## 5. Estimation Results for Threshold MSVAR

## Source details

**Canonical URL:** [5. Estimation Results for Threshold MSVAR](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10270.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10270.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10270.pdf.json)

---

### Overview and purpose
- Aim: examine whether the effects of monetary policy on the real economy in Armenia changed, and whether dollarization helps explain any regime switch, using:
  - a standard structural VAR (for comparison with earlier studies),
  - a Markov Switching VAR (MSVAR) that estimates regime switches endogenously,
  - a Markov Switching VAR with a threshold variable (credit dollarization).
- Data and sample: monthly data spanning 2000M1-2010M5; VAR and MSVAR specifications use real GDP, CPI, the repo rate, monetary aggregates (M1 or M2/M2X as noted), and NEER; exogenous variables include a petroleum price index and the U.S. Federal Funds Rate.

### Key empirical findings — standard VAR and Granger causality
- Granger causality p-values (multivariate and bivariate block tests):
  - Effect on output: Interest Rate (s) 0.00***; Money Supply (m) 0.00***; Exchange Rate (x) 0.05**; Block (p, s, x, m) 0.00***.
  - Effect on prices: Interest Rate (s) 0.96; Money Supply (m) 0.15; Exchange Rate (x) 0.03**; Block (p, s, x, m) 0.14.
  - Effect on NEER: Interest Rate (s) 0.03**; Money Supply (m) 0.06*; Block (p, s, x, m) 0.02**.
  - Note: *, **, and *** denote rejection of the null hypothesis at the 10, 5, and 1 percent significance levels, respectively.
- Standard VAR impulse responses:
  - Interest and money shocks have persistent effects on output.
  - Effects on prices are small and statistically insignificant in the baseline VAR estimation.
- Interpretation: the baseline evidence points to a weak direct interest-rate-to-inflation transmission; the exchange rate channel has a significant link to inflation.

### MSVAR (endogenous regime switching) — setup and results
- Model setup:
  - MSVAR estimated with 2 regimes using endogenous variables y (real GDP), p (CPI), s (repo rate), m (M2), x (NEER).
  - Sample for MSVAR estimation described: 2006M1-2010M5.
  - Identification: Choleski decomposition; estimation via EM algorithm (Oxmetrics MSVAR package).
- Selected estimation statistics (from Table 4):
  - Log-likelihood (non-linear) 1263.6902; linear model log-likelihood 1069.4012.
  - AIC (non-linear) -18.5641; AIC (linear) -16.4130.
  - Linearity LR test: 388.5780; Chi(60) =[0.0000] **; Chi(62)=[0.0000] **; DAVIES=[0.0000]**.
- Transition matrix (estimated):
  - Regime1→Regime1: 0.9873
  - Regime1→Regime2: 0.01266
  - Regime2→Regime1: 1.862e-010
  - Regime2→Regime2: 1.000
- Regime timing and characterization:
  - Smoothed regime probabilities point to a regime switch in mid-2006.
  - Regimes are highly persistent per the transition matrix.
  - Coefficients differ between regimes: in regime 2 (post-switch) the repo rate has a significant effect on inflation and the exchange rate; in regime 1 the repo rate and exchange rate changes cause no significant effect on inflation or output.
- Interpretation: MSVAR supports a structural break around mid-2006 with a strengthening of the transmission mechanism thereafter.

### Threshold MSVAR (dollarization as threshold) — method and dating
- Threshold variable: d_t = ratio of foreign currency loans to total loans (loan dollarization).
- Threshold estimation: MSVAR estimates the threshold value D that maximizes conditional log-likelihood and divides regimes by d_t ≤ D versus d_t > D.
- Data transformations: variables transformed to ensure stationarity (logs and first differences); impulse responses computed regime-dependently assuming the threshold is exogenous.
- Estimated threshold and regime dating (Table 5):
  - Estimated threshold: 51
  - Periods of high dollarization: 2000:3 – 2006:11 and 2009:7 – 2010:5
  - Period of low dollarization: 2006:12 – 2009:6
  - Log likelihood: 1238.35
  - LR linearity test: 329.84; p-values (adjusted chi-squared) [0.0000]
- Interpretation of timing:
  - Threshold D = 51 percent implies the economy moved from high dollarization to low dollarization around late 2006, remained in low dollarization until mid-2009, and moved back to high dollarization afterwards.
  - The threshold MSVAR captures the mid-2006 switch and also identifies a second switch in mid-2009 that the MSVAR without threshold did not pick up (likely due to few observations in that final regime in the non-threshold MSVAR).

### Regime-dependent impulse responses — main quantitative findings
- Repo rate shock:
  - A 100 basis point increase in the repo rate reduces GDP growth in both regimes; the impact under low dollarization is nearly 10 times stronger than under high dollarization.
  - The repo rate has a negative impact on inflation; under low dollarization the response of inflation is much larger but remains insignificant.
  - Conclusion: interest rate channel strengthened after dedollarization but remains weak in affecting inflation directly.
- Money growth shock:
  - A 1 percentage point increase in money growth produces the expected sign for GDP growth (negative for a tightening) and is stronger under low dollarization but not significant in either regime.
  - Impact on inflation is significant under both regimes and stronger under low dollarization.
  - Interpretation: transmission from money to prices strengthened after end-2006; this suggests the inflation targeting framework remained rudimentary and money growth retained relevance.
- NEER shock:
  - Exchange rate pass-through to growth and inflation appears stronger in the low dollarization period (2006–2009) than in the high dollarization periods.
  - Possible explanations:
    - Increased frequency and size of foreign exchange interventions during the low dollarization period may have enhanced the exchange rate’s role in shaping inflation expectations, increasing pass-through.
    - The weaker pass-through in the post-2009 high dollarization episode may reflect recessionary balance-sheet effects: a contractionary depreciation in March 2009 lowered demand and offset the usual inflationary effect of depreciation.
- Diagnostics:
  - Figures show regime probabilities and IRFs with 95 percent standard error bands for responses of GDP growth and inflation to repo, money, and NEER shocks for high vs low dollarization regimes.

### Synthesis of results
- Both MSVAR approaches identify a structural break in the second half of 2006.
  - Threshold MSVAR associates the break with credit dollarization falling below 51 percent.
  - Threshold MSVAR also identifies a return to high dollarization in mid-2009.
- Strengthening of the monetary transmission mechanism after 2006:
  - Evidence of stronger transmission from policy rate and money to output and prices in the low dollarization regime relative to the high dollarization regime.
  - Policy-rate-to-inflation channel strengthened but remains weak overall; money growth appears to have a stronger and significant effect on inflation in the low dollarization regime.
  - Exchange rate pass-through strengthened in the low dollarization period (pre-2009) possibly due to more active FX interventions; pass-through weakened again after mid-2009 amid recession and redollarization.
- Caveats and model assumptions:
  - The threshold MSVAR assumes dollarization is exogenous (shocks to the VAR do not change the regime); this restricts analysis of full endogeneity between monetary policy, macro shocks, and dollarization.
  - Short samples in later regimes limit the ability of unconstrained MSVAR to detect multiple switches.

### Policy implications and recommendations
- Strengthen macroeconomic stability to support dedollarization:
  - A strong track record of macroeconomic stability is crucial to foster dedollarization, which in turn strengthens the MTM.
  - Given vulnerabilities (for example, a current account deficit of more than 15 percent noted in the analysis), measures to reduce external imbalances and fiscal consolidation to keep debt in check are necessary to support dedollarization efforts.
- Pursue consistent monetary and exchange rate policies with clear communication:
  - The central bank should reaffirm its commitment to price stability as the primary objective and improve communication of its framework to enhance credibility.
  - Excessive emphasis on the nominal exchange rate (for example via interventions) reinforces the exchange rate channel and may impede development of interest-rate and credit channels.
  - Active liquidity management is recommended so that the policy (repo) rate becomes more relevant and market short-term rates are less volatile.
  - When policy actions affect the exchange rate, messaging should clarify that the ultimate objective is inflation control, not the exchange rate per se.
- Structural reforms to deepen financial intermediation:
  - Continued development of local-currency money and securities markets, improvements in creditor protection and corporate governance, and stronger banking supervision will help strengthen interest-rate and credit channels over time.
  - Dedollarization, combined with deeper and more liquid domestic currency markets, will improve transmission from the policy rate to lending and deposit rates.

*Source: IMF working paper section “5. Estimation Results for Threshold MSVAR” (content provided).*

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

### _wp10270 - References .............................................................................................................

### Figures
- 1. Stylized Representation of the Transmission Mechanism ..............................................4
- 2. Impulse Responses for the Standard VAR Model .........................................................13
- 3. Regime Probabilities for MSVAR .................................................................................17
- 4. Regime Probabilities for Threshold MSVAR ................................................................19
- 5. Threshold MSVAR: Response to a Repo Shock ...........................................................22
- 6. Threshold MSVAR: Response to a Money Shock ........................................................22
- 7. Threshold MSVAR: Response to a NEER Shock .........................................................23

### Tables
- 1. Effects of Recent Pre-Crisis Developments on the MTM .............................................10
- 2. Effects of Recent Crisis and Post-crisis Developments on the MTM ...........................10
- 3. Multivariate and Bivariate Block Granger Causality Tests ...........................................12
- 4. Estimation Results for MSVAR.....................................................................................16

### Appendix
- Appendix ....................................................................................................................................28

### References
- References ..................................................................................................................................26

*Source: _wp10270 - References — https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10270.pdf*

### 5. Estimation Results for Threshold MSVAR ...................................................................19

### 5. Estimation Results for Threshold MSVAR

### Overview and purpose
- Aim: examine whether the effects of monetary policy on the real economy in Armenia changed, and whether dollarization helps explain any regime switch, using:
  - a standard structural VAR (for comparison with earlier studies),
  - a Markov Switching VAR (MSVAR) that estimates regime switches endogenously,
  - a Markov Switching VAR with a threshold variable (credit dollarization).
- Data and sample references used in the analysis: monthly data spanning 2000M1-2010M5; VAR and MSVAR specifications use real GDP, CPI, the repo rate, monetary aggregates (M1 or M2/M2X as noted), and NEER; exogenous variables include a petroleum price index and the U.S. Federal Funds Rate.

### Key empirical findings (standard VAR and Granger causality)
- Granger causality p-values (multivariate and bivariate block tests):
  - Effect on output: Interest Rate (s) 0.00***; Money Supply (m) 0.00***; Exchange Rate (x) 0.05**; Block (p, s, x, m) 0.00***.
  - Effect on prices: Interest Rate (s) 0.96; Money Supply (m) 0.15; Exchange Rate (x) 0.03**; Block (p, s, x, m) 0.14.
  - Effect on NEER: Interest Rate (s) 0.03**; Money Supply (m) 0.06*; Block (p, s, x, m) 0.02**.
  - Note: *, **, and *** denote rejection of the null hypothesis at the 10, 5, and 1 percent significance levels, respectively.
- Standard VAR impulse responses:
  - Interest and money shocks have persistent effects on output.
  - Effects on prices are small and statistically insignificant in the baseline VAR estimation.
- Interpretation: the baseline evidence points to a weak direct interest-rate-to-inflation transmission; the exchange rate channel has a significant link to inflation.

### Evidence from Markov Switching VAR (MSVAR without explicit threshold)
- Model setup:
  - MSVAR estimated with 2 regimes using endogenous variables y (real GDP), p (CPI), s (repo rate), m (M2), x (NEER).
  - Sample: 2006M1-2010M5 for MSVAR estimation described.
  - Identification: Choleski decomposition; estimation via EM algorithm (Oxmetrics MSVAR package).
- MSVAR estimation results (selected statistics from Table 4):
  - Log-likelihood (non-linear) 1263.6902; linear model log-likelihood 1069.4012.
  - AIC (non-linear) -18.5641; AIC (linear) -16.4130.
  - Linearity LR test: 388.5780; Chi(60) =[0.0000] **; Chi(62)=[0.0000] **; DAVIES=[0.0000]**.
  - Transition matrix (estimated):
    - Regime1→Regime1: 0.9873
    - Regime1→Regime2: 0.01266
    - Regime2→Regime1: 1.862e-010
    - Regime2→Regime2: 1.000
- Regime timing and characterization:
  - Smoothed regime probabilities point to a regime switch in mid-2006.
  - Regimes are highly persistent per the transition matrix.
  - Coefficients differ between regimes: in regime 2 (post-switch) the repo rate has a significant effect on inflation and the exchange rate; in regime 1 the repo rate and exchange rate changes cause no significant effect on inflation or output.
- Interpretation: MSVAR supports a structural break around mid-2006 with a strengthening of the transmission mechanism thereafter.

### Evidence from Threshold MSVAR (dollarization as threshold)
- Method:
  - Threshold variable d_t = ratio of foreign currency loans to total loans (loan dollarization).
  - Threshold MSVAR estimates the threshold value D that maximizes conditional log-likelihood and divides regimes by d_t ≤ D versus d_t > D.
  - Variables transformed to ensure stationarity (logs and first differences); impulse responses computed regime-dependently assuming the threshold is exogenous.
- Estimated threshold and regime dating (Table 5):
  - Estimated threshold: 51
  - Periods of high dollarization: 2000:3 – 2006:11 and 2009:7 – 2010:5
  - Period of low dollarization: 2006:12 – 2009:6
  - Log likelihood: 1238.35
  - LR linearity test: 329.84; p-values (adjusted chi-squared) [0.0000]
- Interpretation of timing:
  - Threshold D = 51 percent implies the economy moved from high dollarization to low dollarization around late 2006, remained in low dollarization until mid-2009, and moved back to high dollarization afterwards.
  - The threshold MSVAR captures the mid-2006 switch and also identifies a second switch in mid-2009 that the MSVAR without threshold did not pick up (likely due to few observations in that final regime in the non-threshold MSVAR).
- Regime-dependent impulse responses (summarized results):
  - Repo rate shock:
    - A 100 basis point increase in the repo rate reduces GDP growth in both regimes; the impact under low dollarization is nearly 10 times stronger than under high dollarization.
    - The repo rate has a negative impact on inflation; under low dollarization the response of inflation is much larger but remains insignificant.
    - Conclusion: interest rate channel strengthened after dedollarization but remains weak in affecting inflation directly.
  - Money growth shock:
    - A 1 percentage point increase in money growth produces the expected sign for GDP growth (negative for a tightening) and is stronger under low dollarization but not significant in either regime.
    - Impact on inflation is significant under both regimes and stronger under low dollarization.
    - Interpretation: transmission from money to prices strengthened after end-2006; this suggests the inflation targeting framework remained rudimentary and money growth retained relevance.
  - NEER shock:
    - Exchange rate pass-through to growth and inflation appears stronger in the low dollarization period (2006–2009) than in the high dollarization periods.
    - Possible explanations discussed in the analysis:
      - Increased frequency and size of foreign exchange interventions during the low dollarization period may have enhanced the exchange rate’s role in shaping inflation expectations, increasing pass-through.
      - The weaker pass-through in the post-2009 high dollarization episode may reflect recessionary balance-sheet effects: a contractionary depreciation in March 2009 lowered demand and offset the usual inflationary effect of depreciation.
- Additional quantitative and graphical diagnostics from the threshold MSVAR:
  - Figures show regime probabilities and IRFs with 95 percent standard error bands (responses of GDP growth and inflation to repo, money, and NEER shocks for high vs low dollarization regimes).

### Synthesis of results
- Both MSVAR approaches identify a structural break in the second half of 2006.
  - Threshold MSVAR associates the break with credit dollarization falling below 51 percent.
  - Threshold MSVAR also identifies a return to high dollarization in mid-2009.
- Strengthening of the monetary transmission mechanism after 2006:
  - Evidence of stronger transmission from policy rate and money to output and prices in the low dollarization regime relative to the high dollarization regime.
  - Policy-rate-to-inflation channel strengthened but remains weak overall; money growth appears to have a stronger and significant effect on inflation in the low dollarization regime.
  - Exchange rate pass-through strengthened in the low dollarization period (pre-2009) possibly due to more active FX interventions; pass-through weakened again after mid-2009 amid recession and redollarization.
- Caveats and model assumptions:
  - The threshold MSVAR assumes dollarization is exogenous (shocks to the VAR do not change the regime); this restricts analysis of full endogeneity between monetary policy, macro shocks, and dollarization.
  - Short samples in later regimes limit the ability of unconstrained MSVAR to detect multiple switches.

### Policy implications and recommendations (from conclusions)
- Strengthen macroeconomic stability to support dedollarization:
  - A strong track record of macroeconomic stability is crucial to foster dedollarization, which in turn strengthens the MTM.
  - Given vulnerabilities (for example, a current account deficit of more than 15 percent noted in the analysis), measures to reduce external imbalances and fiscal consolidation to keep debt in check are necessary to support dedollarization efforts.
- Pursue consistent monetary and exchange rate policies with clear communication:
  - The central bank should reaffirm its commitment to price stability as the primary objective and improve communication of its framework to enhance credibility.
  - Excessive emphasis on the nominal exchange rate (for example via interventions) reinforces the exchange rate channel and may impede development of interest-rate and credit channels.
  - Active liquidity management is recommended so that the policy (repo) rate becomes more relevant and market short-term rates are less volatile.
  - When policy actions affect the exchange rate, messaging should clarify that the ultimate objective is inflation control, not the exchange rate per se.
- Structural reforms to deepen financial intermediation:
  - Continued development of local-currency money and securities markets, improvements in creditor protection and corporate governance, and stronger banking supervision will help strengthen interest-rate and credit channels over time.
  - Dedollarization, combined with deeper and more liquid domestic currency markets, will improve transmission from the policy rate to lending and deposit rates.

*Source: IMF working paper section “5. Estimation Results for Threshold MSVAR” (content provided).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10270.pdf_