## _wp07123

## Source details

**Canonical URL:** [_wp07123](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp07123.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp07123.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp07123.pdf.json)

---

### I. Introduction and research question
- Post-apartheid (post-1994) South Africa experienced substantial changes in the framework for monetary policy formulation, including increased monetary policy transparency and the adoption of an inflation targeting framework in 2000.
- Research question: what are the economic benefits of greater policy transparency, measured by the accuracy of private sector financial forecasts as an indicator of expectations formation?
- Approach: follows Swanson (2006) focusing on private sector ability to forecast interest rates using monthly data for 1997–2006.
- Central conclusions summarized:
  - Accuracy of private sector interest rate forecasts increased significantly during 1997–2006.
  - Financial market surprises around MPC announcements were reduced in the more recent period and dispersion across private sector forecasts fell.
  - Some improvement in accuracy and dispersion of inflation forecasts was observed.
  - Improvements in private sector interest rate forecasts exceeded those in output forecasts, suggesting a role for increases in reserve bank transparency.

### II. Increases in SARB monetary policy transparency (selected timeline)
- 1996: SARB acknowledged existence of a forward book in foreign exchange and thereafter published information on the book; details of net open forward position regularly published.
- October 1999: Establishment of a Monetary Policy Committee (MPC). MPC to meet every 6–8 weeks; policy statements published.
- February 2000: Adoption by government of inflation target for the SARB to implement (3–6 percent annual average by 2002).
- February 2000: Introduction of monetary policy forums convened by SARB; held twice yearly.
- March 2001: Introduction of Monetary Policy Review (MPR), published twice yearly; MPR includes SARB’s inflation forecast fan chart.
- 2002: Frequency of MPC meetings reduced to 4 times yearly.
- May 2003: Frequency of MPC meetings increased to 6 times yearly.
- November 2003: Refinement of inflation target from annual (average) basis to continuous basis (12-month rate).

### III. Accuracy of financial market forecasts — methodology
- Interest-rate forecast accuracy measured by deviation between actual outturn of the spot rate in a given month and the forward rate forecast made at a previous date (e.g., 3-month-ahead error = realized spot value minus the 3-month forward rate that prevailed three months previously).
- Jibar = Johannesburg Interbank Average Rate; data are monthly averages calculated from daily spot and forward rates.
- Cross-sectional dispersion measured as maximum minus minimum forecast in the Reuters monthly survey of 14 financial market institutions.

### IV. Interest-rate forecast errors (1997–2006) — key statistics and trends
- Absolute forecast errors declined over time at 3-month, 6-month, and 12-month horizons.
- Average 3-month-ahead forecast error:
  - 126 basis points during 1997–2000
  - 47 basis points during 2001–2006
- Average 6-month-ahead forecast error:
  - 213 basis points during 1997–2000
  - 84 basis points during 2001–2006
- Average 12-month-ahead forecast error:
  - 326 basis points during 1997–2000
  - 159 basis points during 2001–2006
- Decline interrupted in 1998 and 2002 by significant exogenous shocks; 1998 increase larger than 2002.
- For 1999–2000 (excluding 1998 large errors), forecast errors were:
  - 72 basis points (3m)
  - 166 basis points (6m)
  - 342 basis points (12m)
- Time trend (1997–2006):
  - Statistically significant negative trend at 3-month and 6-month horizons.
  - Not statistically significant at 12-month horizon.
  - Dummy variable for post-2001 not statistically significant (transparency changes part of an ongoing process, not discrete).

### V. Surprise component of MPC announcements (1999–2006)
- Surprise measured by change in market interest rates between the day immediately after and the day immediately before MPC meetings; sample: 1999–2006 (47 policy announcements).
- Panels 4a and 4b (3-month and 1-month Jibar rates) show slight upward trend overall attributable to two large surprises in 2003 (June and October).
- Excluding those two surprises:
  - Average surprise for 1-month Jibar: 12 basis points during October 1999–March 2003 and 13 basis points during December 2003–December 2005.
  - Average surprise for 3-month Jibar: 14 basis points during October 1999–March 2003 and 14 basis points during December 2003–December 2005.
- For full 2003–2005 period, surprise component trend was markedly negative.
- Similar pattern when using Jibar forward rate instead of spot.
- Statistical significance:
  - Negative trend in surprise component during 2003–2006 statistically significant at 10-percent level for the 3-month Jibar rate.
  - Negative trend statistically significant at 5-percent level for the forward rates.
  - Trend for 1-month rate negative but not statistically significant.
  - Note: small sample size (24 MPC announcements in the 2003–2006 period).

### VI. Dispersion of interest-rate forecasts
- Cross-sectional dispersion (max minus min forecast of the prime rate in the Reuters monthly survey) declined over time at both 1-quarter-ahead and 4-quarters-ahead horizons.
- Negative trend in dispersion statistically significant at the 5-percent level.

### VII. Inflation trends, volatility, and forecast performance
- CPI inflation developments:
  - CPI inflation declined markedly over the past decade; average CPI inflation rate in 2006 (4½ percent) lower by one-third against its 1996 level.
  - CPIX inflation (computed since 1997, basis for SARB’s official inflation target) also fell markedly.
  - Sharp rise in inflation in 2002 reflecting earlier easing of monetary conditions and pass-through of a sharp exchange rate depreciation in late 2001; rise was temporary.
  - Increase in monetary policy transparency since 2000 coincided with a statistically significant negative trend in the inflation rate despite the 2002 spike.
- CPIX inflation forecast errors (quarterly Reuters mean forecasts):
  - Forecast errors declined markedly in size during 2003–2006.
  - 2000–2006: 1- and 4-quarter forecast errors have roughly flat and slightly negative trends respectively; trends not statistically significant.
  - 2003–2006: negative trends in forecast errors, with stronger decline and statistical significance at 1-quarter horizon than at 4-quarter horizon.
  - Market forecasters tended to over-predict inflation during the last few years of sample: consistently negative forecast errors from Q3 2003 until Q3 2006.
- Role of exchange rate volatility:
  - At 1-quarter horizon, a 1-percentage point decline in exchange rate volatility associated with a less than 0.1 percentage point decline in forecast error.
  - At 4-quarter horizon, impact of exchange rate volatility not statistically significant.
  - Including exchange rate volatility reduces statistical significance of trend in forecast errors at both horizons; exchange rate volatility is only one factor and may itself be influenced by monetary policy framework.
- Dispersion of inflation forecasts:
  - Cross-sectional dispersion (max minus min forecast of CPIX) declined over time at 1-quarter and 4-quarter horizons.
  - Negative trend in dispersion statistically significant for 2003–2006 period; not significant for full 2000–2006 period due to turbulence when forecasts performed poorly and were disparate.

### VIII. Exchange rate volatility and inflation volatility
- Exchange rate volatility (6-month standard deviation of nominal effective exchange rate changes):
  - 1998–2006: slight upward trend owing to three consecutive quarters of higher volatility starting in Q4 2002.
  - December 2003–December 2006: exchange rate volatility nearly 20 percent lower than March 1998–September 2003.
  - Inflation volatility increase since 1998 attributable to 2002–part of 2003 turbulence; inflation volatility declined during 2003–2006.

### IX. GDP forecast accuracy and comparison with interest-rate/inflation forecasts
- If general forecasting skill improved, real GDP forecasts should also show improvements; evidence is less convincing than for interest rates.
- Absolute GDP forecast errors (1- and 4-quarter horizons) over 2000–2006:
  - Errors did not change appreciably at either horizon.
  - Slight negative trend at 1-quarter horizon and slight positive trend at 4-quarter horizon; neither trend statistically significant.
- During 2003–2006, forecast errors declined at both the 1- and 4-quarter horizons for GDP (the trends were negative) although the decline was not statistically significant.
- Cross-sectional dispersion of GDP forecasts declined at both horizons, and the decline was statistically significant, suggesting increased unanimity among forecasters.
- Overall: improvements in interest rate forecasts exceeded those in GDP forecasts.

### X. Conclusions, interpretation, and research recommendations
- Main findings:
  - Private sector financial forecasts in South Africa improved in recent years.
  - Accuracy of interest rate forecasts increased; forecasters were less surprised by MPC announcements; cross-sectional dispersion of interest rate forecasts declined.
  - Inflation forecasts became more accurate and less dispersed, particularly during 2003–2006.
  - Improvements in interest rate and inflation forecasts exceeded improvements in real GDP forecasts.
- Interpretation:
  - Greater improvements for interest rates and inflation (relative to GDP) suggest monetary policy transparency likely played a role because inflation and interest rates are closely related to the central bank’s policy instrument while GDP is influenced by many other short-run factors.
- Recommendations for further research:
  - Examine the role of alternative factors, such as exchange rate volatility and monetary variables, in accounting for the decline in inflation forecast errors over time.
  - Examine trends in forecast quality for other real sector variables (manufacturing, industrial production, retail sales); these series were not available in the Reuters survey used for the current analysis.

### XI. Data sources and limitations
- Jibar and FRA data from J.P. Morgan Data Query, compiled from daily closing rates.
- Cross-sectional dispersion and private-sector forecasts for CPIX inflation and GDP growth from the Reuters monthly survey of 14 financial market institutions; dispersion computed as max minus min forecast.
- Forecast series for other real sector variables were not available in the Reuters survey used.

*Source: _wp07123 — Selected Changes in SARB Transparency, 1994–2006 (excerpt).*

### 1. Selected Changes in SARB Transparency, 1994–2006 .......................................................10

### 1. Selected Changes in SARB Transparency, 1994–2006

### Covered sections and tables
- 1. Selected Changes in SARB Transparency, 1994–2006 .......................................................10
- 2. Time Series for Absolute Forecast Error in JIBAR .............................................................10
- 3. JIBAR Forward Market Forecast Errors, 1997:1-2006:12 ..................................................12
- 4. Surprise Component of MPC Announcements, 2003-06 ....................................................14
- 5. Forecast Dispersion, March 2000-December 2006 .............................................................15
- 6. Inflation Trends, 2000-06 ....................................................................................................16
- 7. Inflation Forecast Errors and Exchange Rate Volatility, 2000-2006...................................20

### Figures listed (titles and scope)
- Figure 1. JIBAR 3-Month-Ahead Forecast Errors, absolute value, in basis points, 1997-2006 .........11
- Figure 2. JIBAR 6-Month-Ahead Forecast Errors, absolute value, in basis points, 1997-2006 .........11
- Figure 3. 12-Month-Ahead Forecast Errors, absolute value, in basis points, 1997-2005 ...................11
- Figure 4. Fluctuations in the JIBAR Rate Around MPC Announcement Dates, 1999-2006   ...........13
- Figure 5a. Cross-Sectional Dispersion in Interest Rate Forecasts (1-quarter-ahead forecasts) ..........14
- Figure 5b. Cross-Sectional Dispersion in Interest Rate Forecasts (4-quarters-ahead forecasts).........14
- Figure 6. Inflation Trends (12-month change, in percent) ..................................................................15
- Figure 7. Inflation (CPIX) Volatility, 1998-2006 ...............................................................................16
- Figure 8a. CPIX Inflation: Forecast Errors, 2000-06 ..........................................................................17
- Figure 8b. CPIX Inflation: Forecast Errors, 2003-06..........................................................................18
- Figure 9. Exchange Rate Volatility .....................................................................................................19
- Figure 10. Dispersion in Inflation Forecasts, 2000-06 ........................................................................20
- Figure 11. GDP Forecast Errors, 2000-06...........................................................................................21
- Figure 12. GDP Forecast Errors, 2003-06...........................................................................................22
- Figure 13. GDP Forecast Dispersion, 2000-06 ...................................................................................23

### Time coverage and frequency highlights
- JIBAR forecast error series: 1997-2006 (3-month, 6-month) and 1997-2005 (12-month).
- JIBAR forward market forecast errors: 1997:1-2006:12.
- MPC announcement analysis: 1999-2006 (fluctuations around announcement dates) and 2003-06 (surprise component).
- Forecast dispersion and inflation/GDP forecast metrics: primarily 2000-06 and subperiods 2003-06.
- Inflation volatility (CPIX): 1998-2006.
- Inflation trends reported as 12-month change, in percent.

### Key metrics and measures emphasized
- Absolute forecast error in JIBAR, reported in basis points.
- Forward market forecast errors for JIBAR.
- Surprise component of Monetary Policy Committee (MPC) announcements.
- Cross-sectional dispersion in interest rate forecasts (1-quarter-ahead and 4-quarters-ahead).
- Inflation trends and CPIX volatility (12-month changes, percent).
- Inflation forecast errors and dispersion.
- Exchange rate volatility.
- GDP forecast errors and dispersion.

*Source: _wp07123 - 1. Selected Changes in SARB Transparency, 1994–2006 (PDF chapter/section).*

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

### _wp07123 - References

### I. INTRODUCTION
- Post-apartheid (post-1994) South Africa experienced substantial changes in the framework for monetary policy formulation, including increased monetary policy transparency and the adoption of an inflation targeting framework in 2000.
- Research question addressed: what are the economic benefits of greater policy transparency, measured by the accuracy of private sector financial forecasts as an indicator of expectations formation?
- Approach: follows Swanson (2006) focusing on private sector ability to forecast interest rates using monthly data for 1997–2006.
- Central conclusions:
  - Accuracy of private sector interest rate forecasts increased significantly during 1997–2006.
  - Financial market surprises around MPC announcements were reduced in the more recent period and dispersion across private sector forecasts fell.
  - Some improvement in accuracy and dispersion of inflation forecasts was observed.
  - Improvements in private sector interest rate forecasts exceeded those in output forecasts, suggesting a role for increases in reserve bank transparency.

### II. INCREASES IN MONETARY POLICY TRANSPARENCY
- Timeline and institutional changes:
  - 1996: SARB acknowledged existence of a forward book in foreign exchange and thereafter published information on the book.
  - 1999: Establishment of a monetary policy committee (MPC) that meets regularly and publishes conclusions.
  - 2000: Adoption of an inflation target (to become effective in 2002).
  - 2000: Introduction of monetary policy forums at semi-annual frequency for exchanges between the reserve bank and the public.
  - 2001: Introduction of a semi-annual Monetary Policy Review presenting the reserve bank’s views on monetary developments and the inflation outlook.
- Overall: monetary policy transparency increased significantly during the post-apartheid period, particularly after 1999.

### III. ACCURACY OF FINANCIAL MARKET FORECASTS

- Methodology notes:
  - Interest-rate forecast accuracy measured by deviation between actual outturn of the spot rate in a given month and the forward rate forecast made at a previous date (e.g., 3-month-ahead error = realized spot value minus the 3-month forward rate that prevailed three months previously).
  - Jibar = Johannesburg Interbank Average Rate; data are monthly averages calculated from daily spot and forward rates.

- Interest rate forecast errors (1997–2006):
  - Absolute value of forecast errors declined over time at 3-month, 6-month, and 12-month horizons (Figures 1–3; Table 2).
  - Average 3-month-ahead forecast error:
    - 126 basis points during 1997–2000
    - 47 basis points during 2001–2006
  - Average 6-month-ahead forecast error:
    - 213 basis points during 1997–2000
    - 84 basis points during 2001–2006
  - Average 12-month-ahead forecast error:
    - 326 basis points during 1997–2000
    - 159 basis points during 2001–2006
  - Decline in forecast errors interrupted in 1998 and 2002 by significant exogenous shocks; 1998 increase larger than 2002.
  - For 1999–2000 (excluding 1998 large errors), forecast errors were 72 basis points (3m), 166 basis points (6m), and 342 basis points (12m).
  - Time trend in forecast errors during 1997–2006:
    - Statistically significant negative trend at 3-month and 6-month horizons (Table 3).
    - Not statistically significant at 12-month horizon.
    - Dummy variable for post-2001 not statistically significant (transparency changes part of an ongoing process, not discrete).

- Surprise component of MPC announcements:
  - Surprise measured by change in market interest rates between the day immediately after and the day immediately before MPC meetings.
  - Sample: 1999–2006 (47 policy announcements).
  - Panels 4a and 4b (3-month and 1-month Jibar rates) show slight upward trend overall attributable to two large surprises in 2003 (June and October).
  - Excluding those two surprises:
    - Average surprise for 1-month Jibar: 12 basis points during October 1999–March 2003 and 13 basis points during December 2003–December 2005.
    - Average surprise for 3-month Jibar: 14 basis points during October 1999–March 2003 and 14 basis points during December 2003–December 2005.
  - For full 2003–2005 period, surprise component trend was markedly negative (bottom figures of panels 4a and 4b).
  - Similar pattern when using Jibar forward rate instead of spot (panel 4c).
  - Statistical significance:
    - Negative trend in surprise component during 2003–2006 statistically significant at 10-percent level for the 3-month Jibar rate and at 5-percent level for the forward rates (Table 4).
    - Trend in surprise component for 1-month rate negative but not statistically significant.
    - Note: small sample size (24 MPC announcements in the period).

- Dispersion of interest rate forecasts:
  - Cross-sectional dispersion measured as maximum minus minimum forecast of the prime rate in the Reuters monthly survey.
  - Dispersion declined over time at both 1-quarter-ahead and 4-quarters-ahead horizons (Figure 5).
  - Negative trend in dispersion statistically significant at the 5-percent level (Table 5).

- Inflation trends and volatility:
  - CPI inflation declined markedly over the past decade; average CPI inflation rate in 2006 (4½ percent) lower by one-third against its 1996 level (Figure 6).
  - CPIX inflation (computed since 1997, basis for SARB’s official inflation target) also fell markedly.
  - Sharp rise in inflation in 2002 reflecting earlier easing of monetary conditions and pass-through of a sharp exchange rate depreciation in late 2001; rise was temporary.
  - Increase in monetary policy transparency since 2000 coincided with a statistically significant negative trend in the inflation rate despite the 2002 spike (Table 6).
  - CPIX inflation volatility:
    - 1998–2006: 6-month standard deviation of nominal effective exchange rate changes showed slight upward trend (Figure 9), owing to three consecutive quarters of higher volatility starting in Q4 2002.
    - December 2003–December 2006: exchange rate volatility nearly 20 percent lower than March 1998–September 2003.
    - Inflation volatility trend increase since 1998 attributable to 2002–part of 2003 turbulence; inflation volatility declined during 2003–2006.

- Inflation forecast errors:
  - Forecast errors measured by absolute differential between actual CPIX outturn for a period and Reuters mean forecast made at a previous date (quarterly forecasts).
  - CPIX inflation forecast errors declined markedly in size during 2003–2006.
  - Figure 8:
    - 2000–2006: 1- and 4-quarter forecast errors have roughly flat and slightly negative trends respectively; trends not statistically significant.
    - 2003–2006: negative trends in forecast errors, with stronger decline and statistical significance at 1-quarter horizon than at 4-quarter horizon.
  - Market forecasters tended to over-predict inflation during the last few years of sample: consistently negative forecast errors from Q3 2003 until Q3 2006.
  - Role of exchange rate volatility:
    - At 1-quarter horizon, a 1-percentage point decline in exchange rate volatility associated with a less than 0.1 percentage point decline in forecast error.
    - At 4-quarter horizon, impact of exchange rate volatility not statistically significant.
    - Including exchange rate volatility reduces statistical significance of trend in forecast errors at both horizons; exchange rate volatility is only one factor and may itself be influenced by monetary policy framework.

- Dispersion of inflation forecasts:
  - Cross-sectional dispersion defined as maximum minus minimum forecast of CPIX in Reuters monthly survey.
  - Dispersion declined over time at 1-quarter and 4-quarter horizons (Figure 10).
  - Negative trend in dispersion statistically significant for 2003–2006 period.
  - Not significant for full 2000–2006 period due to turbulence when forecasts performed poorly and were disparate.

- Comparison with GDP forecast accuracy:
  - If general forecasting skill improved, real GDP forecasts (less directly related to monetary policy) should also show improvements.
  - Evidence on GDP forecast improvements is less convincing than for interest rates.
  - Figure 11: absolute GDP forecast errors (1- and 4-quarter horizons) over 2000–2006:
    - Errors did not change appreciably at either horizon.
    - Slight negative trend at 1-quarter horizon and slight positive trend at 4-quarter horizon; neither trend statistically significant.
  - Excluding 2002 turbulence and focusing on 2003–2006 changes does not contradict conclusion that improvements in interest rate forecasts exceeded those in GDP forecasts.

*Source: _wp07123 - References (content pages I–III provided).*

### conclusion only slightly with regard to GDP forecasts (Figure 12). During 2003-2006,

### CONCLUSIONS

### Main findings on forecast accuracy
- Private sector financial forecasts in South Africa have improved in recent years.
- Accuracy of interest rate forecasts increased, as reflected in a negative trend in forecast errors.
- Forecasters have been less surprised over time by MPC announcements.
- Cross-sectional dispersion of interest rate forecasts declined (forecasters have become less diverse).
- Inflation forecasts have become more accurate over time.
- The improvements in interest rate and inflation forecasts are greater than those in real GDP forecasts.
- During 2003-2006, forecast errors declined at both the 1- and 4-quarter horizons for GDP (the trends were negative) although the decline was not statistically significant.
- The cross-sectional dispersion of GDP forecasts declined at both the 1- and 4-quarter horizons, and the decline was statistically significant, suggesting an increase in unanimity among forecasters.

### Interpretation and likely drivers
- The greater improvements in interest rate and inflation forecasts relative to real GDP forecasts suggest that monetary policy transparency is likely to have played a role.
- Inflation and interest rates are closely related to the central bank’s policy instrument, whereas GDP/income is influenced by many other factors in the short run; this helps explain why forecast quality improved more for interest rates and inflation than for GDP.

### Policy-relevant observations and recommendations for further research
- It would be useful in future research to examine the role of alternative factors, such as exchange rate volatility and monetary variables, in accounting for the decline in inflation forecast errors over time.
- For completeness, it would be useful to examine trends in forecast quality for other real sector variables, such as manufacturing, industrial production, and retail sales; these series are not available in the Reuters survey used for the current analysis.

### Selected changes in SARB transparency, 1994–2006 (high-level timeline)
- 1996: Existence and extent of SARB foreign exchange forward book disclosed. Thereafter, details of net open forward position regularly published.
- October 1999: Establishment of a Monetary Policy Committee (MPC). MPC to meet every 6–8 weeks; policy statements published.
- February 2000: Adoption by government of inflation target for the SARB to implement (3–6 percent annual average by 2002).
- February 2000: Introduction of monetary policy forums convened by SARB. Held twice yearly in major financial centers and attended by labor, business, government, and academic representatives.
- March 2001: Introduction of Monetary Policy Review (MPR), published twice yearly. MPR includes SARB’s inflation forecast fan chart.
- 2002: Frequency of MPC meetings reduced to 4 times yearly.
- May 2003: Frequency of MPC meetings increased to 6 times yearly.
- November 2003: Refinement of inflation target from annual (average) basis to continuous basis (12-month rate).

### Data sources and limitations (relevant for interpretation)
- Data on the Johannesburg Interbank Rate (Jibar) and forward rate agreements (FRA) were obtained from the J.P. Morgan data source, Data Query, compiled from daily closing rates.
- Data on the cross-sectional dispersion of private sector interest rate and inflation forecasts were obtained from the Reuters monthly survey of 14 financial market institutions. Dispersion was calculated as the maximum minus the minimum forecast for (i) the prime rate and (ii) the CPIX inflation rate, at both the 1-quarter ahead and the 4-quarters-ahead horizons.
- Data on the private sector forecasts of CPIX inflation and GDP growth were taken from the Reuters monthly survey of 14 financial market institutions.
- Forecast series for other real sector variables (manufacturing, industrial production, retail sales) were not available in the Reuters survey used.

*Source: IMF Working Paper conclusions and supporting tables and figures (excerpt).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp07123.pdf_