## wpiea2019277-print-pdf

## Source details

**Canonical URL:** [wpiea2019277-print-pdf](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019277-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wpiea2019277-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019277-print-pdf.pdf.json)

---

### Key findings
- Electricity prices are administered and their inflation (10.7 percent) is the highest among the four components.
- Electricity inflation rose to around 30 percent year on year in the late-2000s and gradually declined.
- Fuel price inflation is the most volatile component (standard deviation of 13.7), strongly affected by global oil prices and rand.

### Data and empirical setup
- Quarterly data for March 2005–June 2018 are used. Core inflation data go back to 2002 but early years removed due to volatility.
- Exchange rate performance captured by NEER. Positive values represent depreciation in year on year terms.
- Foreign inflation controlled by an index of producer price inflation in import origin countries (Foreign PPI).
- Output gap measured from HP filters (percent deviations from trend).
- Wage pressure controlled by unit labor cost inflation (ULC).
- Realized exchange rate volatility represented by the 3-month rolling standard deviation of the daily percent change in the dollar-rand exchange rate (NEER vol.).
- Granger-causality Wald tests (up to 4 lags) indicate NEER depreciation and the output gap strongly impact core inflation at the 1 percent level; core inflation generally does not strongly Granger-cause the explanatory variables.
- To reduce reverse causality, explanatory variables are lagged except foreign PPI (contemporaneous).

### Headline vs core inflation: component statistics
- CPI basket weights:
  - Electricity: 3.8
  - Fuel: 4.6
  - Food: 17.2
  - Core: 74.4
- Inflation performance, 2003–19:
  - Mean:
    - Electricity: 10.7
    - Fuel: 9.0
    - Food: 6.7
    - Core: 4.3
  - Standard deviation:
    - Electricity: 7.5
    - Fuel: 13.7
    - Food: 4.1
    - Core: 2.7
  - Coefficient of variation (b/a):
    - Electricity: 0.7
    - Fuel: 1.5
    - Food: 0.6
    - Core: 0.6
- Correlations (quarterly data 2003–18, log differences):
  - NEER changes vs domestic electricity price inflation: -0.04
  - NEER changes vs domestic fuel price inflation: 0.33
  - NEER changes vs domestic food price inflation: 0.06

### Baseline empirical results for ERPT
- Headline vs core:
  - Yearly ERPT for headline inflation: 9 percent.
  - Yearly ERPT for core inflation: 8 percent.
  - Headline inflation persistence smaller (autoregressive coefficient smaller), yielding LREPRT of 12 percent for headline vs 19 percent for core in one specification.
  - Foreign PPI is an important determinant of headline inflation; unit labor costs and the output gap show limited impact in that specification.
- Baseline core inflation (data from 2005 onward):
  - Estimated yearly ERPT to core inflation is between 2 and 4 percent (declined relative to earlier periods).
  - Long-run ERPT (LRERPT) to core inflation estimated at 6 to 10 percent (below the 19 percent reported when earlier data included).
  - Output gap coefficient significant at the 10 percent level in baseline; one-standard-deviation variation in the output gap is around 1.
- Representative baseline regression coefficients (selected):
  - AR1: 0.588***
  - NEER L1 coefficient: 0.015**
  - Output gap L1: 0.029*
  - N: 78 for the extended headline/core comparison; N: 50 in core-only baseline estimations
  - Adjusted r^2 reported examples: 0.587 (core), 0.551 (headline)
  - Yearly ERPT (core) coef.: 0.078** (core vs full-sample headline table) and other reported yearly ERPT coef. values across specifications

### Time-series behavior of LRERPT
- Rolling 10-year time-series estimates:
  - LRERPT rose from 4 percent in 2014 to 7 percent in 2018.
- Comparison reference:
  - Carstens (2019) reports LRERPT for a set of EMs rose from 10 percent in 2014 to 20 percent in 2018.

### Adding exchange rate volatility (ER vol.) to the model
- Model extension includes exchange rate volatility and its interaction with NEER depreciation (interaction up to four lags tested).
- Key findings:
  - Rise in exchange rate volatility leads to higher core inflation (statistically significant), but economically small in size.
  - In parsimonious specifications, up to 4 percent of exchange rate volatility is passed through to core inflation.
  - Interpretation: a one unit change in the exchange rate volatility data (one standard deviation movement; data normalized as Z-score) affects core inflation by 0.04 percent.
  - Interaction terms (NEER * ER vol) are generally insignificant, suggesting ERPT is not systematically altered in a nonlinear way by the level of exchange rate volatility.
- Selected coefficients from extension models (examples):
  - AR1 values in extensions: 0.756***, 0.579***, 0.599***, 0.611***, 0.586***, 0.588***
  - NEER L1 coef. across extensions: 0.020***, 0.021***, 0.021***, 0.015**, 0.015**
  - ER vol L2 examples: 0.034# (20 percent level), 0.040*** (1 percent level) in parsimonious models
  - Yearly ERPT coef. in extended models: 0.035**, 0.037**, 0.037**, 0.023^, 0.023^
  - Long-run ERPT coef. examples in extended models: 0.083***, 0.092***, 0.095***, 0.056***, 0.056***

### Robustness checks and alternative volatility measures
- Potential endogeneity between NEER depreciation and realized volatility addressed by:
  - Replacing realized exchange rate volatility with residuals from regressions controlling for own lag and NEER depreciation (contemporaneous and first lag).
  - Using implied volatility (rand-dollar, three-month tenor) and residuals from regression on own lag and NEER depreciation.
  - Replacing exchange rate volatility with VIX as a global uncertainty proxy.
- Robustness results:
  - Using the residual of realized exchange volatility: ER vol remains statistically important at the 5–10 percent levels.
  - Implied volatility: coefficient statistically significant at the 1 percent level but pass-through smaller; in some specifications implied volatility somewhat increases ERPT.
  - VIX-based checks: findings remain robust.
- Representative NEER depreciation / ER volatility interaction dynamics (selected):
  - NEER depreciation AR1: 0.294**
  - ER vol AR1: 0.525***
  - NEER depreciation L0 impacts ER vol L0: -1.098*** and ER vol L1: 1.283*** in an example causality table
  - N in causality examples: 53 and r2_a examples: 0.225 and 0.309

### Main conclusions and policy implications
- Core findings:
  - Higher exchange rate volatility lifts core inflation in a statistically significant manner, but with relatively small economic magnitude.
  - The estimated ERPT to core inflation is broadly stable across specifications.
  - The output gap remains an important determinant of core inflation across models.
- Policy interpretation:
  - Results support allowing the rand to float freely and serve as a shock absorber under South Africa’s inflation targeting regime.
  - Large rand volatility has had limited inflationary impact, implying the rand has functioned effectively as a shock absorber for inflation outcomes.
- Suggested extensions:
  - Assess rand volatility effects on other sectors, notably SMEs with limited natural hedges or hedging capacity.
  - Analyze the impact of exchange rate volatility on cross-border bond and equity inflows, given the importance of portfolio investment in South Africa.

*https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019277-print-pdf.pdf*

### 2011. Electricity prices are administered and their inflation (10.7 percent) is the highest

### 2011. Electricity prices are administered and their inflation (10.7 percent) is the highest

### Key findings
- Electricity prices are administered and their inflation (10.7 percent) is the highest among the four components.
- Electricity inflation rose to around 30 percent year on year in the late-2000s and gradually declined.
- Fuel price inflation is the most volatile component (standard deviation of

### Context and implications
- Administered electricity prices can drive headline inflation when they change: the observed 10.7 percent electricity inflation is the largest among the four measured components.
- Historical peak: electricity inflation reached around 30 percent year on year in the late-2000s before declining, indicating substantial past volatility and potential for large swings in energy-related CPI contributions.
- Fuel price inflation is noted as the most volatile component, though the numerical standard deviation value is not completed in the provided text.

*https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019277-print-pdf.pdf*

### 13.7 and coefficient of variation of 1.5), strongly affected by global oil prices and rand

### wpiea2019277-print-pdf - 13.7 and coefficient of variation of 1.5), strongly affected by global oil prices and rand

### Data and empirical setup
- Quarterly data for March 2005–June 2018 are used. Core inflation data go back to 2002 but early years removed due to volatility.
- Exchange rate performance captured by NEER. Positive values represent depreciation in year on year terms.
- Foreign inflation controlled by an index of producer price inflation in import origin countries (Foreign PPI).
- Output gap measured from HP filters (percent deviations from trend).
- Wage pressure controlled by unit labor cost inflation (ULC).
- Realized exchange rate volatility represented by the 3-month rolling standard deviation of the daily percent change in the dollar-rand exchange rate (NEER vol.).
- Granger-causality Wald tests (up to 4 lags) indicate NEER depreciation and the output gap strongly impact core inflation at the 1 percent level; core inflation generally does not strongly Granger-cause the explanatory variables.
- To reduce reverse causality, explanatory variables are lagged except foreign PPI (contemporaneous).

### Headline vs core inflation: component statistics
- CPI basket weights:
  - Electricity: 3.8
  - Fuel: 4.6
  - Food: 17.2
  - Core: 74.4
- Inflation performance, 2003–19:
  - Mean:
    - Electricity: 10.7
    - Fuel: 9.0
    - Food: 6.7
    - Core: 4.3
  - Standard deviation:
    - Electricity: 7.5
    - Fuel: 13.7
    - Food: 4.1
    - Core: 2.7
  - Coefficient of variation (b/a):
    - Electricity: 0.7
    - Fuel: 1.5
    - Food: 0.6
    - Core: 0.6
- Correlations (quarterly data 2003–18, log differences):
  - NEER changes vs domestic electricity price inflation: -0.04
  - NEER changes vs domestic fuel price inflation: 0.33
  - NEER changes vs domestic food price inflation: 0.06

### Baseline empirical results for ERPT
- Headline vs core:
  - Yearly ERPT for headline inflation: 9 percent.
  - Yearly ERPT for core inflation: 8 percent.
  - Headline inflation persistence smaller (autoregressive coefficient smaller), yielding LREPRT of 12 percent for headline vs 19 percent for core in one specification (see Table A2 reference in source).
  - Foreign PPI is an important determinant of headline inflation; unit labor costs and the output gap show limited impact in that specification.
- Baseline core inflation (data from 2005 onward):
  - Estimated yearly ERPT to core inflation is between 2 and 4 percent (declined relative to earlier periods).
  - Long-run ERPT (LRERPT) to core inflation estimated at 6 to 10 percent (below the 19 percent reported when earlier data included).
  - Output gap coefficient significant at the 10 percent level in baseline; one-standard-deviation variation in the output gap is around 1.
- Representative baseline regression coefficients (selected):
  - AR1: 0.588***
  - NEER L1 coefficient: 0.015**
  - Output gap L1: 0.029*
  - N: 78 for the extended headline/core comparison; N: 50 in core-only baseline estimations (varies across tables)
  - Adjusted r^2 reported examples: 0.587 (core), 0.551 (headline)
  - Yearly ERPT (core) coef.: 0.078** (core vs full-sample headline table) and other reported yearly ERPT coef. values across specifications

### Time-series behavior of LRERPT
- Rolling 10-year time-series estimates (last model in Table 3):
  - LRERPT rose from 4 percent in 2014 to 7 percent in 2018.
- Comparison reference:
  - Carstens (2019) reports LRERPT for a set of EMs rose from 10 percent in 2014 to 20 percent in 2018.

### Adding exchange rate volatility (ER vol.) to the model
- Model extension includes exchange rate volatility and its interaction with NEER depreciation (interaction up to four lags tested).
- Key findings:
  - Rise in exchange rate volatility leads to higher core inflation (statistically significant), but economically small in size.
  - In parsimonious specifications, up to 4 percent of exchange rate volatility is passed through to core inflation.
  - Interpretation: a one unit change in the exchange rate volatility data (one standard deviation movement; data normalized as Z-score) affects core inflation by 0.04 percent.
  - Interaction terms (NEER * ER vol) are generally insignificant, suggesting ERPT is not systematically altered in a nonlinear way by the level of exchange rate volatility.
- Selected coefficients from extension models (examples across tables):
  - AR1 values in extensions: 0.756***, 0.579***, 0.599***, 0.611***, 0.586***, 0.588***
  - NEER L1 coef. across extensions: 0.020***, 0.021***, 0.021***, 0.015**, 0.015**
  - ER vol L2 examples: 0.034# (20 percent level), 0.040*** (1 percent level) in parsimonious models
  - Yearly ERPT coef. in extended models: 0.035**, 0.037**, 0.037**, 0.023^, 0.023^ (where “^” denotes 15 percent in some tables)
  - Long-run ERPT coef. examples in extended models: 0.083***, 0.092***, 0.095***, 0.056***, 0.056***

### Robustness checks and alternative volatility measures
- Potential endogeneity between NEER depreciation and realized volatility addressed by:
  - Replacing realized exchange rate volatility with residuals from regressions controlling for own lag and NEER depreciation (contemporaneous and first lag).
  - Using implied volatility (rand-dollar, three-month tenor) and residuals from regression on own lag and NEER depreciation.
  - Replacing exchange rate volatility with VIX as a global uncertainty proxy.
- Robustness results:
  - Using the residual of realized exchange volatility: ER vol remains statistically important at the 5–10 percent levels.
  - Implied volatility: coefficient statistically significant at the 1 percent level but pass-through smaller; in some specifications implied volatility somewhat increases ERPT.
  - VIX-based checks: findings remain robust.
- Representative NEER depreciation / ER volatility interaction dynamics (selected from Table 5):
  - NEER depreciation AR1: 0.294**
  - ER vol AR1: 0.525***
  - NEER depreciation L0 impacts ER vol L0: -1.098*** and ER vol L1: 1.283*** in an example causality table
  - N in causality examples: 53 and r2_a examples: 0.225 and 0.309

### Main conclusions and policy implications
- Core findings:
  - Higher exchange rate volatility lifts core inflation in a statistically significant manner, but with relatively small economic magnitude.
  - The estimated ERPT to core inflation is broadly stable across specifications.
  - The output gap remains an important determinant of core inflation across models.
- Policy interpretation:
  - Results support allowing the rand to float freely and serve as a shock absorber under South Africa’s inflation targeting regime.
  - Large rand volatility has had limited inflationary impact, implying the rand has functioned effectively as a shock absorber for inflation outcomes.
- Suggested extensions:
  - Assess rand volatility effects on other sectors, notably SMEs with limited natural hedges or hedging capacity.
  - Analyze the impact of exchange rate volatility on cross-border bond and equity inflows, given the importance of portfolio investment in South Africa.

*Source: IMF working paper content (Tables, figures, and regression results excerpted from the provided PDF).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019277-print-pdf.pdf_