Delegated private equity investments

Transcription

1 THOMAS R. ARNOLD is deputy global head and head of Americas in the Real Estate and Infrastructure Department at Abu Dhabi Investment Authority (ADIA) in Abu Dhabi, UAE. DAVID C. LING is the McGurn professor of Real Estate in the Department of Finance, Insurance, and Real Estate in the Warrington College of Business at the University of Florida in Gainesville, FL. ANDY NARANJO is the Emerson Merrill Lynch professor of finance and chairman in the Department of Finance, Insurance, and Real Estate at Warrington College of Business at the University of Florida in Gainesville, FL. Waiting to Be Called: The Impact of Manager Discretion and Dry Powder on Private Equity Real Estate Returns THOMAS R. ARNOLD, DAVID C. LING, AND ANDY NARANJO Delegated private equity investments have become a prominent asset class with aggregate fund capitalizations exceeding $5.1 trillion as of 2016 (Burgiss; These private fund vehicles are typically organized as limited partnerships formed by an investment manager to pool capital to invest in a designated strategy. As private equity real estate (PERE) funds play an increasingly important role in real estate capital formation and development, the accurate measurement of PERE performance is central to investment allocations and growth in this sector. However, the research to date examining performance measurement and the influence of fund manager discretion on the reported PERE returns is limited and incomplete. In this article, we examine the return performance sensitivity of PERE funds to management fees, capital deployment speeds, investment horizons (durations), and the opportunity cost associated with reserving capital (i.e., maintaining dry powder for the benefit of the manager) for uncertain capital calls. These factors are largely ignored in traditional PERE performance metrics but are important in measuring PERE performance. Moreover, managers of PERE investments have considerable discretion over the timing of capital calls, deployment speeds, and investment durations. To motivate our analysis, we first provide a series of simulation scenarios using observed PERE data and industry standards to calibrate our input assumptions. We first demonstrate the impact of the pace of investment and the investment holding period of acquired properties on internal rates of return (IRRs) and multiples on invested equity (MOICs), initially ignoring management fees and the opportunity cost of maintaining dry powder. This is accomplished by construction of an analytical model that captures the quarterly inflows and outflows of a set of representative funds with different deployment speeds and investment horizons. We then examine how calculated IRRs and MOICs are affected by management fees. Finally, we incorporate the opportunity cost of maintaining dry powder for the benefit of the manager. This analysis demonstrates the extent to which accounting for the opportunity cost of waiting to be called reduces the reported return performance of PERE funds. Having established through simulations the significant effects of deployment speeds, investment horizons, opportunity costs, and management fees on PERE performance, we turn to PERE fund data to empirically investigate the influence of these factors on reported PERE performance. A growing private equity literature has examined the effects of macroeconomic variables, industry-specific cycles, and fund THE JOURNAL OF PORTFOLIO MANAGEMENT 23

2 strategies on fund performance (e.g., Kaplan and Schoar [2005]; Robinson and Sensoy [2011]; Alcock et al. [2013]; Fisher and Hartzell [2015]). A related literature examines the benefits of investing in private market vehicles instead of the public markets (e.g., Franzoni, Nowak, and Phalippou [2012]; Harris, Jenkinson, and Kaplan [2014]; Sorensen and Jagannathan [2015]). Another branch of the private equity literature focuses on potential agency conflicts within private equity structures, such as manager ownership percentages and contractual fee structures (Robinson and Sensoy [2013]). Our article is most closely related to this third branch of the private equity literature focusing on agency conflicts and managerial compensation. In particular, we focus on the impact on investor returns of managerial discretion over the speed of capital calls and investment, as well as the duration of those fund investments. 1 These managerial decisions dramatically affect the opportunity cost of waiting to be called. Braun and Stoff [2016], Siegel [2008], and Kogelman [1999] identified the cost of maintaining dry powder as an overlooked expense of investing in private equity. Braun and Stoff also identified managerial discretion over deployment speeds as potentially dilutive to investor returns. However, we are the first to quantify the impact of deployment speeds, investment horizons, and committed, but uncalled capital, on PERE fund performance. Importantly, the negative performance effects we document also provide some guidance on the design of improved private equity contracts and structures, such as contractual mandates on longer capital call notices, shorter investment periods, and reduced fees on committed but uncalled capital. E XHIBIT 1 PERE Closed-End Fund Lifecycle THE ECONOMICS OF INVESTING IN PERE FUNDS PERE Background and Timing Commercial real estate investments are lumpy in transaction size; entail significant lead time to structure and close transactions; often require specialized management expertise; and can benefit from economies of scale in terms of financing availability, third-party fees and costs, and operational expertise. These features lend themselves to a pooled investment form of equity ownership, such as a PERE fund. Temporally, the structure has three stages: fundraising, investment, and harvesting. As Exhibit 1 illustrates, the fundraising period (approximately one year) is the period between fund inception and the final close of investor commitments. The manager is permitted to make early capital calls and investments during this time. During the investment period (typically three years), the manager has the discretion to make investments. This period will include capital calls, capital deployment into specific investments, initial distributions of investment income on acquired properties, and proceeds from dispositions of early investments. During this period, investors endure illiquidity and uncertainty around required capital contributions and distributions. During the harvesting period, typically six to eight years, the manager targets liquidation of the entire portfolio of investments and may provide follow-on funding to existing investments but will typically be prohibited contractually from acquiring new assets. A fund s investment horizon is the dollar-weighted window of time over 24 THE IMPACT OF MANAGER DISCRETION AND DRY POWDER ON PRIVATE EQUITY REAL ESTATE RETURNS

3 which the manager deploys capital and keeps that capital invested. This period of capital productivity may start either during the fundraising period when investments are also permitted, subject to the availability of initial capital commitments, or during the investment period. The investment horizon is reflective of actual investment activity, whereas the investment period is a stage of the fund life during which investments are permitted and targeted. PERE fund performance measures include the IRR, the MOIC, and the public market equivalent (PME). Prior to any capital calls or net distributions, these three metrics are estimated by the manager using projected cash flows. These metrics are frequently updated by the manager throughout the life of a fund using a combination of realized and projected cash flows prior to the final liquidation of assets. Fully realized cash outflows and inflows are used to calculate these performance metrics for fully liquidated funds. These metrics may be quoted by a manager or data provider in isolation or used to establish benchmarks and relative manager peer rankings. The most commonly quoted performance metric is the IRR, which is calculated by solving the following equation for r: n [ CFDIST CAPCAL C L t MGTFEEt + OPCOSTt 0 = t ( )], t (1 + r) t= 1 where CFDIST t is the cash flow distributed to investors in quarter t net of any performance fees (carried interest) or other compensation paid to the manager, CAPCALL t is the amount of capital called by the manager in quarter t, and MGTFEE t is the quarterly management fee. OPCOST t is the opportunity cost (in dollars) investors incur by maintaining sufficient liquidity in their portfolios to cover capital calls of uncertain magnitude and timing by the fund manager. OPCOST t is ignored in industry calculations of IRR. The IRR is a dollar-weighted performance metric and thus sensitive to the timing of cash inflows and outflows. When OPCOST t is ignored, fund managers are not penalized for a slow deployment of capital, other than the impact of the quarterly management fee on IRRs. In addition, fund managers may be able to boost realized IRRs by selling assets that are performing well and distributing the proceeds prematurely to investors that is, although potentially inconsistent with the objective of maximizing the investors terminal wealth, managers may be able to boost realized IRRs by delaying the pace of investment and/or by shortening the duration of the investment horizon. 2 The MOIC of a fund investment is calculated as: MOIC = n CFDIST t= 1 t ( CAPCALC L OPCOST ) = t MGTFEE t + 1 t n t For example, if cash distributions of $135 on a cumulative investment of $100 are received by an investor over the life of a fund, the MOIC would be The MOIC captures the magnitude of net cash inflows relative to total capital called; however, the MOIC is not affected by the timing of cash inflows and outflows. Therefore, overall investor wealth may not be maximized by investment in a fund with the highest realized MOIC, with consideration of OPCOST t. The denominator of the standard MOIC does not include the opportunity cost investors incur by keeping a portion of their wealth invested in liquid assets to meet uncertain capital calls. Including an estimate of OPCOST t in the denominator reduces the magnitude of the MOIC. An accurate MOIC calculation includes in its denominator all monies paid in by an investor, whether through capital calls or management fees. 3 Opportunity Cost of Waiting to Be Called In effect, investors in PERE have provided a long call option to the fund manager, who has the contractual right to call capital from the investors on short notice. We argue that the opportunity cost of waiting to be called (i.e., the cost of keeping dry powder) is important. For an institutional real estate investor, the opportunity cost of committing today to invest in the future could include (1) the return anticipated on alternative PERE fund investments of similar risk; (2) the return expected on the investor s overall real estate portfolio; (3) the return expected on the overall multiclass portfolio; (4) the difference between the return expected on alternative PERE fund investments and the short-term risk free rate (such as a two-year Treasury); and (5) some other foregone investment opportunity (such as an equity index or a global real estate investment trust index). Each of these measures of opportunity cost varies over THE JOURNAL OF PORTFOLIO MANAGEMENT 25

4 time, which further complicates the estimation of an accurate opportunity cost proxy. To account for the opportunity cost associated with maintaining dry powder, Mozes and Fiore [2012] proposed that performance be measured from the commitment date rather than the date of the first capital call because ignoring the opportunity cost that occurs between these points inflates measures of private equity performance. The authors further recommended that investment returns reflect committed capital, not just the amount of capital actually called. This is because investors must reserve based on what could be called. A key determinant of OPCOST t is the percentage of committed but uncalled capital for which the investor chooses to reserve. The percentage of the capital commitment reserved by investors is positively related to investors aversion to risking a capital call that they would have insufficient liquid funds to meet. The consequences of a capital call default, a legally binding contractual obligation, may include a punitive interest rate on a manager advance on behalf of a defaulting investor during the cure period; a forced sale of the limited partner s (LP s) interest (potentially at a discount to fair market value); a full or partial dilution of the existing positon; reputational damage to the investor that potentially inhibits future manager access; or a suboptimal sale of other assets by the investor. With perfect foresight, an investor could budget for the amount and timing of each capital call and have these funds available when, but not before, they are needed to meet legal obligations. The design of a cash management strategy to mitigate the opportunity costs of idle capital is influenced by the anticipated size and frequency of future capital calls and the trade-off between the desire to maximize returns from an overall investment strategy and the desire to minimize the risk of failing to meet an unanticipated capital call (Meads, Morandi, and Carnelli [2016]). Investment strategies that address the use of idle but committed capital should account for an investor s risk profile, the size and diversification of the private equity portfolio, and the role of real estate in the investor s overall portfolio. If a fund manager chooses to return capital to investors sooner than investors expect, these early distributions can be used to partially fund remaining capital calls. Thus, the probability of receiving earlier-than-expected capital distributions can potentially reduce the opportunity cost of waiting to be called. However, the extent of this reduction will also vary with the investor s risk aversion. An investor can also use cash distributions from other funds in a diversified private equity portfolio to fund capital calls. Private equity portfolios can be diversified by increasing the number of funds per vintage and/or the number of vintages (Meads, Morandi, and Carnelli [2016]). Opportunity cost is not only time varying and unique to each investor, but it may vary for the same investor under different circumstances (i.e., market conditions, alternative sources of liquidity, the relative maturity of the PERE investment program). The opportunity cost may also depend on whether capital calls must be covered by individuals (e.g., high-net-worth investors), by the real estate department of a multiasset investment entity (e.g., endowments, sovereign wealth funds, pension funds), or by the combined resources of a multiasset investment entity. 4 Kerins, Smith, and Smith [2004] also warned of the implied cost of an underdiversified portfolio in which capital deployment lags capital commitment. Because OPCOST t is difficult to measure and varies with the circumstances and risk aversion of the investor, we use a range of OPCOST t to quantify its impact on calculated returns and equity multiples. 5 SIMULATING PERFORMANCE COSTS OF DEPLOYMENT SPEEDS, INVESTMENT HORIZONS, MANAGEMENT FEES, AND OPPORTUNITY COST OF WAITING TO BE CALLED To examine the impact that deployment speeds, investment horizons, management fees, and dry powder opportunity costs to the investor have on performance, we perform a series of simulations. We first demonstrate the impact of the pace of capital deployment and the length of the investment horizon on IRRs and MOICs, initially ignoring MGTFEE t and OPCOST t. This is accomplished by construction of an analytical model that captures the quarterly inflows and outflows of a set of representative funds with different deployment speeds and investment horizons. We next examine how calculated IRRs and MOICs are affected by the introduction of MGTFEE t and OPCOST t. This analysis demonstrates that OPCOST t materially reduces the reported return performance of PERE funds, especially those with relatively slow deployment speeds and/or short investment horizons. 26 THE IMPACT OF MANAGER DISCRETION AND DRY POWDER ON PRIVATE EQUITY REAL ESTATE RETURNS

5 E XHIBIT 2 Deployment Speed and Investment Horizon Scenarios We consider a range of opportunity costs for idle or suboptimally invested capital. We take a conservative view by assuming a 400 bps return differential between the expected return on a conservative PERE fund or real estate portfolio and the return on a shortterm Treasury security. Neither a conservative investor nor a more risk-tolerant investor will reserve the entire uncalled commitment because the likelihood that a manager will call 100% of the uncalled commitments is very low. Moreover, most PERE funds are never fully called. Accordingly, we assume a conservative investor reserves 70% of the uncalled commitment, given the severity of the consequences of a capital call default. We assume that a more risk-tolerant investor is comfortable with reserving just 35% of the uncalled capital commitment each quarter. These assumptions imply annual opportunity costs of 140 (35% 400 bps) and 280 bps (70% 400 bps), respectively. In the calculation of IRRs, deployment speeds and the life of the fund also interact with management fees and the opportunity cost of waiting to be called. To demonstrate the acute sensitivity of investor IRRs to these factors, we construct four scenarios for hypothetical PERE funds that have received $100 million in capital commitments. These four scenarios are shown in Exhibit 2 and are based on the range of deployment speeds and investment horizons found in our PERE dataset, which we describe in detail in the following. After the initial close, assumed to be the date of legal inception, the fund continues to raise capital for four quarters, after which the final fund closing occurs and no additional investor capital commitments are accepted. In all scenarios, we assume 5% of total committed capital is called in both the second and fourth quarter of the additional capital-raising period; that is, we assume modest investment occurs during the fundraising period. However, time zero in the IRR calculation is the date of legal inception. The fundraising period is defined as the period between the initial and final commitment close. The four-quarter fundraising period is followed by an investment period of three years; the total fund life is nine years, including the fundraising period. Total capital called is equal to 85% of committed capital ($85 million) in each of the four scenarios. 6 One-fourth of the annual 1.5% management fee is paid at the end of each quarter. Until the end of the investment period, the management fee is based on total committed capital. After the end of the three-year investment period, the management fee is based on the amount of capital currently invested in the fund by the LPs, net of any return of capital distributed to investors. The investment period is the contractual period during which the manager may call capital for new investments. In contrast, the investment horizon (a term- or durationlike concept) is the period over which the manager has in fact put the capital to work. The assumption throughout the simulations is that base management fees are calculated on committed capital during the investment period and on invested capital for the remainder of the fund life. These assumptions reflect common practice in the closed-end real estate fund sector (see, e.g., Pension Real Estate Association [2016, pp ]). 7 Variation in this assumption would alter the simulated effects of management fees we report. It is important to note that charging management fees on committed capital during the investment period, followed by charging management fees based on invested capital thereafter, potentially provides an incentive to managers to deploy capital more slowly to maximize the fee base, creating a potential agency issue. Thus, there are two sides to the speed of deployment issue: Slow deployment increases the dilutive effects of MGTFEE and OPCOST; however, managers who deploy quickly or with inadequate investment selection, especially in declining markets, may not optimize investor returns. The assumed cash distributions to the LPs in each scenario, before the deduction of management fees, total $127.5 million over the life of the fund. In other words, we assume the gross multiple in each scenario is the same and only vary the speed of capital deployment and the Quarters to 50% deployment include four quarters of fundraising. THE JOURNAL OF PORTFOLIO MANAGEMENT 27

6 duration of the investment horizon. 8 Finally, we assume investors perceive the risk of each cash flow scenario to be identical. These assumptions allow us to focus on the effects of deployment speeds and investment horizons on performance, holding constant the magnitude and risk of future cash flows distributions across the four scenarios. The first scenario assumes a three-year investment period beyond the final close and a relatively slow deployment of capital, combined with a short investment horizon. Just 50% of committed capital is called by quarter 16; 100% of the capital ultimately called (assumed to be 85% of the total commitment) is called by quarter 22. The quarterly pattern of cash inflows and outflows associated with this and the remaining three scenarios are presented in Exhibits A1 to A4 of the Appendix. The IRR of this scenario, calculated as of the date of the first fund closing, is 18.13% prior to consideration of management fees or the opportunity cost of waiting to be called. The dilutive power of the quarterly management fee on LP IRRs and its interaction with the length of the investment horizon is not well understood. The LP IRR net of the management fee is 13.74%, a decline of 439 bps relative to the IRR, excluding the management fee. The steep decline in the IRR occurs because the management fee is based on 100% of the $100 million of committed capital until the end of the three-year investment period, which far exceeds the actual invested LP capital in the early years of the fund s life. The IRR decline is exacerbated by the assumption that only 85% of committed capital is eventually called. 9 Exhibit 3 provides a tabulation of the simulated effects of management fees and cost of waiting on performance. The base-case IRR for scenario 1 is displayed in the first column of Exhibit 3. The fourth and fifth rows in Exhibit 3 display the effects of including a 1.4% (annual) opportunity cost of waiting to be called. Even with this relatively low assumed value of 1.4% for OPCOST t, the IRR earned by the LPs declines an additional 281 bps to 10.93%. Assuming a 2.8% opportunity cost of waiting (displayed in rows six and seven), the IRR for the slow deployment short horizon scenario declines 532 bps to 8.42%. The bottom panel in Exhibit 3 displays the corresponding MOICs. The introduction of a management fee reduces the MOIC from 1.5 to 1.37 in our slow deployment short horizon (displayed in row nine). The MOIC is further reduced to 1.29 and 1.23 with OPCOST t equal to 1.4% and 2.8%, respectively (rows 11 and 13). 10 The second scenario assumes an investment period of just two years beyond the final close, a relatively fast deployment of capital, and a short investment horizon (duration equal to 9.81 quarters with MGTFEE t equal to 0). The IRR is 17.98% before consideration of the management fee or the opportunity cost of waiting (see column 2 of Exhibit 3). The IRR earned by the LPs net of the quarterly management fee is 14.30%, which is 56 bps greater than the corresponding IRR for the slow deployment short horizon scenario, isolating the effect of deployment speed. The addition of OPCOST t equal to 1.4% to this second scenario reduces the IRR by 239 bps to 11.91%. This reduction is smaller than that observed in the slow deployment short investment horizon scenario, due to the faster speed of deployment. Assuming a 2.8% annual opportunity cost, the IRR for the fast deployment short horizon scenario declines to 9.73%. Although the quarterly magnitudes of the opportunity cost of waiting may seem small relative to the magnitude of invested capital and cash distributions, the cumulative drag of OPCOST t on investor returns is substantial. As expected, the faster deployment of capital slightly reduces the negative effects of MGTFEE t and OPCOST t on MOICs. Overall, faster deployment of capital, when holding total cash distributions and the investment horizon constant, increases IRRs when management costs and opportunity costs are included. Moreover, the dilution of IRRs increases substantially as the magnitude of OPCOST t increases. The third scenario presented in column 3 of Exhibit 3 assumes a three-year investment period beyond the final close and the relatively slow deployment of capital assumed in scenario 1. However, this third scenario assumes a relatively long investment horizon (approximately 16 quarters) with cash f low distributions to investors not beginning until quarter 14. The IRR of this slow deployment long investment horizon scenario is just 7.75%, including MGTFEE t equal to 1.5% and no opportunity cost of waiting. This is a 599 bps (44%) decline from the corresponding slow deployment short investment horizon IRR of 13.74% reported in column Including OPCOST t equal to 1.4%, the LP IRR declines an additional 161 bpss to 6.14%. Finally, column 4 of Exhibit 3 displays the results for our fast deployment long investment horizon 28 THE IMPACT OF MANAGER DISCRETION AND DRY POWDER ON PRIVATE EQUITY REAL ESTATE RETURNS

7 E XHIBIT 3 Simulation Effects of Management Fees and Cost of Waiting on Performance Notes: Slow deployment reaches 50% deployment 12 quarters after final close, 16 quarters after date of legal inception. Fast deployment reaches 50% deployment 8 quarters after final close, 12 quarters after date of legal inception. Short investment horizon is a dollar-weighted investment period of 10 quarters. Long investment horizon is a dollar-weighted investment period of 16 quarters. Row (2) includes a 1.5% annual management fee, which is based on total committed capital during the fundraising and investment periods and based on actual capital deployed thereafter. Row (3.a) reduces the IRR and equity multiple by the inclusion of an opportunity cost of 1.4% of committed but uncalled capital. Row (3.b) reduces the IRR and equity multiple by the inclusion of an opportunity cost of 2.8% on dry powder. scenario, which assumes a two-year investment period beyond the final close. Holding constant the assumption of a long investment horizon, the estimated 6.95% IRR, assuming MGTFEE t equal to 1.5%, OPCOST t equal to 1.4%, and a relatively fast deployment speed, exceeds the corresponding slow deployment IRRs reported in column 3. However, the IRR of this fast deployment long horizon scenario is 496 bps less than the fast deployment short horizon IRR of 11.91%. Overall, the simulations provide important results. First, longer investment horizons are associated with substantially lower IRRs, assuming total cash total distributions to LPs are unaltered by the deployment speed or investment horizon. This reduction reflects the dollar-weighted nature of the IRR calculation and provides an incentive for general partners to distribute cash quickly to investors. Second, management fees and the opportunity cost of waiting to be called are highly dilutive to returns. However, the longer the investment horizon, the less dilutive the investment costs become. Overall, the combination of relatively long investment horizons and the inclusion of OPCOST t are highly dilutive to investor IRRs and should not be ignored when evaluating the absolute and relative performance of PERE funds. Having established through our simulation scenarios the significant effects of deployment speeds, investment horizons, opportunity costs, and management fees on PERE performance, we now turn to a comprehensive PERE fund database to empirically investigate the inf luence of these factors on PERE performance. The Appendix provides additional detail. PRIVATE EQUITY REAL ESTATE FUND DATA Sample Construction Most PERE managers are not required by regulation to disclose their distributions or returns to investors. This is true of most nonpublic investments, and this lack of disclosure greatly impedes the ability of researchers to assess the risk and return performance of nonlisted THE JOURNAL OF PORTFOLIO MANAGEMENT 29

8 E XHIBIT 4 Sample Construction: Private Equity Real Estate Funds investment vehicles. 12 Many PERE data sources are populated with voluntarily provided information, which creates a potential reporting bias in favor of the sponsor/ manager. Phalippou and Gottschalg [2009], for example, found some reporting bias in their analysis of venture capital reporting to the ThompsonOne database. Given the significant potential data problems with PERE funds, we carefully consider the available commercial PERE databases provided by Burgiss, Preqin, and Cambridge Associates (CA; com), as well as a proprietary database. Our objective is to construct the most complete and reliable dataset containing individual fund deployment speeds, IRRs, and MOICs, with sufficient information to calculate the dollar-weighted duration of each fund s investment horizon. We find that CA s dataset provides the most complete data to address our research questions. 13 CA receives fund performance data directly from managers, ensuring high-quality data and a deep time series. Because CA also provides back-office and reporting services for numerous managers, most managers voluntarily provide their operating performance to CA. CA does not make individual fund information publicly available unless a manager is in the market raising a new fund or gives CA permission to unlock their fund data to a subscriber. The confidentiality provided to managers mitigates reporting bias, and there appears to be no selection bias because CA requests information from any PERE fund of which it is aware. Additionally, once a manager s performance data have been included in the CA database, those data remain in the database even if the sponsor suspends reporting. We obtained from CA PERE fund information and quarterly performance metrics (net IRR, DPI, TVPI, and the quartile funded) as of 2016 for the lesser of 40 quarters, or the number of quarters reported by the manager based on the life of the fund. 14 The initial sample over the period consists of 658 funds, sponsored by 226 managers, 15 with $489.8 billion in total assets under management (AUM). 16 CA provided the quartile range funded for each fund but not the exact percentage funded. We delete funds that are less than 50% funded to ensure the reported performance metrics (i.e., realized, projected, or some combination of the two) are based on an identified and largely acquired portfolio of properties. To obtain a clean sample of closed-end, equity real estate funds, we also carefully identify and exclude real estate debt funds; funds providing financing to home builders for lot acquisitions; funds of funds; and funds targeting the infrastructure, agriculture, and health care sectors because they are less representative of general commercial and multifamily real estate exposures. We further delete funds with 1999 or earlier vintages. 17 The impact of these sample construction decisions are summarized in Exhibit 4. Our final sample consists of 497 funds sponsored by 201 distinct managers with a total AUM of $383.9 billion. We perform our primary analysis with the 497- fund sample, although our results are robust to the use of to the 548-fund sample that includes pre-2000 vintages. Exhibit 5 shows significant time variation in both fund count and total AUM by vintage year. It should be noted that 2005, 2006, and 2007 vintages (just before the financial crisis) are disproportionately represented in the sample. Overall, our data provide a comprehensive picture of PERE fund performance over a 14-year period. Descriptive Statistics To evaluate the performance of a fund, investors must construct a benchmark or reference point. In private real estate investment, benchmark indexes inevitably tend to be peer universe based. In this context, a peer universe would consist of all the competing funds for a given style and specialization. An index of returns based on a manager s peer universe is essentially similar to a Consumer Reports of the fund industry 30 THE IMPACT OF MANAGER DISCRETION AND DRY POWDER ON PRIVATE EQUITY REAL ESTATE RETURNS

9 E XHIBIT 5 Total AUM ($ millions) and Fund Count by Vintage Year E XHIBIT 6 Private Equity Real Estate Fund Sample Descriptive Statistics: (497 funds) (Geltner and Ling [2001]). It describes how the returns produced by a specific manager (brand) compare to other similar agents (brands of the same type of product) over a given historical period. An ideal peer universe benchmark for fund investors that would be most ref lective of manager performance would be constructed by aggregating the periodic (quarterly) cash inflows and outflows produced by the set of funds that constitutes the peer universe and calculating the IRR of these aggregate cash flows. In the absence of the underlying fund-level cash flows for the funds in the peer universe, as is typically the case, benchmark returns are constructed by averaging IRRs and MOICs across a representative sample of peer funds. However, averaging IRRs across funds can create a biased benchmark return. Carlson [2015], Phalippou [2008], and Phalippou and Gottschalg [2009] showed that averaging IRRs across different time horizons and different dollar weights is theoretically invalid and upward biased. Absent the ability to aggregate the underlying cash f lows of each fund in the peer sample to compute a benchmark IRR, these authors concluded that the best approximation of the true benchmark IRR is an average IRR that weights IRRs by both the duration (investment horizon) and (dollar) size of each fund. These (duration size) products are summed across all funds in the representative sample or subsample. The weight of each fund in the duration dollar-weighted IRR calculation is determined by its (duration size) divided by the sum of (duration size) for all funds in the sample. Exhibit 6 provides descriptive statistics for various IRR performance measures and the duration distribution of our 497-fund sample. These IRRs include the effects of management fees but not the opportunity cost of dry powder. In the top section of Exhibit 6, we provide the mean, median, 25th and 75th percentiles, and standard THE JOURNAL OF PORTFOLIO MANAGEMENT 31

10 E XHIBIT 7 Deployment Speeds: Quarters to 50% Deployment Notes: Only size tier and risk profile were statistically significant. Statistics for additional characteristics are provided in Exhibit B1 in the appendix. Size Tier: By vintage year, dividing the sample into two equal groups. Risk Profile: low risk includes core, core-plus, and value added; high risk includes opportunistic, development, and distressed.,, significant at 10%, 5%, and 1% levels, respectively. deviations for simple average IRRs, dollar-weighted IRRs, and duration dollar-weighted IRRs as well as (unweighted) durations and dollar-weighted durations. Simple averages reflect equally weighted IRRs. For dollar-weighted IRRs, we use a weighting factor provided by CA based on the relative capital deployed by each fund in relation to the aggregate capital deployed across the entire 497-fund sample. For duration dollar weighted IRRs, we also use each fund s duration (or investment horizon) and the CA weighting factor just described. In the bottom section of Exhibit 6, we provide conditional-duration descriptive statistics by IRR performance, ranking quartiles using simple average and dollar weighting. In the top section of Exhibit 6, we see that average fund performance depends on whether we are using a simple average, as is typically reported, dollar weighting, or duration dollar weighting. The mean IRR of the 497- fund sample is 8.2% using a simple average and rises to 9.3% using dollar weighting. This indicates that larger funds tended to perform better over the sample period. The median and percentile range results confirm the outperformance of larger funds and indicate that this performance is not driven by outliers. The results in Exhibit 6 also show that duration dollar weighting of IRRs has a dramatic effect, lowering the mean IRR of the sample to 4.6% and the median IRR to 5.8%. These results document that the durations of the various IRRranked cohorts have a significant downward effect on performance, with the lowest IRRs having the longest durations. The duration results reported in the top section of Exhibit 6 reflect duration in years at a point in the IRR distribution (i.e., at the 25th percentile, mean, etc.). In contrast, the duration results shown in the bottom section of the exhibit are measured across each IRR performance quartile. More specifically, the 497 funds are sorted into four sorts by IRR performance: 75% 100% is the top quartile of IRR funds, whereas 0% 25% is the lowest quartile of IRR funds. The simple and dollarweighted mean duration in years within each of these IRR performance quartiles is reported. The top part of Exhibit 6 shows a mean duration (observed and dollar-weighted) over the entire sample of approximately four years. However, the mean duration varies by approximately ±1.5 years between the 25th and the 75th percentiles. From the bottom part of Exhibit 6, we observe that the durations of the various IRR cohorts differ significantly. The top-quartile IRRs (75% 100%) have a dollar-weighted duration of only 2.1 years, whereas the lowest-quartile IRRs (0% 25%) have a dollar-weighted duration of 6.4 years. That is, the best-performing funds in effect put capital to work for only 2.1 years, whereas the worst-performing funds deployed capital for 6.4 years. Importantly, the lowest duration dollarweighted IRRs have a weight that is approximately three times the weight of the highest duration dollar-weighted IRRs. All else equal, these results suggest a strong inverse relation between investment horizon and IRR. Moreover, these findings raise serious concerns about the use of simple IRR averages as a performance benchmark. 32 THE IMPACT OF MANAGER DISCRETION AND DRY POWDER ON PRIVATE EQUITY REAL ESTATE RETURNS

11 E XHIBIT 8 Average Deployment Speed by Vintage Year PERE DEPLOYMENT SPEEDS AND PERFORMANCE Evidence on Deployment Speeds In Exhibit 7, we report PERE fund deployment speeds for our sample of 497 funds, where we measure deployment speed as the number of quarters to 50% deployment. We provide the mean, median, and 25th and 75th percentiles as well as standard deviations of the measured deployment speeds. The empirical data shown in Exhibit 7 informed our choice of 12 quarters (fast deployment) and 16 quarters (slow deployment) to 50% deployment in our four simulations. In addition to providing descriptive statistics on deployment speeds for the full sample of 497 funds, we also calculated results by fund size tier, change in fund size, vintage volume, geography, manager ownership, and risk profile. Exhibit 7 illustrates the cross-sectional variability of deployment speeds. The overall sample mean and median are 10.5 quarters and 10.0 quarters, respectively, to 50% deployment. Larger and higher risk funds tend to deploy at modestly slower speeds. However, we find a strikingly large range in deployment speeds within each segment. For example, the 25th percentile for small funds is just 6 quarters; at 14 quarters, the 75th percentile for small funds is substantially longer. Overall, our results suggest that deployment speeds vary significantly across funds as well as by some fund characteristics, such as size and risk. Exhibit 8 provides evidence on deployment speeds by vintage. In recessionary environments, such as and , the average fund deploys capital more slowly. After accounting for this cyclicality, deployment speeds have trended downward over our sample period. In untabulated results, we also find significant time variation in deployment speeds across each of our fund characteristic segments. The Influence of Deployment Speeds, Investment Horizons, and Opportunity Costs on Private Equity Real Estate IRRs and Equity Multiples In Exhibit 9, we sort the PERE funds by investment horizon from shortest to longest quartile. In the top part of the exhibit, we report simple averages, whereas the bottom part of the exhibit reports results that use dollar weighting. The respective columns in Exhibit 9 provide the deployment speeds in quarters, unamortized opportunity costs, mean investment terms in years, IRRs, adjusted IRRs with a 1.4% opportunity cost, amortized opportunity costs, and equity multiples. 18 The results in this exhibit illustrate the relations among deployment speeds, opportunity costs, and IRRs by investment horizon quartiles. We can see in Exhibit 9 that the unamortized opportunity cost is a function of each fund s deployment speed, ranging from 295 bps for the 8.42 quarter deployment speed to 405 bps for the quarter deployment speed for the simple average results. The dollar-weighted results are similar. However, as shown in the annual amortized opportunity cost column, the IRR dilution effect is largest for the shortest investment horizons and smallest for the longest investment horizons. That is, adjusting the simple and dollar-weighted IRRs in column 4 by the unamortized opportunity costs in column 2 and amortizing over the mean investment period shown in column 3 dilutes IRRs from a high of 262 bps annually THE JOURNAL OF PORTFOLIO MANAGEMENT 33

12 E XHIBIT 9 Investment Horizon/Duration E XHIBIT 10 Private Equity Real Estate IRR Performance: ,, Significant at 10%, 5%, and 1% levels, respectively. 34 THE IMPACT OF MANAGER DISCRETION AND DRY POWDER ON PRIVATE EQUITY REAL ESTATE RETURNS

13 E XHIBIT 11 Private Equity Real Estate Equity Multiples: ,, Significant at 10%, 5%, and 1% levels, respectively. for the shortest investment horizon quartile to a low of 42 to 50 bps for the longest investment horizon quartile. Clearly, the annual impact of PERE fund opportunity costs is largely a function of the length of the investment horizon over which those costs are amortized. These results are also illustrated in Exhibit 4 by comparing the management fee and opportunity cost effects between long and short investment horizons (column 1 versus column 3 and column 2 versus column 4). In Exhibit 10, we report simple, dollar-weighted, and duration dollar-weighted average IRRs by fund size, change in fund size, vintage volume, geography, manager ownership, and risk profile in the first three columns. These unadjusted IRRs do not include an adjustment for the opportunity cost of dry powder. In columns 4 and 5, we report IRRs that include adjustments for opportunity costs of 1.4% and 2.8% per annum, respectively. Looking first at columns 1 through 3, we find significant differences in unadjusted IRRs within and across each fund characteristic segment. 19 Except for small funds, domestic funds, and low risk funds, dollar-weighted IRRs exceed simple average IRRs. These results indicate that larger funds tend to outperform smaller funds. 20 Looking further down the dollar-weighted IRR column, we see that funds that grow more slowly (small changes in fund size) outperform funds that grow more quickly (large changes in fund size), funds raised in low-capital vintages outperform funds raised in high-capital vintages, domestic funds outperform international funds, and high-risk funds outperform low-risk funds. These figures are after management fees but before any consideration of the opportunity cost of waiting to be called. In column 3 of Exhibit 10, we report IRRs calculated by overlaying duration on dollar weighting. This more comprehensive IRR measure drops dramatically. For example, for the full sample, the simple weighted IRR is 8.2% and the dollar-weighted IRR is 9.3%. The calculated IRR drops to 4.6% when we use duration dollar weighting. Each fund characteristic segment experiences THE JOURNAL OF PORTFOLIO MANAGEMENT 35

14 a similar dramatic drop in IRR when duration weighted. To further understand the drivers of this adjustment, recall that Exhibit 6 showed the significant differences in durations for different dollar-weighted IRR quartiles. For example, the top quartile of dollar-weighted IRRs has a duration of only 2.1 years, whereas the bottom quartile of dollar-weighted IRRs has a duration stretching to 6.4 years. Comparing the top quartile to bottom quartile duration results, the underperformers carry three times the weight of the best performers. In columns 4 and 5 of Exhibit 10, we provide results that account for the effects of annual opportunity costs of 1.4% (higher risk tolerance) and 2.8% (more conservatively reserved for uncalled capital), respectively, on IRR performance. For each of the 497 funds, we determine the speed of deployment by observing the period between the year of legal inception (using a midyear convention) and the first quarter reported to be 50% called. We then impose an opportunity cost on each fund of either 1.4% or 2.8% per year or prorated per partial year and then recalculate each fund s IRR. 21 After adjusting each fund s IRR to account for its speed of deployment or investment lag, we then dollar- and duration dollar weight by the same method used for the unadjusted fund IRRs. After accounting for opportunity costs in columns 4 and 5 of Exhibit 10, we find that the duration dollarweighted IRRs are considerably lower. For example, the full sample simple average IRR of 8.2% in column 1 is reduced to 4.6% with duration dollar weighting (column 3). The IRR is further reduced to 3.6% with a 1.4% opportunity cost (column 4) and to just 2.6% with an opportunity cost of 2.8% (column 5). Looking further down columns 4 and 5, we see that the duration dollarweighted IRRs, accounting for opportunity costs, are often many multiples lower than their corresponding simple average IRRs. The low vintage volume, small change in fund size, and domestic fund characteristic segments are least affected, whereas the large change in fund size, high vintage volume, international, and low-risk fund characteristic segments are most affected. In Exhibit 11, we report PERE equity multiples in a similar format to the IRR results in Exhibit 10. Looking first at columns 1 3, we again find significant differences within and across each fund characteristic segment. In contrast to the earlier IRR results in Exhibit 10, the dollar-weighted equity multiple results do not provide generally higher multiples than the simple averages. However, column 3 of Exhibit 11 shows that when we overlay duration on the dollar weighting, we find that the equity multiple performance metric drops dramatically. Finally, in columns 4 and 5 of Exhibit 11, we find that the duration dollar-weighted equity multiples net of the opportunity cost of dry powder are considerably lower than the unadjusted MOICs reported in columns 1 3. Overall, the results in Exhibits 10 and 11 show that deployment speeds, investment horizons, and opportunity costs have a significant impact on measured PERE IRRs and equity multiples. CONCLUSION PERE funds are an important investment vehicle in real estate capital formation and development, yet there is limited research on the effective performance of PERE and the role of fund managers discretionary timing of investment decisions, the calling and deployment of capital, and fee structures on returns. This study investigates the performance sensitivity of PERE funds to capital deployment speeds, investment horizons, management fees, and investor opportunity costs associated with committed but uncalled capital. These factors are largely ignored in traditional PERE performance metrics and in the literature but are important in measuring PERE performance given the structure of PERE investments. We first provide a series of simulation scenarios demonstrating the significant effects of these factors on PERE performance. We then use a large sample of 497 funds sponsored by 201 managers with an aggregate AUM of $383.9 billion from to empirically investigate performance effects. Our simulations demonstrate that the shorter the investment horizon, the more dilutive the management fee and the opportunity cost applied to the committed but uncalled capital (i.e., dry powder). We also show that the slower the deployment pace and the lower the percentage of committed capital actually called, the more dilutive the management fee and the opportunity cost of maintaining dry powder. Importantly, the degree of dilution based on our simulations exceeds what most investors assume. For example, a 150 bps management fee dilutes IRRs by 234 to 439 bps, depending on deployment speed, maximum percentage of commitment called, and investment horizon. Uncertainty in the timing and amount of capital calls creates opportunity cost and varies based on the investor s alternative 36 THE IMPACT OF MANAGER DISCRETION AND DRY POWDER ON PRIVATE EQUITY REAL ESTATE RETURNS

15 sources of liquidity and tolerance for risk. Because the deployment pace and investment horizon are a matter of manager discretion, the impact of the management fees and opportunity costs are difficult to ascertain ex ante. Using a comprehensive PERE dataset, we find that fund deployment speeds vary significantly across funds and over time. However, little of this variation is incorporated in traditional private equity performance metrics. We find that the dilutive effects of management fees are positively related to the time over which capital is deployed and negatively related to the percentage of net capital called from investors and deployed by the fund manager. Overall, our results show the importance of accounting for the dilutive effects of capital deployment speeds, investment horizons, management fees, and uncalled capital on measured PERE IRRs and equity multiples. Given the heterogeneity of private equity returns and the resulting importance of manager selection within the industry (Kaplan and Schoar [2005]), it is all-important to measure performance and its related costs as precisely as possible. Our findings are also prescriptively important in that they yield the following suggestions to address our E XHIBIT A1 Slow Deployment and Short Investment Horizon documented negative performance effects: (1) longer capital call notice periods (the industry standard is typically 10 days, whereas the manager often has 30 or more days to fund an investment); (2) shorter investment periods that can only be extended by the agreement of a supermajority vote of investors; (3) a cap on the percentage of undrawn capital that a manager may call during a specific time period; (4) a rebate of any management fees based on committed capital not ultimately called; (5) a convention for a return of basis first to reduce the unreturned capital that forms the basis of the calculation of management fees after termination of the investment period; and (6) significantly reduced or waived fees on committed but uncalled capital to shift more of the manager s focus toward performance fees and away from guaranteed fees. A PPENDIX A FOUR SIMULATION SCENARIOS This portion of the appendix provides more detail on the four simulation scenarios. The first scenario assumes a THE JOURNAL OF PORTFOLIO MANAGEMENT 37