## wpiea2022173-print-pdf

## Source details

**Canonical URL:** [wpiea2022173-print-pdf](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022173-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022173-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022173-print-pdf.pdf.json)

---

### Data coverage and construction
- Two complementary quarterly datasets:
  - Broad dataset: covers 39 countries in Europe comprising 23 advanced and 16 emerging economies, starting in 2000. Wages cover as broad industries as possible, usually for the whole economy or all industries excluding construction.
  - Long dataset: covers 24 mostly advanced economies, starting in the 1960s. Wages cover the manufacturing sector only.
- Oil price measure: simple average of three spot crude oil prices (Dated Brent, West Texas Intermediate, and the Dubai Fateh).
- Country lists (as stated):
  - Advanced European countries—Austria, Belgium, Cyprus, Czech Republic, Demark, Estonia, Finland, France, Germany, Greece, Ireland, Israel, Italy, Latvia, Lithuania, Netherlands, Norway, Portugal, Slovak Republic, Slovenia, Spain, Sweden, and United Kingdom.
  - Emerging European countries—Albania, Belarus, Bosnia and Herzegovina, Bulgaria, Croatia, Hungary, Kosovo, Moldova, Montenegro, North Macedonia, Poland, Romania, Russia, Serbia, Türkiye, and Ukraine.
- Sample period for both datasets ends in 2019Q4.

### Stylized facts on oil prices and inflation
- Recent inflation context cited:
  - 8.6 percent in the U.S. in May 2022 (the highest in 40 years).
  - 8.1 percent in the euro area (the highest since the creation of euro).
- Historical oil price inflation (annual, US dollar):
  - Highest observed oil price inflation: 184 percent in 1974.
  - 130 percent in 1980.
  - 120 percent in 2000.
  - Recurrent spikes exceeding 50 percent in 2003, 2004, 2007-08, 2010, and 2021.
- Median inflation dynamics:
  - Median wage inflation peaked at 19 percent in 1975.
  - Median core inflation peaked at 13 percent in 1975.
  - Inflation rates moderated to around 5 percent in 2000 and remained modest thereafter in the long dataset.
  - Median inflation is more volatile in the broad dataset due to inclusion of more emerging economies.

### Labor market and institutional trends
- Labor market liberalization since the 1980s:
  - Measures related to wage-setting flexibility (degree of unionization, coverage of collective bargaining, centralization of collective bargaining) have been liberalized since the 1980s after earlier trends in the 1960s and 1970s.
- Emerging European economies (seven EEs with data):
  - Median union density dropped sharply in the 1990s from nearly 100 percent to around 25 percent in 2000 and has remained lower than the advanced-Europe median by about 5 percentage points thereafter.
- Constructed binary indicator:
  - A combined binary indicator takes the value of 1 if degrees of unionization and centralization are both above their median values; used to capture highly unionized and centralized bargaining regimes.
- Monetary policy credibility:
  - Institutional arrangements for monetary policy strengthened especially in the 2000s.
  - An index of inflation expectations’ anchoring and a central bank independence index both show marked improvements in the 2000s (data coverage 1994-2017).

### Key empirical approach and interpretation
- Methodology:
  - Impulse response functions estimated with panel data using the local projection method proposed by Jorda (2005).
  - Benchmark specification: 4 lags of quarterly growth in wages and oil prices; controls for cyclical position and non-oil terms of trade shocks via quarterly changes in unemployment rates and nominal effective exchange rates (NEER), both up to 4 quarter lags.
  - Long dataset specification: quarterly log differences in industrial production and nominal exchange rates against the US dollar.
  - Results robust to choice of lag lengths.
- Definition:
  - Pass-through reported as the response of cumulative wage growth to a shock to quarterly oil price inflation by 1 percentage point; equivalent to the percent increase in wage level relative to the pre-shock wage level.

### Baseline impulse responses (main estimated responses)
- Broad dataset (2000Q1–2019Q4, 39 countries):
  - Wages rise by about 0.015 percent in 1 quarter after an oil price shock (increase by 1 percentage point).
  - Cumulative response stabilizes around 0.025 percent by 1 year after the shock.
  - Autoregressive terms for wages are statistically significant in the first three lags consistently.
  - Lagged oil price shocks lose statistical significance beyond one lag.
  - Control variables: wages rise with lower unemployment rates and fall with exchange rate appreciation.
- Long dataset (1960Q1–2019Q4, 24 countries):
  - Wages increase by 0.006 percent in one quarter after the shock.
  - Wages increase by 0.03 percent in about 3 years after the shock.
- Subsample and time variation (long dataset):
  - Pass-through coefficients higher in the subsample covering the 1960s and 1970s, with the response in wages exceeding 0.1 percent in 1½ year after the shock.
  - Almost zero pass-through for 1980-1999.
  - Around 0.02 percent pass-through for 2000-2019.
- Advanced economies (AEs) versus emerging economies (EEs) (broad dataset, 2000Q1–2019Q4):
  - AEs: elasticity of wages to oil price shock peaks around 0.02 percent in 3 quarters.
  - EEs: elasticity nears 0.05 percent in 1½ years after the shock.
  - Unemployment rates statistically significant only in the AE sample.
  - NEER statistically significant in both groups, with larger effects for EEs.
- Practical horizon note:
  - Median quarterly wage growth no longer responds to oil price shocks, or slightly declines, after 4 quarters.

### Pass-through and structural factors
- Estimation with structural interactions:
  - Structural variables z_{t,j} converted to binary series (1 if value greater than the median, 0 otherwise) and each estimated separately to preserve country coverage.
- Energy weights in consumption basket:
  - CPI energy weights data available for 34 countries.
  - CPI energy weights distribution: interquartile range of 8.7 and 13 percent.
  - Regression (broad dataset) finds interaction of energy weights with oil price shock is positively and statistically significantly related to wage growth.
  - Average energy weight: EEs 12.7 percent; AEs 10 percent.
  - Energy weights omitted from long-dataset regressions because energy weight data available only from 1999.
- Labor market institutions and wage bargaining:
  - Focus on three variables that remain statistically significant as stand-alone controls: union density, collective bargaining coverage rate, and a combined indicator for highly unionized and centralized wage bargaining.
  - Degree of centralization of bargaining only weakly significant unless combined with union density; combined binary “high” indicates both unionization and centralization above their medians.
  - Although bargaining coverage coefficient often statistically significant, it usually has a negative sign.
- Median values used for binary conversion:
  - Long dataset medians: 35 percent for the union density, 80 percent for the collective bargaining coverage, 2.5 for the bargaining centralization, 0.2 for the monetary policy anchoring index, and 7 for the central bank transparency index.
  - Broad dataset medians: 10.7 percent for the energy weights, 28 percent for the union density, 78 percent for the collective bargaining coverage, 2.4 for the bargaining centralization, 0.22 for the monetary policy anchoring index, and 7.5 for the central bank transparency index.
- Monetary policy credibility:
  - On average higher in AEs than in EEs during the sample period.
  - Pass-through to wages is weaker if inflation expectations are better anchored and central banks are more independent.
  - Weaker pass-through under high values of the anchoring index with statistically significant differences lasting for a year after the shock.
  - Interaction terms with central bank independence are statistically significant three quarters after the shock and thereafter.

### Economic states and heterogeneity
- Pre-shock inflation (broad dataset, 2000-2019):
  - With threshold of 4 percent:
    - Pass-through elasticity ≈ 0.04 in about a year after the shock for high inflation states.
    - Corresponding elasticity ≈ 0.01 for low inflation states.
  - With threshold of 6 percent:
    - Elasticity in high inflation states ≈ 0.07 versus 0.02 in low inflation states (statistically significant).
  - No statistically significant differences when threshold is 2 percent.
- Business cycle phase:
  - Wages respond more strongly for about a year if the economy was in an expansionary phase at the time of the shock (binary indicator based on deviations of unemployment from HP trends).
  - Relationship is short-lived.
- Large oil price booms (annual oil price inflation threshold set at 35 percent):
  - Long dataset (1960-2019): no significant differences in wage reactions during booms.
  - Broad dataset (2000-2019): shocks during large oil price booms have significantly weaker pass-through to wages.

### Transmission via CPI inflation (oil-induced CPI inflation → wages)
- Instrumenting CPI inflation with oil price shocks (lags up to 4 quarters) in broad dataset:
  - Oil-induced rise in CPI inflation leads to a rise in wages of about the same magnitude in two quarters following the rise in inflation.
  - Impacts thereafter ease to about a half in AEs, while wage increases in EEs are twice as large as the rise in CPI inflation in the second year.
- Consistency:
  - Pass-through from oil-induced inflation to wages is higher if the economy is highly unionized, has less-anchored inflation expectations, and has high pre-existing inflation.

### Pass-through to core and expected inflation
- Timing and magnitudes:
  - First quarter after oil shock: pass-through elasticity for core inflation = 0.002 versus 0.015 for wages.
  - Core inflation rises to 0.03 in three years after the shock.
  - Expected inflation peak elasticity = 0.01 in 3 quarters after the shock (about half of the wage elasticity); expected inflation is affected for about two years.
- Heterogeneity:
  - Both core and expected inflation react more strongly in emerging economies (EEs) than in advanced economies (AEs), with greater dispersion among EEs.
  - More independent central banks are associated with smaller pass-through of oil shocks to core and expected inflation.
  - Pre-existing high inflation increases pass-through strength; interaction terms statistically significant for expected and core inflation a year after the shock with a 6 percent threshold.

### Sources of oil price shocks and pass-through
- Decomposition (structural VAR, January 1973–December 2021, aggregated to quarterly):
  - Three shocks: oil supply shocks (innovations to global oil production), aggregate demand shocks (innovations to global real economic activity not explained by supply shocks), oil-specific demand shocks (innovations to real oil price not explained by supply or aggregate demand shocks).
  - Aggregate demand shocks increased in importance in the 2000s and were main driver of real oil prices between 2005 and 2012.
  - Oil supply shocks played a smaller role; oil-specific demand shocks had large impacts in the 2010s.
- Pass-through by shock type (broad dataset, 2000-2019, 2SLS with shocks as instruments):
  - All three shock types trigger a rise in wages.
  - Aggregate demand–induced price rises: pass-through more immediate and stronger, impact ≈ 0.06 in 2 quarters after the shock.
  - Oil-specific demand shocks: elasticity rises to ≈ 0.01 in 2 quarters after the shock (positive and statistically significant).
  - Supply-driven price increases: reactions more gradual, peaking at ≈ 0.06 in about 1½ years after the shock.
- Implication:
  - Similarity in pass-through coefficients across shock types supports treating global oil price changes as exogenous for analysis of mostly oil-importing small-open economies.

### Taking stock — key empirical conclusions
- Risk of wage-price spirals:
  - Results indicate a limited risk of a wage-price spiral and regime shift in inflation expectations in response to supply shocks.
  - No evidence that oil price shocks led to a sustained wage-price spiral that would have pushed up the inflation rate for a sustained period in the decades studied.
  - Pass-through from oil price shocks to wages and core inflation peaked in 1-2 years and largely dissipated in 3-4 years.
  - Evidence holds in both the broad dataset (post-Volker era) and the long dataset (including earlier periods).
  - The bar for a wage-price spiral being unleashed by oil price shocks would be quite high under the current policy framework and labor market institutions.
- Cross-cutting patterns on determinants of pass-through:
  - Pass-through depends on the level of inflation at the time of the shock, the cyclical state of activity, and structural characteristics.
  - Pass-through appears higher when overall inflation is higher at the time of shocks.
  - Pass-through appears higher in emerging than advanced European economies.
  - Pass-through appears lower when central banks anchor inflation expectations well or are more independent.
  - Higher centralized bargaining cum unionization seems associated with a higher pass-through.
  - Headline inflation and wages are found to move by about the same magnitude in response to oil price shocks.

### Caveats, limits, and suggestions for future research
- Caveats and limits:
  - Results are conditional on the underlying policy and institutional frameworks in operation, especially in the post-2000 period; they do not constitute prescriptive evidence for the “optimal” policy responses.
  - Cross-country estimation cannot capture particular episodes in some countries and times that might indicate an elevated risk of a wage-price spiral or unmooring inflation expectations in the current juncture.
  - Authors note a risk that high headline inflation itself could intensify second-round effects and expected inflation (see BIS (2022) for a similar argument).
- Suggestions for future research:
  - Broaden country coverage to provide greater variation in data, acknowledging this may introduce larger heterogeneity complicating econometric investigation.
  - Undertake more analysis of why pass-through varies with the level of inflation and the implications for monetary policy and the choice of inflation targets.
  - Apply different econometric or quantitative methodologies to sharpen some of the results.

*Source: IMF Working Paper — Second-Round Effects of Oil Price Shocks; Annex I. Data Description and excerpts from Sections 4 and 6 (as provided).*

### Annex I. Data Description ..............................................................................................

### Annex I. Data Description

### Data coverage and construction
- Two complementary quarterly datasets:
  - Broad dataset: covers 39 countries in Europe comprising 23 advanced and 16 emerging economies, starting in 2000. Wages cover as broad industries as possible, usually for the whole economy or all industries excluding construction.
  - Long dataset: covers 24 mostly advanced economies, starting in the 1960s. Wages cover the manufacturing sector only.
- Oil price measure: simple average of three spot crude oil prices (see Annex I for sources).
- Country lists (as stated): Advanced European countries—Austria, Belgium, Cyprus, Czech Republic, Demark, Estonia, Finland, France, Germany, Greece, Ireland, Israel, Italy, Latvia, Lithuania, Netherlands, Norway, Portugal, Slovak Republic, Slovenia, Spain, Sweden, and United Kingdom. Emerging European countries—Albania, Belarus, Bosnia and Herzegovina, Bulgaria, Croatia, Hungary, Kosovo, Moldova, Montenegro, North Macedonia, Poland, Romania, Russia, Serbia, Türkiye, and Ukraine.

### Stylized facts on oil prices and inflation
- Recent inflation context cited:
  - 8.6 percent in the U.S. in May 2022 (the highest in 40 years).
  - 8.1 percent in the euro area (the highest since the creation of euro).
- Historical oil price inflation (annual, US dollar):
  - Highest observed oil price inflation: 184 percent in 1974.
  - 130 percent in 1980.
  - 120 percent in 2000.
  - Recurrent spikes exceeding 50 percent in 2003, 2004, 2007-08, 2010, and 2021.
- Median inflation dynamics:
  - Median wage inflation peaked at 19 percent in 1975.
  - Median core inflation peaked at 13 percent in 1975.
  - Inflation rates moderated to around 5 percent in 2000 and remained modest thereafter in the long dataset.
  - Median inflation is more volatile in the broad dataset due to inclusion of more emerging economies.

### Labor market and institutional trends
- Labor market liberalization since the 1980s:
  - Measures related to wage-setting flexibility (degree of unionization, coverage of collective bargaining, centralization of collective bargaining) have been liberalized since the 1980s after earlier trends in the 1960s and 1970s.
- Specific trends in emerging European economies (seven EEs with data):
  - Median union density dropped sharply in the 1990s from nearly 100 percent to around 25 percent in 2000 and has remained lower than the advanced-Europe median by about 5 percentage points thereafter.
- Constructed binary indicator:
  - A combined binary indicator takes the value of 1 if degrees of unionization and centralization are both above their median values; used to capture highly unionized and centralized bargaining regimes.
- Monetary policy credibility:
  - Institutional arrangements for monetary policy strengthened especially in the 2000s (e.g., adoption of the euro in 1999, expanded inflation targeting). An index of inflation expectations’ anchoring (based on deviations of long-term inflation forecasts from central bank targets) and a central bank independence index both show marked improvements in the 2000s (data coverage 1994-2017).

### Key empirical findings on second-round effects and pass-through
- Primary focus: estimate pass-through from oil price shocks to wages (direct measure) and how pass-through varies with states and structural characteristics.
- Main empirical regularities:
  - The pass-through from oil prices to wages is state-dependent.
  - Pass-through is higher when overall inflation is higher.
  - Cyclical conditions affect pass-through, albeit with weak statistical significance.
  - The pass-through has declined over time in Europe (broadly speaking).
  - Pass-through is higher under lower credibility of monetary policy.
  - Pass-through is higher in emerging European economies.
  - Labor market characteristics: no simple characterization; pass-through is often higher when high centralization of collective bargaining is combined with a high degree of unionization.
- Relationship between oil-induced headline inflation and wages:
  - Headline inflation induced by oil price shocks pushes up wages about one-to-one in the near term; the strength of this pass-through is state-dependent.
- Core inflation and inflation expectations:
  - Core inflation and expected inflation respond more slowly to oil price shocks than wages.
  - Strength of responses of core and expected inflation also depends on structural characteristics and economic states.
- Source of oil shocks:
  - Oil price changes have similar effects on wages whether induced by global demand or supply shocks to the oil market, supporting the premise that global oil price changes are exogenous to individual economies.

*Source: IMF Working Paper — Second-Round Effects of Oil Price Shocks; Annex I. Data Description (as provided).*

### 4. Results

### 4. Results

### Baseline impulse responses
- Methodology and sample
  - Impulse response functions estimated with panel data using the local projection method proposed by Jorda (2005); method noted as less sensitive to misspecification and econometrically equivalent to a VAR (Plagborg-Moller et. al 2021).
  - Oil price shocks treated as exogenous to individual economies in the sample.
  - Sample period for both datasets ends in 2019Q4.
  - Benchmark specification: 4 lags of quarterly growth in wages and oil prices; controls for cyclical position and non-oil terms of trade shocks via quarterly changes in unemployment rates and nominal effective exchange rates (NEER), both up to 4 quarter lags.
  - Long dataset specification: quarterly log differences in industrial production and nominal exchange rates against the US dollar (due to data availability before the 1980s).
  - Results robust to choice of lag lengths.

- Definition and interpretation
  - Pass-through (impulse responses) reported as the response of cumulative wage growth to a shock to quarterly oil price inflation by 1 percentage point; this is equivalent to the percent increase in wage level relative to the pre-shock wage level.

- Main estimated responses (broad dataset, 2000Q1–2019Q4, 39 countries in Europe)
  - Wages rise by about 0.015 percent in 1 quarter after an oil price shock (increase by 1 percentage point).
  - Cumulative response stabilizes around 0.025 percent by 1 year after the shock.
  - Autoregressive terms for wages are statistically significant in the first three lags consistently.
  - Lagged oil price shocks do not enter with the expected sign and lose statistical significance beyond one lag.
  - Control variables: wages rise with lower unemployment rates and fall with exchange rate appreciation.

- Long dataset results (1960Q1–2019Q4, 24 countries in Europe)
  - Wages increase by 0.006 percent in one quarter after the shock.
  - Wages increase by 0.03 percent in about 3 years after the shock.
  - Pass-through elasticities from long and broad datasets are of the same order of magnitude as pass-through estimates to consumer price inflation in Choi et al. (2017) and Conflitti and Luciani (2017), though direct comparison limited by differences in sample and variable of interest.

- Subsample and time variation (long dataset)
  - Pass-through coefficients higher in the subsample covering the 1960s and 1970s, with the response in wages exceeding 0.1 percent in 1½ year after the shock.
  - Almost zero pass-through for 1980-1999.
  - Around 0.02 percent pass-through for 2000-2019.
  - Decline in pass-through to wages consistent with evidence of lower economic effects of oil shocks on macro variables since the mid-1980s.

- Advanced economies (AEs) versus emerging economies (EEs) (broad dataset, 2000Q1–2019Q4 for 23 AEs and 16 EEs)
  - AEs: elasticity of wages to oil price shock peaks around 0.02 percent in 3 quarters.
  - EEs: elasticity nears 0.05 percent in 1½ years after the shock.
  - Unemployment rates statistically significant only in the AE sample.
  - NEER statistically significant in both groups, with larger effects for EEs.

- Practical note on horizon reporting
  - Median quarterly wage growth no longer responds to oil price shocks, or slightly declines, after 4 quarters.
  - Impulse response charts truncated at 2 or 3 years after the shock in the paper to avoid difficult interpretation over longer horizons with expanding confidence bands.

### Pass-through and structural factors
- Extended specification
  - Wage regressions augmented with structural variables z_{t,j} and interaction terms between oil price shocks and z_{t,j}.
  - Structural variables available in annual frequency interpolated to quarterly via last observation carried forward and converted to binary series (1 if value greater than the median for that series, 0 otherwise).
  - Each structural variable estimated separately to preserve country coverage.

- Energy weights in consumption basket
  - Hypothesis: pass-through larger where oil represents a larger share in the consumption basket.
  - Country-specific annual CPI weights of fuel and energy used as gauge; CPI weights data available for 34 countries.
  - CPI energy weights distribution: interquartile range of 8.7 and 13 percent.
  - Regression analysis (broad dataset) finds the interaction term of energy weights with oil price shock is positively and statistically significantly related to wage growth.
  - Binary energy weights: regimes with high energy weights have stronger oil price pass-through to wages compared to low energy weight regimes.
  - Average energy weight: EEs 12.7 percent; AEs 10 percent.
  - Energy weights availability limitation: omitted from long-dataset regressions because energy weight data are available only from 1999.

- Labor market institutions and wage bargaining
  - Focus on three variables that remain statistically significant as stand-alone controls: union density, collective bargaining coverage rate, and a combined indicator for highly unionized and centralized wage bargaining.
  - Structural characteristics converted to binary variables; pass-through estimates reported for “high” and “low” values separately (90 percent confidence bands via delta method).
  - Degree of centralization of bargaining only weakly significant unless combined with union density; combined binary “high” indicates both unionization and centralization above their medians, “low” indicates other combinations.
  - Although bargaining coverage coefficient often statistically significant, it usually has a negative sign.

- Median values used for binary conversion (as reported)
  - Long dataset medians: 35 percent for the union density, 80 percent for the collective bargaining coverage, 2.5 for the bargaining centralization, 0.2 for the monetary policy anchoring index, and 7 for the central bank transparency index.
  - Broad dataset medians: 10.7 percent for the energy weights, 28 percent for the union density, 78 percent for the collective bargaining coverage, 2.4 for the bargaining centralization, 0.22 for the monetary policy anchoring index, and 7.5 for the central bank transparency index.

- Additional structural-factor notes
  - Because most structural variables follow a declining trend from the 1980s, the analysis emphasizes variables that retain statistical significance when included as stand-alone controls.
  - Monetary policy credibility cited as another contributing factor to differences between AEs and EEs; on average higher in AEs than in EEs during the sample period.

*IMF Working Paper — 4. Results (excerpt)*

### 2019. It shows that the pass-through is stronger in countries (and periods) with higher unionization and with

### Second-Round Effects of Oil Price Shocks (2019 section on pass-through heterogeneity)

### Labor market institutions and wage pass-through
- Pass-through to wages is stronger in countries (and periods) with higher unionization and with both highly unionized and centralized wage bargaining; interaction terms are statistically significant at the 5 percent level.
- No statistically significant evidence that the coverage of collective bargaining affects pass-through strength.
- Direct effects: wage inflation is higher in countries with higher unionization and centralized bargaining, in addition to higher pass-through effects.
- Control variables (industrial production and exchange rates) enter with expected signs in all specifications.
- These relationships do not hold in more recent data:
  - With the broad dataset covering 2000-2019 (right column of Figure 10), degree of unionization does not significantly affect pass-through strength.
  - Contrast between high unionization cum centralized bargaining and the rest becomes less stark in post-2000 data.
  - Coverage of collective bargaining does not have statistically significant effects on pass-through strength in the post-2000 subsample.
- Possible explanations:
  - In the longer sample (1960-2019) labor market variables may proxy for other drivers of inflation during 1970-2000.
  - Regime change in labor markets with the 1980s as watershed may have reduced variation in labor market characteristics since 2000.

### Monetary policy credibility and wage pass-through
- Using interaction terms with measures of monetary policy credibility (broad dataset):
  - Pass-through to wages is weaker if inflation expectations are better anchored and central banks are more independent.
  - Weaker pass-through under high values of the anchoring index with statistically significant differences lasting for a year after the shock.
  - Interaction terms with central bank independence are statistically significant three quarters after the shock and thereafter.
- Estimation details: regressions include 4 lags for autoregressive terms; shaded areas represent 90 percent confidence bands.

### Economic states (pre-shock inflation, business cycle, large oil booms)
- Pre-shock inflation:
  - Wages react more strongly to oil shocks if inflation was already high.
  - With threshold of 4 percent (broad dataset, 2000-2019):
    - Pass-through elasticity ≈ 0.04 in about a year after the shock for high inflation states.
    - Corresponding elasticity ≈ 0.01 for low inflation states.
  - With threshold of 6 percent:
    - Elasticity in high inflation states ≈ 0.07 versus 0.02 in low inflation states (statistically significant).
  - No statistically significant differences when threshold is 2 percent.
- Business cycle phase:
  - Wages respond more strongly for about a year if the economy was in an expansionary phase at the time of the shock (binary indicator based on deviations of unemployment from HP trends).
  - Relationship is short-lived, likely because expansions at the time of shock are often followed by contraction.
- Large oil price booms (annual oil price inflation threshold set at 35 percent):
  - Long dataset (1960-2019): no significant differences in wage reactions during booms.
  - Broad dataset (2000-2019): shocks during large oil price booms have significantly weaker pass-through to wages.
  - Possible reasons: complex retail energy markets dampen large swings or other policy/economic reactions (e.g., tax cuts).

### Transmission via CPI inflation (oil-induced CPI inflation → wages)
- Instrumenting CPI inflation with oil price shocks (lags up to 4 quarters) in broad dataset:
  - Oil-induced rise in CPI inflation leads to a rise in wages of about the same magnitude in two quarters following the rise in inflation (Figure 15).
  - Impacts thereafter ease to about a half in AEs, while wage increases in EEs are twice as large as the rise in CPI inflation in the second year.
- Consistency: pass-through from oil-induced inflation to wages is higher if the economy is highly unionized, has less-anchored inflation expectations, and has high pre-existing inflation.

### Pass-through to core and expected inflation
- Core and expected inflation respond more slowly to oil price shocks than wages:
  - First quarter after oil shock: pass-through elasticity for core inflation = 0.002 versus 0.015 for wages.
  - Core inflation rises to 0.03 in three years after the shock.
  - Expected inflation peak elasticity = 0.01 in 3 quarters after the shock (about half of the wage elasticity); expected inflation is affected for about two years.
- Heterogeneity:
  - Both core and expected inflation react more strongly in emerging economies (EEs) than in advanced economies (AEs), with greater dispersion among EEs.
  - More independent central banks are associated with smaller pass-through of oil shocks to core and expected inflation.
  - Pre-existing high inflation increases pass-through strength; interacting terms are statistically significant for expected and core inflation a year after the shock with a 6 percent threshold (less clear significance at 4 percent).

### Sources of oil price shocks and pass-through
- Decomposition (structural VAR, January 1973–December 2021, aggregated to quarterly):
  - Three shocks: oil supply shocks (innovations to global oil production), aggregate demand shocks (innovations to global real economic activity not explained by supply shocks), oil-specific demand shocks (innovations to real oil price not explained by supply or aggregate demand shocks).
  - Aggregate demand shocks increased in importance in the 2000s and were main driver of real oil prices between 2005 and 2012.
  - Oil supply shocks played a smaller role; oil-specific demand shocks had large impacts in the 2010s.
- Pass-through by shock type (broad dataset, 2000-2019, 2SLS with shocks as instruments):
  - All three shock types trigger a rise in wages.
  - Aggregate demand–induced price rises: pass-through more immediate and stronger, impact ≈ 0.06 in 2 quarters after the shock.
  - Oil-specific demand shocks: elasticity rises to ≈ 0.01 in 2 quarters after the shock (positive and statistically significant).
  - Supply-driven price increases: reactions more gradual, peaking at ≈ 0.06 in about 1½ years after the shock.
- Implication: similarity in pass-through coefficients between global demand and supply shocks supports treating global oil price changes as exogenous for analysis of mostly oil-importing small-open economies.

*Italic: Source — IMF Working Paper "Second-Round Effects of Oil Price Shocks" (2019 section on pass-through heterogeneity, figures and notes as provided).*

### 6. Taking stock

### 6. Taking stock

### Key empirical findings on pass-through and wage-price spirals
- The results indicate a limited risk of a wage-price spiral and regime shift in inflation expectations in response to supply shocks.
- For the past several decades, including the decades that preceded the wide adoption of inflation targeting in AEs, the authors found no evidence of oil price shocks leading to a sustained wage-price spiral that would have pushed up the inflation rate for a sustained period (nor permanently).
- The pass-through from oil price shocks to wages and core inflation peaked in 1-2 years and largely dissipated in 3-4 years.
- These results are obtained in both:
  - the broad dataset (which falls in the post-Volker era), and
  - the long dataset that includes earlier periods when pass-through and the risk of a wage-price spiral would have been higher under less inflation-vigilant monetary policy regimes and a labor market with higher unionization and more centralized wage bargaining.
- The evidence suggests that the bar for a wage-price spiral being unleashed by oil price shocks (or supply shocks) would be quite high under the current policy framework and labor market institutions.
- The positive pass-through of oil price shocks to inflation expectations has a 2-3 year horizon (these results were based on the post-2000 data).
- The authors note a risk that high headline inflation itself could aggravate supply-driven inflationary pressure—punctuated by the unusually large increase in gas prices in Europe—thereby intensifying second-round effects and pushing up expected inflation. A similar argument has been made by BIS (2022).

### Cross-cutting patterns on determinants of pass-through
- Pass-through depends on:
  - the level of inflation at the time of the shock,
  - the cyclical state of activity, and
  - structural characteristics of an economy.
- Specific patterns identified:
  - Pass-through appears higher when overall inflation is higher at the time of shocks.
  - Pass-through appears higher in emerging than advanced European economies.
  - Pass-through appears lower when central banks anchor inflation expectations well or are more independent.
  - The link between pass-through and labor market characteristics is not clear-cut, except that a higher level of centralized bargaining cum unionization seems associated with a higher pass-through.
- Effects on core inflation and inflation expectations are consistent with effects on wages.
- In response to oil price shocks, headline inflation and wages are found to move by about the same magnitude.

### Caveats and limits of the analysis
- Results are conditional on the underlying policy and institutional frameworks in operation, especially in the post-2000 period; they do not constitute prescriptive evidence for the “optimal” policy responses.
- Cross-country estimation cannot capture particular episodes in some countries and times that might indicate an elevated risk of a wage-price spiral or unmooring inflation expectations in the current juncture.
- The authors’ results point to a risk that high headline inflation could itself intensify second-round effects and expected inflation (see BIS (2022) for a similar argument).

### Suggestions for future research
- Broaden country coverage to provide greater variation in data, acknowledging this may introduce larger heterogeneity complicating econometric investigation.
- Undertake more analysis of why pass-through varies with the level of inflation and the implications for monetary policy and the choice of inflation targets.
- Apply different econometric or quantitative methodologies to sharpen some of the results.

*IMF WORKING PAPERS — Second-Round Effects of Oil Price Shocks; INTERNATIONAL MONETARY FUND*

### Annex I. Data Description

### Annex I. Data Description

### Indicators: definitions and sources
- Wage (broad data)
  - Definition and coverage may vary across countries. Wages and salaries, or gross earnings are used, except the followings: hourly earnings for Denmark and Sweden, earnings excluding bonus for Finland, compensation of employees for Norway, and net earnings for Croatia.
  - Industry coverage: wages for Austria, Belgium, Cyprus, Czech Republic, Germany, Greece, Ireland, Italy, Lithuania, Portugal, Slovakia and Türkiye cover industry excluding construction. Wages for Norway and Sweden cover industries only. Wages for Denmark cover private sector only. Wages for Kosovo cover public sector only. For the remaining countries, wages cover the whole economy.
  - Seasonally adjusted by the authors.
  - Sources: Haver Analytics
- Wage (long data)
  - Hourly earnings in manufacturing.
  - Sources: OECD
- Core inflation
  - Sources: Haver Analytics; OECD
- Headline inflation
  - Sources: Haver Analytics; OECD
- Oil price (in USD)
  - Crude Oil (petroleum), USD, simple average of three spot prices; Dated Brent, West Texas Intermediate, and the Dubai Fateh
  - Source: IMF RES Commodity Database
- NEER
  - Nominal effective exchange rate
  - Sources: IMF IFS; IMF INS
- Unemployment rates
  - For Albania, Bulgaria and Romania, registered unemployment rates are used. Seasonally adjusted by the authors.
  - Sources: Haver Analytics; OECD
- Energy weights
  - Defined as the sum of CPI weights given to “energy” under housing and “fuels” under transport.
  - Source: Haver Analytics
- Union density
  - Defined as the number of net union members (i.e excluding those who are not in the labour force, unemployed and self-employed) as a proportion of the number of employees. Missing observations are interpolated with the last available observations.
  - Source: OECD
- Collective bargaining coverage
  - The adjusted collective bargaining coverage rate is defined as the number of employees covered by a collective agreement in force as a proportion of the number of eligible employees equipped (i.e., the total number of employees minus the number of employees legally excluded from the right to bargain). Missing observations are interpolated with the last available observations.
  - Source: OECD
- Centralization of collective bargaining
  - A composite index which first measures the dominant level of collective bargaining and then considers the presence of additional enterprise bargaining. In a fully decentralized system, all negotiations take place in the enterprise. In a fully centralized system, all wages are negotiated at the central (cross-industry) level without further amendments.
  - Source: OECD
- MP anchoring
  - An index of inflation expectations' anchoring for 45 countries in the period starting in 1989. The index is calculated based on four metrics: deviation of long-term mean inflation forecasts from target, variability of mean long-term inflation forecasts, dispersion of long-term inflation forecasts, and sensitivity of long-term inflation forecasts to inflation surprises.
  - Source: Bems, Mr Rudolfs, et al. Expectations' Anchoring and Inflation Persistence. International Monetary Fund, 2018.
- CB independence
  - Index ranging between 0 (no independence) and 1 (full independence). The index is calculated based on 42 criteria of institutional design as provided in central bank legislation. Available for 155 countries for 1972-2017.
  - Source: Romelli. 2022

### Annex II. Data coverage (selected entries)
- Advanced economies (Broad) (Long) — from / from
  - Austria 2000q1 1961q1
  - Belgium 2000q1 1961q1
  - Cyprus 2000q1
  - Czech Republic 2000q1 1994q1
  - Denmark 2000q1 1974q1
  - Estonia 2000q1 2001q1
  - Finland 2000q1 1961q1
  - France 2000q1 1961q1
  - Germany 2000q1 1961q1
  - Greece 2000q1
  - Ireland 2000q1 1975q3
  - Israel 2000q1 1996q1
  - Italy 2000q1 1961q1
  - Latvia 2000q1 2001q1
  - Lithuania 2000q1 2001q2
  - Netherlands 2000q1 1971q1
  - Norway 2000q1 1961q1
  - Portugal 2000q1 2001q1
  - Slovak Republic 2000q1 1993q1
  - Slovenia 2000q1 1999q1
  - Spain 2000q1 1982q1
  - Sweden 2000q1 1972q1
  - United Kingdom 2000q1 1968q1
- Emerging economies (Broad) (Long) — from / from
  - Albania 2012q1
  - Belarus 2017q1
  - Bosnia and Herzegovina 2017q1
  - Bulgaria 2000q1
  - Croatia 2000q1
  - Hungary 2004q1 1996q1
  - Kosovo 2016q1
  - North Macedonia 2000q1
  - Moldova 2004q1
  - Montenegro, Rep. of 2008q1
  - Poland 2000q1 1996q1
  - Romania 2000q1
  - Russia 2000q1
  - Serbia 2000q1
  - Türkiye 2009q1 1989q1
  - Ukraine 2004q4

### References (selected citations from source)
- Bems, Rudolfs, Francesca Caselli, Francesco Grigoli, Bertrand Gruss and Weicheng Lian, 2018, Expectations’ Anchoring and Inflation Persistence, IMF WP/18/280.
- Boranova, Vizhdan, Raju Huidrom, Sylwia Nowak, Petia Topalova, Volodymyr Tulin, and Richard Varghese, 2019, Wage Growth and Inflation in Europe: A Puzzles? IMF Working paper 19/280.
- Blanchard, Olivier and Jordi Gali, 2007, The Macroeconomic Effects of Oil Shocks: Why Are The 2000s So Different From the 1970s? NBER WP 13368.
- Killian, Lutz, 2009, "Not All Oil Price Shocks Are Alike: Disentangling Demand and Supply Shocks in the Crude Oil Market." American Economic Review, 99 (3): 1053-69.
- Romeli, Davide, 2022, The political economy of reforms in Central Bank design: evidence from a new dataset, Economic Policy.

*Annex I. Data Description (wpiea2022173-print-pdf).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022173-print-pdf.pdf_