## wpiea2020154-print-pdf ## Source details **Canonical URL:** [wpiea2020154-print-pdf](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020154-print-pdf.pdf) ## Other formats - [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020154-print-pdf.pdf.md) - [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020154-print-pdf.pdf.json) --- ### Introduction and scope - Study assesses the speed of exchange rate pass-through (ERPT) in seven CCA countries using monthly data for the period January 1995-May 2020. - Sample covers Armenia, Azerbaijan, Georgia, Kazakhstan, Kyrgyz Republic, Tajikistan, and Uzbekistan. Turkmenistan is excluded due to lack of sufficiently long time series. - Primary estimation method: local projections (Jorda, 2005) with country fixed effects and time fixed effects; Driscoll-Kraay standard errors are used. - Robustness check: bilateral panel VAR with Choleski identification (exchange rate changes ordered first). - Variables: Δlcpi = annual (year-on-year) logarithmic difference in the CPI index; Δler = annual (year-on-year) logarithmic difference in the exchange rate vis-à-vis the US dollar. - Twelve lags of dependent and independent variables are included in regressions to control for omitted variables and overlapping annual growth rates. - Granger-causality test (12 lags): exchange rate changes Granger-cause CPI inflation (p-value=0.00); CPI inflation does not Granger-cause exchange rate changes (p-value=0.10). ### Main quantitative findings (linear ERPT) - Average ERPT for the CCA panel: - "about 10 percent on impact" (impact = period 0; reported as inflation raises by 0.1 percent in response to a 1 percentage point change in exchange rate depreciation). - "about 25 percent after 12 months." - Panel VAR robustness check: - After 12 months, ERPT reaches "18 percent" (confidence interval overlaps with local projections estimate). - Comparative context: - Pass-through in the CCA is somewhat higher than in CIS countries (Comunale and Simola, 2018) and 28 emerging economies (Caselli and Roitman, 2016). - Pass-through in the CCA is lower compared to Sub-Saharan Africa countries (Razafimahefa, 2012). ### Non-linearities: size of exchange rate changes - Large versus small movements defined by a 15 percent threshold: large if annual exchange rate change exceed "15 percent" in absolute terms; small otherwise. - Findings: - No evidence of non-linearities with respect to the size of exchange rate changes. - ERPT in both cases is "about 10 percent on impact" and "about 30 percent after 12 months." - Confidence interval for large fluctuations is wider, indicating more heterogeneity for large changes. - Note: the 15 percent threshold corresponds to one standard deviation of the annual exchange rate growth in the total sample; results qualitatively unchanged for 10 percent, 20 percent, and 30 percent thresholds. ### Non-linearities: sign of exchange rate changes (appreciations vs depreciations) - Dummy d2 = 1 if Δler > 0 (appreciations), 0 otherwise. - Findings: - No significant evidence of asymmetric response to the sign of exchange rate fluctuations. - ERPT is "about 10 percent on impact" for both appreciations and depreciations. - After 12 months, pass-through for appreciations is about "10 percent higher" than for depreciations, but the difference is not statistically significant given wide confidence intervals. ### Exchange rate regimes - De-facto regime classification from IMF AREAER: d3 = 1 for free floating or managed floating with no pre-determined path; 0 for pegs, crawling bands, stabilized arrangements. - Findings: - No significant difference in ERPT between fixed and floating regimes. - In both regime groups ERPT is "about 10 percent on impact" and "about 25 percent after 12 months." ### Energy-exporting versus energy-importing countries - d4 = 1 for energy-importing CCA countries (ARM, GEO, KGZ, TJK); 0 for energy-exporting CCA countries (AZE, KAZ, UZB). - Findings: - Pass-through is lower in energy-exporting countries by about "2 percentage points on impact" and about "5 percentage points after 12 months." - This difference is significant at the "90 percent confidence level" for the first four months of the projection horizon. - Possible explanations: structural differences (economic diversification, reliance on remittances, proportion of imported goods in CPI baskets). ### Pre-crisis versus post-crisis (global financial crisis) comparison - d5 = 1 for September 2008–May 2020; 0 otherwise. - Findings: - ERPT was higher in the pre-crisis period: "13 percent on impact and 35 percent after 12 months." - ERPT is lower in the post-crisis period: "8 percent on impact and 20 percent after 12 months." - The decline (about "5 percentage points on impact" and "about 15 percentage points after 12 months") is statistically significant at the "90 percent confidence level." - Interpretation: lower pass-through in the post-crisis period is consistent with a lower inflation environment and structural improvements in monetary frameworks. ### Implications and conclusions - Summary: - Average ERPT in the CCA: "about 10 percent on impact" and "about 25 percent after 12 months." - No evidence of asymmetric ERPT with respect to the size and sign of exchange rate changes. - ERPT broadly unchanged across fixed versus floating regimes. - ERPT has declined by "5 – 15 percentage points" in the post-crisis period; lower ERPT is more relevant for the current relatively low inflation environment. - Policy implications: - The estimated speed and linearity of ERPT can inform monetary authorities’ inflation projections. - The absence of additional inflationary costs from non-linear ERPT across exchange rate regimes suggests that transition from fixed to floating regimes (including inflation targeting) would not impose additional inflationary costs. - The decline in ERPT following the global financial crisis reinforces the benefits of maintaining a low inflation environment for reducing exchange rate pass-through to consumer prices. ### Summary of empirical findings (selected studies) - Barhoumi (2006): 24 developing countries, 1980-2003(A). Pooled Mean Group (PMG) estimator. - ERPT=0.77-0.82 on average in the long-run. - Lower ERPT for flexible regimes. - Beirne and Bijsterbosch (2011): 9 Central and Eastern European countries, 1995-2008(Q). Cointegrated Vector Autoregression (CVAR). - ERPT=0.5-0.6 on average in the long-run. - Lower ERPT for flexible regimes. - Ben Cheikh (2012): 12 euro area countries, 1975-2010(Q). Logistic Smooth Transition Model. - ERPT=0.04-0.19 (depending on a country) in the short-run. - Higher ERPT for larger ER changes (threshold is estimated). - Mixed evidence: for some countries higher ERPT for larger ER appreciations, the opposite holds for others. - Bussiere (2013): G7 countries, 1980-2006(Q). Panel OLS. - ERPT=0.27 for export prices and 0.55 for import prices in the long-run. - Lower ERPT for larger ER depreciations (import prices). - Higher ERPT for appreciations (export and import prices). - Caselli and Roitman (2016): 28 emerging markets, 1991-2014(M). Local projections. - ERPT=0.2 after 2 years. - Higher ERPT for larger ER depreciations (10 or 20 percent). - Lower ERPT for appreciations. - Lower ERPT for flexible regimes (under IT). - Colavecchio and Rubene (2020): 19 euro area countries, 1997-2019(Q). Local projections. - ERPT=0.04 after 2 years. - Higher ERPT for larger ER depreciations (2.3 percent). - Higher ERPT for appreciations. - Comunale and Simola (2018): 7 Commonwealth of Independent States countries, 1999-2014(Q). Mean group estimator corrected for cross-sectional dependence. - ERPT=0.12-0.13 after 1 year. - Higher ERPT for larger ER depreciations (2 percent). - No evidence for asymmetric sign effects. - Delatte and Villavicencio (2012): 4 major developed countries, 1980-2009(Q). Autoregressive Distributed Lag Model (ARDL). - ERPT=0.07-0.46 (depending on a country) in the long-run. - Higher ERPT for depreciations. - Ito and Saito (2008): 5 East Asian countries, 1994-2006(M). Vector Autoregression (VAR). - ERPT=0.03-0.41 (depending on a country) after 2 years. - Lower ERPT (except Indonesia) after the Asian crisis. - Jasova and others (2019): 22 EMs and 11 AEs, 1994-2017(Q). Panel OLS with fixed effects. - ERPT is lower in AEs compared to EMs, ERPT has been declining in EMs over time. - Higher ERPT for larger ER depreciations. - Lower ERPT for post-crisis period (starting from 2009Q3). - Kilic (2016): 6 OECD countries, 1975-2010(Q). Logistic Smooth Transition Model. - ERPT=0.42-0.91 for import prices (depending on a country) in the long-run. - Higher ERPT for larger ER appreciations (threshold is estimated). - Kohlscheen (2010): emerging economies, 1995-2008(M). Vector Autoregression (VAR). - ERPT=0.09-0.59 (depending on a country) after 1 year. - Higher ERPT for flexible regimes. - Razafimahefa (2012): Sub-Saharan Africa region, 1985-2008(Q). Dynamic panel OLS. - ERPT=0.4 (on average) after 1 year. - Lower ERPT for flexible regimes. ### Non-linearities and asymmetric pass-through (cross-study patterns) - Evidence of non-linearities with respect to the size of exchange rate changes: - Several studies report higher ERPT for larger ER depreciations (e.g., Caselli and Roitman 2016; Comunale and Simola 2018; Jasova et al. 2019). - Other studies find higher ERPT for larger ER appreciations or mixed results depending on country and threshold estimation (e.g., Ben Cheikh 2012; Kilic 2016; Bussiere 2013). - Evidence on sign asymmetry (appreciations vs. depreciations): - Some studies find higher ERPT for depreciations (Delatte and Villavicencio 2012; Jasova et al. 2019). - Some studies find higher ERPT for appreciations (Bussiere 2013; Colavecchio and Rubene 2020). - Other studies report no asymmetric sign effects for certain samples (Comunale and Simola 2018). - Role of exchange rate regime: - Multiple studies report lower ERPT under flexible regimes or inflation targeting (Barhoumi 2006; Beirne and Bijsterbosch 2011; Caselli and Roitman 2016; Razafimahefa 2012). - Kohlscheen (2010) finds higher ERPT for flexible regimes in his sample. ### Data and variable definitions (sample context) - Sample context noted for the accompanying analysis: - The sample includes seven CCA countries (Armenia, Azerbaijan, Georgia, Kazakhstan, Kyrgyz Republic, Tajikistan, and Uzbekistan) over the period January 1995-May 2020. - Key variables and sources: - ER: Exchange rate of the local currency vis-à-vis the US dollar (daily average) — International Financial Statistics, IMF. - CPI: Consumer price index (seasonally adjusted, base month is January, 2010) — International Financial Statistics, IMF. - DEFACTO: De-facto exchange rate classification — Annual Report of Exchange Arrangements and Exchange Restrictions, IMF. ### Main aggregated results (Table 3) - ERPT= 10 percent on impact - ERPT= 25 percent after 12 months - Linear with respect to the size of exchange rate changes. - Linear with respect to the exchange rate regime (fixed versus floating). - Lower in the post-global financial crisis period. - Linear with respect to the sign of exchange rate changes. *Source: IMF staff estimations using International Financial Statistics; estimations performed for Armenia, Azerbaijan, Georgia, Kazakhstan, Kyrgyz Republic, Tajikistan, and Uzbekistan over the period January 1995-May 2020.* ### References ________________________________________________________________28 ### References ________________________________________________________________28 ### Figures - 1. Dynamics of CPI Inflation in the CCA Countries _______________________________14 - 2. Dynamics of Exchange Rate Changes in the CCA Countries ______________________15 - 3. Association Between CPI Inflation and Exchange Rate Changes in the CCA Countries _16 - 4. ERPT: The Linear Model __________________________________________________17 - 5. ERPT: The Panel VAR Model ______________________________________________18 - 6. ERPT: Non-Linearities for the Size of Exchange Rate Depreciation _________________19 - 7. ERPT: Non-Linearities for Exchange Rate Appreciation versus Depreciation _________20 - 8. ERPT: Non-Linearities for Exchange Rate Regimes _____________________________21 - 9. ERPT: Non-Linearities for Pre-Crisis and Post-Crisis Periods _____________________22 - 10. ERPT: Non-Linearities for Pre-Crisis and Post-Crisis Periods ____________________23 ### Tables - 1. Summary of Selected Exchange Rate Pass-Through Studies _______________________24 - 2. Variables and Their Sources ________________________________________________26 *References ________________________________________________________________28* ### 3. Summary of Main Results _________________________________________________27 ### 3. Summary of Main Results ### Introduction and scope - Study assesses the speed of exchange rate pass-through (ERPT) in seven CCA countries using monthly data for the period January 1995-May 2020. - Sample covers Armenia, Azerbaijan, Georgia, Kazakhstan, Kyrgyz Republic, Tajikistan, and Uzbekistan. Turkmenistan is excluded due to lack of sufficiently long time series. - Primary estimation method: local projections (Jorda, 2005) with country fixed effects and time fixed effects; Driscoll-Kraay standard errors are used. - Robustness check: bilateral panel VAR with Choleski identification (exchange rate changes ordered first). - Variables: Δlcpi = annual (year-on-year) logarithmic difference in the CPI index; Δler = annual (year-on-year) logarithmic difference in the exchange rate vis-à-vis the US dollar. - Twelve lags of dependent and independent variables are included in regressions to control for omitted variables and overlapping annual growth rates. - Granger-causality test (12 lags): exchange rate changes Granger-cause CPI inflation (p-value=0.00); CPI inflation does not Granger-cause exchange rate changes (p-value=0.10). ### Main quantitative findings (linear ERPT) - Average ERPT for the CCA panel: - "about 10 percent on impact" (impact = period 0; reported as inflation raises by 0.1 percent in response to a 1 percentage point change in exchange rate depreciation). - "about 25 percent after 12 months." - Panel VAR robustness check: - After 12 months, ERPT reaches "18 percent" (confidence interval overlaps with local projections estimate). - Comparative context: - Pass-through in the CCA is somewhat higher than in CIS countries (Comunale and Simola, 2018) and 28 emerging economies (Caselli and Roitman, 2016). - Pass-through in the CCA is lower compared to Sub-Saharan Africa countries (Razafimahefa, 2012). ### Non-linearities: size of exchange rate changes - Large versus small movements defined by a 15 percent threshold: large if annual exchange rate change exceed "15 percent" in absolute terms; small otherwise. - Findings: - No evidence of non-linearities with respect to the size of exchange rate changes. - ERPT in both cases is "about 10 percent on impact" and "about 30 percent after 12 months." - Confidence interval for large fluctuations is wider, indicating more heterogeneity for large changes. - Note: the 15 percent threshold corresponds to one standard deviation of the annual exchange rate growth in the total sample; results qualitatively unchanged for 10 percent, 20 percent, and 30 percent thresholds. ### Non-linearities: sign of exchange rate changes (appreciations vs depreciations) - Dummy d2 = 1 if Δler > 0 (appreciations), 0 otherwise. - Findings: - No significant evidence of asymmetric response to the sign of exchange rate fluctuations. - ERPT is "about 10 percent on impact" for both appreciations and depreciations. - After 12 months, pass-through for appreciations is about "10 percent higher" than for depreciations, but the difference is not statistically significant given wide confidence intervals. ### Exchange rate regimes - De-facto regime classification from IMF AREAER: d3 = 1 for free floating or managed floating with no pre-determined path; 0 for pegs, crawling bands, stabilized arrangements. - Findings: - No significant difference in ERPT between fixed and floating regimes. - In both regime groups ERPT is "about 10 percent on impact" and "about 25 percent after 12 months." ### Energy-exporting versus energy-importing countries - d4 = 1 for energy-importing CCA countries (ARM, GEO, KGZ, TJK); 0 for energy-exporting CCA countries (AZE, KAZ, UZB). - Findings: - Pass-through is lower in energy-exporting countries by about "2 percentage points on impact" and about "5 percentage points after 12 months." - This difference is significant at the "90 percent confidence level" for the first four months of the projection horizon. - Possible explanations: structural differences (economic diversification, reliance on remittances, proportion of imported goods in CPI baskets). ### Pre-crisis versus post-crisis (global financial crisis) comparison - d5 = 1 for September 2008–May 2020; 0 otherwise. - Findings: - ERPT was higher in the pre-crisis period: "13 percent on impact and 35 percent after 12 months." - ERPT is lower in the post-crisis period: "8 percent on impact and 20 percent after 12 months." - The decline (about "5 percentage points on impact" and "about 15 percentage points after 12 months") is statistically significant at the "90 percent confidence level." - Interpretation: lower pass-through in the post-crisis period is consistent with a lower inflation environment and structural improvements in monetary frameworks. ### Implications and conclusions - Summary: - Average ERPT in the CCA: "about 10 percent on impact" and "about 25 percent after 12 months." - No evidence of asymmetric ERPT with respect to the size and sign of exchange rate changes. - ERPT broadly unchanged across fixed versus floating regimes. - ERPT has declined by "5 – 15 percentage points" in the post-crisis period; lower ERPT is more relevant for the current relatively low inflation environment. - Policy implications: - The estimated speed and linearity of ERPT can inform monetary authorities’ inflation projections. - The absence of additional inflationary costs from non-linear ERPT across exchange rate regimes suggests that transition from fixed to floating regimes (including inflation targeting) would not impose additional inflationary costs. - The decline in ERPT following the global financial crisis reinforces the benefits of maintaining a low inflation environment for reducing exchange rate pass-through to consumer prices. *Source: IMF staff estimations using International Financial Statistics; estimations performed for Armenia, Azerbaijan, Georgia, Kazakhstan, Kyrgyz Republic, Tajikistan, and Uzbekistan over the period January 1995-May 2020.* ### 1. Size of exchange rate ### 1. Size of exchange rate ### Summary of empirical findings (selected studies) - Barhoumi (2006): 24 developing countries, 1980-2003(A). Pooled Mean Group (PMG) estimator. - ERPT=0.77-0.82 on average in the long-run. - Lower ERPT for flexible regimes. - Beirne and Bijsterbosch (2011): 9 Central and Eastern European countries, 1995-2008(Q). Cointegrated Vector Autoregression (CVAR). - ERPT=0.5-0.6 on average in the long-run. - Lower ERPT for flexible regimes. - Ben Cheikh (2012): 12 euro area countries, 1975-2010(Q). Logistic Smooth Transition Model. - ERPT=0.04-0.19 (depending on a country) in the short-run. - Higher ERPT for larger ER changes (threshold is estimated). - Mixed evidence: for some countries higher ERPT for larger ER appreciations, the opposite holds for others. - Bussiere (2013): G7 countries, 1980-2006(Q). Panel OLS. - ERPT=0.27 for export prices and 0.55 for import prices in the long-run. - Lower ERPT for larger ER depreciations (import prices). - Higher ERPT for appreciations (export and import prices). - Caselli and Roitman (2016): 28 emerging markets, 1991-2014(M). Local projections. - ERPT=0.2 after 2 years. - Higher ERPT for larger ER depreciations (10 or 20 percent). - Lower ERPT for appreciations. - Lower ERPT for flexible regimes (under IT). - Colavecchio and Rubene (2020): 19 euro area countries, 1997-2019(Q). Local projections. - ERPT=0.04 after 2 years. - Higher ERPT for larger ER depreciations (2.3 percent). - Higher ERPT for appreciations. - Comunale and Simola (2018): 7 Commonwealth of Independent States countries, 1999-2014(Q). Mean group estimator corrected for cross-sectional dependence. - ERPT=0.12-0.13 after 1 year. - Higher ERPT for larger ER depreciations (2 percent). - No evidence for asymmetric sign effects. - Delatte and Villavicencio (2012): 4 major developed countries, 1980-2009(Q). Autoregressive Distributed Lag Model (ARDL). - ERPT=0.07-0.46 (depending on a country) in the long-run. - Higher ERPT for depreciations. - Ito and Saito (2008): 5 East Asian countries, 1994-2006(M). Vector Autoregression (VAR). - ERPT=0.03-0.41 (depending on a country) after 2 years. - Lower ERPT (except Indonesia) after the Asian crisis. - Jasova and others (2019): 22 EMs and 11 AEs, 1994-2017(Q). Panel OLS with fixed effects. - ERPT is lower in AEs compared to EMs, ERPT has been declining in EMs over time. - Higher ERPT for larger ER depreciations. - Lower ERPT for post-crisis period (starting from 2009Q3). - Kilic (2016): 6 OECD countries, 1975-2010(Q). Logistic Smooth Transition Model. - ERPT=0.42-0.91 for import prices (depending on a country) in the long-run. - Higher ERPT for larger ER appreciations (threshold is estimated). - Kohlscheen (2010): emerging economies, 1995-2008(M). Vector Autoregression (VAR). - ERPT=0.09-0.59 (depending on a country) after 1 year. - Higher ERPT for flexible regimes. - Razafimahefa (2012): Sub-Saharan Africa region, 1985-2008(Q). Dynamic panel OLS. - ERPT=0.4 (on average) after 1 year. - Lower ERPT for flexible regimes. ### Non-linearities and asymmetric pass-through (cross-study patterns) - Evidence of non-linearities with respect to the size of exchange rate changes: - Several studies report higher ERPT for larger ER depreciations (e.g., Caselli and Roitman 2016; Comunale and Simola 2018; Jasova et al. 2019). - Other studies find higher ERPT for larger ER appreciations or mixed results depending on country and threshold estimation (e.g., Ben Cheikh 2012; Kilic 2016; Bussiere 2013). - Evidence on sign asymmetry (appreciations vs. depreciations): - Some studies find higher ERPT for depreciations (Delatte and Villavicencio 2012; Jasova et al. 2019). - Some studies find higher ERPT for appreciations (Bussiere 2013; Colavecchio and Rubene 2020). - Other studies report no asymmetric sign effects for certain samples (Comunale and Simola 2018). - Role of exchange rate regime: - Multiple studies report lower ERPT under flexible regimes or inflation targeting (Barhoumi 2006; Beirne and Bijsterbosch 2011; Caselli and Roitman 2016; Razafimahefa 2012). - Kohlscheen (2010) finds higher ERPT for flexible regimes in his sample. ### Data and variable definitions (sample context) - Sample context noted for the accompanying analysis: - The sample includes seven CCA countries (Armenia, Azerbaijan, Georgia, Kazakhstan, Kyrgyz Republic, Tajikistan, and Uzbekistan) over the period January 1995-May 2020. - Key variables and sources: - ER: Exchange rate of the local currency vis-à-vis the US dollar (daily average) — International Financial Statistics, IMF. - CPI: Consumer price index (seasonally adjusted, base month is January, 2010) — International Financial Statistics, IMF. - DEFACTO: De-facto exchange rate classification — Annual Report of Exchange Arrangements and Exchange Restrictions, IMF. ### Main aggregated results (Table 3) - ERPT= 10 percent on impact - ERPT= 25 percent after 12 months - Linear with respect to the size of exchange rate changes. - Linear with respect to the exchange rate regime (fixed versus floating). - Lower in the post-global financial crisis period. - Linear with respect to the sign of exchange rate changes. *Source: IMF Working Paper (tables and study summaries from the provided content).* --- _Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020154-print-pdf.pdf_