Are Hedge Funds Registered in Delaware Different?

Transcription

1 Are Hedge Funds Registered in Delaware Different? Abstract Over 60% of U.S. hedge funds choose to register in Delaware, even though 95% of those are physically located and managed elsewhere. Delaware hedge funds exhibit significant differences in contractual structure in terms of higher management and incentive fees. Delaware funds are more likely to use high watermark provisions and less likely to invest their personal capital. Both the redemption notice periods and lock up periods are significantly longer for Delaware hedge funds. While Delaware hedge funds do not outperform or underperform funds registered elsewhere, fund flows are more sensitive to Delaware funds prior performance and Delaware funds are more likely to be liquidated due to poor performance. Further, Delaware funds are more likely to increase risk after poor absolute performance. All these results hold even after we control for the endogenous choice of registration. Keywords: Hedge Funds; Delaware; Law and Finance JEL Classification: G23, G24, G28, K22

2 2 1. Introduction Hedge funds are essentially pools of capital that are typically sourced from institutional investors and high net worth individuals (Brown et al., 1999, 2001, 2008; Fung and Hsieh, 2000, 2002, 2006; Hodder and Jackwerth, 2007; Agarwal et al., 2009, 2011). Hedge funds may select any jurisdiction to be registered, although inevitably the jurisdiction chosen will be determined by more practical, operational factors. For example, the scope of investors that may invest in a hedge fund depends on the regulatory landscape in which the fund is registered (Cumming et al., 2013), therefore fundraising will be a factor in jurisdiction choice. While prior work has compared offshore and onshore fund choices (Aragon et al., 2013; Endelman et al., 2013), prior work has not studied legal registration choice within the U.S. This question is important due to the legal implications associated with Delaware limited partnerships and Delaware limited liability companies identified in practice (Schwartz, 2012), and the practical reality that over 60% of U.S. hedge funds are registered in Delaware despite the fact that 95% of these funds are physically located and managed elsewhere, as documented herein. This paper represents a first law and finance analysis of registering in Delaware for hedge funds. While the focus of most Delaware related research has centered on the question of whether state competition has been good or bad for shareholders (e.g., Romano, 1985), in this paper we seek to ask whether state competition has been good or bad for fund investors, more specifically hedge fund investors. We address the question of whether or not there are differences in the structure and performance of hedge funds that incorporate in Delaware and how these differences may exacerbate or reduce agency behaviors of hedge fund managers.

3 3 What makes an analysis of hedge fund incorporation choices unique and important? First, investors into hedge funds are among the most sophisticated in the financial world. To this end, we would not expect hedge funds in Delaware to earn superior returns (even though Delaware companies are valued higher on average; Daines, 2002), because if there is a premium associated with Delaware law, it will be well-known to hedge fund investors, thereby increasing their price and lowering their return. By examining hedge fund performance among Delaware funds versus other funds, we provide a measure of the degree to which sophisticated investors into hedge funds are familiar with Delaware law. Second, and more importantly, the familiarity with Delaware Law means that investors from a diverse set of states and even countries will be on more equal footing and have a more common understanding about the structure and governance of Delaware hedge funds relative to other hedge funds. As a result, we expect that Delaware hedge funds are more likely to have a more pronounced flow-performance relationship. Investors will be more inclined to invest in the future when a Delaware fund does well, and conversely more inclined to withdraw funds in the event of poor performance, because there are more investors that are better aware and more fully informed of the causes and consequences of investing in a Delaware fund relative to non- Delaware funds. At the extreme end of the spectrum, we expect Delaware hedge funds are more likely to be liquidated or liquidated more quickly in the event of very poor performance. Third, because Delaware investors face less legal uncertainty with the legal and governance structure of Delaware funds, we conjecture that such investors will be more willing to invest in funds that provide Delaware managers with higher fixed and performance fees. Relatedly, investors will be less inclined to require Delaware managers to invest their own capital into the fund that they are managing. Also, investors into Delaware hedge funds will be

4 4 more likely to accept longer redemption notice periods and lock-in periods relative to non- Delaware funds given the greater and more transparent legal environment among Delaware funds. We test these three propositions with the Lipper/TASS database over the period We provide evidence that is highly consistent with these three propositions. Our evidence is robust to a number of controls, including but not limited to the endogenous selection of Delaware as the choice of jurisdiction. In modeling the choice of jurisdiction we control for a number of things such as the actual physical location of the fund and market conditions, as discussed herein. The stronger flow-performance relationship and the more manager-friendly fund structures (high incentives, high water mark, less personal capital, longer lockup periods and redemption periods, etc.) for the Delaware hedge funds as documented in this paper have implications for the potential agency conflicts of hedge funds. In the second part of the paper, we examine the implications of Delaware registration for two specific agency conflicts that are well documented in the hedge fund literature, risk shifting (e.g., Brown, Goetzmann, and Park, 2001; and Agarwal, Daniel, and Naik, 2002; Basak, Pavlova, and Shapiro, 2007; Hodder and Jackwerth, 2007; Panageas and Westerfiled, 2009; Aragon and Nanda, 2012) and return manipulation (e.g., Bollen and Pool, 2008, 2009 and Agarwal, Daniel, and Naik, 2011). 1 1 Our analyses likewise builds on a large and growing literature on hedge fund structure and performance (Ackermann, McEnally, and Ravenscraft, 1999; Agarwal and Naik, 2000a, b, 2004; Agarwal, Daniel, and Naik, 2006; Amin and Kat, 2003; Baquero, Horst, and Verbeek, 2005; Brown, Goetzmann, and Ibbotson, 1999, 2001; Brown and Goetzmann, 2003; Brunnermeier and Nagel, 2004; Cremers, Martijn, and Nair, 2005; Edwards and Caglayan, 2001; Getmansky, 2005; Getmansky, Lo, and Makarov, 2004; Liang, 1999, 2000, 2003; Gupta and Liang, 2005, Teo, 2007), as well as hedge fund activism (Brav et al., 2008a,b; Klein and Zur, 2009) and the structure of hedge funds and strategies (Ding, Getmansky, Liang, and Wermers, 2006; Fung and Hsieh, 1997, 2000, 2001; Goetzmann, Ingersoll, and Ross, 2003; Jorion, 2000). Our analyses are also related to analyses of hedge fund share restrictions (Aragon, 2007) and hedge fund registration (Brown, Goetzmann, Liang, and Schwartz, 2008). Our

5 5 Incentives to take risk and manage returns can arise from the strong flow-performance sensitivity as investors direct more money into hedge funds that outperform. In contrast, the higher incentive fee and the more frequent usage of the high watermark provision among Delaware funds indicate less risk-shifting behavior than other funds. Longer lockup periods and redemption periods indicate that investors cannot withdraw their capital immediately in response to funds poor performance. Therefore, Delaware hedge fund managers face smaller pressure in terms of manipulating short-term performance. Our empirical evidence shows that hedge funds registered in Delaware are more likely to increase risk following poor absolute performance. The evidence regarding misreporting, however, is mixed. This paper is organized as follows. A discussion of why Delaware law matters for hedge funds is provided in section 2. Section 3 introduces the Lipper/TASS dataset and provides summary statistics. Multivariate tests and various robustness checks are presented in section 4. Section 5 provides a summary, concluding remarks and suggestions for future research. 2. Why Delaware Law Matters for Hedge Funds Hedge funds are financial intermediaries that seek to attract capital from an investor base and invest the pool of funds as profitably as possible. It is therefore crucial that the jurisdiction chosen to establish the fund be one that is attractive to investors and also one that facilitates investments from a regulatory, legal, tax and operational perspective. Investments are made in a variety of investments including but not limited to equity and debt securities, derivatives, currency and commodities. From a regulatory perspective and operational perspective, investors analyses are further related to recent work on international cross-country law and finance analyses of hedge fund regulation in relation to fund structure and performance (Cumming and Dai, 2010; Cumming et al., 2013) in the spirit of La Porta, Lopez-de-Silanes, Shleifer, and Vishny, (1998, 2002, and 2006).

6 6 seek to invest in hedge funds that are able to execute their mandates within a jurisdiction s regulatory regime and hedge funds require the services of prime brokers and custodians that are able to operate within a jurisdiction. From a tax perspective, both investors and hedge funds seek clear and neutral tax jurisdictions to maximize investment profits. In this section we first review the literature on why Delaware law matters in general for corporations in subsection 2.1. Thereafter in subsection 2.2 we explain the two primary advantages of Delaware law for hedge funds: tax transparency, and a stable and clear legal regime that offers contractual flexibility that enables both limited liability and active management by fund managers, investors and service providers Why Delaware Law Matters In the U.S., states compete for corporate charters to increase their revenue from incorporation fees and franchise taxes when companies choose to incorporate locally (Romano, 1985). Delaware has emerged as the leader, chartering approximately 43% of the New York Stock Exchange firms and 50% of the Fortune 500 firms (Macey & Miller, 1986; Romano, 1985). Extant research has established that Delaware is a particularly attractive state for incorporation and the winner of the incorporation business due to a few specific reasons. First, it is a small state which earns over 15% of its revenues from incorporations (Romano, 1985, 1993), thereby signaling to the marketplace that it is committed to ensuring responsiveness to corporate interests. Second, Delaware s judges are highly specialized experts in corporate law, and have had years of building up precedents in legal cases that are widely accepted and used throughout the U.S. and in other common law countries around the world (Romano, 1985; Roe, 2003). Delaware is thus the most well recognized state among investors into U.S. companies. U.S.

7 7 lawyers typically understand the legal structure in their home state and that of Delaware, but not necessarily that of other states (Romano, 1985; Roe, 2003; Bebchuk and Cohen, 2003). Third, Delaware supports takeover bids in a number of ways: raising fewer obstacles to takeovers than other U.S. states; erecting only minor barriers to hostile acquisitions (Delaware has the shortest delay on hostile bids of any U.S. state) in a way that has not reduced shareholder wealth; and lowering acquisition costs by establishing clear precedents and occasionally prohibiting extreme defensive tactics by targets that would allow management to entrench themselves. There is no local constituency of firms physically located in Delaware to push back against bidder-friendly trends in a way that would reduce bidder profits. Thus, in addition to housing a majority of firms, Delaware law increases firm value by facilitating the sale of public firms (Daines, 2002). Related to this, subsequent work has show that Delaware is attractive to venture capital-backed companies (Boulton, 2010; Broughman et al., 2012), for example. Delaware law also tax benefits for larger corporations with subsidiaries. That is, Delaware increases firm value by facilitating income shifting across subsidiaries to avoid taxes, since not all items are taxed in the same way across states (Dyreng et al., 2013). Such incentives have raised the possibility that the competition for corporate charters might on one hand give rise to a race to the bottom if states change their corporate law in ways that favor managerial interests over shareholder interests in order to attract new incorporations (Winter, 1977; Bebchuk, 1992; Bebchuk and Cohen, 2003). On the other hand, incorporation competition may give rise to a race to the top if states compete in a way that maximizes shareholder value.

8 8 Empirical evidence (Dodd & Leftwich, 1980, Baysinger & Butler, 1985, Romano, 1985; Daines, 2002; Ferris et al., 2006) is consistent with the view that when firms reincorporate from another state to Delaware, share prices seem in the main to go up and not down. This evidence is consistent with the view that the state competition is generally successful in aligning managerial and shareholder interests. It is also consistent with the view that enhanced managerial freedom is a good thing, not a bad thing, for shareholders Delaware Law for Hedge Funds: Tax Transparency and Contractual Flexibility Delaware provides an efficient platform for hedge funds to establish their pool of funds under the Delaware Limited Liability Company Acts (DLLC) and the Delaware Revised Model Uniform Limited Partnership Act (DRULPA) The hedge funds in Delaware are often structured as DRULPAs. As with any limited partnership, it enabled the investors to obtain efficient tax flow through, tax transparency and limited liability, which is crucial in view of the limited oversight investors have over the hedge fund managers investment powers. However, while the funds themselves are structured as DRULPAs, the hedge fund managers typically structure themselves as DLLCs. This is mainly due to U.S. Internal Revenue Service (IRS) allowing DLLCs to simply elect to opt-into the benefits of a flow-through tax treatment (this was available since 1992, but facilitated by the IRS in 1997, as discussed by Schwartz, 2012), formerly a treatment only availed to partnerships, not corporations. The ability of hedge fund managers to structure themselves as DLLCs instead of DRULPAs enabled unprecedented contractual flexibility. 2 Under Delaware law, a hedge fund 2 In re Ford Holdings, Inc. Preferred Stock, 698 A2d 973 (1997), the Delaware Court of Chancery stated that Delaware General Corporation Law is an enabling statute designed to allow cost-efficient contracts and governance terms to be implemented in a way that best suits the needs of the organization. See also Schwartz (2012).

9 9 structured as a DRULPA requires a general partner to assume full liability of the fund as all other partners had limited liability. With the general partner or the hedge fund manager structured as a DLLC instead of another DRULPA the hedge fund manager is able to manage this assumption of full liability as the corporate structure, albeit enjoying the same tax flow-through privileges as a DRULPA. In other words, the general partner hedge fund manager is facilitated more protection through corporate law, instead of partnership law. Under a DLLC, fund management duties are delegated without the ensuing problems related to partnership. Lawmakers in Delaware have provided investors in a DRULPA with limited liability certain measures of protection not normally availed to limited partners of other limited partnerships, allowing them certain actions which do not inadvertently turn them into general partners, which include: acting as a contractor on behalf of the limited partnership; acting as a guarantor of the limited partnership; consulting with or advising a general partner; selling assets of the limited partnership; and making determinations in respect of investments to be made by the limited partnership. 3 With this legal certainty, coupled with the widespread familiarity with Delaware law as discussed above in sections 2, we expect that investors into Delaware hedge funds would pay a premium in terms of more favorable contract terms to hedge fund managers registered in Delaware. Furthermore, we expect that investors will be more inclined to liquidate a Delaware 3 See Delaware Revised Uniform Partnership Act.

10 10 hedge fund as a result of poor performance without fear of being deemed as a general partner, and likewise more inclined to invest new capital into stronger performing Delaware hedge funds after periods of strong performance. There are other benefits for hedge funds to be legally registered in Delaware. First, the use of DLLC as a general partner in a DRUPLA also allows hedge fund managers to determine management and performance fee flow to management staff more efficiently without contract negotiations with limited partners, especially in the case of staff turnover. Second, there are relatively low disclosure requirements as funds and their general managers are not required to file the DLLC or DRULPA Agreements, and they are not required to file annual reports. Third, there is no requirement for managers to maintain a presence or an office or personnel in Delaware. Overall, U.S hedge funds find Delaware an efficient, transparent, low-cost operational environment. In the remaining sections we empirically test the ways in which this affects investors into Delaware hedge funds. 3. Data and Summary Statistics The main database used in our empirical analysis is supplied by Lipper/TASS, a major hedge funds data vendor. Although many funds report a performance history prior to 1994, TASS started collecting hedge fund data only in To avoid survivorship bias, our sample period covers January 1994 through December We restrict our sample to hedge funds domiciled and registered in the U.S. (funds not domiciled in the U.S. rarely register in the U.S. to 4 To avoid backfilling bias (hedge funds voluntary report, and have a tendency to do so after a year of good performance, we likewise exclude the first 18 months of a fund s reporting in the data.

11 11 avoid scrutiny from the Internal Revenue Service and the Securities and Exchange Commission). The final sample comprises a total of 1,714 hedge funds. Majority of the hedge funds are organized either as Limited Partnership (LP) or Limited Liability Companies (LLC). Among the 1,714 hedge funds, 1,078 funds (about 63%) choose to register in Delaware (Table 1 Panel A). Among the 1,078 funds registered in Delaware, only 75 funds also have their business locations in Delaware. In other words, 95% of funds registered in Delaware are physically located elsewhere. Table 1 Panel B shows the popularity with Delaware increased sharply after 1997, consistent with the legal changes discussed above in section 2. [Insert Table 1 about here] As we discussed in section 2, Delaware is attractive as a registration location to hedge funds for many reasons, for instance, lower tax rate, higher quality court, greater flexibility when contracting with investors, among others. In this section we examine whether Delaware registered hedge funds have systematically different structures, performance, risk-taking behavior, and survival. Table 2 provides a first look by comparing the various characteristics of hedge funds conditional on their jurisdictional locations. 5 [Insert Table 2 about here] 5 In other unreported tests, we note that we do not find evidence of other jurisdictions that are significantly different. It is possible that there are rogue jurisdictions for hedge funds. For example, prior work has documented that Nevada law enables firms to maximize agency problems between managers and shareholders (Barzuza and Smith, 2011). In our sample there are only 8 funds registered in Nevada (Table 1 Panel A), and as such inferences cannot be drawn on such a small subsample. Our empirical tests below therefore focus on Delaware versus non-delaware funds.

12 12 As shown in Table 2, hedge funds registered in Delaware present significant differences from other funds in several aspects. First, the incentive structure of funds registered in Delaware is quite different from others. Funds registered in Delaware on average earn a management fee of 1.39% and an incentive fee of 19.20%, which are 0.12% and 1.45% higher than other funds, respectively. Furthermore, about 82% of Delaware hedge funds use high watermark provision, which is almost twice the percentage among funds registered elsewhere. Second, the data are consistent with the view that managers of hedge funds registered in Delaware are able to negotiate more favorable terms for themselves (i.e., terms that are less investor-friendly). For instance, Delaware hedge funds managers are less likely to have fund managers personal capital invested into the fund. Delaware hedge funds have longer redemption notice periods and have longer lockup periods than other funds. Also, Delaware hedge funds are significantly more leveraged. Table 2 also compares the performance of Delaware hedge funds with others, measured by average monthly returns, average excess returns, and annual alpha. The data do not indicate statistically significant differences in performance between Delaware and non-delaware hedge funds (returns in TASS are net of fees). These comparison tests are of course suggestive and not conclusive. In the next section we provide multivariate analyses controlling for other things being equal.

13 13 4. Empirical Analysis 4.1. The endogeneity of choosing Delaware as a registration location While it is not our major interest to explore why Delaware is so popular as U.S. domestic registration location for hedge funds in this paper, the endogenous nature of this choice presents a challenge for our empirical analysis that follows. We adopt the Heckman (1976, 1979) two-stage treatment regression framework to examine the robustness of our major empirical findings throughout the paper. In this section, we discuss this approach in details. Our setup of the first stage regression is as follows: Delaware Dummy = α + β 1 Business Location + β 2 Inception Year Dummies + β 3 Control Variables Where Delaware Dummy is equal to one if a hedge fund is registered in Delaware; Business Location is a dummy variable which is equal to one if the hedge fund is physically located in Delaware; Inception Year Dummies are a set of dummy variables which are equal to one if a hedge fund was founded in a specific year; additional control variables include a Master-Feeder Fund dummy, a LP dummy, a LLC dummy, and dummy variables indicating the investment strategy of the fund. We believe the business location may be correlated with funds choice of the registration location, but do not obviously affect hedge funds flow-performance relation, their risk choices, and their survival, which are the dependent variables of the second stage regressions that we explore in the following sections. Furthermore, as discussed in section 2 and shown in Table 1, the popularity with Delaware increased sharply after 1997 due to an exogeneous shock related to IRS rule change. Thus, we use the inception years of hedge funds as additional instruments in our first stage regression. [Insert Table 3 about here.]

14 14 As shown in Table 3, funds organized as LLC, LP and master-feeder funds are 22.9%, 23.2% and 17.3% more likely to register in Delaware, respectively. Furthermore, we show that Delaware has become the most popular registration location to U.S. hedge funds only after 1998, consistent with the legal changes documented in subsection 2.2 above. Lamda is estimated off the above probit regression and then included in the second stage regressions to adjust for the endogenity that arises from funds self-selection Hedge Fund compensation and Registration Location Table 4 presents OLS and two-step treatment regressions for the determinants of management and performance fees, as well as probit regressions for high watermark provisions. Not controlling for self-selection effects, Models (1), (2) and (3) in Table 4 respectively show that Delaware hedge funds have on average a higher management fee by 0.164%, a higher incentive fee by 1.445%, and are 28.8% more likely to have a high watermark provision. Controlling for self-selection effects, the economic significance of these effects are larger at 0.471%, 3.964%, and 85.9%. With the average management fee at 1.35% in the data and the average performance fee at 18.66%, these higher management and performance fees are not only statistically significant but also economically large. Likewise, these results are robust to different control variables reported in Table 4 (among other specifications considered and available on request). Overall, the data reported in Table 4 are therefore consistent with the theoretical predictions summarized in section 2 above. [Insert Table 4 about here.] 4.3. Flow-Performance Relation and Registration Location The flow-performance relation has been examined in the context of hedge funds in many recent studies. Focusing on flows and their relationship to liquidation of funds, Getmansky (2003), shows that similar to mutual funds, the likelihood of a hedge fund liquidated is decreased since investors are chasing individual fund performance. Agarwal and Naik (2004), Goetzmann

15 15 et al. (2003) and Getmansky (2003) show there are decreasing returns to scale in the performance of hedge funds, and explain their findings by the limited availability of assets that provide superior hedge fund returns. Agarwal et al. (2004, 2006) control for additional factors such as managerial incentives and fees, size and age of funds, that ranking hedge funds based on non-risk adjusted return performance. They find that flows are positively associated with performance, which is different from the findings in Goetzmann et al. (2003). Similarly, Fung et al. (2007) show that the alphas are associated greater and steadier capital inflows of funds. Convexity versus concavity of the flow-performance relationship has been debated in the hedge fund literature. Agarwal et al. (2004) find a convex relationship in hedge fund flowperformance; Getmansky (2005) finds a concave flow-performance relationship; and Baquero and Verbeek (2005) find a linear flow-performance relationship. The results depend on database used, time period analyzed and the frequency of the sample. Ding et al. (2007) reconciles these conflicting findings by showing that hedge funds exhibit a convex flow-performance relation in the absence of share restrictions (similar to mutual funds), but exhibit a concave relation in the presence of restrictions. Further, live funds exhibit a concave flow-performance relation due to diseconomies-of-scale, but defunct funds display a convex relation due to the various reasons that they had become defunct. As shown in our summary statistics (Table 2), hedge funds registered in Delaware exhibit many different contractual features, such as incentive structures, redemption notice period, lockup period, etc. The existing literature (as discussed above) shows that these factors may have an effect on the flow-performance relation. Based on this reasoning, in this section, we explore whether the flow-performance relation is conditional on the registration location of a hedge fund. Our main model is as follows:

16 16 Flow t = α + β 1 Performance t 1 + β 2 Delaware + β 3 Delaware Performance t 1 + β 4 Control Varialbes We measure flows as a proportion of Assets under Management (AUM) by the month t change in net AUM, adjusted for investment returns (Sirri and Tufano, 1998): Flow t = AUM (1 + return ) AUM AUM t t t 1 t 1 We control for prior fund performance with various alternative measures of performance (such as returns and ranked performance), and use different methods to consider nonlinearities, such as with the use of variables for the square of past performance as well as variables that measure tercile performance ranks (as in Ding et al., 2007). For the latter, according to Sirri and Tufano (1998) and Ding et al. (2007), we first estimate the fractional rank for each fund in each period, from 0 to 1 based on the previous month raw return. Then, we construct the fractional terciles ranks as follows: Bottom Tercile Rank: TRank.1 = min (1/3, FRank) Middle Tercile Rank: TRank.2 = min (1/3, FRank-TRank.1) Top Tercile Rank: TRank.3 = min (1/3, FRank-TRank1-TRank2) where FRank is the fractional ranks in each period. The key variable of our interest is the interaction term between Delaware and measures of the fund past performance. From Delaware*Performance t-1, we can infer the incremental effect that being registered in Delaware has on the flow-performance relation.

17 17 We control for other fund specific factors that are important in hedge fund flows, including management fees, incentive fees, high watermark, fund size, the use of leverage, whether fund managers are requested to invest their personal capital, whether the fund accept managed account, redemption notice period, length of lockup period, whether the fund is registered with the SEC, whether the fund is a master-feeder fund, and the organization form of the fund. Finally, we employ a number of dummy variables to control for the hedge fund investment strategies. 6 [Insert Table 5 about here.] The empirical findings are reported in Table 5. In models (1)-(3) we include all hedge funds, where raw returns are used in models (1) and (2) and the fractional terciles ranks are used in model (3). Consistent with the existing literature (Getmansky, 2005; Cumming and Dai, 2009), we find a positive correlation between fund flow and prior return and a significantly negative coefficient on the square term, which indicates a concave relation. The coefficient on the interaction term Delaware*Performance t-1 is significantly positive and significant at at least the 5% level in all of the specifications, suggesting investors are more sensitive to the past performance of funds registered in Delaware. The economic significance is likewise very large, indicating that Delaware funds have a 40-65% steeper flow performance line (Models (1) and (2)). As argued in Ding et al (2007), the flow-performance relation could be different between live funds and defunct funds. Thus, in models (4) and (5), we repeat the regressions for liquidated funds and live funds separately. In both cases, we show a significantly positive coefficient on the 6 There is a possibility that Delaware hedge funds are more likely to engage in illiquid investment strategies which may result in higher return serial correlation (Getmansky et al, 2004). We control for this issue by including investment strategy fixed effect in regressions. Furthermore, in untabulated regressions, we further include performance measures lagged by two periods. The coefficients of interaction terms of Delaware and lagged performances continue to be significantly positive.

18 18 interaction term Delaware*Performance t-1. This result remains robust and shows a more pronounced effect for Delaware for liquidated funds than non liquidated funds. Model (6) reports the results from the Heckman 2-stage treatment regression which adjusts for funds selfselection of Delaware as the registration location. The economic significance in Model (6) is consistent with that for Model (1). In Model (7), we conduct robustness check using semi-annual fund flow instead of monthly flow with the same battery of control variables as in Model (1) in the 2-strage treatment regression framework. Our results continue to hold. 7 This finding supports the hypothesis that investors will be more inclined to invest in the future when a Delaware fund does well, and conversely more inclined to withdraw funds in the event of poor performance, because there are more investors that are better aware and more fully informed of the causes and consequences of investing in a Delaware fund relative to non- Delaware funds, as discussed in detail in subsection 2.2 above. 4.4.Fund Survival and Registration Location Several authors have empirically examined the survival rates of hedge funds (e.g Brown, Goetzmann, and Ibbotson, 1999; Fung and Hsieh, 2000, 2002; Liang, 2000; Gregoriou, 2002; Getmansky, Lo, and Mei, 2004; Baquero, Horst, and Verbeek, 2005; Grecu, Malkiel, and Saha, 2007; Aragon and Nanda 2012). These work document a significant relation between fund survival and fund characteristics including performance, asset size, and investment styles. In this section, we extend these works by considering the relation between fund survival and its registration location. Specifically, our model is as follows: 7 In untabulated tables, we also tried quarterly fund flow and annual fund flow. The results are qualitatively similar.

19 19 Liquidated t = α + β 1 Return t 1 + β 2 Delaware + β 3 Delaware Return t 1 + β 4 Control Variables Where Liquidated t is a dummy variable which is equal to 1 if a fund was liquidated in a specific year. In Lipper Tass database, a fund is moved to graveyard fund database when it stops reporting performance. This can be caused by many reasons, for instance, funds voluntarily stop reporting although the fund is still operating; funds were liquidated; funds were merged to another entity; etc. We define a graveyard fund as liquidated only when it is explicitly reported so. Return t-1 is the lagged annual raw return of the fund. The coefficient of the interaction term Delaware*Return t-1 indicates how being registered in Delaware changes the relation between fund performance and survival. Finally, we include a battery of control variables that are similar to the ones used in earlier tables. [Insert Table 6 about here.] Table 6 reports the results of the above probit regressions. Model (1) does not include the interaction term Delaware*Return t-1, while model (2) does. Model (3) is the Heckman 2-stage treatment regression. The coefficients of lagged return are significantly negative in all models, suggesting poor performance is positively associated with higher probability of liquidation. We do not find registered in Delaware itself has significant implication for the survival of funds, nevertheless, we do find a significantly negative coefficient on the interaction term Delaware*Return t-1, even after we control for the self-selection of registration location in model (3). The negative coefficient of the interaction term Delaware*Return t-1 suggests that the negative association between performance and liquidation is even stronger for funds registered in Delaware. The probability of a liquidating a Delaware fund in the event of bad performance is

20 20 20% higher than a non-delaware fund in Model (2), and 40% higher after controlling for selfselection in Model (3). This result is consistent with our earlier finding from Table 5 that investors (fund flows) are more sensitive to the performance of the fund registered in Delaware. Poor performance is more likely to trigger capital withdraw and therefore liquidation of funds registered in Delaware, as we had expected (subsection 2.2) Funds Risk Choice and Registration Location The more pronounced flow-performance sensitivity for Delaware hedge funds and the threat of liquidation upon poor performance arguably would give Delaware hedge fund managers stronger motivation to increase the risk of their investments or manipulate their returns when the fund performance is poor. In this section and the section follows, we explore whether Delaware hedge funds are more likely to have such agency problems. There is a substantial body of literature that examines the risk choices of fund managers. Central to this literature is the notion that fund managers may have the incentive to choose investment strategies that markedly increase or decrease portfolio risk, so-called risk-shifting. For instance, it is argued that fund managers might be especially concerned about their performance relative to that of other funds (i.e., tournaments), thereby inducing relatively poor performers to increase risk. The risk-shifting behavior may not necessarily be in the interest of fund investors. Most of the existing works focus on risk shifting of mutual funds with a few exceptions. Brown, Goetzmann, and Park (2001) and Agarwal, Daniel, and Naik (2002) report evidence of tournament behavior among hedge funds. Both Hodder and Jackwerth (2007) and Panageas and Westerfiled (2009) argue that hedge fund managers investment horizon affects their incentives

21 21 for risk-shifting. That is, fund managers are more likely to increase risk when they have short investment horizon (for instance, when the fund is likely to be liquidated), and when the fund is below its high watermark. Basak, Pavlova, and Shapiro (2007) and Hodder and Jackwerth (2007) further model that managers incentives for risk-shifting can be mitigated if he or she is exposed to some downside risk, either through a personal capital stake in the fund or through management fees based on end-of-period assets. Aragon and Nanda (2012) find empirical evidence consistent with these predictions. Specifically, they show that high watermark provision is less effective in moderating risk shifting following poor performance when the fund is likely to be liquidated. And they show that managers personal capital stake does reduce their incentives for risk-shifting. The stronger flow-performance relationship and the survival-performance relationship as documented in the previous sections indicate that Delaware hedge fund managers would have greater incentive to increase risk when the fund performance is poor. In addition, the many different contractual features, including the incentive contracts and the request for managers personal capital, between hedge funds registered in Delaware and those registered elsewhere would also have implications for fund managers choices of risks. For instance, the higher incentive fee and the more frequent usage of the high watermark provision among Delaware funds indicate less risk-shifting behavior than other funds. In contrast, the lower percentage of Delaware funds requesting managers personal capital suggests the opposite. To empirically examine the relation between fund managers choices of risks and the registration location of hedge fund, we follow the methodology used in Aragon and Nanda (2012). Specifically, we estimate the following pooled cross-sectional regression:

22 22 Risk = α + β 1 LagPerf + β 2 Delaware + β 3 LagPerf Delaware + β 4 LagRisk + β 5 Control Variables Our focus is how changes in fund risk between the first and second halves of the year are related to mid-year performance and how these patterns interact with the registration location of the hedge fund controlling for the compensation contract and the presence of personal capital request, among others. ΔRisk is the difference between the sample standard deviations of monthly returns in the second and first halves of the year. Similar to Aragon and Nanda (2012), we require that a fund have the full six monthly observations to be included in the estimate of semi-annual standard deviation. LagRisk is the standard deviation of monthly return during the first six months of a specific year. This is included to control for the potential mean reversion in risk changes. For LagPerf, we use both absolute and relative measures of performance of the fund during the first six months. As shown in Table 7, in models (1) (4), we use the buy and hold return of the fund during the first six months. 8 In models (5) (8), we use the fractional rank of the fund s raw return over the first six months relative to other funds during the same period. As found in Aragon and Nanda (2012), incentive fee, high watermark, and the request of personal capital affect hedge fund managers incentives for risk-shifting. Thus we control for these effects by including interaction terms between LagPerf and these three measures. Other control variables include management fees, incentive fees, high watermark, the use of leverage, whether fund managers are requested to invest their personal capital, whether the fund accept managed account, redemption notice period, length of lockup period, whether the fund is registered with the SEC, whether the fund is a master-feeder fund, and the organization form of 8 We also use average monthly return as a robustness check and get qualitatively similar results.

23 23 the fund. Finally, we employ a number of dummy variables to control for the hedge fund investment strategies and years of performance. [Insert Table 7 about here.] We find a significantly negative coefficient on LagPerf (model (1)), suggesting hedge fund managers in general have a propensity to increase risk following poor performance. As shown in model (2), the coefficient on the interaction term between LagPerf and Incentive fee is significantly negative, while that of the interaction term between LagPerf and High watermark is significantly positive. This finding suggests although asymmetric compensation contract (high incentive fee) strengthens fund managers incentive for risk-shifting given poor performance, the usage of high watermark significantly curb this incentive. This is consistent with the finding in Aragon and Nanda (2012). We do not find the request of personal capital to significantly affect fund managers incentive for risk-shifting. 9 In model (3), we add the interaction term between LagPerf and Delaware, which is our key interest. The coefficient of this interaction term is significantly negative and the coefficients of LagPerf*Incentive fee and LagPerf*HWM continue to be significantly positive and negative, respectively. The negative coefficient of LagPerf*Delaware indicates that hedge funds registered in Delaware have stronger incentives to increase risk following poor performance, where the economic significance in Models (3) and (4) is such that a 1% reduction in monthly returns gives rise to a 8% increase in risk changes for Delaware funds relative to non-delaware funds. This cannot be fully explained by the differences in the compensation contract as we discussed above. This result is robust after we control for funds self-selection of Delaware as registration location 9 Aragon and Nanda (2012) use the amount of requested personal capital. We choose to use the request of personal capital (a dummy variable) as we find the amount of personal capital is missing for many hedge funds.

24 24 (model (4)). This finding supports the notion that the more pronounced flow-performance relationship for Delaware hedge funds has a significant effect on fund managers incentive for risk-shifting. To examine whether Delaware hedge funds are more likely to increase risk following poor relative performance, the so-called tournament behavior, in models (5) (8), we measure fund performance using the fractional rank of the fund s raw returns during the first half year relative to other funds during the same time period. We find similar results for LagPerf, LagPerf*Incentive fee, and LagPerf*HWM. Interestingly, the coefficient on the interaction term LagPerf*Delaware is positive, although significant only at the 10% level. This finding indicates Delaware hedge funds exhibit weaker incentives for the tournament behavior. Together, we show hedge funds registered in Delaware are more likely to increase risk given poor absolute performance during the first half year. This result holds after controlling for the contractual differences between Delaware hedge funds and non-delaware ones. We conjecture the stronger relation between performance and fund flow as well as survival for Delaware hedge funds likely incentivizes managers to increase risk upon poor performance Return Misreporting and Registration Location In this section, we address whether Delaware hedge funds are more likely to misreport returns in order to attract investors. Bollen and Pool (2008, 2009) and Agarwal, Daniel, and Naik (2011) show that investors direct more capital to funds that feature a higher number of months with positive returns. Since managerial compensation is a direct function of the quantity of assets under management, some managers may misreport returns in order to attract and maintain their investor base. Incentives arise from flow-performance sensitivity, as investors direct more

25 25 money into hedge funds that outperform (e.g., Agarwal, Daniel, and Naik, 2004). The stronger flow-performance sensitivity for Delaware hedge funds suggests that Delaware hedge fund managers have greater incentive to misreport returns. On the other hand, lockup and restriction periods act as disciplining mechanisms which allow investors to withdraw their capital in response to hedge fund poor performance. The shorter the lockup and restriction periods, the more quickly investors can withdraw their capital. As documented in Section 3, Delaware hedge funds often have longer lockup period and restriction period, which suggests that Delaware hedge fund managers face less short-term pressure. Therefore, they have smaller incentives to manipulate returns. We utilize the flags developed in Bollen and Pool (2009) and Agarwal, Daniel, and Naik (2011) to empirically examine this issue. According to Bollen and Pool (2009), there is a returns discontinuity in that comparatively fewer hedge fund returns are reported as being zero or marginally negative relative to the frequency of returns reported as marginally positive (Bollen and Pool, 2009). Bollen and Pool define marginally positive and marginally negative by minimizing the mean square error (MSE) using Silverman (1986) approach, and conclude the appropriate bin width is to We use this bin width as the starting point and then assess robustness to alternative bin widths, and explicitly show the results for bin widths of to and to For each of these alternative bin widths, we define dummy variables equal to one for the (marginally) positive returns, and dummy variables equal to zero for returns that are zero or (marginally) negative. Models (1)-(3) in Table 8 report probit estimates of the probability that returns are reported as marginally positive, as opposed to zero or negative. The explanatory variables include the Delaware dummy, fund size, and a battery of variables that describe fund

26 26 characteristics, similar to those used in Tables 5 and 6. The coefficients of the Delaware dummy in models (1) (3) are all significantly positive indicating Delaware funds are approximately 2.3%-3.0% more likely to misreport. The result becomes marginal significant when we apply the Heckman 2-stage treatment regression framework, as shown in model (6). Among the control variables, we find that the probability of return misreporting is positively correlated with incentive fee and the request of personal capital, which makes intuitive sense. When fund managers compensation and personal capital are at risk, they have greater incentives to manipulate returns. Agarwal, Daniel, and Naik (2011) argue that hedge fund managers compensation are determined by their annual performance, which provides incentives for managers to improve performance as the year comes to a close. They show that hedge funds with high incentives and more opportunities (e.g., in-the-money, higher delta, top 20% performers, shorter lockup periods, shorter restriction periods, higher management fees, etc.) to inflate returns show significantly greater December spike the return in December less the average return from January to November after controlling for risk in both the time series and in the cross-section than those with low incentives and less opportunities to inflate returns. To examine whether Delaware hedge fund behave differently in terms of managing their performance at the year end, we modify the approach of Agarwal, Daniel, and Naik (2011) as follows. Return i,m or Residual i,m = α + β 1 Dec + β 2 Delaware + β 3 Dec Delaware + β 4 Control Variables

27 27 where Return i,m is the monthly return; Residual i,m is the residual estimated with the time-series regressions using Fung and Hsieh (2004) 7-factor model. Dec is a dummy variable which is set to equal to 1 if the month of return is December and 0 otherwise. Delaware is a dummy variable which is equal to 1 if the fund is registered in Delaware and 0 otherwise. Dec Delaware is the interaction term between Dec and Delaware. If Delaware hedge funds are more incentivized to manage their returns, we would observe a significantly positive coefficient of the interaction term. The control variables are similar to those used in model (1)- (3) with additional control of year dummies. [Insert Table 8 about here] Models (4) and (5) report the results of the above regressions. The dependent variable in model (4) is the monthly return and the dependent variable in model (5) is the monthly discretionary return or the monthly residual. Models (7) and (8) repeat model (4) and (5) in the Heckman 2-stage regression framework. We find a significantly positive coefficient of the Dec dummy, significantly positive coefficients on management fee, high water mark, lockup period, and redemption period in the regression of returns, confirming the findings in Agarwal, Daniel, and Naik (2011). The coefficient of the interaction term Dec Delaware turns out to be significantly negative, indicating that Delaware hedge funds exhibit a 31.5% weaker December spike. This finding is robust under the Heckman 2-stage framework. In the regression of residuals, we do not find significant coefficients for either the Dec dummy or the interaction term Dec Delaware. 5. Conclusions and Discussions