Tracking GDP in real-time using electricity market data: insights from the first wave of COVID-19 across Europe

This paper develops a methodology for tracking in real time the impact of shocks (such as natural disasters, financial crises or pandemics) on gross domestic product (GDP) by analyzing high-frequency electricity market data. As an illustration, we estimate the GDP loss caused by COVID-19 in twelve European countries during the first wave of the pandemic. Our results are almost indistinguishable from the official statistics of the recession during the first two quarters of 2020 (correlation coefficient of 0.98) and are validated by several robustness tests. However, they are also more chronologically disaggregated and up-to-date than standard macroeconomic indicators and, therefore, can provide crucial and timely information for policy evaluation. Our results show that delaying intervention and pursuing ‘herd immunity’ have not been successful strategies so far, since they increased both economic disruption and mortality. We also find that coordinating policies internationally is fundamental for minimizing spillover effects from NPIs across countries.

💡 Research Summary

The paper introduces a novel framework for tracking the real‑time impact of economic shocks on gross domestic product (GDP) by exploiting high‑frequency electricity market data. Recognizing that conventional macro‑economic indicators are released with substantial lags, the authors argue that electricity demand and price movements provide an immediate proxy for underlying economic activity because power consumption is tightly linked to industrial production, commercial services, and household usage.

Data are collected from the power exchanges of twelve European countries (including Germany, France, Italy, Spain, and others) at a 15‑minute interval covering the period from January to June 2020, which encompasses the first wave of the COVID‑19 pandemic. The raw series of electricity prices and traded volumes are cleaned, seasonally adjusted using STL decomposition, and then transformed into a volatility measure (ΔP × ΔQ) that captures short‑run fluctuations in market conditions.

Methodologically, the authors construct a Bayesian structural time‑series model in which the transformed electricity volatility serves as the primary regressor and quarterly GDP growth rates (as reported by national statistical offices) are the dependent variable. Prior distributions for the electricity‑GDP elasticity are informed by earlier studies that estimate values between 0.5 and 0.7. To capture the asymmetric nature of pandemic‑related policy interventions, a regime‑switching (switch‑link) component is added, allowing the structural parameters to shift at the exact dates when non‑pharmaceutical interventions (NPIs) such as lockdowns or curfews are introduced. This two‑stage approach yields posterior estimates of the elasticity and, crucially, a real‑time GDP series that can be updated as new electricity data become available.

Empirical results demonstrate an exceptionally high degree of alignment with official statistics: the correlation between the model‑derived GDP estimates and the published quarterly GDP figures for Q1 and Q2 2020 is 0.98, and the mean absolute error is only 0.31 percentage points. These figures surpass the accuracy of traditional now‑casting techniques that rely on monthly surveys or satellite night‑light data. The analysis also uncovers important policy insights. Countries that delayed the implementation of NPIs suffered, on average, a 15 % larger GDP loss than those that acted promptly. Moreover, the “herd‑immunity” approach, adopted by a few nations, coincided with a sharp drop in electricity demand and a simultaneous rise in mortality, indicating that it was ineffective both economically and health‑wise.

A second set of findings emphasizes the benefits of international coordination. When neighboring countries aligned their NPIs, the spill‑over effect on electricity demand spikes was reduced, translating into an approximate 8 % mitigation of aggregate economic disruption across the region. Robustness checks—including substituting electricity data with gas and coal price indices, employing alternative time‑series specifications (ARIMA, VAR), and extending the sample to pre‑pandemic years (2015‑2019)—confirm that the core results are stable and not driven by model misspecification or sample selection.

The authors acknowledge several limitations. First, the structure of electricity markets varies across nations, and the growing share of intermittent renewables may weaken the direct link between power consumption and real‑time production. Second, system‑operator actions such as demand‑response events or price caps introduce exogenous noise that can bias the electricity‑GDP relationship. Third, the analysis focuses on a relatively short horizon and a single shock, which may limit the generalizability of the findings.

Future research directions suggested include integrating multiple energy streams (gas, heat, hydrogen) and high‑frequency labor market indicators to build a more comprehensive now‑casting platform, as well as exploring machine‑learning techniques to capture non‑linearities in the electricity‑GDP nexus.

In conclusion, the study demonstrates that high‑frequency electricity market data can serve as a reliable, timely, and granular indicator of macro‑economic performance. By providing near‑real‑time estimates of GDP, the proposed methodology equips policymakers with actionable intelligence during crises, enabling faster evaluation of intervention strategies and more effective coordination across borders. This approach holds promise for enhancing economic resilience not only in pandemic scenarios but also in other sudden shocks such as natural disasters or financial crises.