pith. sign in

arxiv: 2606.05991 · v1 · pith:R36ZYTUZnew · submitted 2026-06-04 · 💱 q-fin.GN · econ.EM

Forecasting of volatility and risk premia in electricity markets

Thomas K. Kloster , Fred Espen Benth This is my paper

Pith reviewed 2026-06-27 23:02 UTC · model grok-4.3

classification 💱 q-fin.GN econ.EM
keywords volatility forecastingelectricity marketsrealized covariationrisk premiaHAR modelrenewable energycovariance forecasting
0
0 comments X

The pith

Incorporating longer time horizons and renewable generation data improves one-week forecasts of electricity market covariation and risk premia.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper constructs a matrix-HAR model to forecast the realized covariation, which represents the latent covariance in electricity prices. It tests this on one-week ahead predictions and shows that longer horizons and renewable information add predictive power. The model also yields better forecasts of spread risk premia in forward markets than methods using only past volatility. A sympathetic reader would care because electricity markets involve high uncertainty from renewables, and better forecasts could aid risk management and pricing.

Core claim

We study forecasting of the realized covariation in electricity markets using a parsimonious matrix-HAR type model. We find that the inclusion of longer time horizons and renewable generation information adds important predictive power to one-week ahead forecasts of the weekly realized covariation. Our variance forecasts provide substantially improved forecasts of spread risk premia compared to standard methods relying on backward looking volatility.

What carries the argument

The matrix-HAR model, a time-series specification that uses multiple lag horizons on the matrix-valued realized covariation to capture dynamics of the latent infinite-dimensional covariance operator.

If this is right

  • Longer time horizons in the model improve one-week ahead covariation forecasts.
  • Adding renewable generation information enhances predictive accuracy.
  • Improved variance forecasts lead to better predictions of spread risk premia in electricity forward markets.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • This framework could apply to forecasting in other volatile commodity markets influenced by weather or renewables.
  • Testing the model on intraday data or different forecast horizons might reveal further improvements or limitations.

Load-bearing premise

The chosen matrix-HAR specification adequately captures the dynamics of the latent covariance operator in electricity markets.

What would settle it

A test showing that one-week ahead forecasts without renewable data or longer horizons match or exceed the full model's accuracy on an extended dataset would challenge the claim of added predictive power.

Figures

Figures reproduced from arXiv: 2606.05991 by Fred Espen Benth, Thomas K. Kloster.

Figure 1
Figure 1. Figure 1: Tr  RCV \ versus Tr (RCV ) over time for five of the forecasting models. To assess the off-diagonal calibration of the models, we consider various “portfolios” that reflect meaningful positions/hedges in electricity markets. The motivation is that electricity producers have very different exposure to price and volatility risk, depending on their production capabilities. For example, photovoltaic producti… view at source ↗
Figure 2
Figure 2. Figure 2: Heatmaps depicting the forecasted versus realized RCV matrix on two distinct days for [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Net-zero volume portfolio profiles. 2021 2022 2023 2024 2025 Date 300 200 100 0 100 200 300 t ( / M W h ) 95% pointwise prediction band Realised t, hour 13 [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Pointwise 95% prediction band for delivery period 12-13 over time based on [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Pointwise versus correlation-adjusted 95% prediction band for [PITH_FULL_IMAGE:figures/full_fig_p015_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Estimated risk premia over time. the best performing model being based on the DRD-HAR-P2 model. In all cases, we find that α >b 0, meaning that the predicted peak-load risk premium is generally higher than the baseload risk-premium. On [PITH_FULL_IMAGE:figures/full_fig_p016_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Predicted versus realized spread risk premium using rolling variances and RCV forecasts [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗
read the original abstract

We study forecasting of the realized covariation in electricity markets. The realized covariation in this context is a matrix-valued representation of the latent infinite-dimensional covariance operator and a parsimonious matrix-HAR type model is constructed to facilitate estimation. We test the model on one-week ahead forecasts of the weekly realized covariation and find that the inclusion of longer time horizons and renewable generation information adds important predictive power. We also investigate the prediction of risk premia in electricity forward markets and find that our variance forecasts provide substantially improved forecasts of spread risk premia compared to standard methods relying on backward looking volatility.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript develops a parsimonious matrix-HAR model for the realized covariation operator (a matrix-valued representation of the latent infinite-dimensional covariance operator) in electricity markets. It reports that one-week-ahead forecasts of weekly realized covariation benefit from the inclusion of longer time horizons and renewable generation information, and that the resulting variance forecasts substantially improve predictions of spread risk premia relative to standard backward-looking volatility methods.

Significance. If the out-of-sample improvements are confirmed with appropriate statistical controls, the work would offer a practical advance in electricity-market volatility modeling by incorporating renewable-generation covariates and a matrix structure for covariation. The approach could aid risk management in forward markets where spread premia are economically relevant.

major comments (2)
  1. [Abstract] Abstract: the central empirical claims of 'important predictive power' and 'substantially improved forecasts' are stated without any quantitative error metrics, confidence intervals, or out-of-sample design details, preventing verification of the magnitude or statistical significance of the reported gains.
  2. [Abstract] Abstract: it is not stated whether the reported 'predictions' are generated from a strictly out-of-sample exercise or include in-sample parameter fitting, which directly affects the validity of the forecasting evaluation and the claim of incremental predictive power from longer horizons and renewables.
minor comments (1)
  1. The weakest modeling assumption—that the chosen matrix-HAR specification adequately captures the dynamics of the latent covariance operator—would benefit from explicit discussion or sensitivity checks in the methods section.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed comments. The two major points both concern the abstract, and we agree that greater specificity is warranted. We will revise the abstract to incorporate quantitative metrics, confidence intervals where applicable, and explicit confirmation of the out-of-sample design. The body of the paper already contains these elements; the revision will make the abstract self-contained.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central empirical claims of 'important predictive power' and 'substantially improved forecasts' are stated without any quantitative error metrics, confidence intervals, or out-of-sample design details, preventing verification of the magnitude or statistical significance of the reported gains.

    Authors: We accept this observation. The revised abstract will report the key out-of-sample metrics (e.g., relative MSE reductions for the matrix-HAR specifications versus benchmarks) together with the evaluation window and any statistical tests employed. These quantities are already computed and presented in Sections 4 and 5 of the manuscript; their inclusion in the abstract will address the concern directly. revision: yes

  2. Referee: [Abstract] Abstract: it is not stated whether the reported 'predictions' are generated from a strictly out-of-sample exercise or include in-sample parameter fitting, which directly affects the validity of the forecasting evaluation and the claim of incremental predictive power from longer horizons and renewables.

    Authors: The forecasts are produced in a strictly out-of-sample rolling-window scheme in which parameters are re-estimated only on data available at each forecast origin. We will add an explicit sentence to the abstract stating this design. No in-sample fitting is used for the reported predictive comparisons. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper constructs a parsimonious matrix-HAR model for the realized covariation operator and evaluates its one-week-ahead forecasting performance, reporting incremental value from longer horizons and renewable data plus improved risk-premia forecasts. This is a standard empirical time-series exercise: parameters are estimated on historical observations and forecasts are generated forward. No equation or claim reduces by construction to its own inputs, no self-definitional loop exists, and the central results rest on out-of-sample evaluation rather than renaming or fitting the target quantity itself. The derivation chain is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

Model rests on standard time-series assumptions plus fitted HAR coefficients and the domain claim that realized covariation matrices proxy the latent operator; no new entities postulated.

free parameters (2)
  • matrix-HAR lag coefficients
    Coefficients at daily, weekly, and monthly horizons are estimated from data.
  • renewable-generation regression weights
    Weights on renewable volume covariates are fitted to improve predictive power.
axioms (1)
  • domain assumption Realized covariation matrix is a faithful finite-dimensional proxy for the latent infinite-dimensional covariance operator
    Explicitly stated in the abstract as the modeling premise.

pith-pipeline@v0.9.1-grok · 5620 in / 1150 out tokens · 26389 ms · 2026-06-27T23:02:06.569327+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

52 extracted references · 22 canonical work pages

  1. [1]

    Representation and approximation of ambit fields in Hilbert space , volume =

    Benth, Fred Espen and Eyjolfsson, Heidar , address =. Representation and approximation of ambit fields in Hilbert space , volume =. Stochastics (Abingdon, Eng. : 2005) , language =. 2017 , doi =

    2005

  2. [2]

    Benth, Fred Espen and Schroers, Dennis and Veraart, Almut E. D. , address =. A feasible central limit theorem for realised covariation of. The Annals of Applied Probability , language =. 2024 , doi =

    2024

  3. [3]

    , address =

    Janczura, Joanna and Trück, Stefan and Weron, Rafał and Wolff, Rodney C. , address =. Identifying spikes and seasonal components in electricity spot price data: A guide to robust modeling , volume =. Energy Economics , language =. 2013 , doi =

    2013

  4. [4]

    2022 , issn =

    A weak law of large numbers for realised covariation in a Hilbert space setting , journal =. 2022 , issn =. doi:https://doi.org/10.1016/j.spa.2021.12.011 , author =

    work page doi:10.1016/j.spa.2021.12.011 2022

  5. [5]

    An introduction to stochastic partial differential equations

    Walsh, John B. An introduction to stochastic partial differential equations. \'E cole d' \'E t \'e de Probabilit \'e s de Saint Flour XIV - 1984. 1986. doi:10.1007/BFb0074920

    work page doi:10.1007/bfb0074920 1984

  6. [6]

    Andersen and Tim Bollerslev and Francis X

    Torben G. Andersen and Tim Bollerslev and Francis X. Diebold and Paul Labys , journal =. Modeling and Forecasting Realized Volatility , volume =. 2003 , doi =

    2003

  7. [7]

    2001 , issn =

    The distribution of realized stock return volatility , journal =. 2001 , issn =. doi:https://doi.org/10.1016/S0304-405X(01)00055-1 , url =

    work page doi:10.1016/s0304-405x(01)00055-1 2001

  8. [8]

    Andersen and Tim Bollerslev , journal =

    Torben G. Andersen and Tim Bollerslev , journal =. Answering the Skeptics: Yes, Standard Volatility Models do Provide Accurate Forecasts , urldate =. 1998 , doi =

    1998

  9. [9]

    Barndorff-Nielsen and Neil Shephard , journal =

    Ole E. Barndorff-Nielsen and Neil Shephard , journal =. Econometric Analysis of Realized Volatility and Its Use in Estimating Stochastic Volatility Models , urldate =. 2002 , doi =

    2002

  10. [10]

    Charles F

    Barndorff-Nielsen, Ole E. and Shephard, Neil , title =. Econometrica , volume =. doi:https://doi.org/10.1111/j.1468-0262.2004.00515.x , url =. https://onlinelibrary.wiley.com/doi/pdf/10.1111/j.1468-0262.2004.00515.x , year =

    work page doi:10.1111/j.1468-0262.2004.00515.x 2004

  11. [11]

    Journal of Financial Econometrics , volume =

    Christensen, Kim and Podolskij, Mark , title =. Journal of Financial Econometrics , volume =. 2012 , month =. doi:10.1093/jjfinec/nbr019 , url =

    work page doi:10.1093/jjfinec/nbr019 2012

  12. [12]

    2010 , issn =

    Pre-averaging estimators of the ex-post covariance matrix in noisy diffusion models with non-synchronous data , journal =. 2010 , issn =. doi:https://doi.org/10.1016/j.jeconom.2010.05.001 , url =

    work page doi:10.1016/j.jeconom.2010.05.001 2010

  13. [13]

    2010 , issn =

    Realised quantile-based estimation of the integrated variance , journal =. 2010 , issn =. doi:https://doi.org/10.1016/j.jeconom.2010.04.008 , url =

    work page doi:10.1016/j.jeconom.2010.04.008 2010

  14. [14]

    1987 , issn =

    Expected stock returns and volatility , journal =. 1987 , issn =. doi:https://doi.org/10.1016/0304-405X(87)90026-2 , url =

    work page doi:10.1016/0304-405x(87)90026-2 1987

  15. [15]

    and Hansen, Peter Reinhard and Lunde, Asger and Shephard, Neil , title =

    Barndorff-Nielsen, Ole E. and Hansen, Peter Reinhard and Lunde, Asger and Shephard, Neil , title =. Econometrica , volume =. doi:https://doi.org/10.3982/ECTA6495 , url =. https://onlinelibrary.wiley.com/doi/pdf/10.3982/ECTA6495 , year =

    work page doi:10.3982/ecta6495

  16. [16]

    The Review of Financial Studies , volume =

    Christoffersen, Peter and Jacobs, Kris and Mimouni, Karim , title =. The Review of Financial Studies , volume =. 2010 , month =. doi:10.1093/rfs/hhq032 , url =

    work page doi:10.1093/rfs/hhq032 2010

  17. [17]

    Stochastic modelling of electricity and related markets , volume =

    Benth, Fred Espen and. Stochastic modelling of electricity and related markets , volume =. 2008 , doi =

    2008

  18. [18]

    2014 , issn =

    Electricity price forecasting: A review of the state-of-the-art with a look into the future , journal =. 2014 , issn =. doi:https://doi.org/10.1016/j.ijforecast.2014.08.008 , url =

    work page doi:10.1016/j.ijforecast.2014.08.008 2014

  19. [19]

    , journal =

    Kloster, Thomas K. , journal =. An ambit field framework for the full panel of day-ahead electricity prices , year =. 2509.17236 , archivePrefix=

    arXiv

  20. [20]

    The Annals of Applied Statistics , number =

    Dominik Liebl , title =. The Annals of Applied Statistics , number =. 2013 , doi =

    2013

  21. [21]

    Ying Chen and J. S. Marron and Jiejie Zhang , title =. The Annals of Applied Statistics , number =. 2019 , doi =

    2019

  22. [22]

    Journal of Business & Economic Statistics , volume=

    Ying Chen and Bo Li , title =. Journal of Business & Economic Statistics , volume =. 2017 , publisher =. doi:10.1080/07350015.2015.1092976 , URL =

    work page doi:10.1080/07350015.2015.1092976 2017

  23. [23]

    2018 , note =

    A space-time random field model for electricity forward prices , journal =. 2018 , note =. doi:https://doi.org/10.1016/j.jbankfin.2017.03.018 , url =

    work page doi:10.1016/j.jbankfin.2017.03.018 2018

  24. [24]

    Measure-valued processes for energy markets , year =

    Cuchiero, Christa and Di Persio, Luca and Guida, Francesco and Svaluto-Ferro, Sara , address =. Measure-valued processes for energy markets , year =. Mathematical Finance , language =

  25. [25]

    and Benth, Fred Espen and Veraart, Almut E

    Barndorff-Nielsen, Ole E. and Benth, Fred Espen and Veraart, Almut E. D. , address =. Modelling Electricity Futures by Ambit Fields , volume =. Advances in Applied Probability , language =. 2014 , doi =

    2014

  26. [26]

    Energy Economics , volume =

    Asymmetric volatility in. Energy Economics , volume =. 2016 , issn =. doi:https://doi.org/10.1016/j.eneco.2016.04.002 , url =

    work page doi:10.1016/j.eneco.2016.04.002 2016

  27. [27]

    Energy Economics , year=2017, volume=

    Bennedsen, Mikkel , title=. Energy Economics , year=2017, volume=

    2017

  28. [28]

    Tests of common stochastic trends , volume =

    Nyblom, Jukka and Harvey, Andrew , address =. Tests of common stochastic trends , volume =. Econometric Theory , language =. 2000 , doi =

    2000

  29. [29]

    and Roberts, Michael R

    Knittel, Christopher R. and Roberts, Michael R. , address =. An empirical examination of restructured electricity prices , volume =. Energy Economics , language =. 2005 , doi =

    2005

  30. [30]

    The Journal of Finance , language =

    Measuring and Testing the Impact of News on Volatility , volume =. The Journal of Finance , language =. 1993 , author =

    1993

  31. [31]

    Energy Economics , language =

    An empirical comparison of alternate regime-switching models for electricity spot prices , volume =. Energy Economics , language =. 2010 , author =

    2010

  32. [32]

    Journal of Financial Econometrics , language =

    A Simple Approximate Long-Memory Model of Realized Volatility , volume =. Journal of Financial Econometrics , language =. 2009 , author =

    2009

  33. [33]

    Journal of Business & Economic Statistics , language =

    Discrete-Time Volatility Forecasting With Persistent Leverage Effect and the Link With Continuous-Time Volatility Modeling , volume =. Journal of Business & Economic Statistics , language =. 2012 , author =

    2012

  34. [34]

    Journal of Financial Econometrics , language =

    A Machine Learning Approach to Volatility Forecasting , volume =. Journal of Financial Econometrics , language =. 2023 , author =

    2023

  35. [35]

    Journal of Econometrics , language =

    Exploiting the errors: A simple approach for improved volatility forecasting , volume =. Journal of Econometrics , language =. 2016 , author =

    2016

  36. [36]

    Nelson , journal =

    Daniel B. Nelson , journal =. Conditional Heteroskedasticity in Asset Returns: A New Approach , urldate =. 1991 , doi =

    1991

  37. [37]

    Journal of Business & Economic Statistics , volume =

    Drew Creal and Siem Jan Koopman and André Lucas , title =. Journal of Business & Economic Statistics , volume =. 2011 , publisher =. doi:10.1198/jbes.2011.10070 , URL =

    work page doi:10.1198/jbes.2011.10070 2011

  38. [38]

    Nonstationarities in Financial Time Series, the Long-Range Dependence, and the IGARCH Effects , urldate =

    Thomas Mikosch and Cătălin Stărică , journal =. Nonstationarities in Financial Time Series, the Long-Range Dependence, and the IGARCH Effects , urldate =. 2004 , doi =

    2004

  39. [39]

    2007 , issn =

    Hourly electricity prices in day-ahead markets , journal =. 2007 , issn =. doi:https://doi.org/10.1016/j.eneco.2006.08.005 , url =

    work page doi:10.1016/j.eneco.2006.08.005 2007

  40. [40]

    Electronic Journal of Probability , publisher =

    Martin Hairer and Jonathan Mattingly , title =. Electronic Journal of Probability , publisher =. 2011 , doi =

    2011

  41. [41]

    Stochastic Analysis and Applications , issue =

    Da Prato, Giuseppe and Gatarek, Dariusz and Zabczyk, Jerzy , title =. Stochastic Analysis and Applications , issue =. 1992 , doi =

    1992

  42. [42]

    Stochastic Equations in Infinite Dimensions , publisher=

    Da Prato, Giuseppe and Zabczyk, Jerzy , year=. Stochastic Equations in Infinite Dimensions , publisher=

  43. [43]

    Journal of Applied Econometrics , volume =

    Creal, Drew and Koopman, Siem Jan and Lucas, André , title =. Journal of Applied Econometrics , volume =. doi:https://doi.org/10.1002/jae.1279 , year =

    work page doi:10.1002/jae.1279

  44. [44]

    and Benth, Fred Espen , journal =

    Kloster, Thomas K. and Benth, Fred Espen , journal =. The fine structure of electricity price volatility , year =. 2605.13320 , archivePrefix=

    Pith/arXiv arXiv

  45. [45]

    Forecasting realized covariances using

    Matias Quiroz and Laleh Tafakori and Hans Manner , year=. Forecasting realized covariances using. 2412.10791 , archivePrefix=

    arXiv

  46. [46]

    Econometrica , volume =

    Archakov, Ilya and Hansen, Peter Reinhard , title =. Econometrica , volume =. doi:https://doi.org/10.3982/ECTA16910 , url =. https://onlinelibrary.wiley.com/doi/pdf/10.3982/ECTA16910 , year =

    work page doi:10.3982/ecta16910

  47. [47]

    Patton , keywords =

    Volatility forecast comparison using imperfect volatility proxies , journal =. 2011 , note =. doi:https://doi.org/10.1016/j.jeconom.2010.03.034 , url =

    work page doi:10.1016/j.jeconom.2010.03.034 2011

  48. [48]

    and Lunde, Asger and Nason, James M

    Hansen, Peter R. and Lunde, Asger and Nason, James M. , title =. Econometrica , volume =. doi:https://doi.org/10.3982/ECTA5771 , url =. https://onlinelibrary.wiley.com/doi/pdf/10.3982/ECTA5771 , year =

    work page doi:10.3982/ecta5771

  49. [49]

    2013 , issn =

    On loss functions and ranking forecasting performances of multivariate volatility models , journal =. 2013 , issn =. doi:https://doi.org/10.1016/j.jeconom.2012.08.004 , url =

    work page doi:10.1016/j.jeconom.2012.08.004 2013

  50. [50]

    Journal of Applied Econometrics (Chichester, England) , language =

    A forecast comparison of volatility models: does anything beat a. Journal of Applied Econometrics (Chichester, England) , language =. 2005 , author =

    2005

  51. [51]

    Lemmon , journal =

    Hendrik Bessembinder and Michael L. Lemmon , journal =. Equilibrium Pricing and Optimal Hedging in Electricity Forward Markets , urldate =

  52. [52]

    2009 , issn =

    Price formation in electricity forward markets and the relevance of systematic forecast errors , journal =. 2009 , issn =. doi:https://doi.org/10.1016/j.eneco.2008.12.001 , url =

    work page doi:10.1016/j.eneco.2008.12.001 2009