شبکه های مالی و معاملات در بازار اوراق قرضه

We examine how financial networks influence asset prices and trading performance. Consistent with theoretical studies on the role of communication networks in information dissemination, we posit that global financial institutions, having more extensive and strategic financial networks, can more efficiently acquire and process information pertaining to asset trading due to their better access to order flows and, thus, have better trading performance than local financial institutions with less extensive and strategic financial networks. Using transaction level Turkish government bond trading data, we find that global financial institutions exhibit a stronger tendency to trade in more liquid bonds and consistently trade at more favorable prices than local financial institutions, suggesting that global financial institutions have an informational advantage. They also enjoy better trading performance on informed trades but this informational advantage tends to decline over time, indicating possible learning by local financial institutions as a result of trading with their global financial counterparts.

Although it is well established that information moves security prices, how information flows through financial markets and is impounded in the prices of financial assets is not as well understood. Traditional asset pricing models assume that individuals behave anonymously with new information becoming known by all the agents in the market simultaneously, thereby making the information common knowledge. As a result, traditional approaches disregard the possibility that agent behavior (individually and collectively) may be influenced by a communication network. Information, however, can also gradually disseminate in a market by word-of-mouth and observational learning. Because of differences in institutional structures and traders' information processing abilities, it is unlikely that information diffusion will be amorphous. Instead, information is likely to spread more rapidly within trading firms than between trading firms, not only because of the presence of an intra-firm network but also because of financial incentives provided to traders that are related to firm profitability. In this paper we examine the role of communication networks used by financial institutions in their trading activities and focus on whether these networks affect asset prices and trading performances. We define a communication network (hereafter to be referred to as a financial network) to be a set of trading platforms linked together by a system that collects and processes relevant information and then disseminates the information within the financial institution that uses these platforms. We refer to a financial institution with trading platform(s) that access only one market as “local” and one with trading platform(s) that access more than one market as “global.” Consistent with the implications of theoretical studies on the role of networks on information dissemination, we posit that global financial institutions, because of their extensive financial network, can more efficiently acquire and process information closely related to asset trading in global financial markets in which they operate than can local financial institutions with financial network restricted to their home market. Such an advantage may result in global financial institutions pursuing different trading strategies and outperforming local financial institutions in the local market in which they both compete. Models of trading dynamics recognize the role of asymmetric information. The distinction between informed and uninformed traders leads to a number of useful insights. For instance, informed traders tend to respond more quickly to news, to trade in more liquid markets, and to show better performance than uninformed traders. Yet it is not entirely clear who the informed traders are or how they become informed. In this regard, several empirical studies show that individuals who reside or work in the same location tend to make similar financial decisions, which suggests the presence of some type of internal group communication.1 The idea is that traders who are spatially or electronically close are exposed to similar information that is diffused via networks within the same group once the information is received by one or more of the traders. For example, anecdotal evidence suggests that Twitter, a social network, plays a role in assessing the markets in the agricultural commodity sector (Berry and Rees, 2009). Existing research on whether certain types of traders are more informed than others often focuses on whether foreign or domestic traders are more informed using data on equity market trading. The logic favoring domestic traders being more informed than foreign traders is that they may be able to gather more timely and accurate information about the prospects of a company through their formal and informal local networks, be more familiar with local laws and information disclosure policies, and be able to avoid information distortions caused by linguistic and cultural differences. Supporting the position that the foreign investors are more informed is that these investors may be able to exploit their prior investment experience and expertise, as well as their superior (supposedly) knowledge of international business conditions. Foreign investors may also employ locals who are familiar with the domestic market, thereby partially offsetting domestic advantages. The empirical evidence regarding which group is more informed, however, is mixed. Several empirical studies involving a variety of equity markets suggest that domestic investors are more informed than their foreign counterparts (e.g., Dahlquist and Robertsson, 2004 and Lee et al., 2004; Choe, Kho, and Stulz, 2005; Dvorák, 2006) while others report the opposite (e.g., Grinblatt and Keloharju, 2000; Bacmann and Bollinger, 2003; Bailey, Mao, and Sirodom, 2004; Huang and Shiu, 2005). These inconsistent results suggest that extant conclusions are market specific or that the foreign–domestic classification provides only a partial explanation. We address this inconsistency by examining the trading behaviors in government bonds of financial institutions with different financial networks. We choose government bond markets because, as Biais and Green (2005) point out, they typically provide little pre-trading transparency, and, compared to equities, the information driving government bond prices is more likely to be eventually known by the public. Moreover, as documented by Driessen, Melenberg, and Nijman (2003), Barr and Priestley (2004), and Kim, Moshiran, and Wu (2006), among others, bond markets are integrated in the sense that news that emanates in one market affects that market and other bond markets as well, although not necessarily at the same intensity or at the same time. Together these attributes define a venue where information channels are potentially important and better informed traders may be able to exploit their superior information. We conduct an extensive empirical investigation of trading in bond markets populated by financial institutions with local and/or global financial networks. Our local bond market is the Turkish Bonds and Bills Market, which is located in Istanbul and hereafter usually referred to as the Istanbul market. The Istanbul market is suitable for our study for four main reasons. First, this market is well-developed, easily accessible to the global investment community, and operates with limited government interference. Because it is not a financial hub, information affecting short-term movements in bond prices worldwide, such as order flows in the major international bond markets, tends to flow to the Istanbul market and not from it. Second, participants in this market are typically financial institutions such as banks or brokerage firms, either acting on their own behalf or at the behest of others, as opposed to the mixture of financial and non-financial institutions as well as households often found in equity markets. This improves the likelihood that the characteristic that distinguishes the performances of different financial institutions is indeed their financial network structure. Third, the financial institutions that trade in the Istanbul market differ in domicile, asset size, trading volume, and scope of financial network. Financial institutions domiciled in countries other than Turkey trade in numerous markets located outside of Turkey, including those located in major financial centers. A few Turkish financial institutions trade in these markets as well. Finally, from a practical perspective, detailed time-stamped price and volume transaction data, including the identities of the transacting counterparties, are available. Empirical research supports the notion that order flow is an ongoing aggregate measure of the market participants' interpretation of the news that affects the prices of financial assets. Motivated by Kyle's (1985) seminal microstructure model, numerous empirical studies show that order flow explains an economically significant portion of the movement in the market price of a financial asset in the very short run. Examples of this phenomenon on bond pricing include Massa and Simonov (2003), Bessembinder, Maxwell, and Venkataraman (2006), among others. In this paper we document that order flow explains, on average, 43% of bond price changes in the Istanbul market, which is not an atypical percentage when compared to other studies. Hasbrouck (1991) argues that the delayed price impact of a trade measures the informativeness of the trader, and in a foreign exchange rate context, Lyons (1995) and Love and Payne (2008) show that even public information is partially impounded into prices through order flows. Intuitively, participants whose trading platforms are seamlessly connected to multiple markets are aware of the order flow details of these markets more quickly than those who do not have access. The more markets that are included in their trading network and the more actively they trade, the more informed they will be about global events, particularly if the markets are located in leading financial centers. This suggests that sophisticated participants who observe and process the order flows in their own and other markets will, on average, engage in more profitable trading than those who operate only in their own market and may employ different trading strategies. To approximate the information flow that a financial institution collects from trading we create two measures of order flow: one for the local market and one for the global market. Both measures are based on trading volume, although they are constructed differently to reflect the observation that in our dataset all of the participants in the local market are known but are not known in the global market. This is because the global market participants include financial institutions that do not trade in the Istanbul market. Our measure for local order flow for a financial institution is its quarterly market share, i.e., the portion of the Istanbul's total quarterly volume associated with that financial institution. Our global order flow measure is also a quarterly market share metric. For each global financial institution, global order flow is its bond trading volume in the global market, which is defined not to include the Istanbul market, scaled by the corresponding volume of the financial institution that traded the most in the global market. We choose to express order flow in market share and not the level because it is more plausible that market share is the relative amount of information that affects the trading behavior that we are investigating. We consider three different definitions of global order flow. The first defines the global market as all of the markets in which our global financial institutions trade, while the second and the third definitions consider the global market contains only those markets located in and outside the nine leading financial centers as defined by Cassis (2010), respectively. We split the global market into two parts to acknowledge the possibility that the quality and quantity of information emanating from the leading financial centers may be greater than that from the rest of the global market. In doing so, we recognize that there may be agglomeration economies in information gathering and dispersal as pointed out by McGahey, et al. (1990, ch. 2), Tschoegl (2000) and Poon (2003), among others. Our investigation reveals that global financial institutions do indeed perform differently than local financial institutions, after controlling for various measures of size. We find that global financial institutions tend to trade more heavily than their local counterparts in the liquid (active) portion of the bond market. Our finding echoes Chowdhry and Nanda's (1991) notion that informed investors are more likely to trade liquid assets, and it supports the view that global financial institutions enjoy an informational advantage and strategically use more liquid bonds to conceal their superior information. We also find that the average delayed price impact of trades initiated by global financial institutions is consistently larger than that initiated by their local counterparts. To illustrate, for every hypothetical one million Turkish Lira (TL) traded, an average global financial institution earns 1.92 basis points more than if the trade was executed by a typical local financial institution. This finding lends support to our conjecture that financial institutions with a global financial network have an informational advantage because they can consistently buy at a lower (or sell at a higher) price than their local counterparts. These results are robust to limiting our attention to only liquid bonds, where liquidity is defined either by trading activity or Amihud's (2002) liquidity measure. Thus, the preference for trading more liquid bonds by global financial institutions is likely to be at least partially due to an informational advantage rather than only liquidity concerns. Building upon the findings on the informativeness of trades for different financial institutions, we find that global financial institutions earn higher profits on informed trades than their local counterparts. For instance, an average global financial institution earns 1.5 basis points higher trading profit per trading cycle on informed trades than a typical local financial institution for daily complete trading cycles and earns 4.3 basis points more trading profit for informed trades for weekly complete trading cycles. This is consistent with the prediction that financial institutions with more extensive global financial networks have an informational advantage and are likely to perform better than others. We also find that trading profitability exhibits some persistence for all financial institutions, while higher interest rate volatility reduces trading profitability and participation. Finally, we find that price impact of trades by global financial institutions declines over time. Consistent with the decreased price impact of these trades, although global financial institutions still perform better on informed trades in bond trading than their local counterparts, their superior performance declined over the same time period. Using Seru, Shumway, and Stoffman's (2010) approach, we identify significant learning by local financial institutions as a result of gaining trading experience with better-informed global financial institutions. Before delving into the details of our analysis, it is useful two raise two concerns relating to the interpretation of the empirical results. First, because the majority of the global financial institutions in our dataset are not domiciled in Turkey, is it possible that our findings only reflect the previously documented foreign versus domestic dichotomy? To answer this question, we exclude trades involving foreign financial institutions from our analysis. Our findings concerning the impact of information qualitatively remain the same, thereby supporting the notion that our classification of the scope of financial networks (global versus local) does not economically correspond simply to the foreign versus domestic categorization of traders. Second, it is commonly thought that the asset size of a firm and its profitability are positively related. The reasons suggested are varied and include economies of scale (Hall and Weiss, 1967), superior management (Demsetz, 1973), use of strategic groups (Porter, 1979), and lower cost of capital (Meyer, 1967). It has also been suggested that the profitability of trading firms is related to the volume of their trades (e.g., Demsetz, 1968, Warga, 1991 and Schultz, 2001), with volume arguably being another dimension of size in the most general sense. Therefore, could our order flow measures proxy for global financial institutions simply being larger than local ones? Unfortunately, we cannot completely disentangle this relationship because volume and asset size are highly correlated in our sample. Our empirical results concerning local order flow and global order flow, however, suggest that if order flow is a proxy for size, it is a nuanced one. For example, in various tests, global order flow provides significant explanatory power over and above that given by local order flow. In addition, the global order flow definition that contains only trading in the leading financial centers performs significantly better than the version that uses only trading in the markets located outside the leading financial centers. The latter version provides little if any explanatory power. This indicates that information benefits accruing to global trading only pertain to a small set of financial markets. Thus, it is neither the number of financial markets nor their combined trading volume that is economically important; it is the whether the financial markets are important providers of information. We organize the remainder of our paper as follows. Section 2 describes the Turkish government bond market and its trading system. Section 3 defines our classification of traders as global versus local financial institutions, introduces our global order flow measure, and presents summary statistics. Section 4 provides our empirical analyses and discusses their results. Finally, Section 5 contains concluding remarks.