This action might not be possible to undo. Are you sure you want to continue?
Jan Simona, Yuval Millo*b, Neil Kellardc, Ofer Engeld
Abstract Financial risk management focuses on the potential outcomes of investment decisions, but pays less attention to the social and organizational processes that underpin the decision making process. This paper is the first to analyze hedge funds’ decision making in detail, examine the risks inherent to these practices and reveal their contribution to the emergence of a market crisis: the Volkswagen-Porsche crisis of October 2008. We collect and triangulate data from interviews and field observations in addition to mapping and analyzing social networks. We conducted interviews with 60 hedge fund managers and brokers and held field observations over a period of 18 months, between December 2007 and June 2009. We investigated 26 hedge funds and 8 brokerage firms in Europe, the United States and Asia. The hedge funds analyzed controlled 15% of all assets managed by hedge funds. We find that decision making in hedge funds relies crucially on an elaborate two-tiered structure of connections. First, competing hedge fund managers share and discuss among themselves detailed, timely and sensitive information on a regular basis. Second, hedge funds and brokers maintain communicative connections, but unlike the hedge fund-to-hedge fund connections, only general information and broad investment ideas are exchanged. Following an examination of these decision-making practices, we focus on the events of October 2008. Our findings indicate that the Volkswagen-Porsche crisis is tightly related to the social structure and the sets of practices through which investment decisions are made. We show that the nature of connections among hedge fund managers, and between them and their brokers, contributed to a situation whereby, once hedge funds collectively accepted an investment idea and invested according to it, they ‘locked in’ on it, ignoring warning signs. These findings have implications for our theoretical understanding on how financial risk events emerge as well as for financial risk managers.
We thank Daniel Beunza, Bino Catasús, Tony Davila, Sarah Hall, Kari Lukka, Gustav Johed, Emmanuel Lazega, Vincent Lepinay, Andrew Leyshon , Andrea Mennicken, Anette Mikes, Martha Poon, Mike Power, Hanna Silvola, Wim Van der Stede, Hendrik Vollmer, Klaus Weber and participants in seminars at the University of Nottingham, SCORE (University of Stockholm), the Social Studies of Finance Annual Conference, the American Sociological Association and the University of Turku for their helpful comments. Any remaining mistakes are ours. This working paper is distributed for purposes of comment and discussion only. It may not be reproduced without permission of the corresponding author. a IESE Business School * b London School of Economics, Department of Accounting (corresponding author) c Essex Business School d London School of Economics, Department of Management
Before the financial crisis, hedge funds played a central role in many financial markets. According to Greenwich Associates, a prime financial services research engine, in 2007, 30% of U.S. fixed income, 20% of global foreign exchange, 95% of distressed debt, 61% of high-yield credit derivatives, 60% of structured credit and 55% of leveraged loans volume was traded by hedge funds. However, even in spite of a sharp decline during 2009, in 2010 hedge funds still ‘remain key players in U.S. fixed-income markets’ (Greenwich Associates, 2010). 2 Hedge funds have also been associated with important market events, such as the devaluation of the British Pound on September 16th, 1992 and the subsequent withdrawal of Britain from the European Exchange Rate Mechanism or the record levels of volatility in August and September 1998, which were related to the collapse of the hedge fund Long Term Capital Management (Booth, 1998; President’s Working Group on Financial Markets 1999). In spite of their centrality, hedge funds are among the most secretive and least accessible financial institutions and, as a result, very little is known about the practices, conventions and organizational modes of operation through which they make their investment decisions. Motivated by the lack of knowledge in the academic literature on this influential market actor, we investigate the practices, conventions and organizational modes of operation through which hedge funds develop and make their investment decisions. Between December 2007 and June 2009, we interviewed 60 hedge fund managers, brokers, analysts and traders from 26 hedge funds and 8 brokerage firms in Europe, the United States and Asia and conducted fieldwork in a ten of
Data for 2007 is from http://www.greenwich.com/.
3 these hedge funds and brokerage firms. We find that a crucial part of the hedge funds’ decision making process relies on initial information ‘raw’ investment ideas that hedge fund managers receive from brokers and on interpretative analysis of this information, which takes part among hedge fund managers. These organizational practices, in turn, are preformed through a network of social connections among hedge funds and between them and brokers. The practices and their accompanying structure of connections are an inherent part of the way hedge funds operate and have a crucial affect the strengths and vulnerabilities of their investments. The risks embedded in the way hedge fund make their decision were illustrated vividly in a market crisis that took place during our data collection period, the Volkswagen-Porsche crisis of October 2008. To analyze the practices and norms that underpin hedge fund’s decision-making and its related risks, we develop a theoretical framework that ties the inter-organizational activity to theories of risk emergence in other organizations and to relevant concepts in economic sociology (section 2). Following a methods section (section 3), which discusses our use of qualitative and quantitative methods, we examine the connections and the practices of information sharing and decision making among hedge funds and brokerage firms under normal conditions (section 4). In this section, we also examine how the nature and quality of information shared by the actors are affected by the technological intermediaries used. Then, in section 5, we corroborate the qualitative empirical findings by constructing a map of the connections, calculating relevant measures and then test the statistical significance of the relation between the observed network and the actors’ and the institutions’ attributes and their network-structural positions. Following this analysis of decision making in hedge funds, we focus on specific phenomena directly related to the crisis. First, we analyze the ‘consensus trade’ phenomenon – the popular trading position (Section 6) and then we look at the VWPorsche trade, one of the popular consensus trades and study in detail the decision making in
Section 8 concludes the paper. along with another 20% of VW stock that were held by the government of the Federal German state of Niedersachsen (Lower Saxony). sell their Porsche holding and make a profit. Borrowers. 13% of VW stock was borrowed and Porsche’s announcement.1%) of the stock of another car marker: Volkswagen (VW). which meant that maximum 5% of the stock was available for trading. sent its price to record heights of more than 1. who rushed to buy VW stock and cover their obligations. deliver it to the lender.000 Euros per share. motivated lenders to ask for the stock to be returned to them immediately (a step known as a ‘short squeeze’). If the price difference would decrease. they intended to buy back VW stock. This large holding. By late October 2008. sold the assets.4 hedge funds regarding this position in the weeks leading to the crisis of October 2008 and its aftermath (Section 7). The realization that such a large part of VW stocks were ‘out of the market’ had dramatic implications for investors who had ‘short sold’ VW stock. The statement revealed that Porsche had a much larger shareholding in VW than investors and analysts had estimated previously. These investors expected the price difference between VW and Porsche’s stocks to decrease and to utilize this situation. meant that about 95% of VW stock were not freely traded. carmaker Porsche announced that it had control of nearly three quarters (74. Theoretical framework On October 26. had borrowed VW stock. more than 6 times its average price in the 12-months period before the crisis . 2. and simultaneously bought Porsche stock. 2008.
L. . which made it into a Black Swan event. Furthermore. (Bloomberg. 2008 came as a surprise. the German federal government amended its securities laws making the stealth accumulation of stocks more difficult. Our theoretical framework for understanding the VW-Porsche crisis stems from the fact that our main goal was to describe and analyze the way hedge funds make their investment decisions and the VW-Porsche crisis of October 2008 took place when we were about to finish our observations and interviews. According to the claim. Nonetheless. Porsche AG. An equally crucial factor contributing to the impact of the announcement was the fact that hedge funds borrowed (‘sold short’) VW stocks to the tune of billions of Euros. More recently. and others vs.P. January 2010: 2). 28 October 2008). These market movements resulted in heavy losses to investors who had to meet their lending obligations. This explanation is also the rationale behind a lawsuit. low-probability events – and is therefore vulnerable to such surprises. when we examined our findings from the 16-month period leading to the crisis. where the car maker was blamed of breaking American law by falsely denying its true intentions and by covering its tracks when building its large shareholding in VW stock (Elliot Associates. referring to the main thesis in Nassim Taleb’s book of the same name (2007) that argues that financial risk management is not prepared for ‘surprises’ – high-impact. The media classified the crisis largely as a ‘Black Swan event’. It is true that Porsche’s announcement on October 26. but this announcement was not the sole reason for the crisis. brought by 18 hedge funds against Porsche and its CEO and VP of Finance in January 2010. we could not agree with the conclusion that the crisis was simply an outcome of hedge funds being taken by surprise.5 and more than 4 times the price the day before the announcement. our findings indicate that hedge fund managers knew that the price discrepancy between VW and Porsche stocks lasted. was the fact that Porsche accumulated so much of VW stock and did so unnoticeably. the risky surprise at the core of the VW-Porsche crisis.
If this is so. italics added). effectively. Put differently. into a ‘consensus trade’ among hedge funds. our material showed that the routine practices hedge funds used. This realization made us look beyond the immediate (and widely accepted) Black Swan explanation and search for a theory that explains crises not as a stark break from normal organizational and inter-organizational activities but as a development of these activities that leads actors to assuming increasing degrees of risk. Instead. then how can we identify dynamics that lead to the emergence of risk in organizations? . adverse. are the breeding ground of both positive and negative outcomes of organizational norms and practices. and of high social impact’ (Vaughan. but continued to invest in the trade. was the factor that made them vulnerable to Porsche’s accumulation of VW’s stock. turning it. but developed gradually while the hedge funds. ignoring the risks that their own practices generated. Organizations. with outcomes that either in the fact of their occurrence or consequences are unexpected. organizational deviance is not different in the ways it emerges from normative and beneficial organizational practices. frame and enable the professional socialization that constitutes the actors who operate in them and. especially the ways they communicated with brokers and with other hedge funds. The initial impression from our retrospective examination of findings was that there were no actors who blatantly took irresponsible risks or exhibited exuberant greed and thus led to the crisis. hence. 1999:293. learned to ‘live dangerously’. a process that results eventually in a crisis. that vulnerability did not emerge overnight. Moreover. for many months and yet they did not regard this as a warning sign.6 unexplainably. our analysis indicated. according to this conceptualization. This description corresponds directly with Diane Vaughan’s concept of organizational deviance (1999) that refers to ‘organizational-technical failures that include acts of omission or commission by individuals or groups of individuals acting in their organization roles.
Vaughan shows. Uzzi. argues that clues for the emergence of risky practices can be found in the ways the organization communicates. in contrast. Under such a condition. leading to relative ignorance and to the possible emergence of deviant organizational practices. Hirschman (1970). and Lawrence et. These two legs are (1) the connections between hedge funds and brokers. The VW-Porsche crisis. Embedded ties are used for the transfer of complex knowledge (Hansen. who studied the disaster of the Challenger space shuttle (1996). timely information is exchanged. was developed for the analysis of a single organization and its connections. In contrast. 1999. both internally and externally. Uzzi (1999) and Hansen (1999) show that straightforward. flow of information and knowledge between different subunits is limited. The structural secrecy concept. That is. Alluding to the paths of information exchanged. we refer to these sets of practices as legs. 1997). however. if structural secrecy did develop in this case. then it evolved predominantly through inter-organizational connections rather than internal ones. al. where typically only general.7 Vaughan. Vaughan analyses the communication within NASA and shows that a condition of ‘structural secrecy’ developed in the organization. where a great degree of involvement (cognitive-professional and emotional) is invested and sensitive. public and unambiguous information is most efficiently exchanged and distributed in an environment rich in arm’s length ties. tacit knowledge and proprietary expertise (Larson. How can inter-organizational connections lead to structural secrecy? We identify two sets of practices and conventions that together contribute to structural secrecy and to the emergence of vulnerability to financial risk among hedge funds. 2005). This configuration of connections correlates with the distinction that structural economic sociology makes between two types of inter-organizational or interpersonal ties: embedded and arm’s length ties. evolved through a dynamic that had no single organizational focal point. This line of . 1992. descriptive information is shared and (2) the connections among hedge fund managers.
The networked nature of risk emergence that is implied in the concepts of structural secrecy and over-embeddedness is complemented by examining.8 work also proposes that clusters where there exists a high density of connections among the actors may develop over-embeddedness. an inadvertent result whereby risk is amplified. a computerized risk assessment model. Hardie and MacKenzie describe an environment rich in material artifacts that take an active part in the evaluation of information and arriving at an investment decision. prices. where risk emerge from the structure and nature connections among actors. in detail. who deduct from the model’s results how competitors behave. . a situation where the actors circulate among themselves a limited set of ideas and thus become effectively insular from developments in other pars of the network. played in increasing the similarity between the trading positions of different hedge funds. who show how a mathematical model serves as common point of reference for hedge funds. This model-mediated. they infer the existence and effectiveness of such inter-organizational networks on the basis of data collected. effectively. in both research cases. MacKenzie (2003) also studied the technological inter-organizational connections among hedge funds. Examining the case of the hedge fund Long Term Capital Management that collapsed in the summer of 1998. Similar picture is presented by Beunza and Stark (2010). a situation that exacerbated the risk and precipitated the crisis. indirect communication between market actors brings about. Whilst both MacKenzie and Beunza and Stark describe networked phenomena. Hardie and MacKenzie (2007a). MacKenzie explored the role that a technological artifact. information about what major categories are doing. and so on’. news. through which flow research reports. who conducted fieldwork at a London hedge fund argue that a hedge fund ‘is part of a rich network of inter-personal and inter-organizational connections. the techno-social environments where investment decisions are being made.
the models are only one of many components in a wider valuation process. for producing investment decisions. evaluation and. relying heavily on the social influences of mathematical models limits the relevance of the research. ultimately. Event-driven hedge funds choose their targets of investment based on the announcement and materialization of certain events (e. a network that was used for obtaining information and for its analysis. while these papers represent pioneering and groundbreaking work. they cultivated a dense network of interpersonal connections. Therefore. In fact. This is true especially when we examine hedge funds. combined. typically taking a long position (buying and holding) in one asset and a short position (borrowing and selling) in another.3% of all assets under . relies on sophisticated mathematical models when making investment decisions. None of the hedge funds we studied (all of which were leading institutions). the models they use are relatively straightforward valuation models. represent the biggest single group of strategies in the hedge funds world (38.9 within a single organization. Instead. In turn. 3.g. the hedge funds we observed do not use models as techno-social ‘mirrors’ that help them to discover what competitors or the market as whole do. a merger/acquisition or an asset sale after bankruptcy procedures). Moreover. Methods We focus mostly on the families of trading strategies known as ‘long-short’ and ‘event-driven’. Our choice is motivated by the fact that these strategies. Long-short hedge funds invest by taking positions in different groups of assets. this empirical limitation motivated this research strand to focus primarily on the technological devices (the mathematical models) and treat them as ‘super-nodes’ that connect all other market actors. of the type taught commonly in MBA programs. Crucially.
Organizations such as mutual funds or pension funds rarely. So I can be the 10th or the 50th one in a trade and still make some decent money.4 In other words. 4 3 .10 management)3 and that both strategies typify elements that distinguish hedge funds from most other investment vehicles: their ability to go short and their focus on arbitrage-like opportunities. Actually we are a minority. Hedge fund managers. 31. most of them can not go short. selling them in the market and returning the borrowed assets at a later date. A representative example comes from HF25. Also. hold a ‘short’ position. analysts and traders) and 24 representatives of the brokerage side (Appendix Table 1). the asset will be there in abundance. We interviewed 36 hedge fund professionals (managers. Another general characteristic of the long-short trading strategy is important for our understating of how hedge fund managers see each other and how they regard other investors. field observations and social network analysis in the research on hedge funds.5 We conducted interviews in As of Dec.. if ever. tend to be certain that when they short an asset. This research is also the first of its kind in terms of global reach and scope of coverage. What percentage of total assets is held by hedge funds? 2%? [M]uch of the price discovery in those stocks is done by traditional mutual funds and pension funds who look at it totally differently than we do. 5 As of Dec. 31st. The hedge funds in our dataset manage 15% of global hedge funds’ assets under management. For the same reason. when the hedge fund managers need to buy back the asset and return it to the lender. their position is reciprocated many times over by long positions held by institutional investors. Barclays Hedge data. is a technique that virtually separates hedge funds from other financial organizations of similar magnitude. Our paper is the first to triangulate interviews. 2007. a hedge fund manager in the long-short strategy: You see although I am competing against them for investors’ money. Short selling. which consists of borrowing assets. 2007. hedge fund managers are less reluctant to share investment ideas that include a short position with other hedge fund managers: the assets themselves are not a scarce resource. being aware that they are a minority in the financial world. we are not the only type of market participants. We conducted 60 interviews between December 2007 and June 2009.
second opinion or selective contribution’. we conducted observatory fieldwork at eight hedge funds and two brokerage houses servicing hedge funds. 2) observing what information is shared and 3) triangulation of questioning. All interviews were taped and transcribed. we analyzed connections between hedge funds and between them and brokers and constructed a social network. it had to be confirmed independently by both parties. while for brokers it was defined as: ‘hedge fund managers: a) with whom consider to have a good relationship. the relationship was not taken into account. 6 . and b) would belong to your top 20 clients or top client list. At our request. In addition to the interviews. at most sites a rotation system6 was organized and some informal ‘debriefing’ sessions were held outsides the offices of the hedge fund or brokerage firm (often held at coffee shops or at a local bar/pub) to follow up issues that raised during the observations. The observations were held typically in blocks of two to five days and. idea sharing.11 New York. were repeated at different times. where possible. Geneva. or indirectly through e. Thus. Hong-Kong. To construct the network we asked our informants to give us the names of the people with whom they have relevant professional interactions. we incorporated brokers and hedge fund managers that belonged to the Long-Short and the Risk Arbitrage strategies and had at least USD 5 billion under management. Madrid and a fourth European city that cannot be identified because of anonymity considerations. if you would have one. 7 Relevant for hedge fund managers was defined as: ‘have influence on the investment decision. The purpose of this is three–fold: 1) understanding how the different functions connect. 25 confirmed independently of their relationship and also agreed to provide detailed information about their past employment and their personal connections. Of these actors.g.g. 7 For each dyadic relationship to be taken into account. five worked on the brokerage side (all with intense Rotations consisted of spending between a half a day and two days with different professionals at a same firm. be it directly through e. Following our qualitative data gathering. and were conducted on the basis of strict anonymity. For our network analysis. Of the 60 people we interviewed and observed. if informant A told us they have a relevant professional relationship with B but B did not mention A. London.
the hedge fund manager makes the final decision on the composition of the fund’s portfolio of 8 We identified only one broker-broker connection in our network. to their connections with hedge fund managers.g. or in other ways (e. Connections and communicative practices between hedge funds and brokers and among hedge funds 4.8 We constructed the questions aimed at identifying the network of connections on the basis of the practices we witnessed during the observations and interviews.1. In the case of hedge fund managers. Connections between hedge funds and brokers To analyze the practices and conventions that perform the structural secrecy among hedge funds we first need to discuss the main actors. in effect. hedge fund managers are partners to the initial capital collected during the set up of the fund and they are frequently also the founders. and (b) who would belong to your top 20 clients or top client list. either in the form of idea sharing. Typically. the question referred to: ‘hedge fund manager with whom (a) you have a good relationship. as was also indicated in our qualitative data. This role centrality is reflected in the decision making process.12 contacts with hedge funds. 9 . Hedge fund managers are the most central functionaries in the hedge funds we studied. four in London and one in the European city) and 20 in hedge funds (14 in London. Hence. emotional support)’. if you have one’. the question referred to ‘influence on the investment decision. one in Geneva and two in the European city).9 4. We constructed the network by asking all respondents to name others with which they have relevant professional interactions. offering interpretations. three in New York. the questions to brokers referred. Almost without exceptions. In the case of brokers.
Hedge fund managers are often assisted by analysts10 (i. the ones belonging to hedge funds are known as ‘buy-side’. which assets to buy and which to borrow and sell).13 holding (e. the brokerage firm’s salesperson is the immediate contact person for the hedge funds and our informants frequently referred to these salespersons as ‘brokers’ (we use both terms).g. The major task of analysts is to develop investment ideas through the assessment of the countries. A salesperson would normally provide the hedge fund with investment ideas and may also be involved in organizing meetings between hedge fund managers and executives from companies or institutional investors (this area of activity is known commonly as ‘corporate access’). the timing of the execution. each.e.. To distinguish between analysts and traders at the brokerage side. Brokerage firms assist hedge funds with their investment decisions. Most commonly. perform the executions of the trades as well as provide operational support for trade executions and they may also provide hedge funds with additional capital with which they can leverage their positions or provide liquidity by buying or selling the requested assets using the brokerage firm’s own account. between one and four analysts assisting them. That is. the information provided by the trader does not tend to change the hedge fund manager’s view on the valuation of the opportunities in a security. which are. typically. The last function holder we typically encounter in hedge funds is the trader. for example. similar to the ones in hedge funds are The hedge fund managers we observed and interviewed had. industries. sectors or companies on which they focus. who executes the trading orders of the hedge fund manager. while their counterparts at the brokers are referred to as ‘sell-side’. ‘buy-side analysts’11). but only influences. The brokerage firm’s analyst produces research reports with recommendations and occasionally also meets with hedge fund managers to elaborate their views. sub-units in an investment bank or a bank holding. While the analysts and the hedge fund managers tend to take a long-term investment horizon. traders in brokerage firms. Finally. Hedge funds interact frequently with brokerage firms. the trader in the fund focuses typically on the short term. 11 10 .
Our observations indicate that hedge fund managers or hedge fund analysts seeking flow information was the single most frequent type of phone calls or emails that brokers received from hedge funds. Flow information. a hedge fund manager we observed. is context-specific information about the conditions surrounding a possible investment action. by the brokers. called a local broker who had ‘a good understanding of the intentions of major holders in the stock’. by brokers conducting some investigation and returning to the hedge fund managers with specific details and notes about the ‘atmosphere in the market’. how big certain orders are and many other factors. When evaluating the merits of a possible investment idea. What information and ideas do hedge funds ‘buy’ from the brokers? Our observations show that the bulk of the communication between hedge funds managers and brokers revolves around the transmission. the Spanish multinational company whose stock is traded in the Bolsa de Madrid (the Madrid Stock Exchange). of ‘flow-information’. For example. and b) connections among hedge fund managers. or ‘market color’ or ‘flow color’. We observed two main types of connections that hedge fund managers maintain: a) connections between brokers and hedge fund managers. as it is commonly known. for example.14 responsible for the actual execution of trading orders in the markets on behalf of the brokerage firm’s clients. who was developing an investment idea that included buying Telefónica stock. Such requests for information were followed up. answers questions such as whether there are more buyers than sellers for certain assets. the type of institutions that are interested in buying or selling. We find that hedge fund managers and their analysts rely heavily on investment ideas and support from outside their hedge funds and on a network of connections through which this information is communicated. The . Flow information. typically. hedge fund managers frequently called brokers asking for flow information.
the intention of this question. the notion of ‘smart’ incorporates important meanings in the hedge fund world. hedge funds (‘fast money’/’smart money’) and central banks. it can be safely expected to hold the assets for a prolonged period.15 broker. These hedge funds want to get an idea who is active. which are likely to sell it within a short period. for instance. In this particular context. have a keen interest in finding out about the types of investors involved in the market. up to the minute information about the activity in the Telefónica stock. Is it the real money. as the salesperson explained to us. is: ‘Are they smart?’ As we will see later. that is. in the words of another hedge fund manager is information ‘not found on the tape’. whom we interviewed four days later. BR10. unlike some hedge funds (fast money). Let us look at flow information in more detail and characterize the types of information exchanged. a distinction is made between traditional asset management (‘real money’). corporations. is it corporate flows or central banks? That kind of information is clearly sought after from the hedge fund community. contacted his Madrid connections and provided to the hedge fund manager an assessment of the expected flows as well as a detailed. is to assess . They want to get a flavour of what the real money [traditional asset manager] or other hedge funds are doing as well as what is driving the price action. for example. not included in the price and volume information. when the hedge fund managers hear that the buyer/seller is a hedge fund. explains: What is absolutely crucial is what is called flow information. A typical follow-up question to this line of enquiry about active market actors. a senior salesperson to hedge funds in a brokerage firm. The distinction between real money and fast money is important.’ Hedge fund managers. as we saw frequently in our observations. Typically. Flow information. is a buyer. is it the faster money. when the hedge fund manager wants to asses how stable current price levels of the specific asset are likely to be: when a mutual fund (a real money player).
16 what is the likelihood that the trade will be a profitable trade. a hedge fund manager active in foreign currencies if he would be aware when central banks intervene in the currencies market. and people would inform us immediately of the intervention. provide the information. hedge fund managers rely on the superior number and variety of connections that brokers have and use them. as their ‘ears and eyes in the market’. on their screens. Flow information. he replied: ‘We get to know this very quickly through our brokers. for instance. has the significant limitation. Hedge funds are eager to establish the intentions of other actors with whom they share the market and brokers. than trades that were deemed to have been set-up by managers that were not qualified as such. sovereign wealth funds and central banks. who knows about the hedge fund’s interest. specific and frequently available. Moreover. in spite of it being timely.’ HFM7 describes a situation we observed frequently: the hedge fund manager and the analyst notice. In foreign exchange markets. as one hedge fund managers put it. A few minutes later a broker. we conclude that trades that were reported to have been set-up by ‘smart money’ were far more likely to be studied and thus imitated. From our observations. The distinction between different players is of crucial importance in many markets. special attention is paid to actions by macro hedge funds. in effect. calls with flow information that typically includes details about the trading activity that caused the price change. who know that such information may lead to more trade orders. when we asked HFM7. For example. imposed by the fact that brokers are required to maintain the anonymity of their customers and prospective customers and are . We would see it as something that’s going on because of the price action. a price change in an asset they have been following. […] We see the price action go. This distinction is facilitated frequently by exchanging information between brokers and hedge fund managers.
. The general nature of flow information is even more pronounced when the brokers disseminate information more widely. the distinctions discussed above. we can see how brokers divulge important flow information while concealing the names of their customers. During our observations at the hedge fund where HFM7 is a partner and manager we noticed that much of the flow communication came in via Bloomberg or IB Chats (Bloomberg messenger).17 prohibited from disclosing their identities. 2009. In fact. May 19th. using emails or instant messaging. On this screen. between real money and fast money. In following Bloomberg screens. These messages are sent to prepared lists of hedge funds the broker believes may be interested in the information. the broker sending the message notifies HFMs that hedge funds with a very good past track record (‘quality names’) as well as long-term investors (which means that the buyers are less sensitive to short-term prices movements) are buying Euros (EUR) against US Dollars.12 Figure 1: Bloomberg terminal screen shot from HFM7. ‘Trichet comments’ in the third paragraph refers to the potential effect that a press conference by 12 The screen has been cropped for anonymity reasons. follow the prohibition on stating explicit names of customer.
the trader at the hedge fund would call the broker asking if a certain flow indicated by the latter was ‘real’ or if they were just ‘fishing’. You pay for research where they scan companies and they filter all the valuation cases for you’ . Indeed. but more sought after type of information are the investment ideas. the head of the European Central Bank. emails or instant messages) at much higher rate than the latter seek their information. While flow information forms much of the volume of the information flowing in the broker – hedge fund manager connection (often via his trader). It is true that hedge fund managers call brokers and ask for specific ‘market color’. That is basically what you pay for. important details and in many cases is superfluous for decision making. It was not uncommon for us to see hedge fund managers or analysts who deleted such messages after looking at them very briefly or even without reading them at all. may have on the markets. but for each of these there are many unsolicited phone calls and Bloomberg messages (known colloquially as ‘Bloomies’) sent to hedge fund managers. a less common. The information that hedge fund managers are used to receive from brokers is frequent. context-specific. and the broker is notifying that hedge funds (‘leveraged players’) were buying some short dated protective put options against a possible weakening of the Euro following this press conference. The ‘coded’ language used in the message reflects an important dimension of the informational exchange between brokers and hedge fund managers. it was apparent in our observations that brokers initiate communication with hedge fund managers (be it via phone calls. but lacks. in some occasions brokers send out ‘indications of interest’ in the hope that it might generate a client order. a long-short London-based hedge fund manager explains: ‘The way I see brokers is a process of scanning for money making ideas. HFM9.18 Jean-Claude Trichet. For example. At times. frequently.
the lay-offs clearly followed a distinction between brokers who had high fee-paying customers and those who had not. We also witnessed brokerage firms where a bell rang in the open-space floor each time a salesperson obtained a 13 14 1 Bp.e.’ 14 This monetary incentive is deeply entrenched in the organizational structure and practices of the brokerage firm. 8 for execution. which equates the role of brokers to a source of investment ideas.19 This view. The reason is that the brokers in our sample provide frequently their on capital. 1 basis point.000 of the value of the transaction. they provide liquidity to the hedge funds by buying and selling from them directly. as BR7. An estimated 7 Bps represents that value of broker-supplied information and ideas. is prevalent among hedge fund managers and it is also supported by an economic infrastructure. Most brokers we questioned on the subject explained that although they were not paid only on the basis of fees earned for the brokerage house. The hedge funds we researched pay normally between three and eight basis points (Bps)13 in fees to brokers who provide only trading execution service and no investment ideas. recruitment managers at brokerage houses stressed that ‘strong commission generators’ would get access to larger customers and would advance faster. brokers who provide execution as well as investment ideas charge most hedge funds between 15 Bps and 20 Bps. . equals 1/10. In addition. explains: ‘If a hedge fund pays you 20 basis points. what percentage of that is for research and ideas? We looked at it. the charge for execution is higher. That is what they pay for. we talked to our main hedge fund clients and the answer is 12 for research access and ideas. Others explained that in present market conditions. In comparison. more fees meant higher bonuses. Note that the 8Bps for execution is higher than for the execution-only brokers. in one of the hedge funds we observed promotions and year-end bonus were a direct function of commissions generated. i. For example. in the hope that the ideas would be acted upon and generate executions (and commissions). as well as the drive to present investment ideas to hedge funds. i.e. the head of hedge fund sales in a brokerage firm. Since they put their capital at risk.
virtually. but the incentives and the relative positions in the social fabric discourage the sharing Incentives installed by clients could consist of one or more of the following methods: (1) percentage of commissions paid based on number of money-making ideas during a certain period. We also noticed how new investment ideas. This information tends to be general. after these have been acted upon) as an alternative form of currency. We saw how hedge funds offer occasionally investment ideas generated in-house as an alternative form of payment to brokers. These economic and social practices prevent. or interpretation to existing ones.i. 15 . TRB1 explained how his hedge fund had tried to compensate the drop in cash commissions paid to brokers by sharing with them some of the fund’s ideas: ‘Since business has been slow and we have not been able to pay our brokers the way we should.’ The information exchanged between hedge funds and brokers is one ‘leg’ of the informational exchange circle that makes up structural secrecy. (2) direct pay-back mode . a trader at an event-driven hedge fund. (2) broker rankings .20 large order or where the head of the trading floor would pay an ostensible congratulatory visit to the salesperson who had just ‘printed’ a big order. from timely. is being disseminated) use investment ideas (frequently.15 In addition to being a valuable service. A representative example comes from TRB1. detailed and interpretative information to be exchanged between hedge fund managers and brokers. investment ideas are also used as a form of currency. We have been giving some of them some ideas we were looking at.e. were exchanged for other ideas.at the end of a term hedge fund managers inform the broker’s supervisors how they ranked versus their competitors. It is not that such information is not available to the actors involved. a useful idea would be executed via the broker who transmitted it. frequently. we have engaged more with them on the ideas’ side. not directed towards specific trades and framed within a strict transactional background: brokers disseminate information to attract execution while hedge funds (whose information.
I mean they are starving for stories. knowing that these connections serve as the basis for making revenue. Brokers want to create and maintain as many connections as possible with hedge funds. they pass it on. It has to be noted that along with such expressions of restrained and controlled relationships. At the same time. they will tell me what other strategies or other hedge funds are doing. hedge fund managers were eager to hear from brokers about what other hedge funds were doing. non-interpretative observations of the market. This logic of connectivity has a direct impact on the quality of information from hedge funds to which brokers are privy. both from executions and from selling investment ideas.21 of anything apart from general insights about the market or pinpointed. This is illustrated by HFM2 and HFM16 who are senior hedge fund managers at two of the biggest hedge funds in the world: ‘The sales side people [brokers] are just desperate to print tickets. Once they do. On the one hand. and so if they hear a good story [i. They engage in what I call parasitic behavior. we observed that brokers and hedge fund managers often spoke with each other . This set of conventions also has a crucial affect on the brokers’ motivations for creating and maintaining connections. [HFM16] Interviewer: HFM2: What is the perception of brokers by hedge fund managers like you? In general they are good people. an interesting idea]. but you should be weary of them. but on the other hand they frowned on what one hedge fund manager called: ‘parasitic behavior’. They try to know or understand what we do. We encountered an ambiguity in the hedge fund managers’ attitude vis-à-vis brokers.e. They do not care how [or] who with. they will use that to generate business from another hedge fund.
4. In fact. share detailed internal information The majority of the hedge funds in our sample receive an administration fee of 2% and a performance fee of 20% of assets under management. for example. a common feature in the daily routines of virtually all hedge fund managers we observed is that they communicate frequently with other hedge fund managers who trade within the same strategy and who are. had the phone numbers of four of his competitors hedge fund managers programmed into his speed-dial phone system. Our observations reveal that this is not an exception and that many of the hedge fund managers talk several times a day with one or more of their competitors when they discuss potential investment ideas. However. but distinctly different in the type of information shared and the underpinning sets of practices. in general. where the social engagements is a function of commissions paid and where hedge funds would only divulge to a broker information he does not mind being disseminated widely. such communication is so common that. For all practical purposes. when we asked senior salespeople in brokerage houses and hedge fund managers about how close the ties are.16 In spite of this fact. had meals together and shared pastime activities such as going to sporting events. They compete for capital and return-generating ideas. exists among hedge funds. they described most ties between brokers and HFMs as governed by a ‘business reality’. therefore. HFM9. Connections among hedge funds An equally active set of connections. their competitors. hedge funds are competitors. a Londonbased long-short hedge fund manager. which contribute directly to their performance and amount of assets under management on which their compensation is based.2. report on success or failure of existing positions and. which consisted of 16 preprogrammed phone numbers (HFM6 and HFM16 being two of them).22 several times a day. 16 .
tends to support the exchange of different quality of informational items. Asking HFM7 what was the basis for such frequent contacts. a London-based hedge fund manager confirms: ‘You try to share information and ideas. insightful feedback on ideas. He used to be a proprietary trader ten years ago and a colleague of mine. or a ‘you scratch my back and I scratch your back’ type of an implicit agreement.’ A similar explanation was offered by HFM3. You would be surprised how relationships endure over time. It is reciprocity. a ‘quid-pro-quo’. HFM15. You will not keep those people as friends if you don’t have something else to offer. hedge funds professionals explained that information sharing is a ‘two-way-street’. manager of a New York-based convertible arbitrage fund: ‘Between hedge funds a lot of it is just your personal contacts. actually. you have great personal contacts with hedge fund managers at other large funds with whom you exchange ideas. In all cases where we discussed this practice of information exchange. moral support or other valued assistance. In some of the small funds. very big American hedge fund. he answered: ‘I know those people from working in the same financial institutions.23 related to the running of the fund. a strong norm of informational reciprocity also establishes the communicative norms of the connection. it was mentioned repeatedly that investment ideas and insights are shared with the expectation that the ‘acquirer’ of information would ‘pay back’ the favor in the form of offering insights or information of their own.’ The reciprocity among hedge funds. In our conversations with hedge fund managers. One guy that I know is head of a very. and when observing their regular discussions with other hedge fund managers.’ While common biographical history serves as a basis for the connections. unlike the one existing between hedge funds and brokers. The .
but in almost all of these communicative exchanges. Hedge fund managers expect other managers with whom they communicate to offer insights. HFM10. but to shed new light. whether they were face to face. We witnessed many conversations that focused on specific issues relevant to trading positions.’ The discussion to which HFM10 refers was about the tax implications (exemptions) of a transatlantic merger. explore different dimensions or scrutinize existing or contemplated investment ideas. to exhaust. by phone or by email. answered the following when asked why he discussed in detail his position with a competitor: ‘I speak to these guys because I know they have a very specific knowledge in that area. Although HFM10 and the fund he ran were very knowledgeable about European tax law and mergers. it would be helpful to have the input of a competing hedge fund manager that he considered very . he felt that for mergers where US tax law might apply. but also includes a crucial interpretative dimension. who manages an event-driven fund. which was noticeable among hedge fund managers. I know some of the guys I speak to although their funds might be similar. it would seem. have very specialist knowledge and that can be very helpful. This motivation justified to many hedge fund managers the exposure required in the reciprocal information exchange. product strategies or implications of regional law. For example. all possible angles of inquiry when evaluating a possible trading position. issues such as composition of boards of directors. This interpretative dimension is related to the motivation. the goal of the conversations was not to find out about a new investment idea.24 information exchanged among hedge funds is not only timely and detailed. commentary or criticism during the discussions.
he explained: ‘I just do not want to be wasting time but I think analysts [in brokerage houses]. but still hedge fund managers we observed clearly preferred to approach another hedge fund manager. say. they sometimes simplify their job a lot. offers an explanation for this preference. because that just simplifies their life. ‘ The quote above. Hedge fund managers communicate with each other not only to share information. When asked about how he evaluates the likelihood of two companies to merge. HFM11 knows . They will have spoken to lawyers and spoken to advisers and spoken to consultants because that is what we focus on. who focuses on investing in announced mergers or acquisitions. put a 50-50 probability on each [company] and that gives them a target [price]. This changes the probabilities. but. which represents many situations we witnessed. The motivation to seek out interpretations and analysis is related directly to the discursive nature of the communication among hedge fund managers.[…]They will. HFM11. which was a position he was examining at the time. This is notable when considering the fact that brokerage firm have their own expert-analysts. over contacting an analyst at a brokerage firm when a difficult question about a trading position arose. highlights another set of motivations for the hedge fund managers’ collaborative process of interpretation. primarily.25 knowledgeable about this specific issue. That is just very different from putting 50-50 on it. a competitor. but the quality of interpretative knowledge they gained from talking with competitors and sharing with them their ideas. in the quest to solve specific problems. they will have done a hell of a lot of work on that. This rationale was presented to us frequently: hedge fund managers were fully aware of the competition among them. views and market positions was worth the exposure. a manager of an event-driven fund. […] But if I speak to someone else who is an event-driven investor.
26 that probabilities should be assigned to the possibility of a merger and he even has an opinion about which probabilities are appropriate. Nonetheless, he wants to share his views with someone who is equally knowledgeable so as to test the reliability and validity of his ideas. To find such conversation partners, HFM11 is reaching out to other hedge funds. As he stated, HFM11, like many other hedge fund managers, believes that analysts in brokerage houses are not as knowledgeable as other hedge fund managers (or their analysts) on specific issues simply because hedge funds tend to specialize in one type of trades, while brokers cater to a wider variety of trading strategies. The shared information and the assessment of investment ideas indicate the level of disclosure that exists among hedge funds. In some cases, sharing of investment ideas can lead to coordination, as HFM9, a London-based long-short hedge fund manager testifies: ‘I was in a meeting recently where an executive of this hedge fund had invited 10 or 15 other hedge funds he knew well to discuss a potential position. They were very open and they said basically they wanted to buy this company and split it up. They put forth their entire valuation models and we discussed it together. They then explained how they were going to unlock the value in it. They then asked: what do you think? Will you help us doing it?’ HFM9 The connections among hedge funds, we realized, in addition to being based on ties originating in shared biographies (in particular, previous working places) and following norms of reciprocity, also exhibit a strong element of trust. Hedge fund managers told us numerous times that unless they trusted the other person, there would be no point in having a relationship with another hedge fund manager and in exchanging information. The words friends, friendship, trust, personal relationship were mentioned frequently when hedge fund managers spoke about other managers with whom they communicated regularly and some used the words
27 brothers and brotherhood. We tried to understand the meaning of trust in this context for hedge fund managers and the role that this notion plays in the connections. We saw that hedge fund managers have two separate, but related, sets of meanings that they associate with what they define as trust. First, they recognize trust in the competence of other hedge fund managers: ‘First there is intellectual trust, people where you trust their judgments. By discussing with them, you know that you are discussing with people who have done their work, have a sharp view on numbers, cannot be messed around with by management, are themselves smart investors, have a smart way of looking at things, a certain vision and so and so.’ [HFM1]
By ‘intellectual trust’ HFM1, manager of a long-short fund, expresses a view we heard repeatedly from hedge fund managers who maintain connections and evaluate information on the basis of their intellectual appreciation of the other party. ANAB3, a senior analyst and a partner in a different long-short hedge fund describes this practice in action when talking about a competing hedge fund and his willingness to share investment ideas:
‘I trust their opinion about stocks. I have had recently a situation where we were short one stock and the guy at [name of a competing hedge fund] was long. So we met up inside our offices with him to discuss why we had different opinions about the stock. He is very smart, so I wanted to pick his brains and share my views to see who was missing what.’ The last quote exemplifies how the concept of ‘trust in competence’ helps in shaping decision making among hedge funds. During the discussion described above, ANAB3 shared with the manager from the competing hedge fund the rationale behind his trading position, the valuations that motivated it and the extensive research work that supported the decision. Following this, the visitor presented his fund’s position and the arguments supporting it. During
28 that presentations, ANAB3 asked many challenging questions and a discussion developed around the different views. This exchange, which is typical to many of the discussions among hedge funds, lasted about two hours and was informative and open and amounted, eventually, to collaborative decision making, as both ANAB3 and the other analyst gained new perspectives regarding their trading positions. The other meaning of trust we encountered is related to the sensitive nature of the shared information. Earlier we saw that the willingness to expose such information is justified by the potential return the hedge fund managers believe they will receive (e.g., in the form of expert opinion), but this willingness is also explained in trust about the intentions of the parties to the communicative ties. The following three quotes explore this meaning of trust among hedge fund managers:
‘And then equally moral trust, people that you would trust sharing your positions with and so on, you know. While you are not discussing anything that’s illegal you’d still know that it’s information that is reasonably confidential and should stay so.’ [HFM1]
‘It is mainly cross checking intelligence with somebody you trust, somebody who actually shows integrity and he does want to compare the notes with you.’ [HFM10]
‘What does trust mean? […] I suspect that each one of us has a set of values that we operate under and I suspect that trust really means that the person that you are interacting with shares those same values as you or has even higher values.’ [HFM22]
The following quote from HFM14. However. This meaning of trust also adds another dimension to the information segregation that forms the other ‘leg’ of structural secrecy in the hedge fund world. Hedge fund managers tend to suspect the safety of information shared with brokers because an inherent part of the broker’s practice is the re-distribution of information.29 We encountered the phrases ‘integrity’. hedge fund managers explained to us. Because what I tell to a broker. he might easily tell other hedge funds to create goodwill. former colleague or not. in one of the conversations at the end of a trading day. brokers simply cannot be trusted with information the same way other hedge fund managers can be. Interviewer: Even compared to former colleagues who are still brokers? HFM14: Of course. definitely. ‘a shared set of values’ and ‘honesty’ when hedge fund managers expressed their belief that others would not abuse the sensitive information that is given through the sharing practices. This means that I will not really trust . The person who mentioned it said that ‘everyone knows about them and now no one he knows talks to them’. a hedge fund was mentioned that used information to spread false rumors and to inflate prices. who was previously an equity hedge fund salesperson at a brokerage house. When we asked about cases when hedge fund managers did take advantage of such information it was obvious that the topic made our informants uneasy and they were reluctant to speak about such instances. represents nicely the communicative practices and their justification: distrust and the distant ties that hedge fund managers maintain with brokers: ‘ Interviewer: Do you find that those people [hedge fund managers] are closer to you than brokers? HFM14: Yes. Because of this inherent tendency.
they still restricted the type of information they shared.’ HFM11 The quote mentions the importance of information sharing. Such conversation. whose funds are invested in the hedge fund. . This will even increase our bond. a salesperson will not. It is all these factors which make for hedge fund managers to be closer to each other than to brokers.’ The quote indicates that the effect of the position of the actor in the network of connections is so strong that even in the cases where the hedge fund manager knew the broker from a previous joint working place (a basis for many of the connections. This is not the case with other hedge fund managers that I trust. also contained. an element of emotional support: ‘There is also empathy. joint analysis of the trading moves and their rationale. as we saw). On the other hand. experience monetary losses personally and thus empathize with other managers. So there will always be some distance. but because it is part of his job. who are employees of the brokerage house. Another hedge fund manager will understand how I feel when I lose money. beyond a detailed. Not because it is him. not a broker. almost without fail. Another set of practices that indicate of the strength and viability of the connections among hedge fund managers relate to giving and receiving emotional support. again. On numerous occasions. it is that we trust each other. The only thing which holds us together is not business. So when I am down I might talk to another hedge fund manager about it. Hedge fund managers. for strengthening interpersonal connections and it also distinguishes and emphasizes. including unpleasant and bad news. the differences between hedge fund managers and brokers. we witnessed hedge fund managers who shared with competitors details about trades in which they lost money. brokers.30 him.
we can summarize the empirical findings in this section by looking at a common perception among hedge fund managers. I think they know the sort of concerns we have because they share those concerns. being referred to as ‘smart’ implied that the person discussed is a member of good standing in the community. Having reviewed many of the communicative practices that take part in the investment decision-making. Implicitly. .’ [ANAB3] ‘You typically find smarter people as a percentage of people employed in hedge funds than in investment banks.31 and whose commissions are paid regardless of the fate of the trading positions are seen as ones who cannot relate to market losses to the same degree as hedge fund managers. When asked what differentiated an analyst at a hedge fund from one at brokerage house. or that you regularly take part in mutually beneficial exchange of interpretative information. as ‘smart’. that your expertise and analytical abilities are appreciated. In many respects. on average. whilst the definition does not exclude completely brokers from being referred to as ‘smart’. usually ones who work at hedge funds.’ [BR4] These comments capture many of the motivations for the particular communicative practice we discussed so far. In many conversations. a salesperson to hedge funds in a brokerage house responded: ‘I actually tend to think the people in hedge funds. we had with them. are smarter. ANAB3. Being ‘smart’ at the hedge fund world meant one or more of the following: that your fund performs well. hedge fund managers and brokers referred to certain people. an analyst at event-driven fund analyst and BR4. someone who takes an active part in important discussions and contributes to them. Let us look at two examples. it is much more common to hear it attributed to a hedge fund manager than a broker.
combined with its wide dissemination. many of which are not connected directly to each other. underpinned by arm’s length ties and embedded ties is the social and organizational arena where decision making in hedge funds take place. while being effective and beneficial for decision making most of the time. proved to lead to destructive outcomes in the case of the VW-Porsche trade. lead hedge fund managers to assign low priority to this type of information and. knows everyone else and from which brokers are excluded. to ignore it altogether. using some of the information from brokers as a basis. virtually.32 In this section we examine two sets of ties that make up the network of connections between hedge fund managers and brokers and together form the two ‘legs’ of the interorganizational structural secrecy – the brokers-hedge fund managers ties and the ties among hedge fund managers – and the different practices and norms associated with the network. on the other hand. This constellation. The combined set of practices and conventions brings about and maintains inter-organizational structural secrecy. partial (sometimes ‘coded’). The differences in the qualities of information exchanged are reflected also in the structure of connections. . The specific content included in much of the flow information. hedge fund managers conduct a consultative process where more detailed and timely information is collected and investment ideas are examined and evaluated. connected only by highly restrictive gateways. a constellation whereby the different types of information and knowledge circulate in separate paths. Following a discussion of our quantitative findings. On the one hand. serve as informational hubs: they are connected to many different hedge funds. These two sets of ties are a fundamental feature of investment decision making in hedge funds. frequently. This analysis reveals that the combination of the two legs. hedge fund managers share and discuss information in small groups within which everyone. flow information. Brokers broadcast initial investment ideas and pinpointed. Brokers. In contrast. we will focus on this trade and analyze the emergence of the crisis.
they maintain. hedge fund managers tend to be selective in the connections they keep. BR7 and TRS1 did have a connection. The five brokers in the sample. BR7 had even introduced TRS1 to some of his customers. Brokers aim to have a large number of connections with hedge fund managers. the average number of connections (degree) that a broker has should be higher than that of the 17 Although competing brokers do not tend to have professional relationships with each other.17 Moreover. these connections allow brokers to transmit information and investment ideas between hedge funds. The managers tend to exchange information and discuss trading ideas in small and dense clusters. Our examination follows the general hypothesis that hedge fund managers and brokers are motivated by two different kinds of ‘logics of connectivity’. are connected directly to twenty hedge fund managers. We examine the resulting networks to see if the evidence from the interviews and participant observations are corroborated by another mode of enquiry. then we should expect them to be reflected in several descriptive network measures. As discussed earlier. Network Findings In addition to the semi-structured interviews and observations. we also constructed a network of the connections among hedge fund managers and between them and brokers. The two informed each other of their best ideas and big orders they recieved. In contrast. If these two logics of connectivity are expressed not only in the actors’ conversations and in interviews. as these connections provide them opportunities for generating fees.33 4. . all known to each other.3. First. connections with different hedge funds allow the brokers to spread information and generate trading orders. each working at different brokerage firm. but also in the aggregate map of connections. made up exclusively of hedge fund managers. These five brokers are all Managing Directors at their firms and responsible for the coverage of the most senior hedge funds managers.
on average. Breiger. where each of them has connections with many hedge fund managers. while having about half the level of dyadic constraint of hedge fund managers. brokers aim to position themselves at the centers of star-like network patterns. on average. 2004).34 hedge fund manager’s. Brokers are less constrained by virtue of having more connections and occupying more central positions in the network: brokers obtain higher betweeness values and higher eigenvalue centrality values. weighted by the importance of the connections for the actor.19 Table 1 reports these measures for brokers and hedge fund managers. almost twice as many direct connections (or ‘degrees’) as hedge fund managers have. 18 while the brokers having higher betweeness centrality. 1979) is based on the number of shortest paths between pairs of nodes in the network on which the measured node is located. 1992. Second. The higher betweeness centrality testifies that brokers hold the network together and that their removal would disintegrate the network into separate components.94 Hedge fund 5. a low dyadic constraint is related to increased brokerage opportunities (Burt. We expect this difference in preferences to be reflected in hedge fund managers having.05 0. where information can be verified and triangulated easily. while incomplete triads gives one actor potential brokerage opportunity (as that actor connects the two others). Their higher eigenvalue centrality indicates that.215 Average betweeness centrality 34.17 managers Table 1: Descriptive network statistics for the hedge funds – brokers’ network The measures indicate that brokers have. on average. According to this rationale. Network measure Average degree Actor type Brokers 9.24 6.473 0.83 Average eigenvalue centrality 0. higher aggregate dyadic constraint than brokers do. Complete triads impose constraint on the actors connected in them (none of them can broker between the other two).4 Average aggregate dyadic constraint 0. The aggregate constraint on an actor is the sum of the dyadic constraints that actor has as a result of the actor’s membership in triads. 18 . the more brokerage opportunities that actor would have. 19 The measure of betweeness centrality (Freeman 1977. while hedge fund managers prefer to be part of higher density patterns of connection. The measure of dyadic constraint is based on the triads to which the measured actor belongs. The rationale behind the measure is that the more such shortest paths ‘cross’ the measured actor.
Brokers are responsible for the flow of ideas in the network. brokers’ centrality tends to be greater than the centrality measure of hedge fund managers. The network analysis reinforces the insight that the ideas are only relatively new: they are known to one part of the broker’s network of connections. at the centre of star-like patterns of connections. since they are connected to actors that are connected to each other directly. brokers have more connections and despite their small number they are instrumental in holding the network in one large component. These findings corroborate the picture emerging from the data collected in the participant observations and interviews. but they have many more connections so that the amount of resources invested in each connection is more limited. Brokers are found. However. Eyeballing the figure confirms the results shown in table 1. The network is presented in figure 1. there is an additional insight to be gleaned from the figure: besides their connections to brokers. but not to another and the brokers exploit this asymmetry. Those specializing in short-long strategies are represented by grey squares and positioned below the brokers. This structure. when working as a group. They have many more connections and a structural position that is instrumental when offering new trading ideas to hedge fund managers for generating trading orders.35 brokers are connected to more central actors in the network (Bonacich 1972). The size of the node represents its betweeness centrality. Furthermore. Hedge fund managers. validate it and develop a richer understanding of the market. Hedge fund managers are represented by squares: those specializing in event driven strategies are represented by black squares and are placed above the line of brokers. to cross check their information. otherwise disconnected actors. in contrast. to compare and contrast different sources. hedge fund managers connect to . are placed in positions where there are less brokerage opportunities. typically. enables hedge fund managers. however. The five brokers are displayed by five circles placed in a horizontal line at the upper-middle part of the figure. an advantageous position for brokering between other.
36 other hedge fund managers that specialize in a similar trading strategy: those specializing in long-short strategies tend to connect between each other. and the same goes to those specializing in event driven strategies. Managers working for the same hedge fund are grouped in a rectangle. squares are hedge fund managers). The node’s size represents the betweeness centrality of the node. There are only two ties that connect hedge fund managers that specialize in different strategies (namely the ties of H20 with H24 and H17). and brokers are in the position to coordinate this division of labour. This indicates that hedge fund managers are reaping economies of specialization. H18 H23 H20 H10 H11 TRS6 BR3 BR6 TRS1 T5 H19 H12 H24 H17 T7 H9 H2 H16 H6 H25 H1 Figure 1 Network of hedge fund managers and brokers. The node’s shape represents its role (circles are brokers. . The node’s fill represents its dominating strategy (grey is longshort strategy. black is event driven strategy).
Finally. Frank and Strauss 1986). they provide tools for comparing between different models that consist of different sets of effects. we aim to capture tendencies in the network that explain global properties of the observed network in its current form. Each of the estimated parameters is associated with a ‘network statistic’. still do not show that the observed structures are not the result of random associations between the actors. For example. we use the estimated parameters of the network statistics to generate a population of networks and compare between these networks and the observed network in terms of new network statistics whose parameters we did not estimate in the model. Finally. to test the goodness of the model’s fit. whereas regions of clumps is measured by network statistics known as “alternating k-triangles” and “alternating 2-paths” (Robins. To determine whether or not the patterns of connectivity revealed in the descriptive measures also reflect network tendencies or effects. ERG models tell us which effects are important in the generation of a given network and what are the relative contributions of the different effects. we simulate a population of networks that are generated by the same mechanisms we identified. 2006). By developing and fitting a model. and Wang 2009. and compare the simulated networks to the observed one. Then. Snijders et al. we may compare the networks in terms of various centrality measures or the . They also indicate to what extent the data supports our conjecture that a certain network effect is indeed at work. whilst confirming the structural properties of the hypothesised ‘logics of connectivity’. The skewedness of the degree distribution is measured by a network statistic known as “alternating k-stars”. we use Exponential Random Graph Models (ERG models) for social networks (Snijders et al 2006. Pattison.37 The network analysis measures presented so far. we can first estimate parameters that reflect the tendency of the observed network to form clumps and the tendency of its degree distribution to be skewed. For example.
The estimation of the edge parameter has a relatively high standard error.46) 4.98)* 1.21 (0.71)* -3. and is therefore unreliable in the first model.37 (1. The results for the two models are presented in table 2.63)* -9.26)* -0. and Pattison 2006).11) 2.95 (1. A negative value for the alternating star effect indicates that there is no tendency towards skewed network degree .69 (3.44 (3. Such a comparison can tell us how well a parsimonious model captures diverse network characteristics.17 (0.12 (0.01 (2.39)* -6.84 (0.14 (0.38 frequency of various types of network configurations.53 (1.35)* 2. For each network statistic in table 2. we report the estimated value of the associated parameter and its standard deviation.23 (1.09)* Model 2: estimate (standard error) 8.65)* 5.05) -0. Asterisks indicate effects for which absolute value of estimates are more than twice the standard error. Three models were fit using the software PNET (Wang. The edge effect in sparse networks determines the marginal log-likelihood of observing a tie between two random nodes.98) -0. Robins. Model 1 is relatively straightforward and common in exponential random graph applications.69) -3.14 (0. This program estimates the model parameters using a Markov-Chain Monte-Carlo (MCMC) Gibbs sampling technique.55 (0.17) 0.74)* edge effect alternating k-stars (lambda=2) alternating k-triangles (lambda=2) alternating two-path (lambda=2) triangle two-triangle hedge fund manager triangles (hfm_t3u) hedge fund manager activity hedge fund manager interaction long-short activity long-short interaction Table 2: Two exponential random graph models of the hedge-fund broker network.15 (0.95)* -0. Model 1: estimate (standard error) 6.08)* 0.17 (0.76 (1.
To correct the model. This is not surprising considering that despite more hedge fund managers than brokers (20 hedge fund managers compared to 5 brokers).23 for hedge fund managers and an additional significant negative interaction effect of -6. etc.39 distribution. . we observe a significant positive activity effect of 4. These statistics count the number of triangles that consist of a single type of actor: the statistic ‘HFM_t3u’ counts the number of triangles consisting of three hedge fund managers. Hedge fund managers ‘attract’ brokers more strongly than they ‘attract’ each other.23 to the loglikelihood that a tie is formed. this means that the log-likelihood for a tie between a broker and a hedge fund manager is about double the log-likelihood for a tie between two hedge fund managers. these statistics are underestimated by the model. Appendix 3 presents the goodness of fit of this model.76 between two hedge fund managers. 2009).76. 45 of the latter). Finally. but the statistics that fit least are the t3u statistics. Most of the network statistics fit well. we need to account for an additional mechanism that explains why different kinds of actors tend to form triangles. over and above what is expected by the first model. whereas the statistic ‘ls_t3u’ counts the number of triangles consisting of three long-short hedge fund managers. The next two parameters can be interpreted together: a significant alternating triangle effect together with a non-significant alternating two-path effect indicates that nodes tend to ‘clump’ into dense regions of connected triangles (Robins et al. Taken together. the log-likelihood that a tie is formed between two hedge fund managers is penalized by -6. This means that each hedge fund manager in each dyad contributes 4. In all these cases. However. fewer ties form between pairs of hedge fund managers than between pairs of hedge fund managers and brokers (28 of the former vs.
as well as a slight but significant difference between different types of hedge fund managers. However. As before. These findings correspond well with the qualitative data exhibited earlier. These connections require more commitment from the parties involved than connections between hedge fund managers and brokers. as can be seen from the comparison in annex 3. which improves the goodness of fit. forces of homophily and heterophily. hedge fund managers are likely to realize ties between each other.40 To achieve this aim. various triangle statistics are added to the model. we see a positive activity and a negative interaction effect of hedge fund managers. as well as activity and interaction effects for hedge fund managers who specialize in the long-short trading strategy. . the ERGM identify that hedge fund managers and brokers attract each other (‘heterophily’). but are more likely to do so with brokers. The result is model two. and that hedge fund managers attract each other (‘homophily’). These effects have been completely replaced by activity and interaction effects. First. but the ERGM also shows that the 'force' that attracts hedge fund managers to brokers is stronger than the force that attracts hedge fund managers to each other. As before. that is. These two tendencies are not surprising. Connections between hedge fund managers demand more resources and put the parties at risk when discussing private information. there is an important difference between the two models: the network-wide tendencies to form triangles (captured by the diverse triangle and alternating two-path statistics) have become unreliable in the second model. it is therefore reasonable to see hedge fund managers being more selective about contacts with other hedge fund managers than with brokers.
41 Finally. BR7. that broker-to-broker connections are exceptional. During our fieldwork. we do not know if brokers ‘repel’ each other: try to avoid making connections. we noticed that hedge fund managers and brokers frequently referred to certain trading positions as ‘consensus trades’. which were held by many hedge funds. 5. In the second model. We were told by numerous hedge fund managers that at any given time there were a few similar or even identical consensus trades. the small number of brokers limits our certainty about whether or not two hedge fund managers connected to the same broker are less likely to know each other directly. we analyze the risks embedded in the investment decision-making process. a broker. BR7 emphasised that it was highly exceptional that a broker would share such information with another broker. let us now focus on a phenomena that is more directly related to the VW-Porsche crisis. The small number of brokers in the network means that these conclusions include certain caveats. however. the specialization of hedge fund managers fully explains the clustering of actors together. We have qualitative evidence. told us that he found out that a hedge fund was using his investment ideas. BR7 learned about this because the broker who executed the trades was his good friend and shared this information. For example. The relation between consensus trades . Consensus trades were trading positions that were popular among hedge funds. In this section. how investment ideas gain popularity and. Second. second. we see that the type of strategy hedge fund managers specialize in is a key factor in explaining their ties. concentrating on the VW-Porsche case. First. first. we describe. Consensus Trades Having discussed the general structure of social ties and sets of norms and conventions that govern decision making in hedge funds. but executed the trades through a cheaper broker.
Finally. cannot explain the dissemination of the ideas and their turning into popular consensus trades. Because if one hedge fund manager knows that something is cheap he is likely to let another hedge fund manager know it is cheap. you might talk with a couple of your friends at other hedge funds. go through the critical issues you are not sure of. HFM9’s description best encompasses our observations about the dynamics that lead to the emergence of a consensus trade: In general. . you like it and invest in it. So to be sure. detailed discussions among hedge fund managers were limited to small groups of trusted individuals. see if you are not missing anything. both when talking on the phone when in their offices and during social events.42 and the ties among hedge fund managers was explained by many of the hedge fund managers and analysts with whom we spoke. some brokers are seeing that hedge funds are [executing the trade] and start telling other similar hedge funds. however.. interpreted and scrutinized within the clusters of trusted hedge fund managers. By now. In contrast. especially amongst hedge funds. there are many people that have similar kind of trades. Decision-making on its own. The other hedge fund managers are doing the same. You discuss it. You look at it and [a certain stock] looks dirt cheap. Investment ideas were discussed.’ [PBS1] We witnessed many times how hedge fund managers introduce to each other investment ideas. brokers were motivated specifically towards disseminating investment ideas and their wide variety of contacts enabled them to do so effectively. So somebody must have been the first one to come up with it. I would say that it starts with an idea. everyone has those trades. As the findings indicate. the arena where investment decisions were made.. People share information. Here is a typical explanation: ‘Yes. There is a certain universe of consensus trades.
The arm’s length ties. The information disseminated by brokers provides an outline of the trade. This description captures the two types of information exchange that underpin the general form of decision making in the hedge fund world. etc. If it does.. Again. are crucial for the diffusion of the initial investment ideas across the hedge funds’ networks. and if they give you the why. talk to other hedge fund managers. We witnessed these dynamics repeatedly during our fieldwork. these findings portray a picture similar to the one described by Uzzi (1997): both embedded ties and arm’s length ties play a crucial role in the making of a consensus trade. but does not develop a rationale and a detailed trading strategy. Because brokers will probably only mention what other hedge funds are doing but not why. other hedge fund managers and even [name of a television host on investments]. whilst the embedded tie. are the ties through which specific know-how or tacit knowledge are explored and are vital for the assessment and evaluation of the information. and if it makes sense. the brokers. it will be very general. between hedge fund managers and brokers. So these other hedge funds will be doing their own research. among hedge fund managers. These other hedge fund managers will analyze it. you start having a consensus trade since at that stage everybody is talking about it: you. invest in it.43 That is where I think it becomes critical. your friends. Many of the hedge fund managers explained and demonstrated in their actions how their common background underpins their connections. An argument we heard less frequently was that the common educational and occupational background of the hedge fund managers also contributed to the emergence of . The latter is developed through the discussions among hedge fund managers.
Instead. however. The significance of consensus trades. ideas and concepts that the hedge fund managers learned at previous joint work experiences. The practices and conventions applied when collecting and evaluating information are similar. BR3 and HFM9 point out that analyses are made and compared on the basis of ‘evaluative frames’ (Beunza & Garud. and this for the simple reason that such communication was so frequent and pervasive. whether a trade is adopted by many hedge funds or not. as BR3. For example. we did not witness many hedge fund managers who develop their investment ideas in complete secrecy. explains: ‘It is a small village. You probably have a big chance that you are going to look at similar things in a similar way. they preferred to share such ideas with their competitors. 2007): valuation methods and conventions. so you come to the same conclusion in a similar timeframe. cannot be isolated from the one discussed earlier. What is interesting is at the end of the day.’ This explanation according to which hedge fund managers develop similar strategies independently is feasible. or because joint discussion of the ideas helped in solving queries and problem. an experienced broker. using the same models. therefore. where hedge fund managers communicate their ideas in detail. are not fundamentally different from any other investment idea that hedge fund managers decide to adopt. we all come from a similar background. is in their volume. either because they expected some reciprocal return. especially when taking into account the high degree of homogeneity in the occupational background among hedge fund managers and brokers. In fact. Consensus trades.44 consensus trades by encouraging the use of similar cognitive and analytical patterns. The proposed causal mechanism. however. A trading . we probably studied very similar things and often have worked together doing valuations or what have you together.
One is to buy Porsche stock (known as ‘going long’). 6. the factor affecting the profitability of the trade is the difference between the prices of VW and Porsche.5% of VW stocks. when taking into account the accumulated VW stake by Porsche. this amount rose to 27.1. the carmaker. composed of two trading actions. To take advantage of the pricing discrepancy while not having exposure to the overall market direction.4% and rose again to 31% in 2007. This. or. In 2005. The rationale behind the idea was that Porsche had been buying VW stock for some time and had by then accumulated a significant position. In contrast. we first heard about the investment idea of the VW-Porsche trade. may have near-systemic implications. according to the rationale. the more profitable the trade would be. the smaller it becomes. in effect. as a hedge fund manager put it: ‘you can buy VW by going long in Porsche and you get Porsche. Porsche held 18. In 2006. The underlying logic behind the long-short position is that it is isolated. from the risk of price changes in the market in general. for free’. bringing their prices more in line with one another as well as recognize the value of Porsche as a carmaker. In this case. The stock is bought back later and returned to the lender. the market valued the rest of Porsche close to zero. The . a failing consensus trade. The second action is to borrow VW stock and sell it immediately at the market.45 position gone wrong held only by a single hedge fund would cause a loss. However. the rationale continues. the hedge fund manager chose a long-short trading position. because it is adopted by many hedge funds. Instead. The VW-Porsche crisis 6. made Porsche’s stock cheap in relation to VW’s stock. which would motivate market participants to sell VW and buy Porsche stock. Increasing popularity of the trade In January 2008.
This leads to a sharp price increase of the stock. The terms of a short sale include typically a set time for returning the borrowed stock. a lot of hedge funds have that trade on now. which can be extended. we heard it discussed in conversations between brokers and hedge funds numerous times. Since their profitability depends on market conditions. but the lender is also given the right to ask back for the borrowed stocks before the end of the set period. The early recall of a borrowed stock is taken. long-short positions can be held for weeks or even months until they are unwound: the ‘long’ part is sold and the ‘short’ part is bought at the market and returned to the lender.. In March 2008.You know Porsche got a chunk of Volkswagen. mentioned to us . So a lot of people make valuation of these two and then strip out one to see what the rest is worth. the first example given to us was that of the VW-Porsche trade: Interviewer: Are you familiar with the term ‘consensus trade’? HFM8: Sure. the rationale behind the trade – the value discrepancy between the two companies – is easy to communicate and. the big one now is Porsche-Volkswagen. also holds: the larger the difference between the prices of the two stocks.. We noticed that the VW-Porsche trade was mentioned in discussions among hedge fund managers and in conversations between them and brokers at increasing frequency from the early months of 2008. But indeed. when we asked about the concept of consensus trades.46 opposite. This practice can lead to what is known as a ‘short squeeze’: the aggressive buying by hedge funds that have to cover their ‘shorts’. a long-short hedge fund manager. indeed. the popularity of the trade rose and in April 2008 HFM16. In the following months. As seen in the quote. the larger the losses would be. usually. if there are serious concerns about the ability of the borrower to complete their part of the transaction. clearly.
When asked whether they (or their analysts) came up with the idea or if they were introduced to the idea by a broker . When splitting the interpersonal connections according to trading strategies. the VW-Porsche trade was a long-short trade and there were hedge funds that specialized in such trades and brokers who catered for these hedge funds. Most hedge fund managers only talk to people within the same strategy. Also. we discussed the VW-Porsche trade with ten long-short managers of which eight admitted to be either invested in it. The connections among these particular actors were the most active in establishing the trade: Interviewer: Do these[consensus] trades travel across strategies or within strategy? HFM8: Almost always within. or having been invested in it. What do I have to say to an emerging markets guy? Don’t really know what he does and vice versa. we understand that when HFM8 stated that the VW-Porsche trade is ‘brokered by everybody’ it was not merely a figure of speech. The concentration along strategy lines was not limited only to this trade. During our field research. However. what does an emerging market hedge fund care in Porsche? So his broker will not even talk to him about it. we find that only seven of the 101 ties crossed strategies.47 that: ‘I tell you something like Volkswagen and Porsche is brokered by everybody. ‘ These quotes illustrate the concentration and homogeneity of expertise that typify the clusters of hedge fund managers we observed and that play a role in the evolution of the consensus trade. virtually everyone was active in this trade. Knowing this.’ The popularity of the trade. as was reflected in the brokers’ activity was also accompanied by similar activity among hedge funds. Among the other hedge fund managers HFM8 was in daily touch with. in spite of its seemingly general popularity. Those trades are mostly a function of hedge fund managers talking to other hedge fund managers and brokers talking to the same hedge fund managers.
it was speculated that the European Court of Justice would abolish this requirement. to the implications that the ‘VW Law’ may have on Porsche’s intentions. for example. assessments of the likelihood of different scenarios of takeover or merger were done and exact details from the profit and loss accounts of the trade for each of the hedge funds shared. thereby opening the route to hostile takeover of VW. However. Discussion also looked at the section of the ‘VW law’ that required approval by holders of at least 75% of the company’s stocks before domination of a buyer over the company can be established. we witnessed numerous discussions about the VW-Porsche trade among hedge fund managers. It is safe to assume that no broker among the ones we observed and with whom we spoke would spare the amount of hours hedge fund managers dedicate to analyzing the finer points related to the trade or would ask his analysts to devote their time completely to one trade . Again. The discussions within these dense clusters were indeed extensive. the European Commission declared that the law violated EU legislation and it was speculated that the European Court of Justice would invalidate it. different trading strategies had little in common with each other and thus tended not to maintain the embedded ties through which the detailed information was exchanged and joint problem solving took place. where arcane sections of German and EU law were analyzed. During our fieldwork. As HFM8 explains. the ‘VW Law’ capped voting rights of any shareholder in VW to 20%. however. is not a result of simple mimicking behavior. Prior to 2008. all hedge fund managers admitted that the idea had initially come from outside the hedge fund. were virtually inconceivable. regardless of their actual size of holding. Conducting such detailed and lengthy discussions between hedge fund managers and brokers. The topics discussed covered a wide range: different valuations of both companies were presented. This finding. Particular attention was given.48 or a different hedge fund manager. The discussions did not rely only on financial and accounting expertise.
percentage of daily volume) is applied in all cases. in the fear of it being becoming widely available. we asked all hedge fund managers if they treat popular trades any differently from other trades when they assess their risks. When we asked how he made this decision. Following this conversation. The economic incentives that the brokers face patently discourage such level of involvement with a single trade and called. Of the eight hedge fund managers who were actively involved in the trade between January 2008 and October 2008 only one hedge fund manager (HFM17) told us that he decided to unwind his position and terminate the trade because he ‘felt’ that too many hedge funds had the same trade on.49 in order to produce the focused background material. The combination of these factors contributed to the fact that although many hedge fund managers learned about the VW-Porsche trade from a person outside their networks of trusted competitors – usually a broker – the discussions where the information was evaluated and investment decisions were made took place firmly within densely connected groups of trusted competitors. for distributing trading ideas among many prospective clients. it would be safe to say that no hedge fund manager would agree to expose such detailed and sensitive information to a broker. instead. The answers in all cases were similar: no special treatment is given and the same set of measurements (VAR. typically large banks. provide credit to hedge funds and frequently finance their trades. The increasing popularity of the VW-Porsche trade that we witnessed during the first six months of 2008 led us to investigate the potential risks involved in holding such a popular trade. he noted that he simply spoke with several of his competitors and that he did not use any formal risk assessment to come to this decision. We received similar replies from two risk managers who work at two of the largest global prime brokers. scenario analysis. Prime brokers. In addition. These prime brokers also did not distinguish between consensus trades and less popular ones .
September 15th 2008. On Monday. from the way you look at other trades. which is one of the three largest equity long-short hedge funds in . C is probably too. might follow suit and some even feared an immanent collapse of the financial system. PBS1. What we have not looked at so far is if there are certain clusters of clients who tend to have the same trade on and hence if A and B are in a trade. when you asses somebody’s risk? PBS1: That is a good question. what could happen to us if that particular security went to zero. yes and no. as a risk manager. the ‘clusters of clients who tend to have the same trade on’. we were in the offices of HF1. a Hong-Kong based risk manager working for one of these prime brokers explains: Interviewer: Do you. This event led to speculation that Merrill Lynch. Morgan Stanley. 6. these clusters also contributed to the unfolding of the crisis of the VW-Porsche trade.2 Hedge funds’ behavior as the crisis unfolds Let us focus on the behavior of several hedge fund managers. On that morning. a London-based hedge fund. look differently at those [consensus trades]. From a client to client perspective— we treat them exactly the same way using the same kind of parameters as with other trades. but also from our perspective. we do aggregate all the exposures to see what would happen to all different books.’ Our findings indicate that this exact potential risk factor that is not examined yet. 2008. who were deeply involved in the VW-Porsche trade during the week of September 15th. but C might have put on the trade via another prime broker. played a crucial role in the decision making process leading to the adoption of the trades. This knowledge or at least the knowledge of the probability might indeed affect financing or even repo decision. As we see below. Goldman Sachs and AIG. among others. to all different accounts. Lehman Brothers filed for bankruptcy.50 when lending stocks or providing capital on credit.
was very similar to the ones we heard from other hedge fund managers. The price differences of VW and Porsche had been increasing. The rationale behind the trade. . making this trade. HFM16 is a senior hedge fund manager in the firm. HFM16 spent 10 minutes with two of his analysts discussing specific stocks (in the banking and automotive sectors). During the regular 7. After the meeting. Sitting at his desk.51 Europe. as he presented it that morning. One of the large positions in HFM16’s portfolio was the VW-Porsche trade and he monitored it closely.00 o’clock morning meeting. paying particular attention to American banking system and its potential effects on financial markets. notably that Porsche’s valuation compared to VW was unjustified by fundamental market variables and that eventually the relative difference between the VW and Porsche stocks would be smaller. analysts and traders. A senior economist gave a quick briefing on his views on the macroeconomic situation. after which we followed HFM16 to his desk. HFM16 examined the relative prices of VW and Porsche on his Bloomberg screen (Figure 2). in general. a losing trade. at that moment. HFM16 met with other managers.
Looking at the brokers’ reports that were waiting on his desk and in his email inbox. 01 July 2008 – 15 Sept. Most brokers’ reports suggested to continue selling stocks of UK banks that had exposure to CDOs and real estate-based securities (RBOS and HSBC were mentioned) as prices were likely to fall further. 2008.52 Figure 2: BLOOMBERG: Stock price VOW GY versus PAH3 GY. HFM16 did not telephone any of the brokers whose reports he had received (and to which the hedge fund was subscribed). putting the pile of reports to one side and . he saw that brokers gave their views on the potential implications of the collapse of Lehman Brothers on financial markets. The burning question on HFM16’s mind was what was driving this joint movement of the stocks and how the day’s events were likely to influence it. HMF16 flicked through the reports quickly and scanned the list of email without opening them. Instead. That morning.
may be by asking his lawyers or his own prime brokerage?’ HFM16: If I do this. Toward the end of the conversation. that he holds the VW-Porsche trade and that he is losing money on it. HFM6 also held the VWPorsche trade and had similar concerns about the trade. who is a former colleague of HFM16 and was working at a competing long-short hedge fund. HFM16 and HFM6 discussed the likelihood of this happening as well as possible action routes to avoid further losses. Such . However. HFM16 called HFM6. Their main concern was that assets held by Lehman Brothers. a major lender of assets to short sellers. In their talk that morning.53 taking a notepad from the far side of the desk. but he could have easily asked one of the brokers with whom he had connections to look into the matter and return with answers. which was the first of several that day. leading to some assets lending being recalled. they will use it as an argument to other hedge funds to close their positions. we asked HFM16 about this query: Interviewer: ‘Couldn’t one of your brokers look this information up. in effect. HFM16 would be disclosing. which would push the prices of VW stock higher and causing more loses in the VW-Porsche trade. HFM16 and HFM6 discussed the impact of Lehman Brothers’ likely bankruptcy on borrowed VW stock. would be frozen. by directing the query to a broker. HFM16’s queries were not trivial. generating commissions and increase my losses. HFM16 asked HFM6: ‘who could be in the know about that’? When the conversation ended. HFM16 took notes on his notepad. which lasted more than half an hour. referring to one of the details discussed. A quick glance at one of the pages revealed that many sentences were followed by question marks. During the conversation. as we witnessed hedge fund managers do many times.
but also from the institutions that had lend their stock to Lehman Brothers and would not get it back as Lehman Brothers’ assets would be frozen under the British bankruptcy protection laws. who was the source of the interpretation and discussed the matter in more depth. HFM16 continued working on other positions in his portfolio. was someone HFM16 trusted. would use this information to persuade others to withdraw from their own VW or Porsche positions.54 an admission would give the broker a valuable piece of information: a direct indication that HF1. saying he had just spoken with HFM2 about the question HFM16 and HFM6 had discussed in the morning. HFM6. Between the telephone calls. one of the largest hedge funds in Europe. HFM16 predicts. Following this conversation. borrowings recalled and not given back to the lenders because British law. . HFM16 received a telephone call from HFM6. Even before HFM16 heard what HFM2 had to say. HFM16 and HFM2 had all worked together at the same investment bank and knew each other well. Immediately after this call finished. In the afternoon. HFM16 called HFM2. unlike a broker. Such eventuality might trigger not only buying activity from the hedge funds who would have to cover their short position. it was clear that he was relieved to hear the identity of the person with whom HFM6 shared the query. if circulated widely enough. The broker. different functionaries in the hedge fund were drifting in and out of the office. HFM2. but would also. which would be applied in the case of borrowings initiated by Lehman Brothers’ UK branch. HFM6 related to HFM16 that HFM2 had told him that their concerns were justified: there was a risk that Lehman Brothers’ assets will be frozen. increase the losses of HFM16 and everyone else holding the position. is involved in the VW-Porsche trade and that it is likely that the position is at a loss. did not allow ‘ring fencing’ of customers accounts in the case of bankruptcy. collecting the hurriedly written notes where HFM16 asked for more information and reporting on the progress and set backs in other trades. a step that would generate execution fees for the broker.
during one of the most dramatic and volatile days in financial markets in recent history. In summary. while opening their considerations to further examination and scrutiny. The above description of the day captures a focused. He believed that a recession was now unavoidable. ‘micro’. HFM16 called back to HFM6. HFM16 was less inquisitive. see both hedge fund managers still holding (a slightly reduced) VW-Porsche trade at the end of the day. one that he believed was supported by what he heard from HFM2. The outcomes of the decision making process.55 Following the conversation with HFM2. We see that the hedge fund managers develop the evaluative framework jointly. version of the phenomena that were described in the previous section from a broader perspective. although they differ for HFM16 and HFM6. Brokers. in contrast. He chose to withdraw from part of the position. HFM6 disagreed with this course of action. they also share and examine potential reactions to the issues they identified and then make decision. HFM16 listed to HFM6 several steps that he thought were appropriate and asked for his opinion. again. This time the conversation was of a slightly different nature. although he did so at a considerable loss. Following this joint process. managers of some of . the unresolved queries he and HFM6 had in the morning were now answered and it was time to choose and implement a course of action. HFM16’s preferred move. are strictly excluded from participating in the discussions and even information produced by brokerage houses (the analysts’ report) is given only superficial attention. was to buy some call options on the VW stock. This decision making ‘forum’ is composed exclusively from competing hedge fund managers who trust each other and share detailed and sensitive information. through their discussions they decide what factors are relevant for assessing the risk embedded in the trade. These options would pay if VW stock continued to rise and would thus compensate for the losing short position. the rising VW price would soon reverse and that VW and Porsche would continue to move in the same direction. That is.
intensified the crisis? To answer this question we need to understand what information could benefit hedge fund managers when adopting and holding the VW-Porsche trade. as a result. which leaves borrowers (short sellers) with many opportunities to buy them back. we still need to ask if this is indeed a case of over-embeddedness and structural secrecy at work that led to ignoring important information. as the analysis of the events of 15 September 2008 show. bought back at a later stage and returned to the lender. while the number of stocks accumulated by Porsche and were put ‘out of the market’ played a crucial role in determining the second amount. the popularity of the long-short trade determined the number of VW stocks that were tied in a short sale. did brokers circulate relevant information or interpretative frameworks that were ignored or overlooked by hedge fund managers and. However.6 million to more than 5 million shares in VW. That is. A crucial factor for assessing and making decisions regarding a long-short trade is the difference between two amounts: the volume of stocks that are ‘free floating’ – available for trading in the market – and the volume of stocks that are tied to a short trade. the ‘short’ side of the position is based on borrowed stocks. In the case of the VW-Porsche. between June and September 2008 the amount of VW stocks that were tied to short sales rose more than three-fold. hence. since most investors tend to hold the stock. which are sold. Corresponding with our findings about the increasing popularity of the trade. A hedge fund manager involved in the trade. from about 1. As explained above. the VW-Porsche trade had several unusual characteristics. While this analysis demonstrates the relative isolation in which decision making in hedge funds takes place. At that point.56 the leading long-short hedge fund managers made investment decisions while focusing exclusively on information and advise from a small group of trusted competitors. Such a possibility is remote. would be interested in information about the possibility that there will not be enough stocks available at the market and would prevent him from unwinding the trade. 13% of .
This amount meant that. The amount of Volkswagen stock borrowed. the highest ratio among the 30 stocks in the DAX Index. In practice. However. 2008 . June-August 2008 (Source: Data Explorers). only a fraction of the free float is for sale since many shareholders do not have their holdings up for sale. in its semi-annual report from March 4th.57 VW shares were tied to short sales. June-August 2008 Number of stocks 6000000 5000000 4000000 3000000 2000000 1000000 0 30/06/2008 14/07/2008 28/07/2008 11/08/2008 25/08/2008 08/09/2008 22/09/2008 FIGURE 3: Amount of VW stock in short sales. which held 20%. theoretically. Determining exactly how much of VW stock was free floating was difficult to establish. at least 13% of the VW stock had to be available for trading for all investors holding the VW-Porsche trade to be able to unwind their positions. with 31% of stocks and the German state of Lower Saxony. VW’s two largest shareholders were Porsche.
These two announcements were interpreted by many of the hedge fund managers we observed as indication that an immanent takeover of VW by Porsche is not likely and. by a corporate statement where it was clarified that Porsche did not seek a domination position in VW and therefore the probability of Porsche raising its take in VW to 75%. Using options to buy shares may help buyers to avoid reporting an increase in holding. that the risk of not having enough free floating stock to unwind the long-short position was low. was ‘very small indeed’. This announcement was accompanied. Less than a month later. however. a share size required for obtaining a domination agreement with VW. stated: Our impression is that VW is a consensus short in the auto sector. published on February 26th. the same analyst restricted the implied prediction he provided before and predicted that Porsche was not likely to increase its holding in VW beyond 51% (VW Focus 070308).58 Porsche announced its intention to ‘acquire the majority shareholding in Volkswagen’. analysts working for brokerage firms speculated about the ways in which Porsche built its stake in VW and offered interpretations about its implications. while going long or . As early as February 2008. Given this. an analyst at Citi Investment Research (VW Note 190208). noted that the price movements in VW stock corresponded with an options’ buying program that Porsche initiated in September 2005 and hinted at the possibility that Porsche was using options to gain control over VW stock. we examined messages circulated by brokers in the 10 months leading to the crisis. John Lawson. 2008 by Lothar Lubinetzki at MainFirst Bank AG. as a result. as the options do not constitute an actual stock transaction. a leading car industry analyst. did information or interpretations regarding the amount of free-floating VW stock was available to hedge fund managers in the months leading to October 2008? To answer this question. a week later. but only a potential one in the future. Another report.
can be tied to the accumulation of VW by Porsche and to the risk of not having enough free-floating stocks. dropping dramatically below the number of shares already borrowed on loan. Borrowing VW ords [ordinary shares] does not seem to be difficult.59 over weighting Porsche also seem to be a trade investors like to do. the report speculates. but even at the time. Brokers’ reports were not the only source for warning signs about holding a short position in VW stock. there would be a substantial risk of a short squeeze. such a dramatic decrease in stock availability was a clear warning signal to anyone holding a short VW position indicating that the level of risk associated with that position had increased significantly. from about 42 million shares to 33 million shares. this price discrepancy. 2008. The market research company Data Explorers notified to its subscribers that the supply of VW stock available for trading (or lending) diminished at September 10th. However. we do not believe that anybody except Porsche really understands why VW’s share price is so stable at EUR 150. The main motivation for the VWPorsche trade lies in the unexplained fact that the price of the VW stock was ‘stable’ at 150 Euros. The question to be asked is who is lending out the shares? It is just a possibility. (VW_Q4 2007 preview 26_02_08 shorting VW ords too risky) This report ties together the two factors that stand at that basis of the VW-Porsche trade: the arbitrage opportunity and the trade’s major source of risk. In retrospect. However. . it is obvious that these shares were bought by Porsche or on its behalf. but how would the picture change if Porsche or banks who are supposed to hold VW ords on behalf of Porsche decided to make some extra money by lending out VW shares? If this was the case and if Porsche decided to call in its VW ords. we believe that shorting VW ords could be very risky. […]To be frank. because institutional investors began selling shares to institutions that kept them out of the market (Figure 4).
Many of the analysts who published reports during this period. hedge fund managers paid dearly for their decision-making practices. A much more reasonable explanation.60 Figure 4: VW shares: lendable and on loan. As it turned out. is that hedge fund managers ignored the reports and excluded them from their analyses when assessing the merits and risks of the trade. are prominent car industry analysts whose predictions and reports are circulated widely. including all the analysts cited above. April – October 2008 (Source: Data Explorers) The reports presented above are a small sample from a much larger dataset of documents and messages in which brokers (or analysts working in brokerage houses) communicated about potential risk involved in holding short positions in VW stock. . it is inconceivable that all this information was simply overlooked by hedge fund managers. considering our other findings about structural secrecy that governed the exchange of information between hedge fund managers and brokers and the fact that the VW-Porsche trade rose in popularity exactly at the period when information about the trade’s risk was abundant. Given this fact.
. whilst VW prices reached record heights. 2008). the total amount of shares borrowed stood at 13%. like many other longshort hedge funds. as. resulting in the price of VW stocks rising more than 6-fold in a few days (figure 5). its announcement led to the realization of the risky scenario described above.6% of VW shares and that it had acquired options for additional 31. the implication of the announcement was that only 5.61 On October 28th. As mentioned above. which meant that many of the investors who held short positions would not be able to return the shares to the lenders. given the scarcity of VW stock. increased the concerns and drove even more lenders to ask for their VW shares. Although Porsche stated that it made the announcement to give investors “the opportunity to close their positions unhurriedly and without bigger risk” (Story et. drove the prices up. 2008 Porsche announced that it owned 42.1% of the shares. lost a substantial amounts of money due to the reactions to Porsche’s announcement.5% of the shares. in turn. concerned about the ability of the borrowers to return the stocks under these distressed conditions asked for the stock to be returned immediately. Porsche’s price stayed relatively stable and the discrepancy between the prices of the two stocks grew. 74. Adding to this figure the 20.1% of shares owned by the German state of Lower Saxony. Porsche controlled. These requests. al. The hedge funds we observed and that were involved in this trade. Lenders of the VW stock.8% of the shares were available for trading. effectively. This. Together. were they asked to do so.
62 FIGURE 5: BLOOMBERG: Stock price VOW GY versus PAH3 GY. It is all exploding in our face’. every hedge fund manager I know is screaming for [Volkswagen] stock and just can not get any. 01 July 2008 – 28 Oct. We were present at one of the hedge funds that held the VW-Porsche trade in late October. 2008. when it became painfully apparent that the impact of the crisis was related directly to the structure of connections among hedge fund managers: ‘The problem is that we are all positioned the same way. [HFM16] .
Yet. many warning signs existed about the potential risks of the trade. analyzed and scrutinized in the dense clusters of hedge fund managers. but shallow. but a fundamental component of the decision making process that enables consensus trades. when these frameworks were accepted and. Our observations indicate that these groups were effective in developing and honing interpretative frameworks that justified trades. the emergent agreement made it easy for hedge fund managers to reject and ignore . there was an element of a ‘Black Swan’ surprise (Taleb. However. in this respect. as seen earlier. sharing the same strategy and relying on common occupational backgrounds where information was filtered and assessed. The relative homogeneity that HFM16 observes (“…every hedge fund manager I know…”) is not simply a characteristic related to the popular trade.63 It is true that Porsche’s share holding size of VW. crucially. but as the evidence shows. as it was revealed in late October 2008 was unexpectedly large and. belonged to tightly-knit cluster of hedge fund managers who created and maintained. hedge fund managers knew that a fundamental uncertainty shadowed over the VW-Porsche trade during the year when it gained popularity: it was not clear why the discrepancy between the stock prices of VW and Porsche remained in spite of it being a glaring arbitrage opportunity. the adopted trade proved to be profitable. Why did sophisticated and knowledgeable investors like HFM16 miss this information? HFM16. Hedge fund managers created homogenous groups. Porsche accumulated the stock secretly and even hid its tracks. leading to the making of decisions about the desirability of investment ideas. The initial investment idea is transported by the efficient. 2007) in the situation: an extraordinary event that defies the conventional thinking. a distributed decision making process regarding the VW-Porsche trade. Porsche’s announcement resolved the mystery: VW’s price did not drop because Porsche’s acquisitions gradually decreased the supply of stocks. in effect. hub-like connections of the brokers and then it is discussed.
2007): low-probability. surprise organizations and these surprises are the fundamental source of the high impact of such events. in advance.g. but that they construct decision making structures and practices that frame. Black Swans. 7. information that can warn against the negative events. in advance. especially when brokers were the source of that information. The recent financial crisis brought to the fore intense criticism of investment decision making in financial institutions. amplified the impact of Porsche’s announcement. One of the popular arguments is that financial organizations tend to rely on stylized representations of the past (e. Such planning can lead to disastrous consequence because financial markets are prone to Black Swans (Taleb. the ways they evaluate information. consequently.64 conflicting information about the trade.. and when these do occur. Conclusion Can the VW-Porsche crisis be understood as a case of organizational deviance? We believe so. Organizations do not identify or develop. that financial decision makers find themselves surprised because of a lack of relevant information. hedge funds) do not simply fail to identify and gather relevant information. The structural informational segregation between hedge fund managers and brokers contributed to the increasing popularity of the VW-Porsche trade and. normal distributions) when planning future scenarios. the resulting damage is significant. is partial. This conclusion has wider implications than analyzing decision-making practices among hedge funds. thus. Hedge fund managers preferred to include into their . high-impact negative events that challenge such pre-existing assumptions. Our findings indicate that financial institutions (in our case. Our analysis shows that the Black Swan’s fundamental hypothesis.
we show that in case of hedge funds this ‘agencement’ might very well consist of human bodies. 2003. . tools. etc. then.. that it is also distributed over the control centers of other ships.65 assessments information and analysis from other managers over information from brokers. Whilst Hardie and MacKenzie (2007b) show that a hedge fund is comparable to Hutchins’ control centre at a US warship where cognition is distributed over the navigators. This paper contributes to the sociology of finance and especially to the stream within it that focuses on the importance of inter-organizational connectedness (MacKenzie. the decision making process leading to choosing a specific trading position is not seen as a potential source of risk and is not incorporated into the risk assessment. Furthermore. While Callon (2005) states that an actor ‘… is made up of human bodies but also of prostheses. Beunza and Stark. the source of financial risk is the interaction between the trading position chosen by the investor and the market’s behavior. technical devices. to continue the metaphor. email messages and telephones located at different and competing hedge fund or brokerage firms. although the latter contained vital details about the risks of the trading position. That is. According to this paradigm. skippers. Our analysis. equipment. our findings inform financial regulation and lawmaking.20 We show that the inter-organizational nature of financial decision-making can be a major source of risk and thus. algorithms. 2010) by demonstrating the intertwined nature of decision-making and the structure of connections among financial actors. this paper also expends the notion of materiality of markets. plotters and charts. we show. We see that decision making in hedge funds is unbounded by the contours of the single hedge fund. criticizes the leading paradigm in today’s financial risk management. etc.’. Finally. analyzing and understanding this arena is vital for assessing the risks facing financial organizations. while regulation 20 Possible exception to this view within mainstream financial risk management is operational risk.
but equally on the networks to which these hedge funds belong.. such as the one analyzed in this paper call for thinking in this direction. and Adler. (1990) ‘Market Networks and Corporate Behavior’. C. Anderson. Baker. being ‘too big to fail’ could be a function of (belonging to) a network. 37-66. the properties of the actors in the network and the properties of the investment when assessing risks. Risk-aware financial regulation. therefore. ‘A p* primer: Logit models for social networks’. initiatives such as the Hearing on Regulation of Hedge Funds’ organized by the U. S. House Oversight and Government Committee on November 13 2008. Operating such regulatory framework seems unrealistic today. For example. References Adler P. P. (1984) ‘The Social Structure of a National Securities Market’. but the emergence of networked risks. Thus.S..66 tends to focus almost exclusively on the single financial institution as its unit of analysis and enforcement. The focus of the said regulatory framework would not only be on hedge fund managers. 589625. are necessary but fail to incorporate the risks introduced by the network structures. W. . supervise them and where necessary intervene in them. 775-811. (1984) The Social Dynamics of Financial Market’. & Crouch. Baker. should incorporate the properties of the network. Failing to refer to the inherently networked nature of investment decision making ignores the potential systemic risks that these networks entail. JAI Press. (1999). Social Networks. the entities and people which service them or relationships between these financial entities and people. American Journal of Sociology. The regulator should then be equipped with rules that allow it to uncover these networks. Greenwich. hedge funds. Wasserman. B. 21. W. American Journal of Sociology.J. not necessarily of the size of the financial organization.
P. Law. U. Greenwich Associates. B. Ronald S.(1979). report published August 2. G. Ronald L. I. 82-111. A. and Systemic Risk: A Critique of Behavioral Finance’ (April 19. and MacKenzie. Industrial and Corporate Change. Cambridge: Harvard University Press. Sociological Review. (1970) Exit. Burt. and Bruegger. Lemmings or Frame-makers? The Intermediary Role of Securities Analysts in Market Devices’ Ed. Latour. Beunza. Porsche AG. Organizations. Boston: MIT Press. Institutional Investor. A. Action and Belief. Granovetter. Latour. 2010). (1986) ‘Some Elements of a Sociology of Translation: Domestication of the Scallops and Fishermen of St Brieuc Bay’. (2002) ‘Global Microstructures: the virtual societies of financial markets’. (2007b) ’Constructing the Market Frame: Distributed Cognition and Distributed Framing in Financial Markets’. M. and others vs. Hirschman. Princeton: University Press. 905-950. Routlage and Kegan Paul. London: Sage . and Garud. January 25. Administrative Science Quarterly. Daniel and Stark. Callon. 215-239. 1-58. D. Millo and Muniesa: Blackwell Publishing. 505–526 in Handbook of Data Analysis. Ed. Hardie. (1986) Laboratory of Life: The Construction of Scientific Facts. (1977) A set of measures of centrality based on betweenness. D. and Woolgar. I. Elliot Associates. Callon. 2010. (1992) Structural Holes. 35--41. and MacKenzie. S. “Hedge Funds: The Comeback”." Pp.com/abstract=1285054 Booth. or Feeble?’. Ed. (1999) ‘The Search-Transfer Problem: The Role of Weak Ties in Sharing Knowledge across Organization Subunits’. and Guattari. (1992). M. Centrality in social networks: Conceptual clarification. Social Networks. (2007) ‘Calculators. Callon. 1463-1484. D. A. R. F. Deleuze. . and Stark. Reflexivity. 389-403. Available at SSRN: http://ssrn. (1995) Cognition in the Wild. Callon. American Journal of Sociology. By Callon. M. 57-80. E. "The Analysis of Social Networks. edited by Melissa Hardy and Alan Bryman. K. Sociometry 40. MA: Harvard University Press. (2007a) ‘Assembling an Economic Actor: the Agencement of a Hedge Fund’. Economic Sociology: European Electronic Newsletter 6/2. 2010. US District Court for the Southern District of New York. (1992) ‘Network dyads in entrepreneurial settings: A study of the governance of exchange processes’. Destructive. 2004. D. MA: Harvard University Press. Hardie. Cambridge. Administrative Science Quarterly. 12–16. 485-510. (2005) Why Virtualism Paves the Way to Political Impotence: A Reply to Daniel Miller’s Critique of The Law of Markets. in The Law of the Markets. 369-400. Voice and Loyalty: Responses of Decline in Firms. L. in Power. 1(3). (1998) ‘Introduction: the Embeddedness of Economic Markets in Economics’. (1987) Science in Action. Journal of Economic Literature. (1982) ‘Rival Interpretations of Market Society: Civilizing.67 Beunza. New Political Economy. Blackwell publishers. Larson. (1980) ‘Mille Plateaux (Capitalisme et Schizophrenie 2)’. Beunza. C. M. Structural Holes: The Structure of Competition. MA: Harvard University Press. (1985) ‘Economic Action and Social Structure: the Problem of Embeddedness ’. M. Freeman. David. London. Hansen. and States. . Knorr Cetina. Breiger. (2004) ‘Tools of the Trade: the Socio-Technology of Arbitrage in a Wall Street Trading Room’. R. D. Tamzin (1998) ‘Why Costas Kaplanis had to sell’. 76-104. ISBN 0-674-84372-X Burt. B. 196-233. Hirschman. L. ‘Models. American Journal of Sociology. Cambridge. Paris : Editions de Minuit. Cambridge. Hutchins. 13-39.
D. B. Y. Snijders. Uzzi. 28 October 2008. Callon. P E Pattison. Blackwell Publishing. American Journal of Sociology. Uzzi. (2003) ‘Relational Embeddedness and Learning: The Case of Bank Loan Managers and Their Clients’. 383-399. (2005) ‘Managing Your Portfolio Connections. and Fowler. Annual Review of Sociology. How Financial Models Shape Markets. J. Millo. Blackwell. D. and Schoorman. T. C. Story. Vaughan D. Misconduct. 2006. (2003) ‘Constructing a Market. 349-380. G Robins.”. Naming and Topology’. (1952) ‘An Integrative Model of Organizational Trust’. B. Economy and Society. Robins. 1-14. G L Robins. and Millo. R. Chicago: Univ.68 Law. ed. and Dougherty. Lawrence..(2007) ‘Making Things Deliverable: The Origins of Index-Based Derivatives’. Miller. and Disaster. de la Merced. “Pnet: a program for the simulation and estimation of exponential random graph models. Morse E. MacKenzie. November 2-4. Boston: MIT Press. 709-1995. in Actor Network Theory and After. Chicago Press Vaughan.. T A B. and Lancaster. New York Times. Wang. The Challenger Launch Decision. D. “The Dark Side of Organizations: Mistake. Uzzi. (2008) ‘Institutional Entrepreneurship and the Emergence of Hedge Funds: The Rich Historical Case. (1999) ‘Embeddedness in the Making of Financial Capital: How Social Relations and Networks Benefit Firms Seeking Financing’. 2006. leverage. S. 2009. American Sociological Review’. . “Panicked Traders Take VW Shares on a Wild Ride”. Performing Theory: The Historical Sociology of a Financial Derivatives Exchange’. R. (1999) After ANT: Complexity. Oxford. Law and Hassard.” University of Melbourne. Millo and Muniesa. J. report of the President’s Working Group on Financial Markets. MacKenzie. in Market Devices. “New Specifications for Exponential Random Graph Models. Mayer. 35-67. (2000) ‘For a Sociology of Worth’. working paper. (1997) ‘Social Structure and Competition in Interfirm Networks: The Paradox of Embeddedness. 25: 271-305. MacKenzie.” Sociological Methodology 36:99-153. Administrative Science Quarterly. Y. President’s Working Group on Financial Markets (1999) ‘Hedge funds. (2003) ‘Long-Term Capital Management and the Sociology of Arbitrage’. 674-698. P. (1996). B. Davis. 59-65. G. L. Not a Camera. 107-145. Management Science. Berlin. (1996) ‘The Sources and Consequences of Embeddedness for the Economic Performance of Organizations: The Network Effect’. 481-505. and the lessons of Long-Term Capital Management’. American Sociological Review. 196-214. M. connectivity and degrees: New specifications for exponential random graph (p*) models for directed social networks. “Closure.. (1999). P Pattison. D. Stark. J.” Social Networks 31:105– 117. Academy of Management Review. D. Uzzi. B. Ed. Keynote address for the Meetings of the European Association of Evolutionary Political Economy. and P Wang. and M S Handcock.. (2006) An Engine. and P Pattison. MIT Sloan Management Review. D.
B TR/HFM S.6 5.69 Appendices Appendix 1: Details of persons who were interviewed or observed WHO PBS1 TRS1 HFM1 TRB1 HFM2 BR1 BR2 ANAS1 HFM3 TRS2 TRS3 HFM4 HFM5 BR3 HFM6 TRS4 BR4 HFM7 TRB2 HFM8 HFM9 TRB3 BR5 TRB4 HFM10 HFM11 HFM12 HFM13 HFM14 BR6 PBS2 BR7 BR8 HFM15 PBS3 HFM16 HFM17 SELL/BUY WHAT SIDE PB S TR S HFM B TR B ANA/BR/HFM S.B BR S TR B HFM B HFM B TR/HFM S.B TR/BR S.B BR/PB S.B BR S BR S ANA S ANA/HFM B TR S ANA/TR S TR/HFM B ANA/HFM S.5 5 5 9 2.12 10 15 10.6.13 15.6 11 15.10 7.10 5 18 15 17 15.S.S TR/HFM S.7 10 2.4 11.10 12.2 9.3.S PB S TR/BR S.10 13 4.6 11.S BR S TR/HFM S.B YEARS EXP 10 17 10 3 4.B TR/HFM S.6 12.S.B BR/HFM S.B BR S TR/HFM S.B TR/HFM S.B TR S.B TR S.B ANA/BR/HFM S.B TR S ANA/BR S TR/HFM S.6 8 14.7 .12 5.10 6.5 10.10 5.
B S S.B S.B B.: Years of experience in this function . ANA: Analyst.B B S.B S.4 3.5 Explanation: Who: Code What: PB: Prime Broker.5 5.10 6.B S.B 16.5 9 10 10 6 6.S B S B S.10 10 10. TR: Trader. HFM: Hedge Fund Manager.70 TRB5 TRB6 HFM18 HFM19 HFM20 BR9 HFM21 ANAS2 ANAB1 BR10 HFM22 ANAB2 ANAB3 BR11 TRS5 BR12 PBS4 HFM23 HFM24 TRB7 ANAB4 TRS6 HFM25 TR/TR TR/TR HFM TR/HFM ANA/HFM BR ANA/HFM TR/ANA ANA BR TR/HFM ANA ANA BR TR BR PB ANA/HFM TR/HFM TR/TR ANA TR BR/HFM S.3 4 7 10. BR: Sales person or salestrader.8 17.4 13.B S.B S S S S S.5 5.B B S S. Sell/Buy: S: Sell-side. B: Buy-side Years Exp.B S.5 5 7 20.3 10.6 23 10.
) 02-Jan-08 16-Jan-08 30-Jan-08 13-Feb-08 27-Feb-08 12-Mar-08 26-Mar-08 09-Apr-08 23-Apr-08 07-May-08 21-May-08 04-Jun-08 18-Jun-08 02-Jul-08 16-Jul-08 30-Jul-08 13-Aug-08 27-Aug-08 10-Sep-08 24-Sep-08 08-Oct-08 22-Oct-08 05-Nov-08 19-Nov-08 03-Dec-08 0 50 100 150 200 250 300 350 400 450 500 Closing Price Turnover Share Price and Stock Turnover. (Source: Share Turnover (%) 71 . 2008.Volkswagen Appendix 2: Share Price and Stock Turnover – Volkswagen.000 100 200 300 400 500 600 700 800 900 0 DataStream.Price (Euros) 1.
446 0.256 -0.679 0.057 -0.061 0.011 0.175 0.307 18.362 646.148 2823.896 with a standard deviation of 10.613 39.747 91.023 197.273 483.415 106. the better the model fit.032 0. 000 simulations and a sample of 1.642 3.071 stddev 10.631 409.032 -0.298 0.849 91.649 68.00) AC(2.558 7. alternating stars.286 0.00) AT(2.608 0.498 139.917 22.401 48.746.038 0.746 134.132 0.143 0.972 1024.229 -0.619 -0.706 4.736 77.393 0.321 0.472 1301.53 2. The measure which has the worst t-value is HFM_t3u.036 0.958 6. the number of edges in the observed network is 70 and the mean of the MCMC samples is 69.089 115.275 0.201 0. the measure observed in the network should be close to the mean of the sample.135 2.095 0.851 253.343 0.00) HFM_t3u HFM_t2u HFM_t1u HFM_o3u HFM_o2au HFM_o2bu HFM_o1au HFM_o1bu HFM_interaction HFM_activity Std Dev degree dist Skew degree dist Global Clustering Mean Local Clustering Variance Local Clustering observed 70 435 997 1803 2585 50 12 0 0 0 161 351 2409 184 187.138 0. The estimated parameters of the model (such as edge.174 -0.336 17.001 0.102 209.661 0.72 Appendix 3: Goodness of fit tests for ERGM Model 1: Goodness of fit is based on one million Monte-Carlo simulations with a burn-in of the first 100.72 6.313 214.456 3078. The t-ratio is the difference between the observed value and the mean divided by the standard deviation.032 0. alternating k triangles etc) are emphasized in the table and are expected to be below 0.697 93.015 -0. For example.657 -0.1.013 0.246 12 278 8 59 101 64 238 273 209 673 23 92 3.453 .01 -0.00) AET(2.631 300.473 263. effects edge 2-star 3-star 4-star 5-star triangles 4-clique 5-clique 6-clique Isolates Triangle2 Bow_tie 3Path 4Cycle AS(2.345 0.056 0.148 7.706 1947. The smaller the t-value.613 0.714 0.893 24.198 186.00) A2P(2.742 0.274 91.79 77.151 -0.706 0.825 64.534 21.005 0.13 188. For a model to fit well.09 47.896 436.111 -0.906 300.238 0.148 2589.056 1189.722 0.021 t-ratio 0.181 0.91 91.000 networks.295 33.256 335.002 0.371 -0.057 46.081 mean 69.679 14.404 0.61 0.627 0.183 3. This is the measure of the number of triangles in the network that consist of exactly three hedge fund managers.024 0.
453 190.81 675.00) AC(2.1 1126.064 -0.00) A2P(2.509 0.526 24.041 0.962 306.089 -0.158 108.079 -0.393 0.745 17.000 networks.00) AT(2. 000 simulations and a sample of 1.073 1925.00) AET(2.324 0.081 mean 70.906 300.73 Model 2: Goodness of fit is based on one million Monte-Carlo simulations with a burn-in of the first 100.274 91.097 0.648 0.167 0.124 -0.061 0.136 426.864 1891.253 -0.187 278.986 10.174 49.345 0.419 86.035 -0.114 -0.496 7.655 981.4 0.176 2849.043 -0.00) HFM_t3u HFM_t2u HFM_t1u HFM_o3u HFM_o2au HFM_o2bu HFM_o1au HFM_o1bu HFM_interaction HFM_activity Std Dev degree dist Skew degree dist Global Clustering Mean Local Clustering Variance Local Clustering observed 70 435 997 1803 2585 50 12 0 0 0 161 351 2409 184 187.456 7.649 221.564 31.563 0.141 0.681 125.118 -0.739 60.174 0.022 t-ratio -0.072 0.627 0.91 76.067 17.14 0.307 93.083 -0.371 0.195 -0.389 196.001 0.228 -0.175 0.439 41.698 60.889 7.744 23.127 -0.651 1.328 0.147 -0.246 12 278 8 59 101 64 238 273 209 673 23 92 3.011 268.046 -0.559 92.201 0.009 0.194 7.239 -0.114 0.057 -0.349 39.009 0.084 -0.894 449.368 155.922 191.627 0.014 -0.016 0.454 102. The t-ratios have now improved by a factor of 10 from the goodness of fit values of model 1. effects edge 2-star 3-star 4-star 5-star triangles 4-clique 5-clique 6-clique Isolates Triangle2 Bow_tie 3Path 4Cycle AS(2.088 408.075 stddev 10.044 0.081 10.584 317.118 0.042 -0.137 240.243 3.381 -0.489 -0.888 60.341 0.286 1049.094 105.629 123.774 43.698 7.244 .587 91.793 2665.