## wpiea2023229-print-pdf

## Source details

**Canonical URL:** [wpiea2023229-print-pdf](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023229-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023229-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023229-print-pdf.pdf.json)

---

### Database construction and coverage
- Constructed a large database with almost1,500monthly macroeconomic and financial indicators, encompassing 19 euro area countries.
- Sample starts from January 2003 with data coming from: Eurostat, OECD, ECB SDW, and BIS.
- Dataset includes both country-specific data and euro area-wide data.
- Country-specific indicators categorized into four groups:
  - Real economic activity: industrial production and its components, car registration (including different types of vehicles), total turnover index (in real term) (with different type of retail sales), unemployment rates, hours, and trade with world (i.e., import and export for different types of goods).
  - Price movements: producer price index and HICP (and their components), nominal effective exchange rate, long term rate, and wages.
  - Survey data and leading indicators: economic sentiment indicator, confidence indicators components, and leading indicators.
  - Financial data: loans, deposits, debt securities (at different maturities), share prices, and spreads.
- Euro area-wide variables included: euro area bond yields and money market rates at different maturity, Dow-Jones euro area Stoxx index (and its components), monetary aggregates M1, M3, and currency in circulation, and several bilateral exchange rates with Euro.
- The computed macroeconomic uncertainty index for the euro area considers inner heterogeneity of member countries and common components.
- Exact dataset counts and composition:
  - "In total, the dataset has 1470 variables, with 1445 country-specific variables and 25 euro area-level variables."
  - After further cleaning (dropping one of any two variables with absolute correlation larger than 0.95 for each country), "the final dataset has a total of1330variables."
- Country coverage remarks:
  - Most countries have about or more than 60 variables from the 122-variable maximum list.
  - Exceptions: Cyprus (CY, 29/122 variables), Greece (GR, 30/122 variables), Malta (MT, 36/122 variables).

### Country-specific MEU construction note
- Country-specific MEU indices are calculated as average of only the specific variables’ uncertainties for that country and euro area common exchange rates (coverage varies across countries).
- Joint estimation across all variables allows identification of common factors; country-specific measures capture both country-specific and common uncertainty.
- Country-specific MEUs are relatively close on average, between 0.81 and 0.93, though differences increase during times of stress.
- Volatility within countries spans from 0.64 to 1.70.

### Methodological approach to MEU measurement
- Follows Jurado et al. (2015) to construct an objective, data-driven measure of euro area macroeconomic uncertainty, accounting for heterogeneity.
- Two-step construction:
  - 1) Measure the h-period ahead uncertainty for each monthly variable y_it as the conditional volatility of the unforeseeable component:
    - U_it(h) = E[(y_it+h − E(y_it+h | I_t))^2 | I_t]
  - 2) Aggregate individual variable uncertainty measures across the database into an area-wide measure:
    - U_t(h) ≡ plim_{N_y→∞} (1/N_y) Σ_{i=1}^{N_y} w_i U_it(h) ≡ E_w[U_it(h)]
    - Baseline aggregation uses a simple average; robustness check uses the first principal component.

### Time-series specifications and volatility modeling
- Forecastable component removed via an approximate factor structure as in Bai and Ng (2006):
  - y_{it+1} = β_i(L) y_it + α_i(L) Z_t + v_{it+1}, with v_it = σ_it ε_it and ε_it ∼ N(0,1).
- Common factors Z_t assumed serially correlated and represented by univariate AR(1):
  - z_t = φ z_{t−1} + v_zt, with v_zt = σ_zt ε_zt and ε_zt ∼ N(0,1).
- Stochastic volatility for idiosyncratic shocks:
  - log(σ_{it+1})^2 = γ_i + φ_i log(σ_it)^2 + τ_i η_{it+1}, η_{it+1} ∼ N(0,1).
- Stochastic volatility for common shocks:
  - log(σ_{zt})^2 = γ_z + φ_z log(σ_{zt−1})^2 + τ_z η_{zt}, η_{zt} ∼ N(0,1).
- Conditional volatility for h-period-ahead errors obtained recursively from one-step-ahead prediction errors.

### BVAR application: setup and identification
- Use a Bayesian VAR (BVAR) to assess impacts of uncertainty on macro variables, with special interest in the COVID-19 period.
- Two Cholesky orderings used:
  - Baseline (Bloom, 2005) ordering: [log(Stock price), Uncertainty, Interest rate, log(Price), Unemployment rate, log(Industrial production)]
  - Alternative (Christiano, Eichenbaum, and Evans, 2005) ordering: [log(Industrial production), Unemployment rate, log(Price), Interest rate, log(Stock price), Uncertainty]
- Monthly data for these series used in BVAR exercises: 2003m6-2021m11 (with checks compared to 2019 and post-first wave up to end-2022).
- Use 3 months-period ahead uncertainty measures (other horizons checked).
- Apply 4 lags; constant included; distribution is Normal-Wishart.
- Robustness checks and variants:
  - Shadow rates (Krippner) up to 2021m9 instead of Euribor 1y.
  - Setup with core HICP or without unemployment.
  - BVARX including Kilian’s global economic activity index as exogenous variable.
  - Inclusion of debt-to-GDP ratio (or real debt) ordered last in Cholesky identification for fiscal implication exercises.
  - Analysis extended to Emerging European economies (EMEU) with block exogeneity (EMEU consists of 6 countries: Bulgaria, Czechia, Croatia, Hungary, Poland, and Romania).

### MEU dynamics and comparative statistics
- MEU shows spikes driven by the GFC and partially by the sovereign debt crisis, but those spikes reflect larger uncertainty in financial markets rather than macroeconomic variables.
- MEU exposes a large macroeconomic uncertainty spike due to the first wave of COVID-19 in 2020.
- MEU moved from around 0.95 to 1.33 between January 2020 and April 2020, declining only in October 2020 to pre-pandemic level.
- Historical average before 2020 was 0.86; between mid-2007 and the end of 2010 it was 0.93.
- Country heterogeneity examples:
  - Greece’s MEU spiked more than aggregate during sovereign debt crisis.
  - Cyprus showed idiosyncratic spike during 2012-13 banking crisis.
  - During COVID-19, Malta and the Baltic states initially less affected (smaller MEU). In the Baltic States MEU approximately same level as during the GFC.
- MEU in 2020 peaked in April and ranged from 0.94 to 1.64 across member states; euro area MEU was 1.33.
- Correlations: 3-period ahead MEU correlations with 1-period, 6-period, and 12-period ahead MEU are 0.97, 0.98, and 0.89.

### Comparison with other uncertainty and stress measures
- Financial-oriented indicators:
  - EAVSTOXX (stock market volatility)
  - CISS (Composite indicator of systemic stress)
  - CLIFS (Duprey and Klaus, 2015)
- Economic Policy Uncertainty (EPUEA) captures policy-related economic uncertainty from newspaper coverage.
- Observations:
  - EAVSTOXX had the biggest spike during the GFC.
  - CISS and CLIFS react more during the sovereign debt crisis.
  - EPUEA is more volatile and captures diverse events (e.g., Brexit).
  - All measures spike during the first wave of COVID-19, but magnitude is much larger for MEU, supporting interpretation of COVID-19 shock as largely macroeconomic.

### Impact of MEU shocks on the macroeconomy (BVAR results)
- Objective: quantify macro impacts of an unexpected increase in macroeconomic uncertainty (MEU) using BVAR.
- Key results for full sample to end-2021:
  - An increase in MEU by one standard deviation has a large and significant negative impact on Industrial Production (IP) in the short-run, up to 1 year after the shock, and relatively short-lived.
  - The shock subtracts around 0.82 percentage points from euro area IP at its maximum impact (after 3-4 months from the shock).
  - The cumulative response is -2.3 percentage points in the first quarter.
  - For the full sample to 2021, the magnitude of the MEU shock almost doubles compared to pre-COVID sample.
  - The 1-year Euribor rate does not react to the shock in the full sample, but decreases if pre-COVID data are used.
  - Stock prices, HICP, and unemployment IRFs are almost not significantly different from zero between 2019 and 2021; slight increase in unemployment in full sample.
- Robustness checks confirm main results:
  - Shadow rates instead of Euribor 1y (up to 2021m9).
  - Core HICP instead of headline HICP.
  - BVARX including Kilian index: maximum impact of MEU shock on IP is 0.78 with the inclusion of the Kilian index, compared to 0.82 baseline without it.
- Confidence intervals for IRFs reported at 95 percent.

### Alternative specifications and uncertainty measures
- IRFs to MEU shocks are robust to alternative identification (Christiano, Eichenbaum, and Evans, 2005).
- IP reacts less to shocks in financially-driven measures (CISS, EAVSTOXX) and to EPU than to MEU.
- CLIFS produces IRFs more similar to MEU only when 2020 and 2021 are included.

### Contribution of uncertainty shocks to industrial production dynamics
- Historical decomposition (baseline identification):
  - During the GFC MEU shocks contributed to IP decline but were relatively small compared to other factors.
  - Pre-COVID lower MEU had a positive effect on IP.
  - During 2020, MEU was a major contributor to the dip in IP.
- COVID-19 episode (2020m3 to 2020m8):
  - MEU shocks explain a major part of the decline in IP.
  - Alternative uncertainty measures do not make substantial contributions in this episode.
  - During 2021 recovery, both MEU and EAVSTOXX shocks continued to impact IP negatively, while CLIFS indicated favorable impacts supporting IP growth.
  - Policy measures in late 2020 and 2021 lowered uncertainty overall, and the “other” component began contributing positively to IP changes.
- Historical-decomposition magnitudes:
  - MEU contribution to first wave peak in the historical decomposition of IP peaked at -15 during COVID-19 (comparison in Online Appendix).
  - Post-2021 historical decomposition shows MEU peaked at -2.7 after 2021.

### Impact on public debt
- Key findings on debt dynamics:
  - A shock to MEU increases debt over GDP with the peak reaching after approximately 1 year, and this response is very persistent over time.
  - The MEU seems to be a major component to debt dynamics both in good times (when debt over GDP decreased) and bad times. This is especially relevant after 2016.
  - Robustness check: using total debt produces results that are in line with the baseline finding.
- Context during COVID-19:
  - Discretionary measures together with automatic stabilizers were put in place during COVID-19 times; public deficit and debt soared as governments massively supported their economies.
  - Specific figures: debt reached 98 percent for the euro area at the end of 2020 from 84 percent in 2019.
- Implications:
  - In response to MEU shocks, public debt goes up and only fades away gradually, implying persistent fiscal implications following macroeconomic uncertainty shocks.
  - The MEU index can be instrumental for policy assessment and added to a rich toolkit of composite indicators for the euro area.

### Related results and post-COVID-19 observations
- Impact on Emerging European economies (EMEU): block-exogeneity BVAR finds similar impacts for EMEU, with a slightly larger response of industrial production; HICP responses almost mimic those for the euro area.
- COVID-19 in VARs: explicitly modeling increased uncertainty (via MEU) explains much of the IP variation during COVID-19; omitting an uncertainty measure places the contribution on IP shocks and partially on stock prices.
- Post-COVID-19 macro uncertainty (sample extended to 2022m12):
  - The estimate of MEU in pre-COVID-19 and during COVID-19 remains similar, indicating robustness and stability.
  - After falling quickly from the 2020m4 peak, the MEU bounces back at the end of 2021 and lingers at a level much higher than the pre-pandemic historical average.
  - The increase of MEU in 2021 appears to be mainly driven by increasing inflation volatility in the euro area (producer price indices and HICP components).
  - Euro area’s HICP inflation reached above 4 percent in 2021m10, continuing rising to more than 10 percent in 2022m10 before slowly decreasing.
  - The constructed MEU for the euro area and Jurado et al. (2015) macroeconomic uncertainty index for the USA coincide from the COVID-19 outbreak onwards and show similar post-COVID-19 dynamics.

### Policy considerations and suggested research extensions
- MEU is a data-rich measure of macroeconomic uncertainty for the euro area that is more persistent and spikes mainly in important events, especially during the first phase of COVID-19.
- Potential future research directions:
  - Study heterogeneous impacts of uncertainty shocks on euro area member countries.
  - Investigate non-linear/asymmetric effects of uncertainty shocks.
  - Disentangle and evaluate the impact of euro area regional, country-specific, and global uncertainty, and between euro area real versus nominal uncertainty.

*Source: wpiea2023229-print-pdf - 3.1  A large monthly database on euro area; 5.3  Impact on public debt*

### 3.1  A large monthly database on euro area

### 3.1  A large monthly database on euro area

### Database construction and coverage
- We first construct a large database with almost1,500monthly macroeconomic and financial indicators, encompassing 19 euro area countries.
- Sample starts from January 2003 with data coming from: Eurostat, OECD, ECB SDW, and BIS.
- Dataset includes both country-specific data and euro area-wide data.
- Country-specific indicators are categorized into four groups:
  - Real economic activity: industrial production and its components, car registration (including different types of vehicles), total turnover index (in real term) (with different type of retail sales), unemployment rates, hours, and trade with world (i.e., import and export for different types of goods).
  - Price movements: producer price index and HICP (and their components), nominal effective exchange rate, long term rate, and wages.
  - Survey data and leading indicators: economic sentiment indicator, confidence indicators components, and leading indicators.
  - Financial data: loans, deposits, debt securities (at different maturities), share prices, and spreads.
- Euro area-wide variables included: euro area bond yields and money market rates at different maturity, Dow-Jones euro area Stoxx index (and its components), monetary aggregates M1, M3, and currency in circulation, and several bilateral exchange rates with Euro.
- The computed macroeconomic uncertainty index for the euro area considers inner heterogeneity of member countries and common components.

### Country-specific MEU construction note
- Country-specific MEU indices are calculated as average of only the specific variables’ uncertainties for that country and euro area common exchange rates (coverage varies across countries).
- Joint estimation across all variables allows identification of common factors; country-specific measures capture both country-specific and common uncertainty.
- Country-specific MEUs are relatively close on average, between 0.81 and 0.93, though differences increase during times of stress.
- Volatility within countries spans from 0.64 to 1.70.

---

### 3.2  Measuring euro area macroeconomic uncertainty

### Methodological approach
- Approach follows Jurado et al. (2015) to construct an objective, data-driven measure of euro area macroeconomic uncertainty, accounting for heterogeneity.
- Construction includes two steps:
  1. Measure the h-period ahead uncertainty for each monthly variable y_it as the conditional volatility of the unforeseeable component:
     - U_it(h) = E[(y_it+h − E(y_it+h | I_t))^2 | I_t]
  2. Aggregate individual variable uncertainty measures across the database into an area-wide measure:
     - U_t(h) ≡ plim_{N_y→∞} (1/N_y) Σ_{i=1}^{N_y} w_i U_it(h) ≡ E_w[U_it(h)]
     - Baseline aggregation uses a simple average; robustness check uses the first principal component.

### Time-series specifications and volatility modeling
- Forecastable component removed via an approximate factor structure as in Bai and Ng (2006): y_{it+1} = β_i(L) y_it + α_i(L) Z_t + v_{it+1}, with v_it = σ_it ε_it and ε_it ∼ N(0,1).
- Common factors Z_t assumed serially correlated and represented by univariate AR(1): z_t = φ z_{t−1} + v_zt, with v_zt = σ_zt ε_zt and ε_zt ∼ N(0,1).
- Stochastic volatility for idiosyncratic shocks:
  - log(σ_{it+1})^2 = γ_i + φ_i log(σ_it)^2 + τ_i η_{it+1}, η_{it+1} ∼ N(0,1).
- Stochastic volatility for common shocks:
  - log(σ_{zt})^2 = γ_z + φ_z log(σ_{zt−1})^2 + τ_z η_{zt}, η_{zt} ∼ N(0,1).
- Conditional volatility for h-period-ahead errors obtained recursively from one-step-ahead prediction errors.

---

### 3.3  BVAR application

### Model setup and identification
- Use a Bayesian VAR (BVAR) to assess impacts of uncertainty on macro variables, with special interest in the COVID-19 period.
- Two Cholesky orderings used for comparison:
  - Baseline (Bloom, 2005) ordering:
    - [log(Stock price), Uncertainty, Interest rate, log(Price), Unemployment rate, log(Industrial production)]
  - Alternative (Christiano, Eichenbaum, and Evans, 2005) ordering:
    - [log(Industrial production), Unemployment rate, log(Price), Interest rate, log(Stock price), Uncertainty]
- Monthly data for these series used in BVAR exercises: 2003m6-2021m11 (with checks compared to 2019 and post-first wave up to end-2022).
- Use 3 months-period ahead uncertainty measures (other horizons checked).
- Apply 4 lags; constant included; distribution is Normal-Wishart.
- Robustness checks:
  - Shadow rates (Krippner) up to 2021m9 instead of Euribor 1y.
  - Setup with core HICP or without unemployment.
  - BVARX including Kilian’s global economic activity index as exogenous variable.
  - Inclusion of debt-to-GDP ratio (or real debt) ordered last in Cholesky identification for fiscal implication exercises.
  - Analysis extended to Emerging European economies (EMEU) with block exogeneity (EMEU consists of 6 countries: Bulgaria, Czechia, Croatia, Hungary, Poland, and Romania).

---

### 4  MacroEconomic Uncertainty for the euro area (MEU)

### MEU dynamics and comparative statistics
- MEU shows spikes driven by the GFC and partially by the sovereign debt crisis, but those spikes reflect larger uncertainty in financial markets rather than macroeconomic variables.
- MEU exposes a large macroeconomic uncertainty spike due to the first wave of COVID-19 in 2020.
- MEU moved from around 0.95 to 1.33 between January 2020 and April 2020, declining only in October 2020 to pre-pandemic level.
- Historical average before 2020 was 0.86; between mid-2007 and the end of 2010 it was 0.93.
- Country heterogeneity: Greece’s MEU spiked more than aggregate during sovereign debt crisis; Cyprus showed idiosyncratic spike during 2012-13 banking crisis.
- During COVID-19, heterogeneity visible: Malta and the Baltic states initially less affected (smaller MEU). In the Baltic States MEU approximately same level as during the GFC.
- MEU in 2020 peaked in April and ranged from 0.94 to 1.64 across member states; euro area MEU was 1.33.
- Correlations: 3-period ahead MEU correlations with 1-period, 6-period, and 12-period ahead MEU are 0.97, 0.98, and 0.89.

### Comparison with other uncertainty and stress measures
- Financial-oriented indicators: EAVSTOXX (stock market volatility), CISS (Composite indicator of systemic stress), CLIFS (Duprey and Klaus, 2015) capture financial market stress and co-movement across market segments.
- Economic Policy Uncertainty (EPUEA) captures policy-related economic uncertainty from newspaper coverage.
- Observations:
  - EAVSTOXX had the biggest spike during the GFC.
  - CISS and CLIFS react more during the sovereign debt crisis.
  - EPUEA is more volatile and captures diverse events (e.g., Brexit).
  - All measures spike during the first wave of COVID-19, but magnitude is much larger for MEU, supporting interpretation of COVID-19 shock as largely macroeconomic.

---

### 5  Impact of MEU in BVAR applications

### 5.1  Impact of MEU shocks on the macroeconomy
- Objective: quantify macro impacts of an unexpected increase in macroeconomic uncertainty (MEU) using BVAR.
- Key results for full sample to end-2021:
  - An increase in MEU by one standard deviation has a large and significant negative impact on Industrial Production (IP) in the short-run, up to 1 year after the shock, and relatively short-lived.
  - The shock subtracts around 0.82 percentage points from euro area IP at its maximum impact (after 3-4 months from the shock).
  - The cumulative response is -2.3 percentage points in the first quarter.
  - For the full sample to 2021, the magnitude of the MEU shock almost doubles compared to pre-COVID sample.
  - The 1-year Euribor rate does not react to the shock in the full sample, but decreases if pre-COVID data are used.
  - Stock prices, HICP, and unemployment IRFs are almost not significantly different from zero between 2019 and 2021; slight increase in unemployment in full sample.
- Robustness checks confirm main results:
  - Shadow rates instead of Euribor 1y (up to 2021m9).
  - Core HICP instead of headline HICP.
  - BVARX including Kilian index: maximum impact of MEU shock on IP is 0.78 with the inclusion of the Kilian index, compared to 0.82 baseline without it.
- Confidence intervals for IRFs reported at 95 percent.

### 5.1.1  Alternative specifications and uncertainty measures
- IRFs to MEU shocks are robust to alternative identification (Christiano, Eichenbaum, and Evans, 2005).
- IP reacts less to shocks in financially-driven measures (CISS, EAVSTOXX) and to EPU than to MEU.
- CLIFS produces IRFs more similar to MEU only when 2020 and 2021 are included.

### 5.2  Contribution of uncertainty shocks on recent industrial production dynamics
- Historical decomposition using baseline identification:
  - During the GFC MEU shocks contributed to IP decline but were relatively small compared to other factors.
  - Pre-COVID lower MEU had a positive effect on IP.
  - During 2020, MEU was a major contributor to the dip in IP.
- COVID-19 episode (2020m3 to 2020m8):
  - MEU shocks explain a major part of the decline in IP.
  - Alternative uncertainty measures do not make substantial contributions in this episode.
  - During 2021 recovery, both MEU and EAVSTOXX shocks continued to impact IP negatively, while CLIFS indicated favorable impacts supporting IP growth.
  - Policy measures in late 2020 and 2021 lowered uncertainty overall, and the “other” component began contributing positively to IP changes.

*Source: wpiea2023229-print-pdf - 3.1  A large monthly database on euro area*

### 5.3  Impact on public debt

### 5.3  Impact on public debt

### Key findings on debt dynamics
- A shock to MEU increases debt over GDP with the peak reaching after approximately 1 year, and this response is very persistent over time.
- The MEU seems to be a major component to debt dynamics both in good times (when debt over GDP decreased) and bad times. This is especially relevant after 2016.
- Robustness check: using total debt produces results that are in line with the baseline finding.

### Context during COVID-19
- Discretionary measures together with automatic stabilizers were put in place during COVID-19 times; public deficit and debt soared as governments massively supported their economies.
- Specific figures: debt reached 98 percent for the euro area at the end of 2020 from 84 percent in 2019.

### Historical decomposition and attribution
- Historical decompositions (sample references including 2020m3-2020m8) indicate that MEU contributed substantially to fluctuations in industrial production during the first wave of COVID-19.
- During COVID-19 the MEU contribution to the first wave peak in the historical decomposition of industrial production was large (comparison in Online Appendix: peaked at -15 during COVID-19 for IP).
- Post-2021 historical decomposition shows MEU contributes negatively to changes in industrial production, with a smaller magnitude than during the first wave of COVID-19 (peaked at -2.7 after 2021).

### Implications
- In response to MEU shocks, public debt goes up and only fades away gradually, implying persistent fiscal implications following macroeconomic uncertainty shocks.
- The MEU index can be instrumental for policy assessment and added to a rich toolkit of composite indicators for the euro area.

---

### Related results in adjacent sections (summarized)
- Impact on Emerging European economies (EMEU): using a BVAR with block exogeneity, similar impacts are found for EMEU, with a slightly larger response of industrial production; HICP responses almost mimic those for the euro area.
- COVID-19 in VARs: explicitly modeling increased uncertainty (via MEU) explains much of the IP variation during COVID-19; omitting an uncertainty measure places the contribution on IP shocks and partially on stock prices.
- Post-COVID-19 macro uncertainty (sample extended to 2022m12):
  - The estimate of MEU in pre-COVID-19 and during COVID-19 remains similar, indicating robustness and stability.
  - After falling quickly from the 2020m4 peak, the MEU bounces back at the end of 2021 and lingers at a level much higher than the pre-pandemic historical average.
  - The increase of MEU in 2021 appears to be mainly driven by increasing inflation volatility in the euro area (producer price indices and HICP components).
  - Euro area’s HICP inflation reached above 4 percent in 2021m10, continuing rising to more than 10 percent in 2022m10 before slowly decreasing.
  - The constructed MEU for the euro area and Jurado et al. (2015) macroeconomic uncertainty index for the USA coincide from the COVID-19 outbreak onwards and show similar post-COVID-19 dynamics.

### Policy considerations and suggested research extensions
- MEU is a data-rich measure of macroeconomic uncertainty for the euro area that is more persistent and spikes mainly in important events, especially during the first phase of COVID-19.
- Potential future research directions:
  - Study heterogeneous impacts of uncertainty shocks on euro area member countries.
  - Investigate non-linear/asymmetric effects of uncertainty shocks.
  - Disentangle and evaluate the impact of euro area regional, country-specific, and global uncertainty, and between euro area real versus nominal uncertainty.

*Source: wpiea2023229-print-pdf - 5.3  Impact on public debt*

### References

### wpiea2023229-print-pdf - References

### References
- The References section contains 35 numbered citations ([1] through [35]) covering literature on economic uncertainty, measurement methods, uncertainty shocks, COVID-19 effects, sectoral uncertainty, indicators of systemic stress, and related empirical and methodological work.
- Notable recurring authors and topics include: Altig et al. (2020) on economic uncertainty and COVID-19; Baker, Bloom, Davis on economic policy uncertainty and COVID-induced uncertainty; Bloom (2009) on uncertainty shocks; Jurado, Ludvigson and Ng (2015) on measuring uncertainty; Kilian on oil-price and global real activity measures; and recent work on firm-level inflation uncertainty and sectoral uncertainty.
- The References list also includes working papers, ECB and central bank working series, CEPII policy brief, OECD and IMF working papers, and journal articles from Journal of Public Economics, Econometrica, American Economic Review, European Economic Review, Oxford Bulletin of Economics and Statistics, Journal of Monetary Economics, Journal of International Money and Finance, Journal of Econometrics, and Journal of Economic Literature.

### Appendix A: Dataset — coverage and composition
- Dataset scope:
  - "almost 1,500 monthly macroeconomic and financial indicators" encompassing 19 euro area countries.
  - The maximum list of country-specific variables includes 122 variables (presented in Table 1).
  - The dataset includes 30 financial variables at the euro area level (described in Table 2).
- Variable groupings (7 groups):
  - Industrial productions components
  - Labor market indicators: Index of employment, Unemployment rates, hours, and wages
  - Prices: Producer Price Index, HICP, and Import price index
  - Car registration, Total Turnover Index, and Building Permits
  - Trade with world: Import and Export for different types of goods
  - Sentiment, confidence, business tendency, and leading indicators
  - Financial data: Nominal effective exchange rate, long-term rate, loans, deposits, debt securities (at different maturities), share prices, spreads
- Euro area-level financial variables (summary):
  - euro area bond yields and money market rates at different maturity
  - Dow-Jones euro area Stoxx components (various sector indices)
  - Monetary aggregates M1, M3, and currency in circulation
  - Several bilateral exchange rates with Euro (UK pound sterling/Euro; US dollar/Euro; Japanese yen/Euro; Chinese yuan renminbi/Euro; Swiss franc/Euro)
- Exact dataset counts and composition:
  - "In total, the dataset has 1470 variables, with 1445 country-specific variables and 25 euro area-level variables."
  - After further cleaning (dropping one of any two variables with absolute correlation larger than 0.95 for each country), "the final dataset has a total of1330variables."
- Country coverage remarks (availability differences):
  - Most countries have about or more than 60 variables from the 122-variable maximum list.
  - Exceptions with lower availability: Cyprus (CY, 29/122 variables), Greece (GR, 30/122 variables), and Malta (MT, 36/122 variables).

### Appendix A: Examples of country-specific variable entries (from Table 1)
- Industrial production examples:
  - Industrial Production Index, Total Industry - NACE Rev2; Eurostat; Working day and seasonally adjusted
  - Industrial Production Index, CONSTRUCTION - NACE Rev2; Eurostat; Working day and seasonally adjusted
  - Industrial Production Index, MANUFACTURING - NACE Rev2; Eurostat; Working day and seasonally adjusted
- Labor market indicators examples:
  - Index of Employment, MANUFACTURING - NACE Rev2; Eurostat; Working day and seasonally adjusted
  - Hours worked, MANUFACTURING - NACE Rev2; Eurostat; Working day and seasonally adjusted
  - Standardised unemployment, Rate, Total (all ages); Eurostat; Seasonally adjusted; percentage of civilian workforce
- Prices examples:
  - Producer Price Index, total, MANUFACTURING - NACE Rev2; Eurostat; Neither seasonally nor working day adjusted
  - HICP - Overall index, Monthly Index; Eurostat; Neither seasonally nor working day adjusted
  - Import Price Index, MANUFACTURED PRODUCTS; Eurostat; Neither seasonally nor working day adjusted
- Car registration / turnover / permits examples:
  - Car registration, New passenger car; Absolute value; European Central Bank; Working day and seasonally adjusted
  - Building Permits/ Sq. meters of useful floor area, All buildings; Eurostat; Seasonally adjusted, not working day adjusted
  - Total Turnover Index, deflated, Retail trade including fuel, except of motor vehicles and motorcycles - NACE Rev2; Eurostat; Working day and seasonally adjusted
- Sentiment and leading indicators examples:
  - Economic sentiment indicator – Seasonally adjusted data; Eurostat
  - Business tendency surveys (manufacturing) – Confidence indicators – Monthly; OECD
  - Leading Indicators OECD – Confidence indicator – Monthly; OECD
- Financial indicators examples:
  - Long term rate; OECD
  - Nominal Effective exchange rate indices; BIS
  - MFIs excluding ESCB – Loans – Total – Outstanding amounts – Euro; ECB
  - Share Prices – All shares/broad – Index 2015=100 – Monthly; OECD
  - Lending spreads and new business lending/deposit rates with detailed definitions and maturities; ECB

### Appendix A: Euro area level variables (selected entries from Table 2)
- Government benchmark bond yields at maturities: 2-year, 3-year, 5-year, 7-year, 10-year; ECB (monthly, Euro area changing composition)
- Real Euro area 10-year Government Benchmark bond yield – Yield, average of observations through period; ECB
- Money market rates: Euribor 1-month, Euribor 3-month, Euribor 6-month, Euribor 1-year; Reuters/ECB (historical close, average of observations through period)
- Real 3-month Euribor, Eonia rate; ECB
- Equity/index series: Dow Jones Euro Stoxx 50 Price Index and numerous sectoral Euro Stoxx indices; DataStream/ECB
- Monetary aggregates: M1, M3, Currency in circulation; ECB (Working day and seasonally adjusted)

### Appendix B–C and Figures: robustness checks, IRFs, and historical decompositions
- The Appendices contain multiple robustness checks and model variations including:
  - Identification with shadow rates (comparison baseline vs. "With Shadow Rates"); estimation periods: Baseline 2003m6-2021m11 and with shadow rates 2003m6-2021m9.
  - Replacement of HICP with core HICP in identification ("With core HICP"); time period 2003m6-2021m11.
  - BVARX specification including Kilian index (Index of Global Real Economic Activity [IGREA]) as an exogenous variable; time period 2003m6-2021m11.
  - Inclusion of Debt/GDP as an additional variable and analysis of Debt/GDP responses and historical decomposition; Baseline without Debt/GDP vs. "With Debt/GDP"; time period 2003m6-2021m11.
  - Block-exogeneity setup for EMEU variables ("With EMEU"), where EMEU variables cannot influence euro area variables; time period 2003m6-2021m11.
  - Post-COVID-19 comparisons with IRFs for 2021 vs 2022 estimations.
- Figures and diagnostics provided (selected figure identifiers):
  - Figure 12 – Number of available country-specific variables over the sample period
  - Figures 13–26 – comparisons of Euro area versus country-specific indices, IRFs baseline and variants, point estimates for Industrial Production (IP) and HICP responses, historical decompositions (including versions with Kilian index and Debt/GDP), Impact to EMEU, post-COVID-19 decompositions, and MEU: Euro Area versus US series.
- Example notes preserved from figures:
  - "Baseline refers to the identification as in Bloom (2005) with 1-year Euribor."
  - For shadow rates: "The identification for the results 'With Shadow Rates' is the same but with this new variable replacing 1-year Euribor."
  - For Kilian index: "The Kilian index is included as an exogenous variable (BVARX). The index is from Federal Reserve Bank of Dallas, Index of Global Real Economic Activity [IGREA], retrieved from FRED, Federal Reserve Bank of St. Louis; https://fred.stlouisfed.org/series/IGREA."
  - Time periods for estimations specified where provided (e.g., 2003m6-2021m11; 2003m6-2021m9).

*Source: wpiea2023229-print-pdf - References*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023229-print-pdf.pdf_