An Empirical Analysis of the Bid-ask Spread in the Continuous Intraday Trading of the German Power Market.

• Author: Balardy, Clara

1. INTRODUCTION

   Liquidity is the major component of a well-functioning market. More liquid is a market, easier it is for a market participant to find a trading counterpart to match its requirements. A typical proxy for liquidity is the bid-ask spread that is the difference between the lowest price for which a seller is willing to sell a megawatt hour of electricity and the highest price that a buyer is willing to pay for it.

   Market participants gain opportunities by exploiting the bid-ask spread which can be interpreted as a premium for immediate execution (Demsetz, 1968). For example, in a setup where the best sell order is at 35[euro] per MWh and the best buy order is at 32[euro] per MWh, if the best buyer (resp. seller) wants to be immediately executed, she has to increase (resp. decrease) her unit price by 3[euro]. It can also be interpreted as an implicit transaction cost; the smaller the bid-ask spread is, the smaller is the implicit transaction cost for the traders and so the end-consumers. Further, the bid-ask spread is a showcase for the quality of the market.

   The German market is the most liquid continuous power market in Europe where about 53% of the German consumption was traded on in 2015. The same year, almost 20 millions of euros were traded each week on the market. It is particularly interesting to study the bid-ask spread of the German continuous power market because the market faces a growing attention in the public debate. First, the continuous market has been playing a growing role in the integration of the renewable energy sources (RES); thus, the traded volume increased by about 170% from 2012 to 2018. It is then a benchmark for the other European countries with a growing renewable capacity. Second, it is also important to understand the liquidity of the market in the context of the European single intraday coupling (SI DC) project where new countries are adopting continuous trading such as Spain or Italy.

   Despite similarities between the continuous spot power market and the traditional financial markets in their mechanisms, there are some major differences due to the physical aspect of power: it is not storable and market participants are balance responsible. (1) In comparison to the financial markets, the power market has a lower liquidity, a higher volatility due to the renewable production, a higher concentration and a highly inelastic demand (Dupuis et al., 2016). Thus, the one-size-fitsall approach is not straightforward and the results of the so-called microstructure literature may not hold in the context of electricity markets. For example, in the financial literature, the pattern of the bid-ask spread ("L-shaped" versus "U-shaped") is explained by the difference in the market making process (Lhabitant and Gregoriou, 2008); this result cannot be transposed to the power market because it does not have market makers. The reason behind the "L-shaped" pattern of the bid-ask spread in the power market can be explained by the urge to trade close to the delivery: there is a peak of activity during the last hours of the trading session.

   This paper is relevant for the market participants in order to decide when to participate in the market during the trading session in order to reduce their implicit transaction costs. It is also relevant for the exchange as a guidance for market design. An example can be the introduction of official market markers (2) in order to increase the liquidity of the market; thus, what would be a decent level of bid-ask spread to ask market markers if the exchange wants to implement some? Marker making is currently implemented in the power future market in Germany. Last, it is useful for regulators in their understanding of the market to propose adequate monitoring tools. For example, in assessing the impact of the concentration of the market on the liquidity.

   From an academic perspective, the contribution of the paper is in threefold. First, the paper assesses the impact of the renewable and the load forecast errors on the liquidity of the market. Beyond the negative and significant impact of the forecast errors on the bid-ask spread, it shows that the market handles the uncertainty of the supply and the demand: it is more liquid after a forecast error. Second, a unique dataset is used and it allows to compute the concentration of the market as well as assessing the impact of it on the liquidity of the market. While the generation part of electricity is highly concentration in Germany (Amanatidis, 2009), I find that the concentration on the continuous market is moderate. Also, while the negative impact is intuitive, this article quantifies it and highlights that the supply concentration have a larger impact on the liquidity in comparison to the demand concentration. Last, to the best of my knowledge, it is the first paper that studies the bid-ask spread of a power market. The complete order book allows me to reconstitute the best order streams (best bid. best ask, and market depths) each time a new event occurs in the power market (i.e., new/modification/cancellation of an order in the order book). The model could be easily extended to other continuous markets.

   In this article, I first do a dynamic analysis of the bid-ask spread and the market depths over an average trading session at a granular level (microseconds). The market depth is the volume available in the order book. It can be divided into the buy depth and the sell depth. They respectively are the total volume available on the buy side and on the sell side at one moment of the trading session. For example, in the setup proposed in figure 1, the sell depth is equal to 40 MW and the buy depth is equal to 65 MW. Second, I identify the main drivers of the bid-ask spread: the risk, the adjustments' need, the activity and the concentration of the market.

   The study yields three main findings. First, it shows the "L-shaped" pattern of the bid-ask spread during a trading session of a power market. Second, I find an average bid-ask spread of 3[euro] per MWh over the trading session of the German power continuous market. Third, I identify four components of the spread: the risk, the adjustments' need, the activity, and the competition in the market. Using a fixed effect model, I find a positive relation between the risk and the bid-ask spread as well as a negative relation between the bid-ask spread and the adjustments' need, the activity, and the competition in the market.

   The paper is organized as follows: the second section is dedicated to the relevant literature. the third one is an overview of the current spot power market in Germany, the fourth section gives some statistical insights on the bid-ask spread and the market depth in the German intraday power market. The fifth part presents the data and the methodology used. Then, the sixth section displays the empirical results. The last section is the conclusion.

2. RELEVANT LITERATURE

   The present paper straddles two streams of literature: the one related to the electricity markets and the one on market microstructure.

   While the literature on the power markets is dense, the literature on the continuous ones is limited and mainly focuses on two issues: wind generation integration (how to handle forecast errors) and market design. The closest literature to this paper is the one on price formation in the intraday continuous market. Hagemann (2015), Hagemann et al. (2016), Karanfil and Li (2017) and Ziel (2017) model the price of this market. Weber (2010) is the first to address the question of the liquidity of the German continuous power market. He finds that the low liquidity might be the cause of a poor market design and/or the absence of a real need for a continuous market. However, those comments have to be balanced as the paper uses a dataset from 2007 when the yearly volume traded on the continuous market was 1.4 TWh--almost 26 times less than the volume traded in 2015. Also, the level of installed wind capacity more than doubled from 2007 to 2015; it increased the need to adjust the renewable generation's position close to the delivery time in order to be balanced at delivery. Chaves-Avila et al. (2013) explain the low liquidity in the continuous power market as the preference of producers to commit their generation long ahead of time because of ramping-up costs and generation planning. Hagemann and Weber (2013) develop two models to explain the liquidity of the German continuous power market. To the best of my knowledge, the work of Hagemann and Weber (2013) is the first paper that investigates the bid-ask spread in the German continuous power market. However, their work neither uses the order book sent by the market participants or a reconstitution of it as input data for their models: they estimate the bid-ask spread using the transactions data. NeuhofTet al. (2016) study the impact of an intraday auction before the opening of the continuous market. They find a negative relation between volatility and the market depths as well as a positive relation between the liquidity and the market depths of the 15-minute intraday auction in Germany.

   The microstructure can be defined as the branch of the finance that deals with the traders' behavior and the market design. The study of the bid-ask spread is part of the microstructure literature, particularly of the sub-literature on price formation and price discovery.

   Demsetz (1968) initiated the literature on the bid-ask spread. He defines market makers (3) as immediacy providers in which the bid-ask spread is a premium paid by a market participant for immediate execution. The work of Demsetz highlights the negative relation between the volume and the bid-ask spread. It is also raised in the paper of Copeland and Galai (1983) who model the bid-ask spread using the volatility and the level of trading as explanatory variables.

   In the theoretical part of the literature, the bid-ask spread reflects three...