https://orcid.org/0000-0002-5951-4298
![Sample our Engineering & Technology journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5933/TOC-Subject-Sample-2021-Engineering-Tech.jpg"})
ABSTRACT
The European electricity industry has witnessed substantial restructuring over the last decades, aiming to introduce competitive and efficient electricity markets. This study sheds light on the effects of vertical disintegration and privatization on the productivity of European electricity generators. Using a data set of firms, incorporated in 17 European countries, from 2006 to 2013, I estimate the effect of structural reforms, considering their endogeneity, on technical efficiency derived from production instead of cost functions. This study, therefore, fills the academic void of the effects of structural reforms on productivity. Furthermore, I control for weather conditions, being a novelty with respect to the previous literature. The overall results show that vertical separation significantly decreases productivity of electricity generators by 8–14%, depending on the production function chosen. Besides, privatization results in efficiency gains, thereby providing policy lessons on the implementation of structural reforms.
KEYWORDS:
Acknowledgments
I would like to thank Marina Di Giacomo (University of Turin), Carlo Cambini (Polytechnic Turin), Florian Szücs (Vienna University of Economics and Business), Marilin Tilkson (Estonian Competition Authority) and the OECD's PMR division for their support. Additionally, I want to thank Wolfgang Nagl (TH Deggendorf), the participants at the EMEE conference, the seminar participants at the Austrian Institute for Economic Research, especially Agnes Kügler and Kurt Kratena, and an anonymous referee for their comments.
Disclosure statement
No potential conflict of interest was reported by the author.
Notes
1 Orbis considers firms to be ‘large’, when operating revenues ≥10 mio EUR or total assets ≥20 mio EUR or employees ≥ 150. Firms are ‘very large’, when operating revenues ≥100 mio EUR or total assets ≥200 mio EUR or employees ≥1,000 or the company is listed Bureau (Citation2007).
2 While the data set covers electricity generators, incorporated in all European countries, only these countries survive because of missing values in the firm-level data.
3 Observations characterized by negative values of inputs and outputs are dropped.
5 When employing the algorithm by Battese and Coelli Citation1995, the results, however, barely change.
6 This condition holds, if the first differences of the errors are negatively correlated of order one and are uncorrelated of higher orders Roodman (Citation2009b).
8 At this point, I want to thank the OECD's PMR team for sending me the data on the outcomes of the mentioned questionnaires.
9 Although data on the questionnaires' outcomes is only available for the years 1998, 2003, 2008 and 2013, I construct the values for the missing years the following way. First, for the years with available outcomes, I recompute the aggregate employing the OECD's aggregation scheme, as described by Koske et al. Citation2015, and compare it with the value provided in the OECD database. As both aggregates, the one of the OECD and mine, are identical, I check for every country and year whether the OECD's aggregate has changed. If the variable is constant, then the regimes in each segment have not changed as well. If the variable has changed, I compare the relevant value with the outcomes of the previous and subsequent questionnaires to identify the segment driving the change in the aggregate in a given year.
10 ‘No separation’ is employed by the Czech Republic, France, Germany, Luxembourg, the Netherlands, Poland and the Switzerland. ‘Accounting separation’ is applied by Austria, Italy, Slovenia and Spain. Belgium, Estonia, Hungary, Portugal, Slovakia and Sweden use ‘legal separation’.
24 From the published monthly averages of temperature and rainfall, annual averages are computed by taking the means for each year and country.
25 Nevertheless, the results stay the same, when substituting temperature and rainfall with the logged share of hydro power in total electricity generated and treating it as a predetermined variable. Similarly, its coefficient is insignificantly positive, being in line with results by Gugler, Rammerstorfer, and Schmitt Citation2013.
29 As already pointed out, I exclude feed-in tariffs, length of power purchase agreements, shares of manufacturing in GDP, shares of renewables etc. due to multicollinearity, as suggested by the bivariate correlation coefficients, variance inflation factors and correlation matrices of regression coefficients. Besides, some of them and other potential regressors (e.g. patents, distribution losses) can be classified as bad controls in the sense of Angrist and Pischke Citation2009, because they respond to liberalization Erdogdu Citation2011; Marino, Parrotta, and Valletta Citation2019; Nicolli and Vona (Citation2019).
30 The company size is a categorical variable in Orbis, used to filter firms. The considered categories are 'large' and 'very large'.
31 The high correlation coefficients of the estimated firm-level TFP and the industry productivity across the different production function models between 0.94 and 0.99 confirm that the employed methods are robust.