Understanding electricity price evolution: Day-ahead market competitiveness in Romania

Bâra, Adela; Oprea, Simona-Vasilica; Georgescu, Irina Article Understanding electricity price evolution: Day-ahead market competitiveness in Romania Journal of Business Economics and Management (JBEM) Provided in Cooperation with: Vilnius Gediminas Technical University (VILNIUS TECH) Suggested Citation: Bâra, Adela; Oprea, Simona-Vasilica; Georgescu, Irina (2023) : Understanding electricity price evolution: Day-ahead market competitiveness in Romania, Journal of Business Economics and Management (JBEM), ISSN 2029-4433, Vilnius Gediminas Technical University, Vilnius, Vol. 24, Iss. 2, pp. 221-244, https://doi.org/10.3846/jbem.2023.19050 This Version is available at: https://hdl.handle.net/10419/317622 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Copyright © 2023 The Author(s). Published by Vilnius Gediminas Technical University *Corresponding author. E-mail: [email protected] This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons. org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Journal of Business Economics and Management ISSN 1611-1699 / eISSN 2029-4433 2023 Volume 24 Issue 2: 221–244 https://doi.org/10.3846/jbem.2023.19050 UNDERSTANDING ELECTRICITY PRICE EVOLUTION– DAY-AHEAD MARKET COMPETITIVENESS IN ROMANIA Adela BÂRA , Simona-Vasilica OPREA , Irina Alexandra GEORGESCU * Department of Economic Informatics and Cybernetics, Bucharest University of Economic Studies, Bucharest, Romania Received 09 February 2023; accepted 20 April 2023 Abstract. The unexpected pandemic eruption in March 2020, the European efforts to diminish the gas house emissions, prolonged drought, higher inflation and the war in Ukraine clearly have had a strong impact on the electricity price. In this paper, we analyze the electricity prices on the Romanian Day-Ahead Market (DAM) along with other variables (inflation, consumption and traded volume of gas on DAM) over the last three and a half years in an attempt to understand its evolution and future trend in the economic and geopolitical context. Autoregressive Distributed Lag models are proposed to analyze the causality among variables on shortand long-term perspective, whereas Quantile Regression (QR) is proposed to enhance the results of the Ordinary Least Squares (OLS) regression. Furthermore, using market concentration metrics– Herfindahl-Hirschman Index (HHI), C1 and C3 ratio, we analyze the competitiveness on the Romanian DAM and correlate it with the electricity price evolution. The concentration indicators on this market reflect the degree of competition manifested between sellers and buyers respectively, their dynamics being able to influence the price level. The higher concentration on the sellers’ side (HHI = 1500) indicates a potential speculative behavior on this market that led to higher prices on DAM. Keywords: market concentration metrics, day-ahead market, electricity price, causality, autoregressive distributed lag, quantile regression. JEL Classification: O13, Q43, C52. Introduction On a timeline, several events and trends have dominated the last three years and significantly lead to today’s energy crisis. The COVID-19 pandemic disease that burst out in March 2020 is not yet eradicated. The consumption flows significantly changed from business and commercial centers to households during the lockdowns, some affairs were closed or drastically diminished their activity. Renewable Energy Sources (RES), especially PV panels, become more attractive for investors and residential consumers and their impact on the electricity 222 A. Bâra et al. Understanding electricity price evolution– day-ahead market competitiveness in... price is indubitable. The recent prices of the CO2 emission certificate that have to be acquired by electricity producers have influenced the electricity price. Moreover, two recent aspects, such as the conflict in Ukraine and the entire geopolitical context have impacted the electricity price. Ultimately, it will be reflected in more expensive products and services and fretful inflation. Therefore, the process of understanding the factors that influence the electricity price and its future evolution are tremendously important especially for policy makers to reduce inflation (Raimi etal., 2022); (Fernández-González etal., 2022), modeling the covariances of financial assets (Boloș etal., 2023). The gradual elimination of restrictions related to the COVID-19 pandemic means a resumption of economic activities at the global level, which also determines an accelerated demand for natural gas consumption. On top of this cause, in Europe there is also the context created by the prolonged winter 2020–2021, which led to the emptying of natural gas deposits. The fall of last year found several European states, including Romania, with reduced natural gas stocks. In addition, the new policy of the European Union to accelerate measures to reduce carbon emissions has led to the increase in the price of carbon dioxide emission certificates, which doubled in 2021 compared to 2020 for fossil fuel producers. The immediate effect was an increase in the price of raw materials, which was then felt in the invoices. The tense situation between Russia and Ukraine also contributes to the current evolution of natural gas and oil prices. Domestically, the market suffered by the end of last year, on the one hand, due to insufficient natural gas stocks, but also because local production was lower than in 2020. According to data provided by the gas providers, Romania’s imports in 2021 have increased by over 300% compared to 2020, during a period when the purchase price on the Russian market increased. For the natural gas price evolution in 2021 and the beginning of 2022, it was interesting to notice whether the liberalization of the market for all consumers, starting on July 1, 2020, contributed to the increase in prices. However, energy specialists argue that liberalization is not the cause of price increases or decreases, it only establishes the conditions for the formation of fair prices. The increase in natural gas prices in the European Union has also affected countries that have a liberalized natural gas market that has been operating for years, not just Romania. However, the accumulation of external and internal factors made the price increase level to be among the highest at the European level. There are several gaps that motivated us to perform this study. First, no recent studies were performed considering the random events (COVID-19 lingering effects, conflict in Ukraine in the Black Sea region) that took place in the recent years and led to a disruptive economic and geopolitical context. The lack of data sets was also a reason that motivated us to create a relevant data set with economic variables (inflation, interest rates), market prices (for gas, oil, electricity and traded quantities) and understand their relationship. Second, most studies focused on the Western European countries and less on the S-E European region that is near the conflict and equally affected by the inflation and energy market volatility. Considering these reasons, we built up a data set merging several variables from various sources that can be input data for other studies. The input data is publicly available for open research. The interval of data extracted from web is generous and contains records taken before and after COVID-19 and Russian invasion. Journal of Business Economics and Management, 2023, 24(2): 221–244 223 The main contribution of the current study consists of: – Proposing an Autoregressive Distributed Lag (ARDL) model to check the long-run and the short-run causalities between electricity price on DAM and the independent variables consumption, inflation and traded volume of gas on DAM. – Revealing a long-run relationship between electricity price on DAM and the three independent variables. – Using Quantile Regression (QR) to find associations of predictors at 25th, 50th and 75th percentiles of electricity price on DAM. – Revealing that the effect of consumption and inflation on the electricity price increases for higher quantiles, while the effect of traded volume of gas on the electricity price decreases for higher quantiles. – Analyzing the market concentration metrics to estimate the competitiveness on the Romanian DAM and understand their relationship with the electricity price. – Revealing a strong causality between market concentration and the level of prices, as the higher concentration of sellers was followed in July and August 2022 by the higher prices on DAM. This paper aims to analyze the factors that influence the competitiveness on the DAM, electricity price and its trend. The price set for the Romanian electricity DAM by a uniform price auction is considered. The bidding process takes place 24 hours before delivery. The producers and suppliers bid to sell or buy the electricity that was unsold or unbought on the bilateral market on which bilateral negotiated contracts are signed more in advance. Thus, the DAM is much closer to real-time and covers the fluctuations that usually appear in consumption. Furthermore, for RES generators this market is suitable as their forecast improves closer to the delivery time. The data set spans over three and a half years. We started to collect data from 1st of January 2019, before the COVID-19, to end of August 2022. This generous time span for data collection also covered the first shock waves generated by the Russian aggression towards Ukraine. The remainder of this paper is structured as follows: in Section 1, a brief literature review is presented, in Section 2, the input data is described emphasizing on the data pre-processing and analytics. Section 3 is dedicated to competitiveness assessment on DAM, Section 4 to methodology, Section 5 to the results, Section 6 includes discussions and the last Section consists in drawing conclusion and providing several implications. 1. Literature review 1.1. Electricity price fluctuations on DAM and Renewable Energy Sources (RES) RES forecast and demand are significant determinants of electricity prices using enhanced regressive models (K. Wang etal., 2022; Ma etal., 2015; Chaikumbung, 2021). These forecasts are helpful for spot and balancing markets and may result in an increase of revenues (Maciejowska etal., 2021). Extensions to basic structural models are often performed including the solar and wind power or CO2-emissions as in Carmona etal. (2013), Hildmann etal. (2015). Scientists underlined that electricity prices show typical characteristics regardless of where the electricity 224 A. Bâra et al. Understanding electricity price evolution– day-ahead market competitiveness in... is traded. The results of Forbes and Zampelli (2014) offered evidence of a significant relationship between the DAMelectricityprice, natural gasprice and hourly load. This correlation allowed a better prediction with ARMA than the Californian Independent System Operator. Several factors are considered to predict the spikes of the electricity price using classification and decision tree (Fragkioudaki etal., 2015). Electricity price fluctuations on DAM are interesting for optimization and storage purposes and the electricity price forecast plays an important role. Ziel and Steinert (2016) emphasized on the five types of electricity price models: fundamental (based on market fundamentals), multi-agent (based on supply and demand), statistical, reduced form and computational intelligence. Except the first two types, they are based on the price, renewable energy and/or electricity demand. They propose a model (known as X-model) using real auction data considering both bids and asks, or sale and purchase curves, and identifying the extreme price or spikes. One of the proposed models is based on a recent novel X-model that captures price spikes. An adjustment to the X-model to handle data from the Great Britain market is proposed. This approach could be translated to the local energy markets that require probabilistic price forecasting models to show the price uncertainty and future trends (Haben etal., 2021). A methodology using the Artificial NN (ANN) is proposed to predict the electricity prices (Keles etal., 2016). Load, fuel prices, RES generation and available capacity are considered as fundamental factors. The selection and preparation of fundamental data that have a noticeable impact on electricity prices gained focus. This was performed using cluster algorithms, but also by comparing the results. 1.2. Electricity price fluctuations on DAM and Analytical Models Numerous visions regarding electricity price forecast exposed the complexity of the concept (Lago etal., 2021). Contradiction among supporters of statistical and Machine Learning (ML) methods, small size data sets that are not relevant for this type of analysis, lack of replicability and publicly available data sets and various software implementation are only a few aspects that make the comparison of these methods very difficult. Statistical methods, such as linear regression, Least Absolute Shrinkage and Selection Operator (LASSO), elastic net, Lasso Estimated Autoregression, ensemble methods, including clustering analysis, Generalized Autoregressive Conditional Heteroskedastic (GARCH), ARIMA, ML (Multi-Layer Perceptron– MLP, Support Vector Regressor– SVM), deep learning (Convolutional Neural Networks– CNN, Long Short-Term Memory– LSTM) as well as hybrid methods were identified in the context of the electricity price forecast. A framework architecture with numerous statistical and ML models (especially automatic ML– H2O AutoML and TPOT), exogenous features, several time series decomposition methods and time series analysis methods were applied to the Spanish wholesale market, proving good accuracy on mean absolute error and mean absolute scaled error (Beltrán etal., 2022). Romero etal. (2019) analyzed the electric market in Spain to understand prices and market participants that can make prices vary. They noticed the close relationship between the objective variable and the electric demand. A normalization of the price variable due to Journal of Business Economics and Management, 2023, 24(2): 221–244 225 a strong seasonal component and usage of several models of ML, such as: ridge regression, k-nearest neighbors, support vector machines, Neural Networks (NN) and random forest were proposed, identifying random forest as providing the best results without normalizing the objective. 1.3. Electricity price fluctuations on DAM and Geopolitical Overview The disruptive economic and geopolitical context dramatically influenced the electricity markets. As resulted from data analytics, inflation, interest rates and other resources prices (such as oil and gas) increased in the last two years influencing the electricity prices in the European Union countries. This also affected forecast accuracy that recorded higher errors due to higher market volatility and RES share. The prices for CO2 emissions and prolonged drought in the European countries additionally increased electricity prices. Therefore, the most recent studies in the electricity price forecast field also showed a lower accuracy that is provoked by the unstable geopolitical context. Several similar studies were performed before and after the random events that influenced the electricity markets. The weather parameters influence on the DAM in Italy is considered in Bigerna (2018). The data set of hourly market variables and temperature variables was considered. A new econometric estimation showed some evidence of different effects of temperature and provided a more accurate estimation of the hourly prices. The results offered welfare-improvement and policy implications to handle extreme weather conditions. NN are investigated to forecast day-ahead electricity spikes in prices in the electricity market of Ontario, Canada Sandhu etal. (2016). They were trained using a data set consisting of similar price days. The spikes were identified using a spike classification method showing improvements in terms of accuracy. The Singular Spectrum Analysis (SSA) is a technique in time series analysis and forecasting having capabilities in extracting the main structure of the broad classes of the time series. Using SSA, the original electricity price series was decomposed into components: trend, periodic and noisy. The processed price series was considered as input for predicting the day-ahead electricity prices, proving performant for the Australian and Spanish electricity markets (Miranian etal., 2013). In Spain, real-time energy-related data can be extracted to develop algorithms for price forecasting and understand how price vary. For the prediction, a Quantile Regression (QR) model based on Gradient Boosted Regression Trees was proposed in Díaz etal. (2019), improving the accuracy over regression models. Furthermore, it is simpler than ML approaches, providing low prediction errors when using the median as point prediction method. A method of Bootstrap Aggregation (bagging) was proposed in Özen and Yıldırım (2021) to facilitating the traceability of the predictors selection. However, according to B. Wang etal. (2021), the impact of the electricity prices on residential electricity consumption is not that significant. Habits influence households’ electricity consumption. Thus, it is important to create policies to diminish electricity consumption considering the perspective of households’ habits rather than electricity prices (B. Wang etal., 2021). 226 A. Bâra et al. Understanding electricity price evolution– day-ahead market competitiveness in... 2. Input data and data analyses For analyzing the electricity price on the DAM in Romania, the following data sets were collected for interval 1st of January 2019– 31st of August 2022: from various websites such as: – OPCOM1– the Romanian Electricity Market Operator from where the DAM data– hourly prices and quantities were extracted using web scraping. – Curs BNR2 from where interest rate (ROBOR 3M) was extracted. – Bursa Romana de Marfuri3 from where the DAM gas price and quantity were extracted. – Institutul National de Statistica4 from where the consumption prices indices are extracted and the inflation index in Romania is calculated. Web scraping was used to obtain the prices for food, non-food products and services. – Rate inflation5 in European Union from where inflation rate was extracted. – Macrotrends6 from where the oil price was extracted. – Danube level7 from where water level was measured in Turnu Magurele, Braila and Tulcea using web scraping. A data sample from OPCOM is presented in Table 1. The data was extracted using BeautifulSoup and Selenium libraries in Python. Table 1. Data sample for DAM prices and traded volumes from OPCOM website Trading area Hour ROPEX_DAM_H Price [lei/MWh] Volume [MWh] Romania 1 326.62 2418.2 Romania 2 288.92 2525.9 Romania 3 259.02 2569.6 Most data sets are time series with two columns (date and value) and they are open source data that can be found visiting the links below. After the data was extracted, as in Figure 1, the pre-processing stage was required as the time granulation was various (hourly, daily, monthly, trimester). From OPCOM, for instance multiple files were extracted using web scraping. 44 files, one for each month, was obtained. To concatenate and map the data into one single data set, pre-defined Pandas methods were used (concat and map). Then, the Date column of each time series was converted to date and time to prepare the data sets for merging. Numerous missing values were encountered due to different recordings, thus backfill and forward fill were necessary. The code for data extraction can be found on github8. 1 https://www.opcom.ro/pp/grafice_ip/raportPIPsiVolumTranzactionat.php?lang=ro 2 https://www.cursbnr.ro/robor 3 https://www.brm.ro/piata-spot-gn/ 4 http://statistici.insse.ro/shop/?page=ipc1 5 https://www.rateinflation.com/inflation-rate/euro-area-historical-inflation-rate/ 6 https://www.macrotrends.net/2480/brent-crude-oil-prices-10-year-daily-chart 7 https://www.cotele-dunarii.ro/ 8 https://github.com/simonavoprea/opcom_pzu_data/blob/main/data_extraction Journal of Business Economics and Management, 2023, 24(2): 221–244 227 The electricity price varied from 0.01 to 4698.99 RON/MWh, whereas the quantity of electricity traded on the DAM was up to 5052.9 MW. Most of the prices vary between 0 and 1150 RON/MWh. The traded electricity on DAM varies between 1250 and 4500 MW that is an important share of the load in Romania. Very interesting insights can be found in Figure 2 that consists in the hourly electricity price average variation on DAM during the analyzed interval. The data based on which the graphical representations in Figures 2 and 3 are shown was extracted from OPCOM website. Two important aspects are identified: the tremendous increase of the price (more than 5 times) from 2020 to 2022 and the price curve shape that is changing from an almost flat curve in 2019 and 2020 to two-humped curve in 2021 and 2022. The standard deviation of the electricity price on the DAM sharply increased from 85.7 in 2020 to 707.9 in 2022. Additionally, we can notice that the electricity was cheaper in 2020 during lockdowns due to COVID-19 pandemic times. Also, interesting insights can be obtained from Figure 3 that depicts the hourly variation on average of the traded electricity on the DAM. The data for Figure 3 is based on the OPCOM website. The request for electricity traded on DAM also increase but this increase is not that impressive and is somehow expected due to the fact that the business activities were restored gradually in 2021 and 2022 after lockdowns (Dobrowolski etal., 2022). In Figure 4, the Pearson correlation indices between the dependent variable (electricity price on DAM) and the other variables are represented. Figure 1. Data processing and analytics flow Web scraping Csv files Input data set Data Pre-processing. Concatenate and map mulple csv files Convert the Date columns to date me. Join data sets using Date column Backfill and forward fill Data analycs (pivong, melng, charts) and processing. (a) (b) Figure 2. Average hourly electricity price curves (a) and price variation (b) 228 A. Bâra et al. Understanding electricity price evolution– day-ahead market competitiveness in... Figure 4. Correlation between electricity price on DAM and other variables Very strong direct correlations are noticed between the electricity price and gas price (0.89), inflation rate in EU (0.81) and in Romania (0.79), oil price (0.68) and ROBOR 3M (0.66). This correlation suggests that variables are highly dependent. The higher the gas price and oil prices, the higher the electricity price on DAM. Furthermore, the inflation in Romania followed the trend of the inflation in EU and it impacted the interest rate as well. Moderate inverse correlations are identified between the electricity price and the Danube water levels (–0.36). Very relevant is that in 2022, the dependency between the electricity price and Danube water levels is increasing to –0.6 showing a strong inverse correlation, that is the more drought, the higher prices. The prolonged drought led to less hydro energy. Therefore, more conventional sources (gas and oil) were consumed to generate electricity and it has influenced the DAM. Furthermore, for the entire interval, relevant is the rather weak correlation between electricity price and quantity or demand (0.19). In 2020, this correlation was higher (0.31), but decreased in 2020 to 0.16. The demand did not follow the price trend, it remained in its usual trend characterized by a small annual increase (up to 5%). (a) (b) Figure 3. Average hourly electricity load curves (a) and total traded electricity (b) on DAM between Jan. 2019– Aug. 2022 Journal of Business Economics and Management, 2023, 24(2): 221–244 235 Variable Coefficient T-statistics Prob. D(El_price_DAM(-2)) 0.0450 2.24 0.02 D(Consumption_ 0.046 23.27 0.00 D(Consumption(-1)) 0.0233 7.87 0.00 CointEq(-1) –0.8576 –25.05 0.00 R-squared 0.53 Adjusted R-squared 0.53 Durbin-Watson statistic 1.99 (2022) ARDL(4,2,0,0) Variable Coefficient T-statistics Prob. D(El_price_DAM(-1)) 0.1484 4.64 0.00 D(El_price_DAM(-2)) 0.2030 7.62 0.00 D(El_price_DAM(-3)) 0.0911 4.37 0.00 D(Consumption) 0.0595 13.43 0.00 D(Consumption(-1)) 0.1682 4.14 0.00 CointEq(-1) –0.9981 –26.31 0.00 R-squared 0.49 Adjusted R-squared 0.48 Durbin-Watson statistic 2.00 The Error Correction Term (ECT) represents for any shocks the speed of adjustment from short-run disequilibrium to long-run equilibrium. ECT is in all 4 years negative, larger than –2 and statistically significant at 5% level of significance, showing that there is evidence of long-run cointegration. The speed of adjustment to long run equilibrium after a deviation has occurred in the short run was 100% in 2019, 97% in 2020, 85% in 2021 and 99% in 2022. The adjustment coefficients indicate how the deviation from the long-term equilibrium is corrected. In 2019, ECT is between –1 and –2, causing dampening oscillations. This means that the error correction process varied in 2019 around the long-run value in a dampening manner, rather than uniformly converging to equilibrium. Coefficients of the first differenced variables are interpreted as the short-run elasticities. The short-run association between the first lag of Consumption and El_price_DAM is positive and statistically significant at 5% significance level. As the number of lags of Consumption increases, this correlation tends to become negative. Each year the variables jointly explain around 50% of the total variation in El_price_DAM as shown by Adjusted R-squared. The Durbin Watson statistic being around 2 indicates no autocorrelation in the residuals. The stability of the model is tested using the CUSUM test (Tanizaki, 1995). The red lines in Figure 5a–d represent the critical bounds at 5% threshold of significance. Since the blue line is situated within the critical bounds, it follows that the parameters are stable each year. It follows that ARDL-ECM has a good fit and the results are reliable for making inferences and policy recommendations. End of Table 5 236 A. Bâra et al. Understanding electricity price evolution– day-ahead market competitiveness in... 5.2. Quartile regression results The second model that we use is the QR model. The following linear model will be estimated in STATA, following the same steps as Fávero and Belfiore (2019): =++ + +ε 01 2 3 _ . __ __ t tt tt El price DAM b b Consumption b Inflation EU b Gas quantity DAM (6) The specific QR equation is: (_ _ | , _, _,_ _ ) tttt t Quantile El price DAM Consumption Inflation EU Inflation EU Gas quantity DAM = + + + +ε 01 2 3 ._ __ t t tt b b Consumption b Inflation EU b Gas quantity DAM (7) The histogram of the dependent variable El_price_DAM is shown in Figure 6. The Q-Q plot of the dependent variable El_price_DAM is shown in Figure 7. From the histogram and the Q-Q plot of the dependent variable (Figures 6 and 7), we can see that El_price_DAM is not normally distributed. We check the existence of skewness in the dependent variable El_price_DAM. c) CUSUM test for January–March 2021 d) CUSUM test for January–March 2022 Figure 5. Plots of CUSUM for coefficients’ stability of ARDL model at 5% level of significance a) CUSUM test for January –March 2019 b) CUSUM test for January–March 2020 Journal of Business Economics and Management, 2023, 24(2): 221–244 237 Table 6. Skewness/Kurtosis tests for normality Variable Pr(Skewness) Pr(Kurtosis) El_price_DAM 0.00 0.00 Since the p-value is less than 0.05 as shown in Table 6, we will reject the null hypothesis which asserts that the distribution of the dependent variable is normally distributed. Together with the Q-Q plot in Figure 7, this points to the estimation of a QR model. First, we examine the standard linear regression in Table 7 and notice that all the regressors are statistically significant. Table 7. Standard linear regression El_price-DAM Coef. Std. Error t P>|t| [95% Conf. Interval] Consumption 0.078 0.00 46.89 0.00 0.07 0.08 Inflation_EU 165.76 0.63 259.71 0.00 164.5 167.01 Gas_quantity_DAM –0.0000153 1.03e-06 –14.79 0.00 –0.0000173 –0.0000132 Constant –424.46 11.66 –36.39 0.00 –447.33 –401.60 We analyze the results of the QR model for q = 25, q = 50 and q = 75 (as in Tables 8–10). Table 8. 0.25 Quantile regression El_price-DAM Coef. Std. Error t P>|t| [95% Conf. Interval] Consumption 0.037 0.00 33.22 0.00 0.03 0.04 Inflation_EU 105.85 0.43 242.23 0.00 104.99 106.7 Gas_quantity_DAM –0.0000105 7.06e-07 –14.89 0.00 –0.0000119 –9.13e-06 Constant –170.10 7.98 –21.3 0.00 –185.75 –154.44 Figure 7. Q-Q plot of the dependent variable El_price_DAM Figure 6. Histogram of the dependent variable El_price_DAM 238 A. Bâra et al. Understanding electricity price evolution– day-ahead market competitiveness in... Table 9. 0.5 Quantile regression El_price-DAM Coef. Std. Error t P>|t| [95% Conf. Interval] Consumption 0.042 0.00 29.24 0.00 0.040 0.045 Inflation_EU 136.48 0.56 243.12 0.00 135.37 137.58 Gas_quantity_DAM –0.0000148 9.08e-07 –16.3 0.00 –0.000016 –0.000013 Constant –165.13 10.25 –16.10 0.00 –185.23 –145.02 Table 10. 0.75 Quantile regression El_price-DAM Coef. Std. Error t P>|t| [95% Conf. Interval] Consumption 0.066 0.00 30.38 0.00 0.06 0.07 Inflation_EU 177.82 0.83 212.45 0.00 176.17 179.46 Gas_quantity_DAM –0.0000213 1.35e-06 –15.70 0.00 –0.0000239 –0.0000186 Constant –252.23 15.29 –16.49 0.00 –282.21 –222.25 We notice that in all three cases of QR model, all coefficients are statistically significant. Summarizing the above results, there are discrepancies between the parameters estimated by OLS and the ones obtained by the QR models (as in Table 11). Table 11. Parameters estimated in each model and their standard errors Variable OLS QREG25 QREG50 QREG75 Consumption 0.078 0.037 0.042 0.066 Standard error 0.00 0.00 0.00 0.00 Inflation_EU 165.76 105.85 136.48 177.81 Standard error 0.63 0.43 0.56 0.83 Gas_quantity_DAM –0.00001526 –0.00001052 –0.00001479 –0.00002125 Standard error 1.03e-06 7.06e-07 4.07e-07 1.35e-06 Constant –424.46 –170.1 –165.13 –252.23 Standard error 11.66 7.98 10.25 15.29 The standard errors of the parameters are lower for the 25th percentile QR model, which indicates that the 25th percentile QR model has a greater precision of the estimation. 1% increase in Consumption has a lower positive impact (0.037%) on El_price_DAM at the 25th percentile than at the 50th percentile, and 75th percentile by 0.042%, respectively 0.066%. 1% increase in Inflation_EU at the 50th, 75th percentile has a greater positive impact on El_price_DAM by 136.48%, respectively 177.81% than at the 25th percentile, by 105.85%. 1% increase in Gas_quantity_DAM at the 50th, 75th percentile has a lower negative impact on El_price_DAM by –0.000014%, respectively –0.000021% than at the 25th percentile, by –0.000010%. In Figure 8, the paneled charts show the difference between the estimators obtained by QR and those obtained by OLS, along with OLS confidence intervals. We also see that the sign of the regressors is the same in all 4 estimation models from Table 11. Journal of Business Economics and Management, 2023, 24(2): 221–244 239 In Figure 8, the quantiles of the regressors are represented on the horizontal axis, while the magnitude of the coefficients is represented on the vertical axis. The coefficient obtained by linear regression is represented on the horizontal line, while the dashed lines delimit the confidence intervals of this coefficient. As Fávero and Belfiore (2019) remarked, the parameters obtained by standard regression and their confidence intervals do not vary with the percentiles. This remark motivates the choice of QR models over OLS models. If the percentile coefficient lies outside the OLS confidence interval, then there are significant differences between the quantile and OLS coefficients. From that we conclude that the percentile coefficients for all regressors are significantly different from the OLS coefficients. The effect of Consumption and Inflation_EU on El_price_ DAM increases for higher quantiles, while the effect of Gas_quantity_DAM on El_price_DAM decreases for higher quantiles, as it was implied from Table 11. 6. Discussions This paper proposed an ARDL model to check the long-run and the short-run causalities between El_price_DAM and the independent variables Consumption, Inflation_EU and Gas_quantity_DAM during January–March 2019–2022. The ARDL bounds test revealed a long-run relationship between El_price_DAM and the three independent variables. From this analysis it follows that Consumption, Inflation_EU and Gas_quantity_DAM are important determinants of El_price_DAM. The correlations between El_price_DAM and its determinants are in accordance with economic theory. Most coefficients of lagged dependent variables are significant at 5% level of significance, meaning that short-term causality also exists. Then, we used QR to find associations of predictors at 25th, 50th and 75th percentiles of El_price_DAM. QR is an effective tool to study how Consumption, Inflation_EU and Gas_ quantity_DAM affect the distribution of El_price_DAM. Our results reveal that the effect of Consumption and Inflation_EU on El_price_DAM increases for higher quantiles, while the effect of Gas_quantity_DAM on El_price_DAM decreases for higher quantiles. Figure 8. Parameters estimated for QR models and by OLS, with their confidence intervals 240 A. Bâra et al. Understanding electricity price evolution– day-ahead market competitiveness in... The two approaches used in this paper can be enriched with the application of the novel dynamic ARDL (DYNARDL) proposed by Jordan and Philips (2018), Philips (2018). The advantage of DYNARDL is that based on the simulations of the parameters of a multivariate normal distribution, it can check the long and short-run causality in level and in difference. At present, 60.2% of the total European energy is low carbon, the primary source being nuclear. The Low-Carbon Energy Observatory (LCEO) has been set up to investigate the innovative measures and policies for low-carbon technologies. Streimikiene etal. (2021) connected low carbon energy transition and poverty energy issues. The policies for climate change mitigation in households regarded according to Streimikiene etal. (2021): Greenhouse Gas (GHG) emissions reduction; energy poverty reduction; the increase of energy efficiency and the expansion of renewable energy share in the total energy consumption. Energy poverty has recently become a multi-parametric issue, having four dimensions in the literature: economic, geographic, in-field and human (Streimikiene & Kyriakopoulos, 2023). The economic dimension refers to energy poverty in different European countries. The in-field dimension takes into account the regional level of analysis, while the human dimension focuses on the population which does not have access to electricity services. The conclusion of this paper is that the heterogeneity of energy and its implications should be analyzed. Streimikiene etal. (2020) developed an integrated framework concerning the measures to reduce energy poverty and increase energy efficiency in energy poor households in the EU. Moreover, the Eastern Europe faces the problem of poor households who suffer from energy poverty. The improvement of energy efficiency in households may be done by means of decarbonization of energy systems and by resorting to renewable energy sources. Energy consumption is on an increasing trend, therefore renewable energy sources diversification and production intensification should be targeted. Solar photovoltaic (PV) energy has become the third source of renewable energy, behind hydroelectric power and wind. In 2021, solar energy represented 3.6% of total electricity generation. Mišnić etal. (2022) made a financial analysis of a project involving the economic viability of a 5MV solar power plant in Montenegro. The electricity price was predicted by ARIMA and neural networks. In the first decade of the 2000s, Romania made a significant progress on the electricity market by de-monopolization and its liberalization (Budulan etal., 2003). Clodnițchi and Chinie (2015) recalled that the European Commission established support systems to be implemented by each EU member state, such as to promote renewable energy. Among these support systems (Clodnițchi & Chinie, 2015) recalled: feed-in-tariffs, premium feed-in-tariffs, contracts for differences, quota obligations and renewable energy certificates, capacity procurement auctions, tax incentives, and hybrid instruments. In Romania quota obligations and renewable energy certificates were implemented with the purpose of stimulating energy production from renewable sources. Conclusions In this paper, we analyzed the market concentration metrics to estimate the competitiveness on the Romanian DAM and understand their relationship with the electricity price. HHI, C1 and C3 are significant in assessing market power in controlling prices and quantities on the DAM. For sellers, HHI indicates a low concentration until May–July 2022, when HHI Journal of Business Economics and Management, 2023, 24(2): 221–244 241 has started to increase to the limit of high concentration. This higher concentration on the sellers’ side negatively influences the electricity market competitiveness, indicating a potential speculative behavior on this market that ultimately led to higher prices on DAM. The sum of market shares of the biggest three market participants (C3) indicated the same trend of low competitiveness. This higher concentration also increases the likelihood of collusion between dominant sellers, resulting in higher electricity pricing on DAM. Therefore, there is a strong causality between market concentration and the level of prices, as the higher concentration of sellers was followed in July and August 2022 by the higher prices on DAM. However, this trend with higher prices took place in all European countries DAM and it was influenced mainly by the economic and the geopolitical context. To satisfy the demand for electricity, it must be invested in the development of the power system and diversity of energy resources. Policies should be adopted to fulfill electricity supply by hydroelectric and nuclear sources. In 2021, in Romania the electricity was provided by hydroelectric power plants by 29% of the total generated electricity, followed by the two nuclear power units (by 18.9%). The government should also attract investments into the electricity sector and create the environment for an efficient and competitive market. Energy policies should be directed towards sustainable development, integrating more RES and electricity storage. The advantage of the ARDL model consists in its suitability for modeling the price fluctuations on the electricity market, characterized by seasonality, volatility and changing dynamics. Possible future directions of this research could be Markov regime switching approach or smooth transition logistic regression model which capture the sensitivity of prices on DAM. As a limitation, bias-corrected bootstrap method may provide better estimations. Thus, the ARDL model could be continued with the bias-corrected bootstrap model, which provides more robust estimates for the long-run coefficients. The most important limitation of QR is that it leads to global effects. An issue to be studied is how the distribution of the dependent variable is affected if the population characteristics change. Future research will be conducted to extend the data set and extract new insights from the economic and geopolitical context that significantly influences electricity prices and markets. Acknowledgements This work was supported by a grant of the Ministry of Research, Innovation and Digitization, CNCSUEFISCDI, project number PN-III-P4-PCE-2021-0334, within PNCDI III. Funding This work was supported by a grant of the Ministry of Research, Innovation and Digitization, CNCSUEFISCDI, project number PN-III-P4-PCE-2021-0334, within PNCDI III. Author contributions Adela Bâra: Conceptualization, Validation, Formal analysis, Investigation, Resources, Data Curation, Writing– Original Draft, Writing– Review and Editing, Visualization, Supervi- 242 A. Bâra et al. Understanding electricity price evolution– day-ahead market competitiveness in... sion. Simona-Vasilica Oprea: Conceptualization, Validation, Formal analysis, Investigation, Writing– Original Draft, Writing– Review and Editing, Visualization, Project administration. Irina Alexandra Georgescu: Methodology, Writing– Original Draft, Writing– Review and Editing, Visualization. Disclosure statement Authors declare that there is no competing financial, professional, or personal interests from other parties. References Beltrán, S., Castro, A., Irizar, I., Naveran, G., & Yeregui, I. (2022). Framework for collaborative intelligence in forecasting day-ahead electricity price. Applied Energy, 306, 118049. https://doi.org/10.1016/j.apenergy.2021.118049 Bigerna, S. (2018). Estimating temperature effects on the Italian electricity market. Energy Policy, 118, 257–269. https://doi.org/10.1016/j.enpol.2018.03.068 Boloș, M.-I., Bradea, I.-A., & Delcea, C. (2023). Modeling the covariance of financial assets using neutrosophic fuzzy numbers. Symmetry, 15(2). https://doi.org/10.3390/sym15020320 Budulan, P., Rugina, V., & Bogzianu, R. (2003). Electricity market development in Romania. 2003 IEEE Bologna Power Tech Conference Proceedings, 4. https://doi.org/10.1109/PTC.2003.1304780 Carmona, R., Coulon, M., & Schwarz, D. (2013). Electricity price modeling and asset valuation: A multi-fuel structural approach. Mathematics and Financial Economics, 7, 167–202. https://doi.org/10.1007/s11579-012-0091-4 Chaikumbung, M. (2021). Institutions and consumer preferences for renewable energy: A meta-regression analysis. Renewable and Sustainable Energy Reviews, 146, 111143. https://doi.org/10.1016/j.rser.2021.111143 Clodnițchi, R., & Chinie, A. C. (2015). Factors of impact on the evolution of electricity markets from renewable energy sources: A comparison between Romania and Germany. Management and Marketing, 10(1), 34–52. https://doi.org/10.1515/mmcks-2015-0003 Díaz, G., Coto, J., & Gómez-Aleixandre, J. (2019). Prediction and explanation of the formation of the Spanish day-ahead electricity price through machine learning regression. Applied Energy, 239, 610–625. https://doi.org/10.1016/j.apenergy.2019.01.213 Dickey, D. A., & Fuller, W. A. (1979). Distribution of the estimators for autoregressive time series with a unit root. Journal of the American Statistical Association, 74(366), 427–431. https://doi.org/10.2307/2286348 Dobrowolski, Z., Sułkowski, Ł., & Panait, M. (2022). Using the business model canvas to improve audit processes. Problems and Perspectives in Management, 20(3), 142–152. https://doi.org/10.21511/ppm.20(3).2022.12 Engle, R. F., & Granger, C. W. J. (1987). Co-integración y corrección de error: representación, estimación y prueba [Co-integration and error correction: Representation, estimation, and testing]. Econometrica, 55(2), 251–276. https://doi.org/10.2307/1913236 Fávero, L. P., & Belfiore, P. (2019). Simple and multiple regression models. In Data Science for Business and Decision Making (pp. 443–538). Academic Press. https://doi.org/10.1016/B978-0-12-811216-8.00013-6 Fernández-González, R., Puime-Guillén, F., & Panait, M. (2022). Multilevel governance, PV solar en- Journal of Business Economics and Management, 2023, 24(2): 221–244 243 ergy, and entrepreneurship: The generation of green hydrogen as a fuel of renewable origin. Utilities Policy, 79, 101438. https://doi.org/https://doi.org/10.1016/j.jup.2022.101438 Forbes, K. F., & Zampelli, E. M. (2014). Do day-ahead electricity prices reflect economic fundamentals? Evidence from the california ISO. Energy Journal, 35(3). https://doi.org/10.5547/01956574.35.3.6 Fragkioudaki, A., Marinakis, A., & Cherkaoui, R. (2015). Forecasting price spikes in European dayahead electricity markets using decision trees. International Conference on the European Energy Market, EEM. https://doi.org/10.1109/EEM.2015.7216672 Haben, S., Caudron, J., & Verma, J. (2021). Probabilistic day-ahead wholesale price forecast: A case study in Great Britain. Forecasting, 3(3), 596–632. https://doi.org/10.3390/forecast3030038 Hildmann, M., Ulbig, A., & Andersson, G. (2015). Empirical analysis of the Merit-Order effect and the missing money problem in power markets with high RES shares. IEEE Transactions on Power Systems, 30(3), 1560–1570. https://doi.org/10.1109/TPWRS.2015.2412376 Johansen, S. (1988). Statistical analysis of cointegration vectors. Journal of Economic Dynamics and Control, 12(2–3), 231–254. https://doi.org/10.1016/0165-1889(88)90041-3 Johansen, S., & Juselius, K. (1990). Maximum likelihood estimation and inference on cointegration– with applications to the demand for money. Oxford Bulletin of Economics and Statistics,52(2), 169– 210. https://doi.org/10.1111/j.1468-0084.1990.mp52002003.x Jordan, S., & Philips, A. Q. (2018). Cointegration testing and dynamic simulations of autoregressive distributed lag models. Stata Journal, 18(4), 902–923. https://doi.org/10.1177/1536867x1801800409 Keles, D., Scelle, J., Paraschiv, F., & Fichtner, W. (2016). Extended forecast methods for day-ahead electricity spot prices applying artificial neural networks. Applied Energy, 162, 218–230. https://doi.org/10.1016/j.apenergy.2015.09.087 Koenker, R. (2004). Quantile regression for longitudinal data. Journal of Multivariate Analysis, 91(1), 74–89. https://doi.org/10.1016/j.jmva.2004.05.006 Koenker, R., & Bassett, G. (1978). Regression quantiles. Econometrica, 46(1), 33–50. https://doi.org/10.2307/1913643 Koenker, R., & Hallock, K. F. (2001). Quantile regression. Journal of Economic Perspectives,15(4), 143– 156. https://doi.org/10.1257/jep.15.4.143 Lago, J., Marcjasz, G., De Schutter, B., & Weron, R. (2021). Forecasting day-ahead electricity prices: A review of state-of-the-art algorithms, best practices and an open-access benchmark. Applied Energy, 293, 116983. https://doi.org/10.1016/j.apenergy.2021.116983 Ma, C., Rogers, A. A., Kragt, M. E., Zhang, F., Polyakov, M., Gibson, F., Chalak, M., Pandit, R., & Tapsuwan, S. (2015). Consumers’ willingness to pay for renewable energy: A meta-regression analysis. Resource and Energy Economics, 42, 93–109. https://doi.org/10.1016/j.reseneeco.2015.07.003 Maciejowska, K., Nitka, W., & Weron, T. (2021). Enhancing load, wind and solar generation for dayahead forecasting of electricity prices. Energy Economics, 99, 105273. https://doi.org/10.1016/j.eneco.2021.105273 Menegaki, A. N. (2019). The ARDL method in the energy-growth nexus field; Best implementation strategies.Economies,7(4), 105. https://doi.org/10.3390/economies7040105 Miranian, A., Abdollahzade, M., & Hassani, H. (2013). Day-ahead electricity price analysis and forecasting by singular spectrum analysis. IET Generation, Transmission and Distribution, 7(4), 337– 346. https://doi.org/10.1049/iet-gtd.2012.0263 Mišnić, N., Pejović, B., Jovović, J., Rogić, S., & Đurišić, V. (2022). The economic viability of PV power plant based on a neural network model of electricity prices forecast: A case of a developing market. Energies, 15(17). https://doi.org/10.3390/en15176219 Narayan, P. K., & Smyth, R. (2005). Electricity consumption, employment and real income in Australia evidence from multivariate Granger causality tests. Energy Policy, 33(9), 1109–1116. https://doi.org/10.1016/j.enpol.2003.11.010 244 A. Bâra et al. Understanding electricity price evolution– day-ahead market competitiveness in... Özen, K., & Yıldırım, D. (2021). Application of bagging in day-ahead electricity price forecasting and factor augmentation. Energy Economics, 103, 105573. https://doi.org/10.1016/j.eneco.2021.105573 Pesaran, M. H., & Shin, Y. (1999). An autoregressive distributed lag modelling approach to cointegration analysis. In Econometrics and Economic Theory in the 20th Century: The Ragnar Frisch Centennial Symposium (pp. 371– 413). Cambridge University Press. https://doi.org/10.1017/CCOL0521633230.011 Pesaran, M. H., Shin, Y., & Smith, R. J. (2001). Bounds testing approaches to the analysis of level relationships. Journal of Applied Econometrics, 16(3), 289–326. https://doi.org/10.1002/jae.616 Philips, A. Q. (2018). Have your cake and eat it too? Cointegration and dynamic inference from autoregressive distributed lag models. American Journal of Political Science,62(1), 230–244. https://doi.org/10.1111/ajps.12318 Phillips, P. C. B., & Ouliaris, S. (1990). Asymptotic properties of residual based tests for cointegration. Econometrica, 58(1), 165–193. https://doi.org/10.2307/2938339 Phillips, P. C. B., & Hansen, B. E. (1990). Statistical inference in instrumental variables regression with i(1) processes. Review of Economic Studies, 57(1), 99–125. https://doi.org/10.2307/2297545 Phillips, P. C. B., & Perron, P. (1988). Testing for a unit root in time series regression. Biometrika, 75(2), 335–346. https://doi.org/10.1093/biomet/75.2.335 Raimi, L., Panait, M., Grigorescu, A., & Vasile, V. (2022). Corporate social responsibility in the telecommunication industry-driver of entrepreneurship. Resources, 11(9). https://doi.org/10.3390/resources11090079 Romero, Á., Dorronsoro, J. R., & Díaz, J. (2019). Day-ahead price forecasting for the Spanish electricity market. International Journal of Interactive Multimedia and Artificial Intelligence, 5(4), 42–50. https://doi.org/10.9781/ijimai.2018.04.008 Sandhu, H. S., Fang, L., & Guan, L. (2016). Forecasting day-ahead price spikes for the Ontario electricity market. Electric Power Systems Research, 141, 450–459. https://doi.org/10.1016/j.epsr.2016.08.005 Streimikiene, D., & Kyriakopoulos, G. L. (2023). Energy poverty and low carbon energy transition. Energies, 16(2). https://doi.org/10.3390/en16020610 Streimikiene, D., Kyriakopoulos, G. L., Lekavicius, V., & Siksnelyte-Butkiene, I. (2021). Energy poverty and low carbon just energy transition: Comparative study in Lithuania and Greece. Social Indicators Research, 158, 319–371. https://doi.org/10.1007/s11205-021-02685-9 Streimikiene, D., Lekavičius, V., Baležentis, T., Kyriakopoulos, G. L., & Abrhám, J. (2020). Climate change mitigation policies targeting households and addressing energy poverty in European Union. Energies, 13(13), 3389. https://doi.org/10.3390/en13133389 Tanizaki, H. (1995). Asymptotically exact confidence intervals of cusum and cusumsq tests: A numerical derivation using simulation technique. Communications in Statistics– Simulation and Computation, 24(4), 1019–1036. https://doi.org/10.1080/03610919508813291 Variyam, J. N., Blaylock, J., & Smallwood, D. (2002). Characterizing the distribution of macronutrient intake among U.S. Adults: A quantile regression approach. American Journal of Agricultural Economics, 84(2), 454–466. https://doi.org/10.1111/1467-8276.00310 Wang, B., Yuan, Z., Liu, X., Sun, Y., Zhang, B., & Wang, Z. (2021). Electricity price and habits: Which would affect household electricity consumption? Energy and Buildings, 240, 110888. https://doi.org/10.1016/j.enbuild.2021.110888 Wang, K., Wang, H., & Li, S. (2022). Renewable quantile regression for streaming datasets. KnowledgeBased Systems, 235, 107675. https://doi.org/10.1016/j.knosys.2021.107675 Ziel, F., & Steinert, R. (2016). Electricity price forecasting using sale and purchase curves: The X-Model. Energy Economics, 59, 435–454. https://doi.org/10.1016/j.eneco.2016.08.008