equations does not change results.
  - * and ** denote significance at the 5 percent and 1 percent level, respectively.

*Source: IMF staff estimates.*

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

### _wp03212 - References

### I. INTRODUCTION
- Georgia stabilized after hyperinflation in 1993–1994: inflation fell from 15,607 percent in 1994 to 163 percent in 1995 and reached single-digit level in 1997.
- Devaluation of the lari after the Russian financial crises raised the price level but did not have a lasting effect on inflation.
- Monetary policy challenges: high dollarization; low monetization; volatile money demand.
- Purpose of the paper:
  - Test for economically interpretable long-run relationships between prices, money, and the exchange rate.
  - Estimate a short-run inflation equation.
  - Test robustness and stability of the inflation specification.
- Key qualitative conclusions:
  - Exchange rate plays a dominant role in inflation dynamics.
  - Inflation persistence in Georgia is low.
  - After the Russian crisis depreciation, inflation rose sharply but was quickly subdued once the authorities maintained the exchange rate at a higher, stable level.

### II. MACROECONOMIC DEVELOPMENTS IN GEORGIA
- Historical policy context:
  - Coupon introduced April 1993; stabilization program mid-1994 ended hyperinflation.
  - New currency (lari) introduced October 1995; lari effectively pegged to the U.S. dollar October 1995–December 1998.
  - NBG interventions in the foreign exchange market and sterilization operations were used to maintain exchange rate stability.
  - In late-1996 NBG provided liquidity through interbank auctions to lower interest rates; strategy was abandoned in second half of 1997 due to adverse reserve effects.
- Russian crisis and aftermath (late 1998–2002):
  - Demand for lari declined sharply in late 1998 (broad money declined by 25 percent in nominal terms from August to November).
  - NBG interventions and tightening measures failed; lari floated December 7, 1998; exchange rate fell by 20 percent and monthly inflation soared to 12 percent.
  - NBG tightened policy thereafter, but government financing pressures in late 1999–first half 2000 increased direct financing and downward pressure on the currency.
  - Fiscal improvement in second half 2000 allowed NBG to restrain net domestic assets and control reserve money growth while rebuilding foreign reserves at year-end.
  - Exchange rate appreciation at end-2002 prompted NBG foreign exchange purchases.
  - Dollarization rose from 73 percent at end-1998 to 86 percent at end-1999 (measured as a ratio of foreign to total deposits), indicating persistent low policy credibility.
- Selected macroeconomic indicators (1995–2002) — annual percent change / percent of GDP / levels preserved exactly as in Table 1:
  - Real sector — GDP volume: 1995: 2.6; 1996: 10.5; 1997: 10.6; 1998: 2.9; 1999: 3.0; 2000: 1.9; 2001: 4.7; 2002: 5.6.
  - Consumer prices, period average: 1995: 162.7; 1996: 39.4; 1997: 7.0; 1998: 3.6; 1999: 19.1; 2000: 4.0; 2001: 4.7; 2002: 5.6.
  - Consumer prices, end-of-period: 1995: 57.4; 1996: 13.7; 1997: 7.2; 1998: 10.7; 1999: 10.9; 2000: 4.6; 2001: 3.4; 2002: 5.4.
  - Government finances: Revenue and grants (percent of GDP): 1995: 10.6; 1996: 14.1; 1997: 14.4; 1998: 15.6; 1999: 15.4; 2000: 15.2; 2001: 16.3; 2002: 15.8.
  - Of which: revenue (percent of GDP): 1995: 7.7 12.2 14.1 14.7 14.6 14.9 15.6 15.6 (as presented).
  - Expenditure and net lending (percent of GDP): 1995: 18.7; 1996: 21.5; 1997: 21.2; 1998: 21.8; 1999: 22.1; 2000: 19.2; 2001: 18.3; 2002: 17.8.
  - Balance (commitments, percent of GDP): 1995: -8.1; 1996: -7.4; 1997: -6.8; 1998: -6.1; 1999: -6.7; 2000: -4.0; 2001: -2.0; 2002: -2.0.
  - Balance (cash, percent of GDP): 1995: -7.0; 1996: -7.2; 1997: -6.1; 1998: -4.9; 1999: -5.0; 2000: -2.6; 2001: -1.6; 2002: -1.9.
  - Money and credit (end-period, percent change / levels): Net domestic assets of the NBG: 1995: 156.5; 1996: 158.5; 1997: 82.2; 1998: 69.9; 1999: 15.3; 2000: 6.9; 2001: -3.9; 2002: 8.8.
  - Credit to the rest of the economy: 1995: 40.0; 1996: -14.0; 1997: 63.4; 1998: 40.3; 1999: 34.3; 2000: 30.9; 2001: 10.8; 2002: 24.7.
  - Broad money including foreign currency deposits: 1995: 135.2; 1996: 41.9; 1997: 45.5; 1998: -1.2; 1999: 20.7; 2000: 39.0; 2001: 18.5; 2002: 17.9.
  - Broad money excluding foreign currency deposits: 1995: 319.7; 1996: 38.2; 1997: 35.4; 1998: -11.5; 1999: 9.6; 2000: 33.5; 2001: 5.7; 2002: 14.5.
  - External sector: Exports (goods): 1995: 362.6; 1996: 417.0; 1997: 493.5; 1998: 478.3; 1999: 477.0; 2000: 527.7; 2001: 473.0; 2002: 468.6.
  - Imports (goods): 1995: -700.1; 1996: -767.9; 1997: -1,052.4; 1998: -1,163.7; 1999: -1,013.0; 2000: -937.0; 2001: -934.8; 2002: -944.1.
  - Trade balance: 1995: -337.4; 1996: -350.9; 1997: -558.9; 1998: -685.4; 1999: -536.0; 2000: -409.2; 2001: -461.8; 2002: -475.6.
  - Current account (including transfers): 1995: -404.8; 1996: -415.4; 1997: -364.1; 1998: -370.3; 1999: -217.5; 2000: -175.6; 2001: -179.2; 2002: -202.5.
  - Official reserves, end-of-period (In millions of U.S. dollars): 1995: 156.7; 1996: 158.0; 1997: 173.3; 1998: 118.4; 1999: 132.4; 2000: 109.4; 2001: 161.1; 2002: 197.7.
  - Official reserves (In months of imports): 1995: 2.7; 1996: 2.5; 1997: 1.5; 1998: 1.0; 1999: 1.3; 2000: 1.0; 2001: 1.4; 2002: 1.8.
  - Memorandum items: GDP, current prices (in millions of lari): 1995: 2,432; 1996: 3,768; 1997: 4,638; 1998: 5,040; 1999: 5,665; 2000: 6,013; 2001: 6,638; 2002: 7,457.
  - Exchange rate, lari/U.S. dollar — Period average: 1995: 1.29; 1996: 1.26; 1997: 1.30; 1998: 1.39; 1999: 2.02; 2000: 1.98; 2001: 2.07; 2002: 2.20.
  - Exchange rate, lari/U.S. dollar — End-of-period: 1995: 1.23; 1996: 1.27; 1997: 1.30; 1998: 1.79; 1999: 1.96; 2000: 1.98; 2001: 2.06; 2002: 2.09.

### III. MODEL OF INFLATION
A. Model Specification
- Theoretical framework:
  - Long-term price level (P) determined by balance between aggregate demand (YD) and supply (YS).
  - Aggregate demand is a function of real money supply (M/P) and the real exchange rate (E/P). In logs:
    - yD = α1 (m – p) + α2 (e – p)    (equation (1))
  - Aggregate supply is exogenously given and equal to aggregate demand in equilibrium:
    - y = yS = yD    (equation (2))
- Market equilibrium implications:
  - If goods, money, and foreign exchange markets are all in equilibrium on average, expect two unique long-run cointegrating vectors among nominal variables (p, m, e) with y treated as exogenous.
  - If money and foreign exchange markets are persistently out of equilibrium, only one cointegrating vector may be found, corresponding to goods market equilibrium (equation (1)).
- Long-run price equation under goods-market equilibrium:
  - p = α1/(α1 + α2) m + α2/(α1 + α2) e – 1/(α1 + α2) y    (equation (3))
  - This equation implies neo-classical dichotomy; testable restriction: coefficients of money and exchange rate sum to one.
- Money demand considerations:
  - In high-inflation economies, expected inflation E(p – p-1) or expected exchange rate depreciation E(e – e-1) may enter money demand.
  - If inflation or exchange-rate-depreciation series contain a unit root, they must be incorporated when estimating cointegrating relationships describing money demand or prices.
- Short-run dynamics:
  - Modeled as an error-correction mechanism; lag structure determined by general-to-specific methodology.

B. Sources, Transformations, and Statistical Properties of the Data
- Data and period:
  - Monthly data for January 1996–February 2003 (post-stabilization period).
  - Domestic CPI and GDP from Georgian State Department of Statistics (SDS); quarterly GDP interpolated under assumption monthly series follows a unit root process.
  - Exchange rate: average lari/U.S. dollar from NBG.
  - Money measured by M2; estimation with M3 also attempted.
  - Fruit and vegetable prices from disaggregated CPI divided by total CPI to obtain relative values.
  - Average oil prices from IMF’s WEO database.
  - All series in logs and seasonally adjusted using a version of the X-12 procedure.
- Time-series properties and stationarity:
  - ADF tests indicate regime change: January 1991–December 1995 sample cannot reject non-stationarity; Jan 1996–February 2003 sample shows logs of CPI, M2, and the lari/U.S. dollar exchange rate are all integrated of order one (see Table A-1 in Appendix I).
  - Modeling the regime change is beyond scope; estimation begins January 1996.

- Empirical implications highlighted:
  - Visual and test evidence suggest that after eradication of hyperinflation and introduction of the lari, inflation and exchange rate depreciation became integrated of order zero, while real money balances remained integrated of order one; this affects specification of long-run cointegrating relations and inclusion of expected inflation or expected exchange rate depreciation where appropriate.

_Italic: Source: _wp03212 - References (IMF staff paper content provided)._

### introduction of the lari to insulate estimates from the pre-lari regime

### introduction of the lari to insulate estimates from the pre-lari regime

### C. Testing and Estimation of Cointegrating Vectors
- Methodology and setup:
  - Johansen (1988) procedure used to test for the number of cointegrating vectors, estimate coefficients, and test for weak exogeneity.
  - Endogenous variables: prices, the exchange rate, and money.
  - Real GDP (log of real GDP) determined outside the system and restricted to enter only the cointegration space; the log of real GDP is interpolated under the assumption that the series follows a unit root process.
  - Two exogenous short-run variables added: percentage changes in relative prices of fruits and vegetables, and percentage changes in oil prices (current values only).
  - Two impulse dummies: December 1998 (month of de facto regime change when the lari started floating against the dollar) and September 1998 (first month after the Russian crises).
  - VAR estimated with six lags of each endogenous variable; effective estimation period is June 1996-February 2003.
  - Sequential testing allows reduction of lag length to four, but six-lag specification retained to avoid underspecification.

- Tests for number of cointegrating vectors (between p, m, e, and y):
  - Results suggest there is only one cointegrating vector.
  - Table 2 reported λmax and λtrace statistics with significance indicators:
    - Rank 0: λmax 39.25 [0.003]**, λmax using T-nk 28.38 [0.003]**, λtrace 30.53 [0.041]*, λtrace using T-nk 22.07 [0.035]*.
    - Rank 1: λmax 10.87 [0.223], λmax using T-nk 9.09 [0.285], λtrace 8.46 [0.425], λtrace using T-nk 7.07 [0.489].
    - Rank 2: λmax 1.78 [0.182], λmax using T-nk 1.78 [0.182], λtrace 1.38 [0.239], λtrace using T-nk 1.38 [0.239].

- Homogeneity and weak exogeneity tests (Table 3):
  - Homogeneity restriction (sum of coefficients of money and exchange rate = 1): χ2 (1) = 2.8404 [0.0919].
  - Weak exogeneity tests (a. no other restrictions imposed):
    - Prices: χ2 (2) = 17.868 [0.0001]**, χ2 (1) = 17.606 [0.0000]**.
    - Money: χ2 (2) = 3.1609 [0.2059], χ2 (1) = 3.0294 [0.0818].
    - Exchange rate (eUSD): χ2 (1) = 0.2559 [0.6130].
  - Weak exogeneity tests (b. with homogeneity restriction imposed):
    - Prices: χ2 (3) = 19.273 [0.0002]**, χ2 (2) = 22.616 [0.0000]**.
    - Money: χ2 (3) = 6.8798 [0.0758], χ2 (2) = 6.7213 [0.0347]*.
    - Exchange rate (eUSD): χ2 (2) = 4.3538 [0.1134].
  - Inference:
    - Homogeneity restriction, weak exogeneity of the exchange rate, and weak exogeneity of money are not rejected at the 5 percent significance level.
    - Weak exogeneity of prices is strongly rejected.
    - Exogeneity of the exchange rate implies prices—rather than exchange rate—adjust to restore equilibrium in the goods market; weak exogeneity of money implies prices adjust to generate changes in real money balances necessary to restore equilibrium.

- Monetary aggregate robustness:
  - Estimation using M3 yields similar results, but weak exogeneity of money is harder to reject with M3.
  - M2 is easier to control by monetary authorities; weak exogeneity of M2 is consistent with no feedback policy rule governing that aggregate.
  - Correlation between log changes in M2 and M3 is 0.88.

- Choice of long-run specification:
  - Equation with imposed restrictions of homogeneity and weak exogeneity of money (M2) and exchange rate chosen for further analysis.
  - Key estimated long-run coefficients (Table 4, specification A):
    - β: p_t = 1.0000; m_t = -0.2530 (0.0457); eUSD_t = -0.5213 (0.0314); y_t = 0.4418 (0.1478).
    - α (adjustment coefficients): p_t = -0.1561 (0.0394); m_t = 0.3271 (0.2015); eUSD_t = -0.0451 (0.0888).
  - Alternative specifications B and C reported with differing β and α; exchange rate coefficient generally higher than money coefficient, both close to one-half in less restricted estimates.
  - Real income coefficient is negative in accordance with theoretical prediction.
  - Recursive estimates of parameters show coefficients are stable over time (Figure 3).
  - Cointegrating vector and long-run components plotted (Figure 4); largest deviation from equilibrium appears after the post-Russian-crisis devaluation—prices did not fully adjust, producing upward pressure on inflation.

- Econometric implications:
  - Given weak exogeneity of exchange rate and money, subsequent error-correction model for inflation conditions on these two variables.
  - Robustness check: model also estimated by instrumental variables using past changes in money and exchange rate as instruments.

### D. Error-Correction Model for Inflation
- Specification search:
  - General-to-specific methodology used.
  - Unrestricted inflation equation included:
    - five lags of inflation;
    - five lagged and current values of changes in log money and log exchange rate;
    - lagged error-correction term from long-run price equation;
    - changes in oil prices and relative prices of fruits and vegetables;
    - dummy variables D1298 and D0998.
  - Restrictions tested with F-tests; information criteria (Akaike, Schwarz, Hannan-Quinn, Final Prediction Error) used to guide reductions.

- Parsimonious final inflation equation (Table 5):
  - ∆p_t = 0.4633 + 0.0935 ∆eUSD_t + 0.0973 1/3 (∆m_t + ∆m_{t-1} + ∆m_{t-2}) + 0.0589 ECM_{t-1}
    + 0.1322 ∆(p_t^{food} - p_t) + 0.0287 ∆p_t^{oil} + 0.0872 D1298
    - Standard errors in parentheses: (0.1078) (0.0274) (0.0253) (0.0137) (0.0225) (0.0076) (0.0081).
    - R^2 = 0.8583; Equation standard error: 0.0054; DW = 2.18.
    - Sample: 1996.6 – 2003.2; Number of observations: 81.
    - Notes: ∆ indicates first difference. D0998 is one for September 1998, zero otherwise, and D1298 is one for December 1998, zero otherwise.

- Key findings from the short-run equation:
  - Inflation is strongly affected by exchange rate changes and pass-through is fast.
  - Changes in money significantly impact inflation, but with a slower transmission than exchange rate changes (moving average of three changes used).
  - Error-correction term is highly significant—prices adjust to their long-run equilibrium determined by money, exchange rate, and output.
  - Lagged inflation terms drop out of the final specification—conditional on exchange rate, money, and relative prices, inflation persistence is very low.
  - Supply shocks from agriculture have a high and significant short-term impact on inflation.
  - Changes in oil import prices have a smaller but significant impact.

- Diagnostics and stability:
  - Final inflation equation passes specification tests (Table 6):
    - AR 1-5 test F(5,69) = 1.5659 [0.1813].
    - ARCH 1-5 test F(5,64) = 0.17575 [0.9707].
    - Normality test χ2 (2) = 1.6839 [0.4309].
    - hetero test F(11,62) = 0.77122 [0.6668].
    - hetero-X test F(21,52) = 0.57618 [0.9165].
    - RESET test F(1,73) = 0.93119 [0.3377].
  - One-step ahead residuals within a 2 standard deviations band; one-step ahead and break-point Chow tests do not reject parameter stability at 1 percent level.
  - Re-estimation on shorter sample (June 1999 to February 2003) yields results remarkably close to full-sample estimates (Table 7, column B).
  - IV estimation (column C of Table 7) using lagged changes of exchange rate and lagged moving average of money changes as instruments yields estimates close to OLS results:
    - OLS (full sample, column A): constant 0.4633 (0.0935); ∆eUSD_t 0.0935 (0.0274); 1/3 MA(∆m) 0.0973 (0.0253); ECM_{t-1} -0.0589 (0.0137); ∆(p^{food}-p) 0.1322 (0.0225); ∆p^{oil} 0.0287 (0.0076); D1298 0.0872 (0.0081).
    - OLS (short sample, column B) and IV (column C) reported with respective standard errors.

### IV. Conclusions
- Main conclusions:
  - It is feasible to estimate robust price and inflation equations for Georgia.
  - Long-run price equation: prices are a function of money, the exchange rate, and real income—interpreted as equilibrium in the goods market.
  - Short-run inflation dynamics are strongly affected by current exchange rate changes, money growth, and changes in relative food prices and oil prices.
  - Estimated long- and short-run relationships are stable and may be useful for policy formulation and evaluation.
  - Inflation in Georgia exhibits very low persistence, possibly owing in part to the use of relatively short-term nominal wage contracts (a legacy of the hyperinflationary period).

- Policy-relevant observations:
  - The National Bank of Georgia (NBG) faces challenges in conducting monetary policy:
    - Public memory of hyperinflation is still fresh; external or internal shocks quickly exert strong pressure on the exchange rate.
    - The NBG's stock of foreign exchange reserves remains small, leaving little scope for leaning against downward pressure on the lari—especially when budget-financing needs complicate monetary tightening.
  - Despite challenges, the NBG has enjoyed substantial success in keeping inflation low and relatively stable.
  - Policy recommendations to strengthen NBG capacity:
    - Further accumulation of foreign reserves.
    - Development of indirect monetary control instruments, facilitated by a deeper treasury bill market.
    - These measures would increase the NBG's capacity to respond to shocks.

### Appendix I. Tests for Stationarity
- Unit root Augmented Dickey-Fuller test statistics (January 1996-February 2003):
  - Seasonally Adjusted Series (lags and test statistics reported):
    - p: Level lag 2 test statistic -2.951; First difference lag 1 test statistic -3.873**.
    - m: Level lag 1 test statistic -3.180; First difference lag 1 test statistic -6.250**.
    - eUSD: Level lag 3 test statistic -1.628; First difference lag 2 test statistic -5.172**.
  - Unadjusted Series:
    - p: Level lag 2 test statistic -2.732; First difference lag 4 test statistic -4.344**.
    - m: Level lag 1 test statistic -3.181; First difference lag 1 test statistic -6.596**.
    - eUSD: Level lag 3 test statistic -1.750; First difference lag 2 test statistic -4.854**.
  - Notes:
    - Null hypothesis is that a series contains a unit root against a stationary alternative.
    - Number of lags chosen based on F-tests and Akaike information criterion, starting from six lags.
    - Equations for levels of unadjusted series estimated with a constant, seasonals and linear trend; levels of seasonally adjusted series estimated with a constant and linear trend.
    - First differences not trended; inclusion of trend into first-difference equations does not change results.
    - * and ** denote significance at the 5 percent and 1 percent level, respectively.

*Source: IMF staff estimates.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2003/_wp03212.pdf_