## wpiea2025007-print-pdf

## Source details

**Canonical URL:** [wpiea2025007-print-pdf](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025007-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025007-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025007-print-pdf.pdf.json)

---

### II. DATA OVERVIEW
- Sample: balanced panel of high-frequency observations covering 24 countries in Europe during the period 2014–2024.
- Price data:
  - Wholesale electricity prices obtained from ENTSO-E in €/MWh at 15-minute frequency for 24 European countries.
  - Volatility variable: natural logarithm of price returns between each 15-minute period; standard deviation of log-returns computed for each country.
- Key descriptive statistics for wholesale electricity price volatility (sample period 2015–2024):
  - Average volatility of wholesale electricity prices: 0.262.
  - Minimum mean value: 0.128 in Greece.
  - Maximum mean value: 0.414 in Finland.
  - Highest standard deviation in the dataset: 0.308 (Finland).
  - Lowest standard deviation in the dataset: 0.043 (Italy).
  - Distributional features: positive skewness across countries; sample kurtosis markedly higher than 3 for some countries (fat tails).
  - Stationarity: unit root tests reject the null at the 1 percent level for all countries (ADF significance at 1 percent marked with *** in Table 1).
- Gravity and control variables:
  - Geographic distance and contiguity from CEPII (great-circle distance in kilometers between capital cities; contiguity = 1 if adjacent border, 0 otherwise).
  - Real GDP per capita and population in origin and destination countries.
  - 15-minute frequency share of renewables in electricity generation from ENTSO-E.
- Descriptive dispersion (selected values from Table 2):
  - Bilateral Electricity Price Volatility Spillovers: Number of observations 876,852; Mean 0.004; Standard deviation 0.189; Minimum -3.277; Maximum 2.855.
  - Distance (in kilometers): Number of observations 803,781; Mean 1,291; Minimum 670; Maximum 553,290.
  - Contiguity: Number of observations 803,781; Mean 0.1; Standard deviation 0.3; Minimum 0.0; Maximum 1.0.
  - Real GDP (in billions) Origin: Number of observations 876,852; Mean 3,253; Standard deviation 9,435; Minimum 254; Maximum 3,604.
  - Share of Renewables (in percent) Origin: Number of observations 875,870; Mean 37.5; Standard deviation 24.7; Minimum 0.0; Maximum 100.0.
  - Share of Renewables (in percent) Destination: Number of observations 872,544; Mean 47.8; Standard deviation 29.7; Minimum 0.0; Maximum 100.0.

### III. EMPIRICAL METHODOLOGY
- Objective: measure volatility connectedness of electricity markets across Europe and decompose contributions of domestic vs. cross-country shocks.
- Primary approach:
  - Spillover index method introduced by Diebold and Yilmaz (2009; 2012; 2014).
  - Use of generalized VAR framework (Koop, Pesaran, and Potter, 1996; Pesaran and Shin, 1998) to obtain variance decompositions invariant to variable ordering.
- Definitions and measures:
  - Total volatility spillover index S(H): proportion of forecast error variance attributable to cross-variance shares.
  - Directional spillovers received by market i from all other markets S_i←j(H) and transmitted by market i to all other markets S_i→j(H) (normalized generalized variance decomposition).
  - Net volatility spillover S_i^net(H) = S_i→j(H) − S_i←j(H).
  - Net pairwise volatility spillover S_ij^net(H) = θ̃_ij^g(H) − θ̃_ji^g(H).
- Estimation details:
  - Full-sample estimation reported in Table 3 (each row sums to 100; off-diagonal entries are cross-variance shares).
  - Rolling-window estimation: 90-day rolling window used to capture time evolution of spillovers.

### IV. EMPIRICAL RESULTS — SPILLOVERS AND INTERACTIONS
- Total spillover index (full-sample result):
  - Total spillover index: 72.8 percent.
  - Interpretation: about 73 percent of the forecast error variation is explained by cross-variance shares; 27 percent attributed to within-country shocks.
- Country heterogeneity in spillovers (selected entries and summaries drawn from Table 3):
  - Lowest contribution from others (Contribution From Others column): Ireland 31.5 percent.
  - Highest contribution from others: Hungary 82.7 percent.
  - Strongest bilateral interactions:
    - Spain → Portugal: Spain explains 32.4 percent of Portugal’s forecast error variance.
    - Portugal → Spain: Portugal accounts for 31.4 percent of Spain’s forecast error variance.
    - Lithuania → Latvia: Lithuania explains 22.1 percent of Latvia’s forecast error variance.
    - Latvia → Lithuania: Latvia accounts for 18.6 percent of Lithuania’s forecast error variance.
  - Very low bilateral spillovers between geographically distant/unconnected markets:
    - Austria ↔ Ireland: Austria accounts for 0.5 percent of Ireland’s forecast error variance; Ireland accounts for 0.3 percent of Austria’s forecast error variance.
- Net contribution (Contribution to others minus Contribution from others; Net Contribution column highlights direction):
  - Largest net transmitters (selected):
    - Latvia: 63.1 percent.
    - Lithuania: 48.0 percent.
    - Estonia: 34.3 percent.
  - Largest net recipients (selected):
    - Greece: -51.8 percent.
    - Austria: -41.8 percent.
  - Additional selected totals:
    - Contribution to others (sum of off-diagonal components) for all countries sums to 1,748.4 (as reported).
    - Contribution including own (row sums) for entire system: country-specific values reported in Table 3 (e.g., AT 58.2; BE 89.6; CH 111.6; ...; Total Spillover Index 72.8 percent).
- Interpretation of net patterns:
  - High net spillover transmission from Latvia, Lithuania, and Estonia is linked to historical integration with the Russia-controlled power grid before Russia’s invasion of Ukraine; these countries cut down energy imports from Russia following the war and are set to desynchronize completely from the Soviet-era electricity network in 2025.
  - Net volatility spillovers reflect factors such as electricity production mix, electricity trade (exports and imports), price formation, and infrastructure quality.

### Dynamics of the Total Spillover Index and High-Frequency Patterns
- Highlighted time period: January 1, 2022 to April 30, 2024.
- Key dynamics and peaks:
  - Total spillover index increases to a peak of 96.3 percent.
  - Sample average rises to 83.4 percent, from 72.8 percent for the entire sample period.
  - Reported dynamic trend in rolling-window estimation: clear upward trend in dynamic volatility spillovers, increasing from an average of 89.9 in 2015 to 91.7 during the period 2019–2021.
  - Additional reported figure: 91.9 after 2022 with Russia’s invasion of Ukraine.
- Identified drivers and episodes of elevated volatility and spillovers:
  - Oil price plunge of 2014-2016.
  - COVID-19 pandemic in 2020.
  - Russia’s invasion of Ukraine after 2022.
  - The unabating rise in renewables as a weighty factor contributing to volatility (reference to Cevik and Ninomiya (2023)).

### Dataset and Modeling FRAMEWORK (Gravity Model and Estimation)
- Dataset:
  - 15-minute frequency dataset with more than 876,582 observations on 576 pairs of countries over the period 2014–2024.
- Gravity model specification (augmented):
  - Dependent variable: bilateral wholesale electricity price volatility spillovers s_ijt.
  - Main regressors: gravity_ijt (income, population, geographic distance, geographical contiguity), renewable_it (share of renewables in origin), renewable_jt (share of renewables in destination).
  - Fixed effects: η_ij (country-pair), φ_it (origin time), μ_jt (destination time).
  - Standard errors clustered at the country-pair level.
- Estimation approach:
  - Poisson Pseudo Maximum Likelihood (PPML) regression to control for heteroskedasticity and correlated errors across countries and over time.

### Key Empirical Results from the Gravity Model (summary of Table 4)
- Sample sizes and model fit:
  - Number of observations: [1] 803,781; [2] 798,724; [3] 871,565.
  - Number of countries: 24.
  - R2: 0.117, 0.119, 0.135 for columns [1], [2], [3] respectively.
  - Country-pair FE: Yes in column [3]; No in [1] and [2].
  - Country-time FE: Yes in all specifications.
- Estimated coefficient directions and statistical significance (reported significance levels retained):
  - Distance: positive and significant (0.000***).
  - Contiguity: negative and significant (-0.098***; -0.103***).
  - Real GDP (Origin): negative and significant (0.000*** or -0.000*** reported).
  - Real GDP (Destination): positive but weak/significance varies (0.000*; 0.000**).
  - Population (Origin and Destination): positive and significant (0.001*** for both).
  - Share of Renewables (Origin): positive and significant (0.066***; 0.095***).
  - Share of Renewables (Destination): positive and significant (0.140***; 0.058***).
- Interpretation of key coefficients:
  - Greater geographic distance is associated with higher bilateral volatility spillovers in the estimations, while direct geographical contiguity exerts a dampening effect.
  - Larger population in origin and destination countries is associated with greater spillovers (capturing electricity demand effects).
  - Higher shares of intermittent renewables (solar, wind) in both origin and destination are associated with greater bilateral volatility spillovers, consistent with Cevik and Ninomiya (2023).

### Aggregate and Country-Level Findings
- Cross-border dominance:
  - About 73 percent of the forecast error variation is explained by cross-variance shares; only 27 percent is attributable to within-country shocks.
  - Cross-border volatility spillovers dominate national electricity market behavior in Europe and have grown over time, especially after Russia’s invasion of Ukraine.
- Country heterogeneity and notable bilateral interactions:
  - Ireland: lowest spillover from others.
  - Hungary: highest level of spillover effects.
  - Portugal–Spain bilateral interaction: Spain explains 32.4 percent of Portugal’s forecast error variance; Portugal accounts for 31.4 percent of Spain’s forecast error variance.
  - Latvia–Lithuania bilateral interaction: Lithuania explains 22.1 percent of Latvia’s forecast error variance; Latvia accounts for 18.6 percent of Lithuania’s forecast error variance.
  - Austria–Ireland: very low bilateral spillovers (Austria accounts for 0.5 percent of Ireland’s forecast error variance; Ireland accounts for 0.3 percent of Austria’s).
- Appendix Table A1 headline statistics:
  - Contribution to others (sum across countries): 2001.5
  - Total Spillover Index (In percent): 83.4

### V. KEY TAKEAWAYS
- Cross-border volatility spillovers dominate national electricity market volatility in Europe: Total spillover index 72.8 percent (cross-variance shares explain ~73 percent of forecast error variation).
- Marked cross-country heterogeneity:
  - Average wholesale electricity price volatility across countries: 0.262; country means range from 0.128 (Greece) to 0.414 (Finland).
  - Some countries are strong net transmitters of volatility (e.g., Latvia 63.1 percent; Lithuania 48.0 percent; Estonia 34.3 percent), while others are strong net recipients (e.g., Greece -51.8 percent; Austria -41.8 percent).
- Geographical proximity and physical interconnection increase spillover intensity (e.g., Portugal–Spain; Latvia–Lithuania), while distant/unconnected markets exhibit much lower bilateral spillovers (e.g., Austria–Ireland).
- Volatility spillovers have increased over time based on 90-day rolling-window estimates (average spillover rising from 89.9 in 2015 to 91.7 in 2019–2021).

### Policy Implications and Recommendations
- Infrastructure and market design:
  - Infrastructure modernization and regulatory reforms can help minimize volatility in wholesale electricity prices during the transition to renewables.
  - Moving from the current zonal system with cost-based redispatch to a nodal pricing system would improve efficient distribution of electricity and dampen excessive price fluctuations, given growing shares of renewables with greater intermittency and close to zero marginal costs of generation.
- Energy security and integration:
  - These reforms would support the transition to low-carbon power generation and aid integration of electricity markets and strengthening of energy security in Europe.

*Source: wpiea2025007-print-pdf (content provided).*

### 44.7 percent in 2023, and it is projected to reach over 70 percent by 2030 (Busch et al., 2023).

### wpiea2025007-print-pdf - 44.7 percent in 2023, and it is projected to reach over 70 percent by 2030 (Busch et al., 2023).

### II. DATA OVERVIEW
- Sample: balanced panel of high-frequency observations covering 24 countries in Europe during the period 2014–2024.
- Price data:
  - Wholesale electricity prices obtained from ENTSO-E in €/MWh at 15-minute frequency for 24 European countries.
  - Volatility variable: natural logarithm of price returns between each 15-minute period; standard deviation of log-returns computed for each country.
- Key descriptive statistics for wholesale electricity price volatility (sample period 2015–2024):
  - Average volatility of wholesale electricity prices: 0.262.
  - Minimum mean value: 0.128 in Greece.
  - Maximum mean value: 0.414 in Finland.
  - Highest standard deviation in the dataset: 0.308 (Finland).
  - Lowest standard deviation in the dataset: 0.043 (Italy).
  - Distributional features: positive skewness across countries; sample kurtosis markedly higher than 3 for some countries (fat tails).
  - Stationarity: unit root tests reject the null at the 1 percent level for all countries (ADF significance at 1 percent marked with *** in Table 1).
- Gravity and control variables:
  - Geographic distance and contiguity from CEPII (great-circle distance in kilometers between capital cities; contiguity = 1 if adjacent border, 0 otherwise).
  - Real GDP per capita and population in origin and destination countries.
  - 15-minute frequency share of renewables in electricity generation from ENTSO-E.
- Descriptive dispersion (selected values from Table 2):
  - Bilateral Electricity Price Volatility Spillovers: Number of observations 876,852; Mean 0.004; Standard deviation 0.189; Minimum -3.277; Maximum 2.855.
  - Distance (in kilometers): Number of observations 803,781; Mean 1,291; Minimum 670; Maximum 553,290.
  - Contiguity: Number of observations 803,781; Mean 0.1; Standard deviation 0.3; Minimum 0.0; Maximum 1.0.
  - Real GDP (in billions) Origin: Number of observations 876,852; Mean 3,253; Standard deviation 9,435; Minimum 254; Maximum 3,604.
  - Share of Renewables (in percent) Origin: Number of observations 875,870; Mean 37.5; Standard deviation 24.7; Minimum 0.0; Maximum 100.0.
  - Share of Renewables (in percent) Destination: Number of observations 872,544; Mean 47.8; Standard deviation 29.7; Minimum 0.0; Maximum 100.0.

### III. EMPIRICAL METHODOLOGY
- Objective: measure volatility connectedness of electricity markets across Europe and decompose contributions of domestic vs. cross-country shocks.
- Primary approach:
  - Spillover index method introduced by Diebold and Yilmaz (2009; 2012; 2014).
  - Use of generalized VAR framework (Koop, Pesaran, and Potter, 1996; Pesaran and Shin, 1998) to obtain variance decompositions invariant to variable ordering.
- Definitions and measures:
  - Total volatility spillover index S(H): proportion of forecast error variance attributable to cross-variance shares (equations provided in text).
  - Directional spillovers received by market i from all other markets S_i←j(H) and transmitted by market i to all other markets S_i→j(H) (normalized generalized variance decomposition).
  - Net volatility spillover S_i^net(H) = S_i→j(H) − S_i←j(H).
  - Net pairwise volatility spillover S_ij^net(H) = θ̃_ij^g(H) − θ̃_ji^g(H).
- Estimation details:
  - Full-sample estimation reported in Table 3 (each row sums to 100; off-diagonal entries are cross-variance shares).
  - Rolling-window estimation: 90-day rolling window used to capture time evolution of spillovers.

### IV. EMPIRICAL RESULTS — SPILLOVERS AND INTERACTIONS
- Total spillover index (full-sample result):
  - Total spillover index: 72.8 percent.
  - Interpretation: about 73 percent of the forecast error variation is explained by cross-variance shares; 27 percent attributed to within-country shocks.
- Country heterogeneity in spillovers (selected entries and summaries drawn from Table 3):
  - Lowest contribution from others (Contribution From Others column): Ireland 31.5 percent.
  - Highest contribution from others: Hungary 82.7 percent.
  - Strongest bilateral interactions:
    - Spain → Portugal: Spain explains 32.4 percent of Portugal’s forecast error variance.
    - Portugal → Spain: Portugal accounts for 31.4 percent of Spain’s forecast error variance.
    - Lithuania → Latvia: Lithuania explains 22.1 percent of Latvia’s forecast error variance.
    - Latvia → Lithuania: Latvia accounts for 18.6 percent of Lithuania’s forecast error variance.
  - Very low bilateral spillovers between geographically distant/unconnected markets:
    - Austria ↔ Ireland: Austria accounts for 0.5 percent of Ireland’s forecast error variance; Ireland accounts for 0.3 percent of Austria’s forecast error variance.
- Net contribution (Contribution to others minus Contribution from others; Net Contribution column highlights direction):
  - Largest net transmitters (selected):
    - Latvia: 63.1 percent (largest net spillover to other countries).
    - Lithuania: 48.0 percent.
    - Estonia: 34.3 percent.
  - Largest net recipients (selected):
    - Greece: -51.8 percent.
    - Austria: -41.8 percent.
  - Additional selected net contributions and totals:
    - Contribution to others (sum of off-diagonal components) for all countries sums to 1,748.4 (as reported).
    - Contribution including own (row sums) for entire system: various country-specific values are reported in Table 3 (e.g., AT 58.2; BE 89.6; CH 111.6; ...; Total Spillover Index 72.8 percent).
- Interpretation of net patterns:
  - High net spillover transmission from Latvia, Lithuania, and Estonia is linked to historical integration with the Russia-controlled power grid before Russia’s invasion of Ukraine; these countries cut down energy imports from Russia following the war and are set to desynchronize completely from the Soviet-era electricity network in 2025.
  - Net volatility spillovers reflect factors such as electricity production mix, electricity trade (exports and imports), price formation, and infrastructure quality.
- Time-varying dynamics (rolling-window estimation):
  - Spillover index estimated on a 90-day rolling window.
  - Reported dynamic trend: clear upward trend in dynamic volatility spillovers, increasing from an average of 89.9 in 2015 to 91.7 during the period 2019–2021.

### V. KEY TAKEAWAYS
- Cross-border volatility spillovers dominate national electricity market volatility in Europe: Total spillover index 72.8 percent (cross-variance shares explain ~73 percent of forecast error variation).
- Marked cross-country heterogeneity:
  - Average wholesale electricity price volatility across countries: 0.262; country means range from 0.128 (Greece) to 0.414 (Finland).
  - Some countries are strong net transmitters of volatility (e.g., Latvia 63.1 percent; Lithuania 48.0 percent; Estonia 34.3 percent), while others are strong net recipients (e.g., Greece -51.8 percent; Austria -41.8 percent).
- Geographical proximity and physical interconnection increase spillover intensity (e.g., Portugal–Spain; Latvia–Lithuania), while distant/unconnected markets exhibit much lower bilateral spillovers (e.g., Austria–Ireland).
- Volatility spillovers have increased over time based on 90-day rolling-window estimates (average spillover rising from 89.9 in 2015 to 91.7 in 2019–2021).

*Source: wpiea2025007-print-pdf (content provided).*

### 91.9 after 2022 with Russia’s invasion of Ukraine. The patter of volatility spillovers with peaks and

### wpiea2025007-print-pdf - 91.9 after 2022 with Russia’s invasion of Ukraine. The patter of volatility spillovers with peaks and

### Dynamics of the Total Spillover Index and High-Frequency Patterns
- Time period highlighted for surge in cross-border spillovers: January 1, 2022 to April 30, 2024.
- Figure 3 and text evidence:
  - Total spillover index increases to a peak of 96.3 percent.
  - Sample average rises to 83.4 percent, from 72.8 percent for the entire sample period.
- Identified drivers and episodes of elevated volatility and spillovers:
  - Oil price plunge of 2014-2016.
  - COVID-19 pandemic in 2020.
  - Russia’s invasion of Ukraine after 2022.
  - The unabating rise in renewables as a weighty factor contributing to volatility (reference to Cevik and Ninomiya (2023)).

### Dataset and Modeling Framework
- Dataset:
  - 15-minute frequency dataset with more than 876,582 observations on 576 pairs of countries over the period 2014–2024.
- Gravity model specification (augmented):
  - Dependent variable: bilateral wholesale electricity price volatility spillovers s_ijt.
  - Main regressors: gravity_ijt (income, population, geographic distance, geographical contiguity), renewable_it (share of renewables in origin), renewable_jt (share of renewables in destination).
  - Fixed effects: η_ij (country-pair), φ_it (origin time), μ_jt (destination time).
  - Standard errors clustered at the country-pair level.
- Estimation approach:
  - Poisson Pseudo Maximum Likelihood (PPML) regression to control for heteroskedasticity and correlated errors across countries and over time.

### Key Empirical Results from the Gravity Model (summary of Table 4)
- Sample sizes and model fit:
  - Number of observations: [1] 803,781; [2] 798,724; [3] 871,565.
  - Number of countries: 24.
  - R2: 0.117, 0.119, 0.135 for columns [1], [2], [3] respectively.
  - Country-pair FE: Yes in column [3]; No in [1] and [2].
  - Country-time FE: Yes in all specifications.
- Estimated coefficient directions and statistical significance (reported significance levels retained):
  - Distance: positive and significant (0.000***).
  - Contiguity: negative and significant (-0.098***; -0.103***).
  - Real GDP (Origin): negative and significant (0.000*** or -0.000*** reported).
  - Real GDP (Destination): positive but weak/significance varies (0.000*; 0.000**).
  - Population (Origin and Destination): positive and significant (0.001*** for both).
  - Share of Renewables (Origin): positive and significant (0.066***; 0.095***).
  - Share of Renewables (Destination): positive and significant (0.140***; 0.058***).
- Interpretation of key coefficients:
  - Greater geographic distance is associated with higher bilateral volatility spillovers in the estimations, while direct geographical contiguity exerts a dampening effect.
  - Larger population in origin and destination countries is associated with greater spillovers (capturing electricity demand effects).
  - Higher shares of intermittent renewables (solar, wind) in both origin and destination are associated with greater bilateral volatility spillovers, consistent with Cevik and Ninomiya (2023).

### Aggregate and Country-Level Findings
- Cross-border dominance:
  - About 73 percent of the forecast error variation is explained by cross-variance shares; only 27 percent is attributable to within-country shocks.
  - Cross-border volatility spillovers dominate national electricity market behavior in Europe and have grown over time, especially after Russia’s invasion of Ukraine.
- Country heterogeneity and notable bilateral interactions:
  - Ireland: lowest spillover from others.
  - Hungary: highest level of spillover effects.
  - Portugal–Spain bilateral interaction: Spain explains 32.4 percent of Portugal’s forecast error variance; Portugal accounts for 31.4 percent of Spain’s forecast error variance.
  - Latvia–Lithuania bilateral interaction: Lithuania explains 22.1 percent of Latvia’s forecast error variance; Latvia accounts for 18.6 percent of Lithuania’s forecast error variance.
  - Austria–Ireland: very low bilateral spillovers (Austria accounts for 0.5 percent of Ireland’s forecast error variance; Ireland accounts for 0.3 percent of Austria’s).
- Appendix Table A1 headline statistics:
  - Contribution to others (sum across countries): 2001.5
  - Total Spillover Index (In percent): 83.4

### Policy Implications and Recommendations (from the paper)
- Infrastructure and market design:
  - Infrastructure modernization and regulatory reforms can help minimize volatility in wholesale electricity prices during the transition to renewables.
  - Moving from the current zonal system with cost-based redispatch to a nodal pricing system would improve efficient distribution of electricity and dampen excessive price fluctuations, given growing shares of renewables with greater intermittency and close to zero marginal costs of generation.
- Energy security and integration:
  - These reforms would support the transition to low-carbon power generation and aid integration of electricity markets and strengthening of energy security in Europe.

*Source: IMF authors’ estimations and analysis (content unit: wpiea2025007-print-pdf).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025007-print-pdf.pdf_