Could Market Making be Profitable in The European Carbon Market?

• Author: Galariotis, Emilios

1. INTRODUCTION

   Investors, every time they enter the market, need to make a decision about the type of order, i.e. limit order versus market order, that better serves their economic interests. When they submit a market order, they face no execution risk, because their order is executed immediately at the opposite side quote. This means, though, that they are exposed to price risk, since their trades might be executed at a price that is not the most favourable, especially when market depth is shallow (e.g. Biais et al., 1995). In contrast, if they submit a limit order, they reduce price risk, because they declare the maximum (minimum) price they are willing to buy (sell) the asset for, but they cannot be sure whether their order will be executed, since prices might move away from their quote.

   Previous literature (e.g. Handa and Schwartz, 1996; Handa et al., 1998) recognizes that market conditions affect the suitability of each order type and suggests that price formation and its impact on volatility and spreads determines the relative merits of each order type to gain economic significance. Intraday price formation is driven by liquidity and information (O'Hara, 1995). Prices are affected temporarily by increasing demand on one side of the spread and therefore, market makers increase the price of liquidity in order to deal with persistent order imbalances (Madhavan, 2000). This would increase the price of liquidity and thus, spreads, which would increase investors' preference for limit orders (e.g. Chung et al., 1999), so they can charge for liquidity and earn the spread. In the opposite case, when the price for liquidity is low and spreads are narrower, investors seem to prefer to consume it and therefore, they reduce the use of limit orders (Biais et al., 1995).

   In parallel, the arrival of information affects prices permanently. Market makers, in fear of trading with better informed agents, charge a fee associated with adverse selection (Kyle, 1985; Glosten and Milgrom, 1985), but they generally lose when they trade with informed agents and recover their losses when they trade with uninformed agents. Limit orders would be rather inappropriate in a market setting where a higher presence of informed agents is expected, because they have option features. Copeland and Galai (1983) suggest that a buy (sell) limit order is like a put (call) option because if prices move against the trader who submitted the limit order, the order will be executed, most frequently at the trader's losses, especially if the order is mispriced (Bae et al., 2003). If prices move in the opposite direction, the order will not be executed. This is clearly an unfavourable condition for limit orders and, although market orders (which face a higher price risk (Peterson and Sirri, 2002)) should be less preferable upon higher volatility (e.g., Foucault, 1999), the literature makes a clear distinction between liquidity (transitory) and information-driven volatility. Higher transitory volatility seems to attract more limit orders (e.g. Handa and Schwartz, 1996; Handa et al., 1998), while limit orders upon a high information component of volatility seem to be associated with informed agents whose information advantage decays slowly (Keim and Madhavan, 1995).

   Consequently, the literature recognizes that order type suitability is not a static concept and it depends on changing market conditions. Empirical market microstructure literature (e.g., Cohen et al., 1981; Chakravarty and Holden, 1995; Harris and Hasbrouck, 1996) elaborates on this and introduces additional factors, e.g. trade initiation and order size, that might affect the performance of limit orders. All empirical studies use order by order data, i.e. the submission of all limit and market orders, and approach the issue from a fully empirical perspective, by counting the frequency or by measuring the relative costs associated with different order types across different levels of price change volatility, implied spread, order size and direction (buy or sell). This approach, although it uses all the information available (all orders), it is rather agnostic and descriptive in its nature. It derives some empirical conclusions based on unconditional statistics of limit orders, and therefore, it is not able to provide a trading rule, conditional on some (model) predictions.

   This is the primary concern of our study, which focuses on a transaction-by-transaction - rather than on an order-by-order - basis, trying to provide an order-type selection rule based solely on transaction level data (e.g. Kalaitzoglou and Ibrahim, 2017). Transactions, as opposed to limit orders, are realized commitments to trade and therefore they convey vital information with regards to realized and expected price changes (Hasbrouck, 1991). We suggest using this information to condition the selection of a suitable order type, based on informed agents' identification and the estimation of time-variant price components, extracted solely by observable transaction data. First, we recognize that the presence of better informed agents might deter investors from submitting limit orders. We classify market transactions into three categories according to their link to private information, namely uninformed, fundamental and informed agents, based on the deviations of the arrival rate of their trading from (a data-driven estimate of) normal levels. Second, we also recognize the impact of liquidity and information price components of intraday formation and their impact on spreads, variance and ultimately, on the selection of order type. Our major difference from previous literature is that we employ an intraday pricing model that estimates time-variant price and variance components based solely on transaction level data. Using these estimates we then develop an order-type selection rule based on the theoretical and empirical propositions of previous literature. Finally, we test the relevance of our conditional predictions by testing their economic performance. Our approach is relevant to investors who want to condition their orders on expected market conditions.

   The second contribution of our study refers to the investigation of the suitability of the order-type selection rule we develop in a market with an increasing importance for the global emission reduction targets (e.g. MacCracken et al., 1999; Klepper and Peterson, 2006), where no prior study exists. We focus our interest in the European Carbon Futures market and especially its most liquid venue; the European Climate Exchange (ECX) in London. This market has undergone a significant development in terms of overall liquidity and maturity (Kalaitzoglou and Ibrahim, 2013a) and previous literature (e.g. Benz and Hengelbrock, 1998) reports that spreads and price-change volatility, identified as the main determinants of the suitability of different order types, exhibit persistent patterns and considerable predictability, mainly due to the cap-and-trade system (e.g. Daskalakis et al., 2009), regulatory announcements (Mansanet-Bataller and Pardo, 2009) and the relative illiquidity (Mizrach and Otsubo, 2014). This predictability could increase execution certainty, which would be beneficial for limit orders, but could also reduce price risk, which would be beneficial for market orders. In general, increased predictability would provide a better estimate of the presence of private information and of price components and, therefore, could render the conditional selection of order type more profitable. All these would lead to a more mature price discovery, which would assist the market in achieving emission reduction targets by assigning a more "faire" price on Carbon; a rather complex task due to the uncertainty about the fundamentals of Carbon emissions (e.g. Manne and Richels, 1991; Alberola and Chevallier, 2009).

   The order selection is an intrinsic issue of intraday trading and the development of a trading rule requires a pricing model with time-variant components. Previous literature in the Carbon Market recognizes liquidity and order fow as the main drivers of intraday price formation. Several studies (e.g. Bredin et al., 2014; Mizrach and Otsubo, 2014) trading intensity, i.e. duration and/or transaction size, as a major element of intraday price formation. They report an increasing price impact of higher trading intensity and, therefore, a stronger predictability of returns. Along the same lines, Kalaitzoglou and Ibrahim (2013b, 2015) report that higher trading intensity is associated with a higher presence of information and, therefore, it induces price uncertainty and thus, trading patterns. These patterns are also observed in order-fow dependency (e.g. Ibikunle et al., 2013; Medina et al., 2014) and higher trading intensity is also found to increase (decrease) the information (liquidity) price and volatility component (Ibrahim and Kalaitzoglou, 2016).

   Following previous literature, we consider trading intensity as a major driver of intraday price formation in the European Carbon Futures market and we use it to identify different agent types, to extract the price components and to generate some conditional predictions for the suitability of the different order types. (1) The empirical findings identify an increasing impact of trading intensity on the information price component due to a stronger link to information, but a decreasing impact on the liquidity component, due to a higher execution probability/lower inventory risk. Consequently, the overall impact on price formation, and subsequently on the profitability of the different order types, depends on the relative magnitude of each effect. When trading intensity is expected to be low, the information price component is also low, but the liquidity component is at its highest level. We find that these are the market conditions that market makers, who prioritize liquidity gains and not execution risk, would prefer...