## 6. Perceived Long-term Effect of a 1 Percentage Point Shock to Other Variables

## Source details

**Canonical URL:** [6. Perceived Long-term Effect of a 1 Percentage Point Shock to Other Variables](https://www.imf.org/-/media/files/publications/wp/2018/wp18163.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2018/wp18163.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2018/wp18163.pdf.json)

---

### Data, method, and sample
- Data source: Consensus Economics vintages of professional long-term forecasts (twice a year since 1989, April and October), supplemented with GDS (October 2017 vintage) for actual output in forecast-efficiency tests.
- Key estimation equation (horizon h):
  - Forecast Revision of ΔYi,t:t+h|t = α + βh Forecast Revision of ΔYi,t|t + εi,t:t+h
- Horizon of interest:
  - h = 10 years (main long-term results).
  - Also estimated for h = 0 to 5 (annual revisions available) and for h = 5 using IMF WEO forecasts.
- Sample:
  - 38 advanced and emerging market countries, 1989-2017.
  - Effective number of equivalent annual observations: 21,280 (computed as 38 × (2017-1989) × 2 × 10).

### Baseline empirical findings on perceived persistence
- Cross-country panel (full sample, 1989-2017):
  - Relation between current-period forecast revision and 10-year-ahead forecast revision:
    - Slope for advanced markets: 1.89 (t-statistic = 13.19).
    - Slope for emerging markets: 1.98 (t-statistic = 13.87).
  - Hypothesis that slope = 1 is rejected at the 1 percent level.
- Country-by-country results:
  - For 36 out of 38 countries, a 1 percent innovation in real GDP typically increases forecasted real GDP in 10 years by more than 1 percent.
  - Range of point estimates:
    - Advanced markets: 0.92 to 2.59.
    - Emerging markets: 1.34 to 4.41.
  - Pesaran-Shin-Smith mean group averages:
    - Advanced markets: 1.78 (t-statistic = 17.80).
    - Emerging markets: 2.30 (t-statistic = 13.53).
- Interpretation:
  - Professional forecasters typically view shocks to current-period output as having permanent or near-permanent effects; a 1 percent surprise in current-period output typically changes the 10-year-ahead output level forecast by about 1–2 percent, with an average near 2 percent.

### Robustness checks
- Outlier trimming (drop largest and smallest 5 percent of revisions):
  - β at 10-year horizon:
    - Advanced markets: 1.57 (t-statistic = 19.63).
    - Emerging markets: 1.93 (t-statistic = 10.72).
- Sample splits:
  - 1989-2003 vs 2004-2017: Larger β estimates for 1989-2003; 2004-2017 estimates remain well above unity; for emerging markets indistinguishable from 2.
- Sign of innovation:
  - For both positive and negative current-period innovations, β estimates are above unity and statistically indistinguishable from 2 for both country groups.
- Controls:
  - Including country and time fixed effects, and lagged revisions, leaves results virtually unchanged.
- Group robustness summary (selected entries from Table 2):
  - Baseline: Advanced 1.89 (0.29) N 902 R^2 0.57; Emerging 1.98 (0.16) N 798 R^2 0.45.
  - Trimmed sample: Advanced 1.57 (0.08) N 769 R^2 0.29; Emerging 1.93 (0.18) N 679 R^2 0.29.
  - 1989-2003: Advanced 2.61 (0.21) N 372 R^2 0.66; Emerging 2.05 (0.34) N 268 R^2 0.44.
  - 2004-2017: Advanced 1.45 (0.10) N 530 R^2 0.55; Emerging 1.93 (0.09) N 530 R^2 0.45.

### Forecast-efficiency tests (do forecasters overestimate persistence?)
- Forecast-error regression:
  - Forecast Error of ΔYi,t:t+10|t = κ + θ Forecast Revision of ΔYi,t|t + εi,t:t+h
  - Data limitation: last usable 10-year-ahead forecast error corresponds to forecasts made in fall of 2007.
  - Standard errors: Newey-West with 10 lags.
- Results (selected entries from Table 3):
  - Advanced markets:
    - Baseline: θ -1.25 (0.39); β 2.55 (0.19); θ + β 1.29 (0.34); N 486.
    - Trimmed sample: θ 0.19 (0.86); β 1.90 (0.20); θ + β 2.10 (0.89); N 437.
    - Trimmed sample, country and time fixed effects: θ 0.15 (0.72); β 1.83 (0.22); θ + β 1.98 (0.79); N 437.
  - Emerging markets:
    - Baseline: θ -1.06 (0.41); β 2.04 (0.30); θ + β 0.98 (0.61); N 380.
    - Trimmed sample: θ -0.16 (1.29); β 3.09 (0.33); θ + β 2.93 (1.26); N 342.
    - Trimmed sample, country and time fixed effects: θ 0.76 (1.31); β 2.72 (0.35); θ + β 3.48 (1.28); N 342.
- Interpretation:
  - Baseline θ negative suggests perceived persistence > realized persistence, but this negative relation is driven by a small number of extreme values (outliers).
  - After trimming outliers, θ is close to zero and statistically insignificant; sums β + θ in trimmed samples imply little systematic overestimation.
  - β estimates for 1989-2007 sample: 2.55 (advanced) and 2.04 (emerging).

### Extensions and heterogeneity
- Supply vs demand shock proxies (Table 4):
  - Supply shock proxy: global oil supply shocks identified via a half-yearly VAR following Kilian (2009).
  - Demand shock proxy: differences between ex-post government consumption and ex-ante forecasts (Auerbach and Gorodnichenko).
  - Large shocks defined as above 75th percentile or below 25th percentile; periods overlapping both proxies excluded.
  - Findings:
    - Perceived persistence (β) substantially larger in periods associated with supply-shock proxies than with demand-shock proxies.
    - Even for demand-shock periods, β > 1 for both advanced and emerging markets.
  - Selected β estimates (Table 4):
    - Large AD: Advanced β 1.70 (0.40) N 111; Emerging β 1.90 (0.11) N 46.
    - Large AS: Advanced β 2.71 (0.36) N 151; Emerging β 2.87 (0.85) N 20.
    - Recession, no crisis: Advanced β 1.51 (0.22) N 69; Emerging β 1.49 (0.21) N 69.
    - Recession, crisis: Advanced β 2.26 (0.61) N 19; Emerging β 2.05 (0.10) N 24.
    - R^2: Advanced 0.501; Emerging 0.621.
- Permanent–transitory decomposition (Table 5; Krane (2011) approach):
  - Model: Yi,t = Pit + Cit with shocks to permanent component (eit level; wit long-term growth) and transitory shock uit.
  - Results (selected variance moments):
    - Advanced markets:
      - Var(y) 1.40
      - Var(w) 0.06
      - Var(e) 0.87
      - Var(u) 0.41
      - Var(e + w)/Var(y) 0.61 (0.10)
    - Emerging markets:
      - Var(y) 3.24
      - Var(w) 0.20
      - Var(e) 1.96
      - Var(u) 1.07
      - Var(e + w)/Var(y) 0.59 (0.26)
  - Interpretation: forecasters attribute more than half of current-period output fluctuations to permanent factors; Var(e) exceeds Var(u) by a factor greater than 2.
- Alternative forecasters: IMF WEO staff forecasts
  - WEO forecasts cover horizon up to h = 5.
  - Results up to h = 5 are remarkably similar: an unexpected change in current-period output normally changes the 5-year-ahead output level forecast by more than one-for-one for both Consensus Economics and IMF WEO forecasts.
- Other macro series (Table 6 highlights):
  - Y (log level of real GDP):
    - Advanced β 1.90 (0.30) N 751 R^2 0.58.
    - Emerging β 1.99 (0.18) N 646 R^2 0.44.
  - C (log level of real private consumption):
    - Advanced β 2.07 (0.22) N 745 R^2 0.35.
    - Emerging β 2.28 (0.25) N 571 R^2 0.32.
  - I (log level of real business investment):
    - Advanced β 1.39 (0.22) N 555 R^2 0.30.
    - Emerging β 1.70 (0.18) N 543 R^2 0.24.
  - π (CPI inflation):
    - Advanced β 0.13 (0.02) N 649 R^2 0.05.
    - Emerging β 0.23 (0.10) N 579 R^2 0.03.
  - i (long-term government bond nominal yield) — advanced markets only:
    - Advanced β 0.31 (0.04) N 569 R^2 0.23.
  - Summary: consumption perceived more persistent than output; investment perceived less persistent but current-period investment shocks are 2-3 times larger in absolute size than current-period output shocks for both country groups. Inflation and long-term yields are expected to fade: effects of a 1 percentage point current-period change reach 0.1–0.3 percentage point in 10 years.

### Synthesis and implications
- Main conclusion:
  - Professional forecasters (1989-2017) typically view current-period output shocks as having permanent or near-permanent effects: a 1 percent shock to current-period output typically changes the 10-year-ahead forecast by about 1–2 percent, with average estimates near 2 percent.
- Mechanisms and heterogeneity:
  - The perceived dominance of permanent factors is consistent with supply-side channels and with hysteresis mechanisms; both may contribute to persistent effects.
  - Perceived persistence is larger following supply-type shocks and during recessions associated with financial crises, though recessions without financial crises still show super-persistence for many cases.
- Policy relevance:
  - Long-term expectations shaped by perceived persistence affect households’ permanent income assessments, firms’ profit expectations, consumption, investment, and asset prices.
  - The common expectation of slow convergence of output growth to pre-shock rates implies recessions can have long-lasting level effects, including when recessions are not associated with financial crises.
- Remaining questions:
  - Further research needed to identify and model the shock processes and mechanisms (e.g., hysteresis vs supply factors) underpinning strong perceived and actual persistence.

*Source — Content unit "6. Perceived Long-term Effect of a 1 Percentage Point Shock to Other Variables ________________ 29" from the provided IMF PDF chapter.*

### References _______________________________________________________________________________________ 18

### References _______________________________________________________________________________________ 18

### Figures

- 1. Long-term Output Forecast Revisions vs. Current-period Output Forecast Revisions __________ 20
- 2. Perceived Impulse Responses: Advanced vs. Emerging Markets _______________________________ 21
- 3. Current-period Output Forecast Revisions vs. Subsequent Forecast Errors ____________________ 22
- 4. Perceived Impulse Responses for Output Level: _______________________________________________ 23
- 5. GDP, Consumption and Investment Level: Perceived Impulse Responses______________________ 24

### Tables

- 1. Individual Country Estimates. Perceived Long-term Effect of a 1 Percentage Point Shock _____ 25
- 2. Robustness ____________________________________________________________________________________ 26
- 3. Forecast Evaluation ____________________________________________________________________________ 27
- 4. Perceived Persistence of Output: Proxies for Demand and Supply Shocks _____________________ 28
- 5. Perceived Transitory and Permanent Shocks to GDP: Variance Decomposition ________________ 28

*References _______________________________________________________________________________________ 18*

### 6. Perceived Long-term Effect of a 1 Percentage Point Shock to Other Variables ________________ 29

### 6. Perceived Long-term Effect of a 1 Percentage Point Shock to Other Variables ________________ 29

### Data, method, and sample
- Data source: Consensus Economics vintages of professional long-term forecasts (twice a year since 1989, April and October), supplemented with GDS (October 2017 vintage) for actual output in forecast-efficiency tests.
- Key estimation equation (horizon h):
  - Forecast Revision of ΔYi,t:t+h|t = α + βh Forecast Revision of ΔYi,t|t + εi,t:t+h
- Horizon of interest: h = 10 years (main long-term results). Also estimated for h = 0 to 5 (where annual revisions available) and for h = 5 using IMF WEO forecasts.
- Sample: 38 advanced and emerging market countries, 1989-2017.
- Effective number of equivalent annual observations: 21,280 (computed as 38 × (2017-1989) × 2 × 10).

### Baseline empirical findings on perceived persistence
- Cross-country panel (full sample, 1989-2017):
  - Relation between current-period forecast revision and 10-year-ahead forecast revision:
    - Slope for advanced markets: 1.89 (t-statistic = 13.19).
    - Slope for emerging markets: 1.98 (t-statistic = 13.87).
  - Hypothesis that slope = 1 is rejected at the 1 percent level.
- Country-by-country results:
  - For 36 out of 38 countries, a 1 percent innovation in real GDP typically increases forecasted real GDP in 10 years by more than 1 percent.
  - Range of point estimates:
    - Advanced markets: 0.92 to 2.59.
    - Emerging markets: 1.34 to 4.41.
  - Pesaran-Shin-Smith mean group averages:
    - Advanced markets: 1.78 (t-statistic = 17.80).
    - Emerging markets: 2.30 (t-statistic = 13.53).
- Interpretation:
  - Professional forecasters typically view shocks to current-period output as having permanent or near-permanent effects; a 1 percent surprise in current-period output typically changes the 10-year-ahead output level forecast by about 1–2 percent, with an average near 2 percent.

### Robustness checks
- Outlier trimming (drop largest and smallest 5 percent of revisions):
  - β at 10-year horizon:
    - Advanced markets: 1.57 (t-statistic = 19.63).
    - Emerging markets: 1.93 (t-statistic = 10.72).
- Sample splits:
  - 1989-2003 vs 2004-2017: Larger β estimates for 1989-2003, but 2004-2017 estimates remain well above unity; for emerging markets indistinguishable from 2.
- Sign of innovation:
  - For both positive and negative current-period innovations, β estimates are above unity and statistically indistinguishable from 2 for both country groups.
- Controls:
  - Including country and time fixed effects, and lagged revisions, leaves results virtually unchanged.

### Forecast-efficiency tests (do forecasters overestimate persistence?)
- Forecast-error regression (Blanchard-Leigh style):
  - Forecast Error of ΔYi,t:t+10|t = κ + θ Forecast Revision of ΔYi,t|t + εi,t:t+h
  - Data limitation: last usable 10-year-ahead forecast error corresponds to forecasts made in fall of 2007.
  - Standard errors: Newey-West with 10 lags.
- Results:
  - Baseline (full dataset): θ is negative for advanced and emerging markets (suggesting perceived persistence > realized persistence).
  - However, negative relation is driven by a small number of extreme values (outliers).
  - Trimmed sample (exclude largest and smallest 5 percent of current-period output revisions): θ is close to zero and statistically insignificant.
- Implications for actual persistence (sums of perceived β and θ):
  - β estimates for 1989-2007 sample: 2.55 (advanced) and 2.04 (emerging).
  - Sum β + θ (full dataset): 1.29 (advanced) and 0.98 (emerging).
  - Sum β + θ (trimmed sample): 2.10 (advanced) and 2.93 (emerging).
  - Overall: limited evidence that actual 10-year persistence is substantially weaker than forecasters expect; after trimming outliers, little evidence of systematic overestimation.

### Extensions and heterogeneity
- Supply vs demand shock proxies:
  - Supply shock proxy: global oil supply shocks identified via a half-yearly VAR following Kilian (2009).
  - Demand shock proxy: differences between ex-post government consumption and ex-ante forecasts (Auerbach and Gorodnichenko).
  - Large shocks defined as > 1.67 sample standard deviations; periods overlapping both proxies excluded.
  - Findings:
    - Perceived persistence (β) substantially larger in periods associated with supply-shock proxies than with demand-shock proxies.
    - Even for demand-shock periods, β > 1 for both advanced and emerging markets.
- Recessions and financial crises:
  - Recessions defined as periods with forecasted current-period real GDP growth negative.
  - Financial-crisis recessions: within 3 years of start of a banking crisis (Laeven and Valencia, 2013).
  - Findings:
    - Recessions coincident with financial crises feature more persistent perceived output effects.
    - Recessions without financial crises still show super-persistence (β above 1), though for emerging markets β may be not significantly different from unity (permanent but not super-persistent).
- Permanent–transitory decomposition (adapted from Krane, 2011):
  - Model: Yi,t = Pit + Cit with shocks to permanent component (eit level; wit long-term growth) and transitory shock uit.
  - Identification: assume revisions to forecast of Y in 10 years reflect perceived permanent shocks; use calibration and regressions across horizons to recover eit and uit.
  - Results (Table 5):
    - Variance share of permanent components:
      - Advanced markets: Var(eit + wit) / Var(uit + eit + wit) = 61 percent (t-statistic = 8.71).
      - Emerging markets: share = 59 percent (t-statistic = 7.38).
    - Var(eit) exceeds Var(uit) by a factor > 2.
    - Interpretation: forecasters attribute more than half of current-period output fluctuations to permanent factors (including long-term growth revisions).
- Alternative forecasters: IMF World Economic Outlook (WEO) staff forecasts
  - WEO forecasts cover horizon up to h = 5 (not 10).
  - Results up to h = 5 are remarkably similar: an unexpected change in current-period output normally changes the 5-year-ahead output level forecast by more than one-for-one for both Consensus Economics and IMF WEO forecasts.
- Other macro series (Consensus Economics forecasts):
  - Private consumption: perceived more persistent than output (β larger than for output).
  - Business investment: perceived less persistent than output (β smaller), but current-period investment shocks are 2-3 times larger in absolute size than current-period output shocks for both country groups.
  - Inflation and long-term nominal government bond yields (interest rates):
    - Forecasters expect effects of a 1 percentage point unexpected current-period change to fade over time, reaching 0.1-0.3 percentage point in 10 years.
    - Interest-rate results available only for advanced markets.
    - For inflation, β point estimate for advanced markets is about half as large as for emerging markets (consistent with better-anchored long-term inflation expectations in advanced markets over 1989-2017).

### Synthesis and implications
- Professional forecasters (1989-2017) typically view current-period output shocks as having permanent or near-permanent effects:
  - A 1 percent shock to current-period output typically changes the 10-year-ahead forecast by about 1–2 percent, with average estimates near 2 percent.
- The perceived dominance of permanent factors is consistent with supply-side channels and with hysteresis mechanisms; both may contribute to persistent effects.
- Policy relevance:
  - Long-term expectations shaped by such perceived persistence affect households’ permanent income assessments, firms’ profit expectations, consumption, investment, and asset prices.
  - The finding that forecasters commonly expect slow convergence of output growth to pre-shock rates suggests that recessions can have long-lasting level effects, including when recessions are not associated with financial crises.
- Remaining questions:
  - Further research needed to identify and model the shock processes and mechanisms (e.g., hysteresis vs supply factors) underpinning strong perceived and actual persistence.

*Italic: Source — Content unit "6. Perceived Long-term Effect of a 1 Percentage Point Shock to Other Variables ________________ 29" from the provided IMF PDF chapter.*

### References

### wp18163 - References

### References
- Aguiar, Mark, and Gita Gopinath, 2007, “Emerging Market Business Cycles: The Cycle Is the Trend.” Journal of Political Economy. Vol. 115, No. 1, pp. 69–102.
- Auerbach, Alan, and Yuriy Gorodnichenko, 2012, “Measuring the Output Responses to Fiscal Policy,” American Economic Journal: Economic Policy. Vol. 4, No. 2, pp. 1–27.
- Auerbach, Alan, and Yuriy Gorodnichenko, 2013, “Output Spillovers from Fiscal Policy.” American Economic Review, Vol. 103, pp. 141–46.
- Blanchard, Olivier J., and Daniel Leigh, 2013, “Growth Forecast Errors and Fiscal Multipliers,” American Economic Review, Vol. 103, pp. 117–20.
- Blanchard, Olivier, Eugenio Cerutti, and Lawrence Summers, 2015. "Inflation and Activity - Two Explorations and their Monetary Policy Implications," NBER Working Paper No. 21726, November.
- Bry, Gerhard and Charlotte Boschan, 1971, Cyclical Analysis of Time Series: Selected Procedures and Computer Programs, New York, NBER.
- Campbell, John and Angus Deaton, 1989, “Why is Consumption So Smooth?”, Review of Economic Studies, Vol. 56, pp. 357-374.
- Campbell, John Y., and N. Gregory Mankiw, 1987, “Are Output Fluctuations Transitory?” The Quarterly Journal of Economics, Vol. 102, No. 4, pp. 857–80.
- Campbell, John Y., and N. Gregory Mankiw, 1989, “International Evidence on the Persistence of Economic Fluctuations,” Journal of Monetary Economics, Vol. 23, pp. 319-33.
- Cerra, Valerie, and Sweta Saxena, 2008, “Growth Dynamics: The Myth of Economic Recovery.” American Economic Review, Vol. 98, No. 1, pp. 439–57.
- Cerra, Valerie, and Sweta Saxena, 2017, “Booms, Crises, and Recoveries: A New Paradigm of the Business Cycle and Its Policy Implications,” IMF Working Paper No. 17/250.
- Fatas, Antonio, and Lawrence H. Summers, 2017, “The Permanent Effects of Fiscal Consolidations,” Journal of International Economics, Vol. 112, May, pp. 238-250.
- Harding, Don and Adrian Pagan, 2002, “Dissecting the cycle: a methodological investigation,” Journal of Monetary Economics, Vol. 49, pp. 365-81.
- Howard, Greg, Robert Martin, and Beth Anne Wilson, 2011, “Are Recoveries from Banking and Financial Crises Really So Different?”, Federal Reserve Board International Finance Discussion Papers No. 1037.
- International Monetary Fund (IMF), 2009, World Economic Outlook: Sustaining the Recovery, October.
- International Monetary Fund (IMF), 2012, World Economic Outlook: Growth Resuming, Dangers Remain, April.
- Kilian, Lutz, 2009, “Not All Oil Price Shocks Are Alike: Disentangling Demand and Supply Shocks in the Crude Oil Market,” American Economic Review, Vol. 99, No. 3, pp. 1053–69.
- Krane, Spencer D., 2011, “Professional Forecasters’ View of Permanent and Transitory Shocks to GDP,” American Economic Journal: Macroeconomics, Vol. 3, pp. 184-211.
- Laeven, Luc, and Fabian Valencia, 2013, “Systemic Banking Crises Database,” IMF Economic Review, Vol. 61, Issue 2, pp. 225–27.
- Martin, Robert, Tenyanna Munyan, and Beth Anne Wilson, 2015, “Potential Output and Recessions: Are We Fooling Ourselves?”, Federal Reserve Board International Finance Discussion Papers No. 1145.
- Newey, Whitney K., and Kenneth D. West, 1987, “A Simple, Positive Semi-definite, Heteroskedasticity and Autocorrelation Consistent Covariance Matrix,” Econometrica, Vol. 55, No. 3, pp. 703–708.
- Obstfeld, Maurice, 2004, “Globalization, Macroeconomic Performance, and the Exchange Rates of Emerging Economies,” Monetary and Economic Studies, December: 29-55.
- Pagan, Adrian, 1997, “Towards an Understanding of Some Business Cycle Characteristics,” Australian Economic Review, Vol. 30, No. 1, pp. 1-15.
- Pesaran, M. Hashem, Yongcheol Shin, and Ron P. Smith, 1999, “Pooled Mean Group Estimation of Dynamic Heterogeneous Panels,” Journal of the American Statistical Association, Vol. 94, No. 446, pp. 621–34.
- Reinhart, Carmen, and Kenneth Rogoff, 2009, “The Aftermath of Financial Crises,” American Economic Review, Vol. 99, No. 2, pp. 466–72.
- Sichel, Daniel E., 1994, “Inventories and the Three Phases of the Business Cycle,” Journal of Business and Economic Statistics, Vol. 12, No. 3, pp. 269–77.
- Smets, Frank, and Rafael Wouters, 2007, “Shocks and Frictions in US Business Cycles: A Bayesian DSGE Approach,” American Economic Review, Vol. 97, No. 3, pp. 586-606.

### Figures: captions and notes
- Figure 1. Long-term Output Forecast Revisions vs. Current-period Output Forecast Revisions (Percentage points)
  - Note: Revision to long-term output level is revision to cumulative growth over years t to t + 10.
  - Plot axis labels: "Revision to forecast for long-term output level" and "Revision to forecast for current-year output level"; series labeled Advanced and Emerging.
- Figure 2. Perceived Impulse Responses: Advanced vs. Emerging Markets (Scaled by standard deviation of perceived shocks; years on horizontal axis)
  - Note: Figure reports point estimates and 90 percent confidence intervals (dashes). Impulse responses estimated for years t to t + 5 and year t + 10. Fine dashes indicate interpolation for years t + 6 to t + 9.
  - Panels include GDP Level and GDP Growth with series Advanced and Emerging.
- Figure 3. Current-period Output Forecast Revisions vs. Subsequent Forecast Errors (Percentage points)
  - Note: Forecast errors are for cumulative growth over years t to t + 10.
  - Plot axis labels and series Advanced and Emerging shown.
- Figure 4. Perceived Impulse Responses for Output Level: Consensus Economics (CE) vs. IMF World Economic Outlook (WEO) Forecasters (Scaled by standard deviation of perceived shocks; years on horizontal axis)
  - Note: Figure reports point estimates and 90 percent confidence intervals (dashes). For CE forecasts, impulse responses estimated for years t to t + 5 and year t + 10; fine dashes indicate interpolation for years t + 6 to t + 9. For WEO forecasts, results available for years t to t + 5 only.
- Figure 5. GDP, Consumption and Investment Level: Perceived Impulse Responses (Scaled by standard deviation of perceived shocks; years on horizontal axis)
  - Note: Figure reports point estimates and 90 percent confidence intervals (dashes). Impulse responses estimated for years t to t + 5 and year t + 10. Fine dashes indicate interpolation for years t + 6 to t + 9.

### Tables: key estimates and statistics
- Table 1. Individual Country Estimates. Perceived Long-term Effect of a 1 Percentage Point Shock to GDP Level
  - Equation: Forecast Revision of ΔY_{i,t:t+h|t} = α + β_h Forecast Revision of ΔY_{i,t|t} + ε_{i,t:t+h}
  - Note: Estimates for each economy indicated by ISO three-letter codes. Mean indicates arithmetic mean of β estimates for each group’s economy, obtained using Pesaran, Shin, and Smith (1999) mean group estimator. Standard errors clustered by year.
  - Selected entries (β (s.e.) N R^2):
    - AUS 1.23 (0.34) 53 0.30
    - CAN 0.92 (0.25) 55 0.31
    - CHE 2.35 (0.17) 38 0.78
    - DEU 1.53 (0.39) 55 0.49
    - ESP 2.21 (0.24) 45 0.44
    - FRA 2.31 (0.25) 55 0.61
    - GBR 1.28 (0.17) 55 0.37
    - JPN 1.86 (0.51) 53 0.52
    - KOR 2.59 (0.48) 45 0.81
    - NLD 2.10 (0.36) 45 0.53
    - USA 2.02 (0.33) 55 0.63
    - ARG 1.66 (0.13) 48 0.61
    - BRA 2.06 (0.42) 48 0.50
    - CHN 3.46 (0.86) 45 0.50
    - IND 4.41 (1.41) 45 0.36
    - MEX 1.85 (0.28) 48 0.62
    - MYS 2.68 (0.37) 45 0.72
    - PER 2.84 (0.28) 40 0.57
    - POL 2.02 (0.29) 39 0.51
    - VEN 1.34 (0.40) 42 0.17
  - Group means:
    - Advanced Mean 1.78 (0.10) N 9020.51
    - Emerging Mean 2.30 (0.17) N 7980.52
- Table 2. Robustness (Group estimates of β for advanced (19) and emerging (19) markets)
  - Equation: Forecast Revision of ΔY_{i,t:t+h|t} = α + β_h Forecast Revision of ΔY_{i,t|t} + ε_{i,t:t+h}
  - Note: Standard errors clustered by year. Variants include trimmed sample, subsamples by period, positive/negative revisions, controlling for lagged revisions, adding fixed effects.
  - Advanced markets (β (s.e.) N R^2) and Emerging markets (β (s.e.) N R^2) entries:
    - Baseline: Advanced 1.89 (0.29) 902 0.57; Emerging 1.98 (0.16) 798 0.45
    - Trimmed sample: Advanced 1.57 (0.08) 769 0.29; Emerging 1.93 (0.18) 679 0.29
    - 1989-2003 sample: Advanced 2.61 (0.21) 372 0.66; Emerging 2.05 (0.34) 268 0.44
    - 2004-2017 sample: Advanced 1.45 (0.10) 530 0.55; Emerging 1.93 (0.09) 530 0.45
    - Positive revisions sample: Advanced 1.97 (0.48) 417 0.32; Emerging 1.88 (0.30) 398 0.12
    - Negative revisions sample: Advanced 1.79 (0.41) 485 0.54; Emerging 1.65 (0.18) 400 0.40
    - Controlling for lagged revisions: Advanced 1.86 (0.27) 873 0.58; Emerging 1.98 (0.14) 761 0.45
    - Adding economy-fixed effects: Advanced 1.89 (0.30) 902 0.58; Emerging 1.99 (0.16) 798 0.45
    - Adding time-fixed effects: Advanced 1.95 (0.30) 902 0.61; Emerging 1.88 (0.17) 798 0.49
- Table 3. Forecast Evaluation. Equations:
  - Forecast Error of ΔY_{i,t:t+10|t} = κ + θ Forecast Revision of ΔY_{i,t|t} + ε_{i,t:t+h}
  - Forecast Revision of ΔY_{i,t:t+10|t} = α + β Forecast Revision of ΔY_{i,t|t} + ε_{i,t:t+h}
  - Note: Trimmed sample middle 90 percent. Standard errors for θ use Newey-West (1987) with 10 lags. Standard errors for β clustered at economy level. Constant included but not reported.
  - Advanced markets (θ, β, θ + β, N):
    - Baseline: θ -1.25 (0.39); β 2.55 (0.19); θ + β 1.29 (0.34); N 486
    - Trimmed sample: θ 0.19 (0.86); β 1.90 (0.20); θ + β 2.10 (0.89); N 437
    - Trimmed sample, country and time fixed effects: θ 0.15 (0.72); β 1.83 (0.22); θ + β 1.98 (0.79); N 437
  - Emerging markets (θ, β, θ + β, N):
    - Baseline: θ -1.06 (0.41); β 2.04 (0.30); θ + β 0.98 (0.61); N 380
    - Trimmed sample: θ -0.16 (1.29); β 3.09 (0.33); θ + β 2.93 (1.26); N 342
    - Trimmed sample, country and time fixed effects: θ 0.76 (1.31); β 2.72 (0.35); θ + β 3.48 (1.28); N 342
- Table 4. Perceived Persistence of Output: Proxies for Demand and Supply Shocks
  - Note: Two regression models estimated jointly for advanced and emerging samples using saturated interactions. AS shocks are global oil supply shocks (VAR identification approach of Kilian (2009) as in Bluedorn and others (2012)). AD shocks measured by differences between ex-post government consumption and ex-ante forecasts following Auerbach and Gorodnichenko (2012, 2013). Large shocks defined as above 75th percentile or below 25th percentile with periods predominantly large AS/AD and not both. Recessions are years with negative growth. Periods of crisis defined as occurring within 3 years of start of a banking crisis identified in Laeven and Valencia (2013).
  - Sample entries (β (s.e.) N) and R^2:
    - Large AD: Advanced β 1.70 (0.40) N 111; Emerging β 1.90 (0.11) N 46
    - Large AS: Advanced β 2.71 (0.36) N 151; Emerging β 2.87 (0.85) N 20
    - Recession, no crisis: Advanced β 1.51 (0.22) N 69; Emerging β 1.49 (0.21) N 69
    - Recession, crisis: Advanced β 2.26 (0.61) N 19; Emerging β 2.05 (0.10) N 24
    - R^2: Advanced 0.501; Emerging 0.621
- Table 5. Perceived Transitory and Permanent Shocks to GDP: Variance Decomposition
  - Note: Table reports estimated variance of current-period output revisions, y, and moments of perceived transitory shocks to output, u, permanent shocks to the output level, e, and shocks to long-term growth, w, as well as bootstrapped standard errors in parentheses (based on 1,000 replications). Sample variances of y and w are fixed and thus have no standard errors. Decomposition based on Krane (2011).
  - Advanced markets:
    - Var(y) 1.40
    - Var(w) 0.06
    - Var(e) 0.87
    - Var(u) 0.41
    - Var(e + w)/Var(y) 0.61 (0.10) (0.00) (0.07)
  - Emerging markets:
    - Var(y) 3.24
    - Var(w) 0.20
    - Var(e) 1.96
    - Var(u) 1.07
    - Var(e + w)/Var(y) 0.59 (0.26) (0.01) (0.08)
- Table 6. Perceived Long-term Effect of a 1 Percentage Point Shock to Other Variables
  - Equation estimated: Forecast Revision of x_{i,t+10} = α + β Forecast Revision of x_{i,t} + ε_{i,t}
  - Note: Y = log level of real GDP; C = log level of real private consumption; I = log level of real business investment. π = CPI inflation; i = long-term government bond nominal yield. For emerging markets, sample for CPI inflation trimmed to discard the 5 lowest and highest observations for year t forecast revisions.
  - Advanced Markets (β (s.e.) N R^2) and Emerging Markets (β (s.e.) N R^2):
    - Y: Advanced 1.90 (0.30) 751 0.58; Emerging 1.99 (0.18) 646 0.44
    - C: Advanced 2.07 (0.22) 745 0.35; Emerging 2.28 (0.25) 571 0.32
    - I: Advanced 1.39 (0.22) 555 0.30; Emerging 1.70 (0.18) 543 0.24
    - π: Advanced 0.13 (0.02) 649 0.05; Emerging 0.23 (0.10) 579 0.03
    - i: Advanced 0.31 (0.04) 569 0.23
    - Table continues with additional rows not fully reproduced in this excerpt.

*Source: wp18163 - References (IMF working paper PDF).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2018/wp18163.pdf_