## wp1839 - 1. Introduction

## Source details

**Canonical URL:** [wp1839 - 1. Introduction](https://www.imf.org/-/media/files/publications/wp/2018/wp1839.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2018/wp1839.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2018/wp1839.pdf.json)

---

### Key findings
- Recessions occur about 10-12 percent of the time across the sample of countries.
- Recessions are rarely forecast in advance: forecasters tend to detect the possibility of a recession but substantially understate its magnitude until the forecast horizon is nearly closed.
- Forecast revisions during non-recession years display considerable rigidity; during recession years forecasts are revised more rapidly but not rapidly enough to avoid large errors.
- The pattern of missed magnitude and sluggish downward revision is shared by private-sector (Consensus Forecasts) and official-sector (IMF WEO) forecasters.
- Prior literature confirms similar findings (Lewis and Pain (2014); Zarnowitz (1991); Loungani (2001); Abreu (2011); González Cabanillas and Terzi (2012); Dovern and Jannsen (2017)).

### Illustrative evidence (selected statistics)
- Average forecasts in April of the year before a recession: 3 percent (both Consensus and IMF).
- Example (United States 2009): realized -3 percent vs. earlier forecasts nearer +3 percent.
- By October of the year of the recession, forecasts were for a fall in output in 118 of the 153 cases.

*Source: IMF World Economic Outlook, Consensus Forecasts, and authors’ estimates (from "wp1839 - 1. Introduction").*

### Interpretation and possible explanations
- Forecasters are generally aware that recession years will differ from other years but miss the magnitude until late in the forecasting horizon.
- Potential reasons (not resolved in this paper):
  - Lack of timely or high-quality data.
  - Economic models may not predict outlier events well.
  - Recessions may be triggered by inherently hard-to-predict events.
  - Incentive problems (e.g., reputational concerns) could discourage early, aggressive downward revisions.
- Evidence excluding crisis-linked recessions (Appendix Figure A1) yields similar results, suggesting crisis vs. non-crisis distinction does not fully explain the failure.
- Similar underprediction occurs for booms: forecasts for boom years start near the unconditional average and remain below actual growth even by December of the boom year (about 1.5 percent lower on average).

### Data (Section 2 — definitions, sources, coverage)
- Forecast target: annual real GDP growth. "Target year" denotes the year forecasted.
- Forecast types:
  - Year-ahead forecasts: made in the year before the target year.
  - Current-year forecasts: made during the target year.
- Consensus Forecasts (private sector):
  - Monthly series of year-ahead and current-year forecasts.
  - First year-ahead forecast: January before the target year.
  - Last current-year forecast: December of the target year.
  - For any target year there is a sequence of 24 forecasts; horizon 'h' takes values from 1 to 24.
- IMF forecasts (official sector):
  - Published every April and October.
  - Available at 4 of the 24 horizons: April(t-1) = h=21; Oct(t-1) = h=15; Apr(t) = h=9; Oct(t) = h=3.
- Sample:
  - 63 countries: 29 advanced economies and 34 emerging economies.
  - Longest period available: 1992 to 2014; some countries start later (unbalanced panel).
  - Sources: Consensus Forecasts and IMF World Economic Outlook; actual real GDP growth from the IMF.
- Recession definition:
  - A year when output growth was negative.
  - Sample totals: 1306 country-year observations; 153 recessions (86 in advanced economies and 67 in emerging markets).
  - Economies are in recession 153 years out of 1306 = 12 percent of the time.
  - In April of the year before the recession, forecasters expected output to fall in only 5 of these 153 cases.
  - By October of the year of the recession, forecasts were for a fall in output in 118 of the 153 cases.

### Evolution of forecasts and forecast errors (Section 3 and 3.1)
- Type 1 error (recession happened but was not forecast):
  - April(t-1): 148 of 153 recessions missed (forecasts were for positive growth).
  - Oct(t): 35 recessions missed.
  - Downward revisions in consensus forecasts during recessions: 134 (Apr[t-1]), 147 (Oct[t-1]), 129 (Apr[t]), 125 (Oct[t]).
  - Corresponding IMF downward revision counts (where reported): 146, 121 (for comparable horizons reported in the table).
- Mean Forecast Error (MFE) during recessions (Consensus Forecasts):
  - All countries: -5.85 (Apr[t-1]), -4.82 (Oct[t-1]), -2.01 (Apr[t]), -0.41 (Oct[t])
  - Advanced: -4.68, -3.75, -1.49, -0.47
  - Emerging: -7.35, -6.20, -2.67, -0.32
- IMF MFE during recessions (reported horizons):
  - All countries: -5.85 (Apr[t-1]), -5.15 (Oct[t-1]), -1.94 (Apr[t]), -0.52 (Oct[t])
  - Advanced: -4.53, -3.84, -1.23, -0.53
  - Emerging: -7.53, -6.82, -2.86, -0.51
- Performance pre- vs. post-Great Recession:
  - Pre-Great Recession (70 recessions): Consensus MFE all countries: -6.31, -5.33, -2.81, -0.52 (Apr[t-1] to Oct[t])
  - Post-Great Recession (83 recessions): Consensus MFE all countries: -5.46, -4.39, -1.33, -0.31
  - General pattern: somewhat better performance post-Great Recession (larger proportion of recessions forecast and smaller MFE), with nuances across advanced vs emerging economies and forecast horizons.
- Type 2 error (recession forecasted but did not happen) — out of 1153 non-recession episodes:
  - Number of false forecasts (Consensus): Apr[t-1]=8, Oct[t-1]=18, Apr[t]=27, Oct[t]=33
  - Number of false forecasts (IMF): Apr[t-1]=14, Oct[t-1]=11, Apr[t]=24, Oct[t]=27
  - MFE for false forecasts (Consensus — All countries): 3.42, 4.11, 2.57, 1.47 (Apr[t-1] to Oct[t])
  - MFE for false forecasts (IMF — All countries): 3.73, 5.71, 3.18, 1.49
- Summary: Type 1 errors (missing actual recessions) are far more common and larger in magnitude than Type 2 errors (falsely forecasting recessions).

### Comparing recessions and non-recession years (Section 3.2)
- Average realized growth in recession years:
  - Advanced economies: about -2 percent.
  - Emerging economies: about -3 percent.
- Unconditional (all-years) forecast path:
  - Starts near 3 percent for advanced economies and near 4.5 percent for emerging economies and is slowly revised down over 24 months.
- Forecasts for recession years:
  - Begin close to the unconditional average in the year preceding the recession.
  - Start to deviate around mid-year before the target year, indicating forecasters become aware of deteriorating prospects.
  - Revisions are serially correlated and too small in magnitude; forecasts only catch up to realized outcomes by December of the recession year.
- Similarity across forecasters and country groups:
  - Consensus and IMF forecasts show similar smooth downward revisions.
  - Pattern holds for advanced and emerging economies alike.
- Evidence on booms:
  - Booms defined as years with growth greater than one standard deviation above the country average.
  - Forecasts for boom years also start at the unconditional average and underpredict actual growth by about 1.5 percent even by December of the boom year.
  - This symmetric underreaction to outliers (both recessions and booms) supports the view that lack of information or model limitations (difficulty producing large swings from steady state) drives part of the forecasting failure.
- Behavioral explanation:
  - People tend to smooth forecasts, incorporating surprises too slowly; this could partly explain slow revision behavior (Nordhaus, 1987).

### Information rigidity around turning points (Section 3.3)
- Efficiency test framework:
  - Under full information rational expectations, forecast revisions should be serially uncorrelated (martingale).
  - Regression used: Rev_it,h = α_h + β_h Rev_it,h+k + μ_i,h + ε_it,h. Null β_h = 0.
  - Dependent variable: revision between Oct[t] and Apr[t] of the current-year forecast.
  - Explanatory variable: revision between Apr[t] and Oct[t-1].
  - Country fixed effects included.
- Main findings (Table 5):
  - Consensus — Lagged Revision (β_h):
    - All: 0.35*** (0.04)
    - Advanced: 0.29*** (0.04)
    - Emerging: 0.38*** (0.06)
  - Consensus — Constant:
    - All: 0.06*** (0.02)
    - Advanced: 0.03* (0.02)
    - Emerging: 0.07*** (0.02)
  - IMF — Lagged Revision:
    - All: 0.21*** (0.04)
    - Advanced: 0.09* (0.05)
    - Emerging: 0.27*** (0.05)
  - IMF — Constant:
    - All: 0.05** (0.02)
    - Advanced: 0.02 (0.03)
    - Emerging: 0.05** (0.02)
  - Sample sizes: N. of Obs. = 1306 (All), 639 (Advanced), 667 (Emerging).
  - R-sq:
    - Consensus: All 0.18, Advanced 0.12, Emerging 0.21
    - IMF: All 0.08, Advanced 0.01, Emerging 0.13
- Interpretation:
  - Coefficient estimates are positive and significantly different from zero for all country groups and both forecast sources; reject full information rational expectations.
  - Serial correlation (information rigidity) is higher for emerging economies than for advanced economies.
  - Serial correlation is higher for Consensus than for IMF forecasts.

### Information rigidity during recessions (augmented regression and Table 6)
- Augmented regression: Rev_it,h = α_h + β_h Rev_it,h+k + γ_h Rec_it + θ_h Rev_it,h+k * Rec_it + μ_i,h + ε_it,h.
  - Negative and significant θ_h indicates relatively lower information rigidity during recession years.
  - Test whether β_h + θ_h = 0 (no rigidity during recessions).
- Key coefficient estimates (Consensus):
  - Lagged Revision: All 0.35*** (0.06); Advanced 0.36*** (0.12); Emerging 0.33*** (0.06)
  - Lagged Rev.*Rec. (θ_h): All -0.26* (0.14); Advanced -0.34 (0.25); Emerging -0.29* (0.16)
  - Recession (γ_h): All -1.59*** (0.32); Advanced -1.10** (0.42); Emerging -2.61*** (0.34)
  - Constant: All 0.15*** (0.02); Advanced 0.10*** (0.03); Emerging 0.21*** (0.03)
- Key coefficient estimates (IMF):
  - Lagged Revision: All 0.13** (0.05); Advanced 0.22** (0.09); Emerging 0.07 (0.06)
  - Lagged Rev.*Rec.: All -0.11 (0.12); Advanced -0.45** (0.21); Emerging 0.00 (0.12)
  - Recession (γ_h): All -1.60*** (0.34); Advanced -1.46*** (0.46); Emerging -2.44*** (0.38)
  - Constant: All 0.16*** (0.03); Advanced 0.12*** (0.03); Emerging 0.20*** (0.03)
- Sample sizes and fit:
  - N. of Obs.: 1306 (All), 639 (Advanced), 667 (Emerging).
  - R-sq:
    - Consensus: All 0.25, Advanced 0.17, Emerging 0.35
    - IMF: All 0.15, Advanced 0.12, Emerging 0.27
- P-values for β_h + θ_h = 0:
  - Consensus: 0.40 (All), 0.90 (Advanced), 0.76 (Emerging)
  - IMF: 0.82 (All), 0.11 (Advanced), 0.34 (Emerging)
- Interpretation:
  - Interaction coefficients (Lagged Rev.*Rec.) are negative, indicating lower information rigidity during recessions relative to other years.
  - For all six tests, p-values lead to failing to reject β_h + θ_h = 0 — conclusion: there is no information rigidity during recessions.
  - Recession dummy coefficients (γ_h) are negative and significant: forecast revisions are relatively larger for recession years than for normal years.

### Comparator analysis: Consensus vs IMF forecasts
- Correlation between forecast errors of the two sources exceeds 0.9.
- Diebold and Mariano (1995) tests (quadratic loss DMS and absolute loss DMA) applied across 63 countries; focus on quadratic loss.
- Example at horizon = 15 months (October of the year before the recession):
  - Consensus Forecasts more accurate for 47 countries, of which 14 are significant.
  - IMF forecasts more accurate for 16 countries, of which 3 are significant.
- General patterns:
  - Consensus Forecasts tend to be more accurate than IMF forecasts for a larger proportion of countries overall.
  - By country group:
    - Advanced economies: IMF forecasts made in April in the year-ahead and current-year are more accurate for more countries.
    - Emerging economies: Consensus Forecasts are more accurate for more countries.

### Overall conclusions (Section 5)
- The ability to predict turning points is limited.
- Forecasts in recession years are revised each month but fail to capture the onset of recessions in a timely way and miss the extent of output decline during recessions by a wide margin; this holds for both private sector (Consensus) and official sector (IMF) forecasts.
- Three non-mutually exclusive classes of theories suggested:
  - Forecasters do not have enough information to reliably call a recession.
  - Forecasters lack incentives to predict a recession (asymmetric loss functions).
  - Behavioral reasons: forecasters hold on to priors and revise slowly and insufficiently in response to incoming information (Nordhaus, 1987).
- Additional empirical notes:
  - Related literature: Mankiw and Reis (2002) sticky information; Sims (2003) and Woodford (2003) noisy information.
  - Empirical findings align with Coibion and Gorodnichenko (2012, 2015) and Dovern (2013) that degree of information rigidity declines during recessions; Dovern et al. (2012) find disagreement in growth forecasts increases in recession years.
  - Forecasts are updated often (e.g., monthly consensus revisions in recession years), but revisions are insufficient to capture the onset of recessions.

*Source: wp1839 - 1. Introduction and 3.3 Information rigidity around turning points*

### 1. Introduction ........................................................................................................

### wp1839 - 1. Introduction ........................................................................................................

### Key findings (from the Introduction)
- Recessions occur about 10-12 percent of the time across the sample of countries.
- Recessions are rarely forecast in advance: forecasters tend to detect the possibility of a recession but substantially understate its magnitude until the forecast horizon is nearly closed.
- Forecast revisions during non-recession years display considerable rigidity; during recession years forecasts are revised more rapidly but not rapidly enough to avoid large errors.
- The pattern of missed magnitude and sluggish downward revision is shared by private-sector (Consensus Forecasts) and official-sector (IMF WEO) forecasters.
- Prior literature confirms similar findings (Lewis and Pain (2014); Zarnowitz (1991); Loungani (2001); Abreu (2011); González Cabanillas and Terzi (2012); Dovern and Jannsen (2017)).

### Illustrative evidence (Figure 1 examples summarized)
- Average forecasts in April of the year before a recession: 3 percent (both Consensus and IMF).
- By October of the year before, forecasts are marked down but still far from signaling a recession.
- In the year of the recession, forecasts call for a recession by April but still understate its magnitude (example: realized -3 percent vs. earlier forecasts nearer +3 percent for the United States 2009).
- Only by year-end do forecasts converge to realized outcomes.

### Interpretation and possible explanations
- Forecasters are generally aware that recession years will differ from other years but miss the magnitude until late in the forecasting horizon.
- Potential reasons (not resolved in this paper):
  - Lack of timely or high-quality data.
  - Economic models may not predict outlier events well.
  - Recessions may be triggered by inherently hard-to-predict events.
  - Incentive problems (e.g., reputational concerns) could discourage early, aggressive downward revisions.
- Evidence excluding crisis-linked recessions (Appendix Figure A1) yields similar results, suggesting crisis vs. non-crisis distinction does not fully explain the failure.
- Similar underprediction occurs for booms: forecasts for boom years start near the unconditional average and remain below actual growth even by December of the boom year (about 1.5 percent lower on average), consistent with model limitations in generating large deviations from steady state.

---

### Data (Section 2 — definitions, sources, coverage)
- Forecast target: annual real GDP growth. "Target year" denotes the year forecasted.
- Forecast types:
  - Year-ahead forecasts: made in the year before the target year.
  - Current-year forecasts: made during the target year.
- Consensus Forecasts (private sector):
  - Monthly series of year-ahead and current-year forecasts.
  - First year-ahead forecast: January before the target year.
  - Last current-year forecast: December of the target year.
  - For any target year there is a sequence of 24 forecasts; horizon 'h' takes values from 1 to 24.
- IMF forecasts (official sector):
  - Published every April and October.
  - Available at 4 of the 24 horizons: April(t-1) = h=21; Oct(t-1) = h=15; Apr(t) = h=9; Oct(t) = h=3.
- Sample:
  - 63 countries: 29 advanced economies and 34 emerging economies.
  - Longest period available: 1992 to 2014; some countries start later (unbalanced panel).
  - Sources: Consensus Forecasts and IMF World Economic Outlook; actual real GDP growth from the IMF.
- Recession definition:
  - A year when output growth was negative.
  - Sample totals: 1306 country-year observations; 153 recessions (86 in advanced economies and 67 in emerging markets).
  - Economies are in recession 153 years out of 1306 = 12 percent of the time.
  - In April of the year before the recession, forecasters expected output to fall in only 5 of these 153 cases.
  - By October of the year of the recession, forecasts were for a fall in output in 118 of the 153 cases.

---

### Evolution of forecasts and forecast errors (Section 3 and 3.1: Type 1 vs. Type 2 errors)
- Type 1 error (recession happened but was not forecast):
  - April(t-1): 148 of 153 recessions missed (forecasts were for positive growth).
  - Oct(t): 35 recessions missed.
  - Downward revisions in consensus forecasts during recessions: 134 (Apr[t-1]), 147 (Oct[t-1]), 129 (Apr[t]), 125 (Oct[t]).
  - Corresponding IMF downward revision counts (where reported): 146, 121 (for comparable horizons reported in the table).
- Mean Forecast Error (MFE) during recessions (Consensus Forecasts):
  - All countries: -5.85 (Apr[t-1]), -4.82 (Oct[t-1]), -2.01 (Apr[t]), -0.41 (Oct[t])
  - Advanced: -4.68, -3.75, -1.49, -0.47 (same horizon ordering)
  - Emerging: -7.35, -6.20, -2.67, -0.32 (same horizon ordering)
- IMF MFE during recessions (reported horizons):
  - All countries: -5.85 (Apr[t-1]), -5.15 (Oct[t-1]), -1.94 (Apr[t]), -0.52 (Oct[t])
  - Advanced: -4.53, -3.84, -1.23, -0.53
  - Emerging: -7.53, -6.82, -2.86, -0.51
- Performance pre- vs. post-Great Recession:
  - Pre-Great Recession (70 recessions): Consensus MFE all countries: -6.31, -5.33, -2.81, -0.52 (Apr[t-1] to Oct[t])
  - Post-Great Recession (83 recessions): Consensus MFE all countries: -5.46, -4.39, -1.33, -0.31
  - General pattern: somewhat better performance post-Great Recession (larger proportion of recessions forecast and smaller MFE), with nuances across advanced vs emerging economies and forecast horizons.
- Type 2 error (recession forecasted but did not happen) — out of 1153 non-recession episodes:
  - Number of false forecasts (Consensus): Apr[t-1]=8, Oct[t-1]=18, Apr[t]=27, Oct[t]=33
  - Number of false forecasts (IMF): Apr[t-1]=14, Oct[t-1]=11, Apr[t]=24, Oct[t]=27
  - MFE for false forecasts (Consensus — All countries): 3.42, 4.11, 2.57, 1.47 (Apr[t-1] to Oct[t])
  - MFE for false forecasts (IMF — All countries): 3.73, 5.71, 3.18, 1.49
- Summary: Type 1 errors (missing actual recessions) are far more common and larger in magnitude than Type 2 errors (falsely forecasting recessions).

---

### Comparing recessions and non-recession years (Section 3.2)
- Average realized growth in recession years:
  - Advanced economies: about -2 percent.
  - Emerging economies: about -3 percent.
- Unconditional (all-years) forecast path:
  - Starts near 3 percent for advanced economies and near 4.5 percent for emerging economies and is slowly revised down over 24 months.
- Forecasts for recession years:
  - Begin close to the unconditional average in the year preceding the recession.
  - Start to deviate around mid-year before the target year, indicating forecasters become aware of deteriorating prospects.
  - Revisions are serially correlated and too small in magnitude; forecasts only catch up to realized outcomes by December of the recession year.
- Similarity across forecasters and country groups:
  - Consensus and IMF forecasts show similar smooth downward revisions.
  - Pattern holds for advanced and emerging economies alike.
- Evidence on booms:
  - Booms defined as years with growth greater than one standard deviation above the country average.
  - Forecasts for boom years also start at the unconditional average and underpredict actual growth by about 1.5 percent even by December of the boom year.
  - This symmetric underreaction to outliers (both recessions and booms) supports the view that lack of information or model limitations (difficulty producing large swings from steady state) drives part of the forecasting failure.
- Behavioral explanation (Nordhaus, 1987):
  - People tend to smooth forecasts, incorporating surprises too slowly; this could partly explain slow revision behavior.

---

*Source: IMF World Economic Outlook, Consensus Forecasts, and authors’ estimates (from the content of "wp1839 - 1. Introduction").*

### 3.3 Information rigidity around turning points

### 3.3 Information rigidity around turning points

### Efficiency test framework and empirical specification
- Under full information rational expectations, a sequence of forecasts for the same target should follow a martingale: forecast revisions should be serially uncorrelated.
- Regression used to test efficiency (equation (1)):
  - Rev_it,h = α_h + β_h Rev_it,h+k + μ_i,h + ε_it,h
  - Null hypothesis: β_h = 0. A positive and significant β_h indicates information rigidity (forecast smoothing).
- Dependent variable used: revision between Oct[t] and Apr[t] of the current-year forecast.
- Explanatory variable used: revision between Apr[t] and Oct[t-1].
- Country fixed effects included.

### Main findings on information rigidity (Table 5)
- Results reported for Consensus and IMF forecasts; samples: All, Advanced, Emerging.
- Coefficient estimates on Lagged Revision (β_h) and constants:
  - Consensus — Lagged Revision:
    - All: 0.35*** (0.04)
    - Advanced: 0.29*** (0.04)
    - Emerging: 0.38*** (0.06)
  - Consensus — Constant:
    - All: 0.06*** (0.02)
    - Advanced: 0.03* (0.02)
    - Emerging: 0.07*** (0.02)
  - IMF — Lagged Revision:
    - All: 0.21*** (0.04)
    - Advanced: 0.09* (0.05)
    - Emerging: 0.27*** (0.05)
  - IMF — Constant:
    - All: 0.05** (0.02)
    - Advanced: 0.02 (0.03)
    - Emerging: 0.05** (0.02)
- Sample sizes and fit:
  - N. of Obs. for each column: 1306 (All), 639 (Advanced), 667 (Emerging) for both Consensus and IMF.
  - R-sq:
    - Consensus: All 0.18, Advanced 0.12, Emerging 0.21
    - IMF: All 0.08, Advanced 0.01, Emerging 0.13
- Interpretation of Table 5:
  - The coefficient estimates are positive and significantly different from zero for all country groups for both forecast sources; the null hypothesis of full information rational expectations can be rejected.
  - Serial correlation (information rigidity) is higher for emerging economies than for advanced economies.
  - Serial correlation is higher for Consensus than for IMF forecasts.

### Information rigidity around recessions — augmented specification
- Augmented regression (equation (2)) to test turning-point behavior:
  - Rev_it,h = α_h + β_h Rev_it,h+k + γ_h Rec_it + θ_h Rev_it,h+k * Rec_it + μ_i,h + ε_it,h
  - Interpretation: a negative and significant θ_h indicates relatively lower information rigidity during recession years; test also whether β_h + θ_h = 0 (i.e., no rigidity during recessions).

### Results for recession episodes (Table 6)
- Coefficients (standard errors in parentheses) and key statistics:
  - Consensus — Lagged Revision:
    - All: 0.35*** (0.06)
    - Advanced: 0.36*** (0.12)
    - Emerging: 0.33*** (0.06)
  - Consensus — Lagged Rev.*Rec. (θ_h):
    - All: -0.26* (0.14)
    - Advanced: -0.34 (0.25)
    - Emerging: -0.29* (0.16)
  - Consensus — Recession (γ_h):
    - All: -1.59*** (0.32)
    - Advanced: -1.10** (0.42)
    - Emerging: -2.61*** (0.34)
  - Consensus — Constant:
    - All: 0.15*** (0.02)
    - Advanced: 0.10*** (0.03)
    - Emerging: 0.21*** (0.03)
  - IMF — Lagged Revision:
    - All: 0.13** (0.05)
    - Advanced: 0.22** (0.09)
    - Emerging: 0.07 (0.06)
  - IMF — Lagged Rev.*Rec.:
    - All: -0.11 (0.12)
    - Advanced: -0.45** (0.21)
    - Emerging: 0.00 (0.12)
  - IMF — Recession:
    - All: -1.60*** (0.34)
    - Advanced: -1.46*** (0.46)
    - Emerging: -2.44*** (0.38)
  - IMF — Constant:
    - All: 0.16*** (0.03)
    - Advanced: 0.12*** (0.03)
    - Emerging: 0.20*** (0.03)
- Sample sizes and fit:
  - N. of Obs.: 1306 (All), 639 (Advanced), 667 (Emerging) for both Consensus and IMF.
  - R-sq:
    - Consensus: All 0.25, Advanced 0.17, Emerging 0.35
    - IMF: All 0.15, Advanced 0.12, Emerging 0.27
- P-Values reported for hypothesis β_h + θ_h = 0:
  - Consensus columns: 0.40 (All), 0.90 (Advanced), 0.76 (Emerging)
  - IMF columns: 0.82 (All), 0.11 (Advanced), 0.34 (Emerging)
- Interpretation of Table 6:
  - Coefficients on the interaction variable (Lagged Rev.*Rec.) are all negative, indicating a lower level of information rigidity during recession episodes relative to other years.
  - For all six tests (three groups × two forecast sources), the p-values lead to failing to reject the hypothesis β_h + θ_h = 0 — conclusion: there is no information rigidity during recessions.
  - The signs of the recession dummy (γ_h) are all negative and significant: forecast revisions are relatively larger for recession years than for normal years.

### Comparator analysis: Consensus vs IMF forecasts
- Correlation between forecast errors of the two sources exceeds 0.9.
- Diebold and Mariano (1995) test applied using quadratic loss (DMS) and absolute loss (DMA); summary focuses on quadratic loss across 63 countries.
- Example summary statistic at horizon = 15 months (October of the year before the recession):
  - Consensus Forecasts more accurate for 47 countries, of which 14 are significant.
  - IMF forecasts more accurate for 16 countries, of which 3 are significant.
- General patterns:
  - Consensus Forecasts tend to be more accurate than IMF forecasts for a larger proportion of countries overall.
  - By country group:
    - Advanced economies: IMF forecasts made in April in the year-ahead and current-year are more accurate for more countries.
    - Emerging economies: Consensus Forecasts are more accurate for more countries.

### Overall conclusions (Section 5)
- The ability to predict turning points is limited.
- Forecasts in recession years are revised each month but fail to capture the onset of recessions in a timely way and miss the extent of output decline during recessions by a wide margin; this holds for both private sector (Consensus) and official sector (IMF) forecasts.
- The paper does not identify a definitive explanation; three non-mutually exclusive classes of theories are suggested:
  - Forecasters do not have enough information to reliably call a recession (economic models insufficient; recessions caused by difficult-to-anticipate shocks).
  - Forecasters lack incentives to predict a recession (asymmetric loss functions; greater loss from incorrectly calling a recession than benefits from correctly calling one).
  - Behavioral reasons: forecasters hold on to priors and revise slowly and insufficiently in response to incoming information (Nordhaus, 1987).
- Additional empirical notes:
  - Related literature: Mankiw and Reis (2002) sticky information; Sims (2003) and Woodford (2003) noisy information.
  - Empirical findings align with Coibion and Gorodnichenko (2012, 2015) and Dovern (2013) that degree of information rigidity declines during recessions; Dovern et al. (2012) find disagreement in growth forecasts increases in recession years.
  - Forecasts are updated often (e.g., monthly consensus revisions in recession years), but revisions are insufficient to capture the onset of recessions.

*Source: wp1839 - 3.3 Information rigidity around turning points*

### References

### References

### Cited works

1. Abreu, I. (2011), “International Organisations’ vs. Private Analysts’ Forecasts: An Evaluation”, Banco de Portugal Working Paper, 21/2011.

2. Batchelor, R. and P. Dua (1995), “Forecaster Diversity and the Benefits of Combining Forecasts”, Management Science, 41(1), 68-75.

3. Coibion, O. and Y. Gorodnichenko (2015), “Information Rigidity and the Expectations Formation Process: A Simple Framework and New Facts”, American Economic Review, 105, 2644-2678.

4. Coibion, O. and Y. Gorodnichenko (2012), “What can survey forecasts tell us about informational rigidities?” Journal of Political Economy 120(1) 116- 159.

5. Diebold, F.X. and R.S. Mariano (1995), “Comparing Predictive Accuracy,” Journal of Business and Economic Statistics, 13, 253–263.

6. Dovern, J. (2013), “When Are GDP Forecasts Updated? Evidence from a Large International Panel”, Economics Letters, 120(3), 521-523.

7. Dovern, J., U. Fritsche, and J. Slacalek (2012), “Disagreement among Forecasters in G7 Countries”, Review of Economics and Statistics, 94(4), 1081-1096.

8. Dovern, J., U. Fritsche, P. Loungani, and N. Tamirisa (2015), “Information Rigidities: Comparing Average and Individual Forecasts for a Large International Panel”, International Journal of Forecasting, 31(1), 144-154.

9. Dovern, J. and N. Jannsen (2017), “Systematic Errors in Growth Expectations over the Business Cycle”, International Journal of Forecasting, 33(4), 760-769.

10. Fintzen, D. and H.O. Stekler (1999), “Why Did Forecasters Fail to Predict the 1990 Recession,” International Journal of Forecasting, 15(3), 309-323.

11. González Cabanillas, L. and A. Terzi (2012), “The accuracy of the European Commission's forecasts re-examined”, European Economy, Economic Papers 476. Brussels: European Commission.

12. IMF, Independent Evaluation Office (2014), “IMF Forecasts: Process, Quality and Country Perspectives”.

13. Laeven L. and F. Valencia (2008), "Systemic Banking Crises: A New Database", IMF Working Paper, WP/08/224.

14. Lewis, C. and N. Pain (2015), “Lessons from OECD Forecasts during and after the Financial Crisis”, OECD Journal: Economic Studies, 2014(1), 9-39.

15. Loungani, P. (2001), “How Accurate are Private Sector Forecasts? Cross-country Evidence from Consensus Forecasts of Output Growth”, International Journal of Forecasting, 17(3), 419-432.

16. Loungani, P., H. Stekler, and N. Tamirisa (2013), “Information Rigidity in Growth Forecasts: Some Cross-country Evidence”, International Journal of Forecasting, 29(4), 605-621.

17. Mankiw, G., and R. Reis (2002), “Sticky Information versus Sticky Prices: A Proposal to Replace the New Keynesian Phillips Curve.” Quarterly Journal of Economics, 117(4), 1295-1328.

18. McNees, S.K. (1991), Forecasting Cyclical Turning Points: The Record in the Past Three Recessions. In K. Lahiri and G.H. Moore, Leading Economic Indicators: New Approaches and Forecasting. New York: Cambridge University Press.

19. Nordhaus, W. (1987), “Forecasting Efficiency: Concepts and Applications”, Review of Economics and Statistics, 69(4), 667-674.

20. Sims, C. A. (2010), “Rational Inattention and Monetary Economics.” In Handbook of Monetary Economics, edited by B. Friedman and M. Woodford. Elsevier-North Holland, vol. 3A, chapter 4, 155-181.

21. Sinclair, T. M., F. Joutz, and H. Stekler (2010), “Can the Fed Predict the State of the Economy?”, Economics Letters, 108(1), 28-32.

22. Woodford, M. (2003), “Interest and Prices: Foundations of a Theory of Monetary Policy”, Princeton: Princeton University Press, 2003.

23. Zarnowitz, V. (1991), “Has Macro-Forecasting Failed?,” NBER Working Paper No. 3867.

*Source: wp1839 - References (https://www.imf.org/-/media/files/publications/wp/2018/wp1839.pdf)*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2018/wp1839.pdf_