“Private Equity Performance: Returns and Benchmarking”
Baladeva Keilman Masters (MSc) in International Finance – Master Thesis Amsterdam Business School, University of Amsterdam Supervisor: Dr Ludovic Phalippou September 2013 Abstract: This thesis compliments the ongoing academic research and question of PE performance and benchmark choices by providing a perspective of returns to the LP investor and supports the conclusion that PE performance is superior against a large cap driven benchmark as shown by many recent academic papers but that this does not hold against more appropriate benchmarks. Using the cash flow data of two large US pensions funds, CalPERS and SBA, over the last twenty years, this thesis demonstrates that the PE performance is the same as what has been found in recent academic papers when benchmarked against the S&P 500. However, large cap stocks are not representative of PE portfolio investments and when benchmarked against a more comparable index, in particular, a leveraged small‐value index, the PE portfolio of CalPERS has not been able to outperform the benchmark whilst SBA has underperformed by 3,2% annually.
Contents
1. BACKGROUND ... 4 2. RESEARCH QUESTION ... 6 3. LITERATURE REVIEW ... 10 4. DATA SOURCES ... 12 4.1 CalPERS ... 12 4.1.1 Background: CalPERS PE Equity Program ... 12 4.1.2 Cash Flow Data ... 13 4.1.3 Net‐Asset Values (NAVs) ... 13 4.1.4 Time Periods ... 14 4.1.5 Fund Types & Geographies ... 14 4.1.6 CalPERS Website Performance Data and Annual Financial and Investment Reports ... 15 4.2 SBA ... 16 4.2.1 Background: SBA PE Investments ... 16 4.2.2 Cash Flow Data ‐ SBA ... 16 4.2.3 Net‐Asset Values (NAVs) ... 16 4.2.4 Time Periods ... 16 4.2.5 Fund Types ... 17 4.3 Benchmarks ... 17 5. DESCRIPTIVE STATISTICS AND DATA ISSUES ... 18 5.1 CalPERS Data Set ... 18 5.2 SBA Data Set ... 19 5.3 Data Discrepancies ... 20 6. DATA ANALYSIS – PE PERFORMANCE AND BENCHMARKING ... 22 6.1 CalPERS Absolute Performance Analysis – Net IRRs and Multiples ... 22 6.1.1 CalPERS Website IRR and Multiples vs Data Set ... 23 6.1.2 CalPERS Data Set IRRs and Multiples by Vintage Years and Periods ... 23 6.1.3 CalPERS Annual Financial Reports – Historical PE Performance as Reported ... 24 6.2 SBA Absolute Performance Analysis – Net IRRs and Multiples ... 25 6.2.1 SBA Reported IRRs and Multiples vs.Data Set ... 25 6.2.2 SBA Data Set IRRs and Multiples by Vintage Years and Periods ... 26 6.2.3 SBA Annual Investment Reports – Historical PE Performance as Reported ... 26 6.3 Duration of PE Portfolios ... 27 6.4 LP Policy Benchmarks Tested Using the Public Market Equivalent (PME) Methodology ... 28
6.5 Relative PE Performance and Choosing an Appropriate Benchmark ... 31 6.5.1 Large Cap & Total Market Indices ... 31 6.5.2 Size Premium ... 31 6.5.3 Value Premium ... 32 6.5.4 Tradability ... 34 6.5.5 Leverage ... 35 6.5.6 PME Performance Results by Vintage Years ... 35 6.5.7 Additional Risks ... 36 6.5.8 Robustness Checks ... 36 7. CONCLUSION ... 37 BIBLIOGRAPHY ... 38 TABLES 1‐10 ... 40 Table 1 – CalPERS PE Investments Descriptive Statistics ... 40 Table 2 – SBA PE Investments Descriptive Statistics ... 42 Table 3 – IRRs and Multiples Analysis ‐ Cash Flow Data Set versus LP Website/Reports ... 44 Table 4a ‐ CalPERS PE Performance vs Policy Benchmark as per Report Year ... 47 Table 4b – SBA PE Performance vs Policy Benchmark as per Report Year ... 48 Table 5 ‐ PE Relative Performance Using LP Policy Benchmarks and the PME Methodology ... 49 Table 6 ‐ PE Performance and the Size Premium ... 50 Table 7 – PE Relative Performance and the Value Premium ... 51 Table 8 – PE Relative Performance versus Mutual Funds ... 52 Table 9 – PE Relative Performance and the Effect of Leverage ... 53 Table 10 – PE Fund Performance by Vintage Year ... 54 Table 11– Robustness Checks ... 56 FIGURE 1 – Size Premium Spread ... 58 FIGURE 2 – Annualized 10 Year Forward Looking Moving Average Returns of Key Indices ... 59 APPENDICES ... 60
1. BACKGROUND Private Equity (PE) has been attracting more attention from academic researchers over recent years given the increased importance of this asset class for institutional investors and the huge increase of capital going into PE funds. Over the last seventeen years annual capital commitments to PE have risen from $38 billion globally per year in 1995 to $682 billion at the peak of the last PE cycle in 2008 and have continued to average close to the $300 billion mark for each of the last four years1. Arguably, one of the underlying causes for the increased interest of investors in this asset class has been the reported superior performance of PE relative to the major public equity indices such as the S&P 5002. For example, a recent article on Reuters.com stated “Private equity has established a track record of outperforming other asset classes. The U.S. private equity index compiled by advisory firm Cambridge Associates LLC shows a net internal rate of return (IRR) of 13.7 percent in the 10 years through September 30, 2012, compared with an 8 percent return by the S&P 500 Index”.3 However, the issue of superior performance is not so clear cut for three principal reasons. Firstly, standard measures of performance used in PE, namely the Internal Rate of Return (IRR) and Cash Multiple, have significant drawbacks which have been well documented in Corporate Finance textbooks and by PE academic researchers4. Phalippou (2007) argues that the performance data provided by GPs prevents relative performance benchmarking across GPs as IRRs are an absolute measure and so are not easily comparable. In addition, the IRR implicitly assumes that returned capital is reinvested at the IRR rate, which means that an early big success in which a small investment makes a large return over a relatively short period of time will result in a high IRR that will overly influence the IRR for the lifetime of the fund irrespective of the results of later investments (Ang and Sorenson, 2013). However, the core issue in the current debate of superior performance is that most comparisons of PE performance to date have tended to benchmark PE returns against large cap stock indices such as the S&P 500 or broad market indices which are value‐weighted and thus driven by the large stocks which make up over 80% of the value. However, as the recent paper by Phalippou (2012) points out, PE funds are more closely related to small‐value indices given the size and type of companies that 1 Global Private Equity Report 2013, Bain & Company 2 For example, the website of CalPERS, the LP investor whose investment history is the basis of this Thesis, states “Historically, private equity has generated superior performance when compared to public equity” (as of 26th June 2013) 3 http://www.reuters.com/article/2013/02/27/us‐privateequity‐conference‐returns‐idUSBRE91Q0DT20130227 4 Phalippou 2008 and 2009b; Ang and Sorenson (2013)
typical PE funds invest in. Phalippou therefore argues that PE performance should really be benchmarked against small‐value stock indices to determine whether true outperformance exists. Indeed, Higson and Stucke (2012, p.25‐26) conclude in their paper on PE performance relative to the S&P 500 with the comment that “hard conclusions about whether or not the U.S. buyout industry has created alpha for its investors requires further research on the appropriate benchmark”. Furthermore, there remain a lot of outstanding questions about the true costs and risks of investing in PE. For example, standard PE performance measures such as IRRs and Multiples and even the Public Market Equivalent (PME) measure do not capture the Liquidity Risk of having capital tied up in Illiquid investments for 6 years on average (Phalippou, 2009) nor the Commitment Risk of providing a credit line for the total capital committed, typically around five years. Then, there is also the issue of leverage as many PE portfolio companies tend to be highly levered which would imply higher levels of market risk (Beta). Returns should therefore be adjusted to account for this higher level of risk when performance is being compared relative to other asset classes. The fact that such important issues around risk and performance remain outstanding around an asset class that has already become an integral part of the asset allocation for institutional investors may seem surprising. However, PE involves long term investments with life‐times of funds running between 10‐13 years and by definition concerns investments in private companies. As such, finding suitable data samples to work with has been one of the most challenging aspects for academics to test the performance claims of the industry. PE funds do not openly share their underlying performance data in terms of cash flows, timings, fees, etc, which are necessary to undertake a thorough performance analysis, whilst investors have also been restricted from sharing information given the strong confidentiality clauses in place when investing in PE funds. As a result, academic research into PE fund performance with suitable data sets has only really been possible over the last ten years or so but is currently an area of increasing interest with many important questions still to be answered, of which PE performance relative to an appropriate benchmark is a key issue.
2. RESEARCH QUESTION This Thesis paper will first review the existing academic research around (relative) PE performance and appropriate benchmarking to understand the latest developments and evidence on this topic. Then, largely following the structure of the research paper of Phalippou (2012) it will use cash flow data from two major US based pension funds who have been investing in PE for over the last 20 years to analyze the net‐of‐fees absolute and relative performance of their investments in the PE asset class using various measures. The analysis will focus on the effects of using different benchmarks to determine over‐ or under‐performance, trying to establish the most appropriate benchmark for the investments and ultimately, whether they have out‐performed this benchmark. The data relate to investments from 1989 until the end of 2012 with a combined total of $65 billion invested over this time period covering 290 PE funds with the first investor (CalPERS based in California) and 145 funds with the second (Florida State Board of Administration “SBA” ‐ based in Florida). We begin by examining the publicly reported performance (net‐IRRs and multiples) and compare this with the cash flow data provided by the two Limited Partner5 (“LP”) investors. Comparing CalPERS value‐weighted net‐IRRs of the current PE funds as published on their website and using a subset of the same matched funds from our data sample (a total of 278 funds versus 290 for the full data sample) we get a net‐IRR of 9,93% for the website versus 9,74% for our data sample (and 9,65% for the full data sample). The cash multiple is also slightly lower with our data set compared to the website, 1,32 versus 1,41 (and 1,32 for our whole data sample). Comparing the value‐weighted net‐IRRs of our cash flow data set from SBA with the investment report we find a net‐IRR of 7,82% versus 6,64% ‐ somewhat lower than CalPERS. The multiples are roughly the same at 1,29 versus 1,30 respectively. However, Phalippou (2009b, p.12) points out that the average or median IRR/multiple is not a good measure for the performance of an asset class “because they cannot be benchmarked and, even as a measure of absolute performance, they are of limited interest. IRR cannot be averaged (since it is a geometric average and investments have different durations) but even if it could be averaged, its implicit re‐investment assumption means that the average IRR can be far away from the overall performance experienced by investors”. Therefore, we move to the key issue of this thesis to determine a calculation methodology that overcomes some of the issues of IRRs and multiples and to 5 Limited Partner (“LP”) refers to the financial investors in private equity fund structures as opposed to the General Partners (“GPs”) who are the managers of the fund and responsible for investing and overseeing the funds.
find a way of benchmarking that will allow us to judge whether the PE portfolios have out‐performed or not. To overcome the problems of IRR and Multiple performance measures, academic research has moved towards the Public Market Equivalent (PME) for measuring PE fund investments. The PME is effectively the sum of the present value of each distribution divided by the sum of the present value of each cash investment (capital call) where the discount rates used are the cumulative returns of a public equity index at time (t) for each cash flow. The index is therefore a benchmark and a result of one tells us that the investments have performed the same as the benchmark whilst a PME of above one tells us the investments have performed better than the benchmark and vice‐versa for a PME less than one. As this is an NPV6 calculation it resolves many of the shortcomings of the IRR method. The key question of this thesis though, is which benchmark is the most appropriate to use for assessing PE investments and then whether PE investments out‐perform or under‐perform such a benchmark? We begin by looking at the ‘policy benchmarks’ used by the LPs themselves to compare performance of their PE investments and then use the PME method to compare results against the reported relative performance of the LPs7. Both LPs use a broad (total) market index that is value weighted – and thus dominated by the performance of the large cap stocks – plus a premium currently of 3% p.a. to account for the different risk characteristics of PE (SBA had a premium of 4,5% until 2010)8. We find that the reported relative performance figures of the LPs are fairly close to our PME results. CalPERS reports a 2,4% annual over‐performance against its’ benchmark on average whilst in our analysis of the data set we find an average PME of 1,064 which is a 6,4% over‐ performance in a 2,93 year average fund holding period9 and is thus equivalent to 2,1% annually. SBA, on the other hand reports a ‐2,3% annual under‐performance against its policy benchmark whilst our PME analysis found an average ‐4,04% under‐performance which, given a 2,32 average holding period, results in a ‐1,76% annual under‐performance. However, part of this is due to the higher premium attached to the SBA’s benchmark until 2010 of 4,5%. When measured against the CalPERS benchmark, SBA outperformed the benchmark by 1,18% annually. Clearly, the issue of choosing an appropriate benchmark and the premium to attach to PE investments is something that LP investors are also struggling with. This can be seen by the numerous changes to the PE policy benchmarks since they began investing. Indeed, in the early years of PE investing, both LPs simply 6 Net Present Value (NPV) 7 As per the LPs annual financial and investment reports published on their websites 8 CalPERS’ benchmark is 67% Wilshire + 33% FTSE AW (ex US) + 3% p/a. (Lagged 1Q). SBAs’ benchmark is currently Russell 3000 +3% although for 2010 and earlier the premium was 4,5%. 9 Average fund holding period, i.e. duration, methodology is explained later on in the performance analysis section
used a hurdle rate as the benchmark – for example CalPERS used a flat hurdle of 15% for their PE investments in 2002 and earlier years (as shown in Table 4a). Recent academic research, however, has tended to use a large cap index (such as the S&P 500) when analyzing PE performance. This is despite the fact that the LPs themselves already recognize that there must be an additional premium added to such a benchmark due to the higher risk factors of PE investing. One additional question is therefore, whether a seemingly arbitrary 3% premium is a correct adjustment or not for PE investments. We continue our performance analysis by largely following the structure of the research of Phalippou (2012). We calibrate the PME to several different benchmarks including the CRSP value weighted index10 and Vanguard S&P500 index mutual fund amongst others to compare with recent academic studies using similar large cap driven indices. We then also analyse performance against the Fama & French size indices11, the Fama French small‐value index12 and the DFA‐value mutual fund13 to capture the size and value premium that Phalippou proposes more closely reflect PE investments. We find that CalPERS outperforms the Vanguard S&P 500 index by 5, 7% annually (given a PME of 1,178 and average holding period of 2,93 years), which is exactly the same as Phalippou’s results and in‐line with other academic research. SBA’s performance is lower as they outperform the S&P 500 by 4,3% annually (given a PME of 1,084 and an average holding period of 2,32 years). However, against the small cap index of Fama‐French, which would fit 95% of PE portfolio companies by market value, performance drops significantly. We find a PME of 1,086 for CalPERS and 1,018 for SBA which are equivalent to an outperformance of 2,9% annually and 0,8% annually for CalPERS and SBA respectively. Against the DFA Value mutual fund index, performance is slightly lower with a PME of 1,0785 and 0,989 for CalPERS and SBA respectively. The PME for SBA is below one which means that SBA has actually under‐performed against the DFA Value mutual fund and a negative t‐stat of ‐ 0,42 means that SBA’s average PME is not different from one at a 5% significance level. Against the Fama French small‐value index, results drop even further though. CalPERS has a PME of 1,037 or 1,2% annually, however, it is not statistically different from one based on the t‐test at a 5% significance level. Meanwhile SBA has underperformed with a PME of 0,967 or ‐1,4% annually. 10 The CRSP indices are downloaded from WRDS and cover the key US stock markets, Dow Jones, NASDAQ etc. 11 The Fama‐French size indices are based on the CRSP stock indices sorted into three size bins based on market equity size of the company. Available from Ken French website 12 The Fama French small‐value index is one of the 6x2 portfolios double sorted by Fama and French based on size (large versus small) and style (growth, mid or value). 13 Dimensional Fund Advisors (DFA) are well known for running low‐cost passive mutual funds in small stocks and value stocks.
Clearly, a small‐value benchmark would show that these LPs have not outperformed with their PE investment. We note that this drop in performance is not as dramatic as per the results of Phalippou (2012), where the PME against the Fama‐French index dropped to 0,96 ( a drop of 24% over the time period compared to a drop of 14% for CalPERS and 11,7% for SBA). The next part of the analysis is to adjust the benchmarks for the higher leverage factor of PE investments compared to the market index benchmark. The PME analysis so far generally assumes a beta of 1 and that PE investments carry the same risk profile as the market. However, PE investments typically involve considerable leverage, endowing portfolio companies with a capital structure containing 50% to 70% debt, which is much higher than comparable public companies. Thus, one should expect risk to also be higher. We test the PME results against a 1,3x and 1,5x leveraged index and find that CalPERS PME against the Fama French small‐value index drops to 1,001 and SBA’s to 0,927. CalPERS PE portfolio has performed in‐line with the benchmark whilst SBA has under‐performed (the large negative t‐stat indicates that the average PME is different from one at the 5% level). Thus, against the leveraged small‐value benchmark, PE has at best equaled its benchmark or more likely under‐performed for these investors. Finally, we carry out some robustness checks to show that our shortened sample start periods (1995) don’t affect the average PMEs compared to the full data samples. We also vary the sample end periods to contain only mature funds and we test sub‐sets of the data sample based on PE fund type to determine how these may affect our results.
3. LITERATURE REVIEW Early evidence on the performance of PE funds showed mixed results. Ljungqvist et al (2003) found that PE funds had generated an excess return of almost 6% per annum compared to the S&P 500 index. On the other hand Kaplan et al (2005) found that PE Buy Out funds had on average underperformed the S&P 500 whilst Phalippou et al (2009) concluded that PE funds had considerably under‐performed by over 3% per annum when benchmarked against the S&P 500 on a net‐of‐fees basis14. In explaining the continued growth of funds into PE despite an apparent underperformance for the LP investor, Phalippou (2007) proposed that selection of poor performing funds could be due to mispricing by the investors. Essentially, investors may have a biased view of PE fund performance because of the often confusing and non‐standard performance reporting procedures. Issues such as reporting returns gross‐of‐fees (which may be 12% higher than net‐of‐fees performance according to Swensen (2000)), aggregation of IRRs and the near impossibility of benchmarking funds’ past performance given limited information, mean that investors may simply not be able to accurately price the asset class. In response to the findings of Kaplan et al (2005), Phalippou et al (2009) and a paper by Stucke (2011) that threw into doubt the reliability of the data sets used in those earlier papers, a number of academic papers were recently published using up‐to‐date, larger and better quality proprietary data samples than had previously been used. These papers determined that PE demonstrates returns (net‐of‐fees) clearly in excess of the S&P 500. Harris et al (2012) used data on 1400 Buy Out and VC funds and found that the average PE Buy Out fund had outperformed the S&P 500 by over 3% per annum (or almost 27% over the average lifetime of the fund) whilst US VC funds had outperformed the S&P 500 in the 1990’s but had underperformed in the 2000’s. Robinson & Sensoy (2011) found that the average Buy Out fund outperformed the S&P 500 by 18% over the lifetime of the fund whilst Higson & Stucke (2012) used a data set that they claimed represented 85% of capital ever raised by US Buy Out funds and determined that they had significantly outperformed the S&P 500 by between 4.5% and 6% per annum. However, Phalippou (2012) questioned the conclusion that PE as an asset class provided superior risk‐adjusted returns. In particular Phalippou (2012) questioned whether the use of S&P500 as a benchmark was appropriate given that there is a size and value premium that should be considered 14 All references to PE returns in this paper refer to returns net‐of‐fees to LPs, i.e. returns to LPs after the GPs have deducted all management fees and carried interest charges.
in PE, as well as noting that there were several other costs that were not included in the analysis such as illiquidity risk, commitment risk and other external costs and fees that investors in PE often must bear. Using publicly available data, Phalippou (2012) was able to recreate the results of the studies of Harris et al (2012) and Higson et al (2012) and found that the average buyout fund outperforms the S&P 500 by about 5.7% per annum. However, Phalippou (2012) firstly showed that 95% of buy‐out investments related to portfolio companies whose market value would fall within the Fama & French small stock index or the small‐cap index of the Dimensional Fund Advisor (DFA) small‐cap mutual fund15 and secondly that PE funds investments were in value companies rather than growth companies. Furthermore, PE transactions also tend to include high amounts of leverage and so when benchmarked against a leveraged small‐value index the average buyout fund underperformed by 3.1% per annum. In essence, he concluded that “the average buyout fund performance is similar to that of small‐cap stock indices, of a passive small‐cap mutual fund (DFA small cap), and of a passive value mutual fund (DFA value).” This Thesis complements these recent academic studies by comparing their results against a more up‐to‐date data set, with cash flows from 1989 to the end 2012, and further testing the conclusions of Phalippou (2012) on the sensitivity of returns to different benchmarks. It also has the added advantage of using the cash flow data from the point of view of the end investor, where fees and carried interest payments will have already been deducted, permitting us to build a view of the true investment performance to the end investors in this asset class. Recent major studies on PE performance have tended to build data sets that cover the entire PE fund industry to estimate the average performance of the asset class for investors. However, there are always a certain percentage of funds that will be missing from such an analysis and there are dangers of selection bias due to reporting bias (where poor performing funds may choose not to report with data aggregators). By looking from the investor perspective we can determine performance to the end investor but of course this suffers from the limitation that two investors (despite being large investors) cannot reliably be used to proxy average PE performance as an asset class. Nonetheless, this thesis will complement existing studies by providing a contrasting perspective that is useful to build a multi‐dimensional view on PE performance returns. 15 The largest stock in the Fama & French small cap index was $1.08 billion and the largest in the DFA small mutual fund was $1.13 billion.
4. DATA SOURCES The main focus of the PE performance analysis of this thesis is based on cash flow data which has been obtained by my thesis supervisor directly from CalPERS and SBA under the Freedom of Information Act. In addition to the cash flow data, we have obtained PE performance data from the respective websites, financial reports and investment reports of the pension funds which we can use to compare against our cash flow data and give a broader perspective of some of the issues around PE performance reporting. Finally, to carry out the performance benchmarking it was necessary to obtain the index returns for each of the benchmarks for the time period in question. This data came from three sources being: Thompson Datastream, Wharton Research Data Services (WRDS) and Ken French’s website. 4.1 CalPERS 4.1.1 Background: CalPERS PE Equity Program CalPERS is the largest US public pension fund and one of the largest pension funds in the world with over $257billion of assets under management16. CalPERS is also one of the largest PE investors in the world. Since it first started investing in PE in 1991 until December 31, 2011, CalPERS has contributed $53.0 billion in PE investments, received distributions of $38.5 billion with a remaining reported value of $31.6 billion, according to statements on CalPERS website. Of the $38.5 billion in distributions, CalPERS states that $20.2 billion represented realized gains and thus profit to CalPERS attributable to their Private Equity Program. As of March 31st 2013 CalPERS PE Investments had a market value of $32 billion, representing an allocation of approximately 12% of their total asset base, slightly under their interim strategic target allocation for the asset class of 14%17. CalPERS claims that their PE goal is “to perform as the investor of choice and leverage marketplace opportunities in order to achieve superior risk‐adjusted returns” which they accomplish through three investment components – Partnership (LPs), Direct investments (including co‐investments), and Fund‐of‐Fund. It is interesting to note their the following statement on their website under Private Equity Performance Comparison “historically, private equity has generated superior performance when compared to public equity” followed by a table quoting the annualized 10 year return of PE (as calculated by Venture Economics – an aggregator of PE industry data) versus the S&P500 as being 6.86% to 2.92% respectively18. 16 As of 31.03.2013 – website http://www.CalPERS.ca.gov/index.jsp?bc=/investments/assets/mvs.xml 17 CalPERS “Facts at a Glance” July 2013 18 http://www.CalPERS.ca.gov/index.jsp?bc=/investments/assets/equities/pe/programoverview.xml
4.1.2 Cash Flow Data The data set supplied by CalPERS contains each cash flow record as either a ‘Capital Call’ (payments by the LP into Private Equity funds to cover Management Fees, Carried Interest Payments or Committed Capital Drawdowns) or a ‘Distribution’ (cash returns from the PE fund from successful exits of portfolio companies usually via an IPO or trade sale). Each cash flow records the date of the cash inflow/outflow and the name of the PE fund it relates to. The Distributions as obtained from the data source are the net‐of‐fees returns to the LP as the GP has deducted all management fees and carried interest payments. The exact amount of fees and carried interest payments made by CalPERS cannot be determined from this data set as they are included within Capital Calls and not separately identified. Another point to note is that the Distributions may potentially include shares of an underlying portfolio company that underwent an IPO or was acquired by an existing stock listed company. The data does not distinguish between share or cash distributions. This can create potential issues as the LP may not be able to directly obtain the market value of the stock depending on the specific circumstances. For example, there may be a lock‐in period for a newly listed portfolio company and the LP may have to hold the shares until the expiry of this period. However, for simplicity, we will treat any potential share distributions as if they were cash distributions received on the date recorded by the LP and any difference should be relatively small. 4.1.3 Net‐Asset Values (NAVs) It will be necessary to supplement the data set to include the Net‐Asset Value for all the funds that are not yet fully liquidated as a cash Distribution as of September 2012 (being the end of our sample period). The NAVs should represent the value of the investments still at work in the fund but they are determined and reported by the GPs themselves and are therefore open to variation in valuation methods across GPs as well as the potential for biased reporting, for example, reporting a higher NAV to support the performance track record of the GP whilst raising a new fund. Whilst it may be appropriate to run some robustness checks on the NAV at 75%, 100% and 125%, due to the limited scope of this paper, in terms of length, we will assume that the reported market values of the holdings are the actual value and are treated as single cash distributions as of the last date of the data set. For the CalPERS data set we have to make a choice over which NAVs to use. Our cash flows run until the end of Q3 2012 whilst there are two possible sources for the NAVs but both with different dates.
The CalPERS website reports the NAVs of open funds on their website and in their annual investment report but as of 31st December 2012 – meaning we would be missing one quarter’s cash flows and there may be a mismatch with our data set. Alternatively, the 2011‐2012 Investment Report also posted on the CalPERS website contains the NAVS for their current active funds as of June 30th 2012. If we use these NAVs then we will discard the Q3 cash flows but then our data set would be complete and more accurate. Ultimately, we chose for the latter option, using the Investment Report NAVs. 4.1.4 Time Periods Whilst the first PE investment of CalPERS was in mid‐1991, in the early years CalPERS was very cautious in their investments in this new asset class. Thus, between 1991 and 1994, they only invested in five funds – which five funds would represent 0,5% of their total invested capital up to 201219. However, in terms of carrying out the Benchmark analysis there are a number of Benchmarks for which I could only obtain data (via Datastream) going back to the beginning of 1995. In particular, the FTSE All World excluding USA index which is a component of the benchmark that CalPERS itself uses to measure its own PE performance. Therefore, the large majority of the performance analysis will only use vintage years from 1995 onwards, representing 285 funds and 99,5% of the total invested capital. For robustness of the sample, an analysis on the full data period for several key benchmarks was also undertaken and showed was virtually no change in the results versus the 1995‐2012 sample (Table 11). It should also be noted that the selection made earlier regarding NAVs means that we won’t use the Q3 cash flows and that we also exclude any funds with vintages in Q3 2012. This brings the number of funds down from 293 to 290. This will have no effect on our results as funds with vintages in 2012 are not relevant from a performance perspective as they will have only just started investing and are very unlikely to have received any distributions at all given the long investment cycle of PE. Indeed, for this reason we run a number of robustness checks over different time periods to only include funds that are running for some period of time and including ones that are at least ten years old and so should be either fully liquidated or close to realizing all their investments. This is explained in more detail later on. 4.1.5 Fund Types & Geographies The data supplied does not indicate the type of funds included, specifically whether they are Buy Outs, Mezzanine, Expansion or Venture Capital Funds. However, from the CalPERS website there is a description for current active PE funds which can be used to cross reference against the data set to 19 See Table 10
get an indication of the proportion of PE fund types. In addition, the annual Investment Reports also list the PE funds according to their category (Table 1, Panel B). As can be seen, 59% of their PE investments relate to Buy Out funds, 31% to other categories such as Credit Related and Special Situations, whilst 10% relates to VC. We note that much of this thesis and recent academic research has been focused on Buy Out related PE. Therefore, we will run some robustness checks on these subsets to determine the impact on our overall conclusions. Regarding the geographical location of the PE funds, we are not provided with this information in the cash flow data and nor is it available in the recent financial and/or investment reports. From the PE Policy Benchmark that CalPERS uses, we can assume that roughly 67% of their PE investments are in US based funds whilst 33% are in other countries. 4.1.6 CalPERS Website Performance Data and Annual Financial and Investment Reports There are two sources of PE Performance data provided by CalPERS on their website. The first is a webpage under the PE Performance Review20 and the second is from their Comprehensive Annual Financial Reports and annual Investment Reports (which will be discussed in the following sub‐ section). CalPERS website provides a lot of information about their investments in PE with information explaining the features of this asset class, such as the long investment cycle of 10 years or more and the ‘J‐Curve’ effect21. CalPERS also posts a detailed list of all their current LP stakes in PE funds, detailing the capital committed, cash‐in, cash‐out, total NAV plus cash out, the net IRR and Multiple for each fund. This information is quite useful to cross‐reference against our cash flow data set and get a feel for how consistent the two data sets are. With respect to relative performance of their PE investments, CalPERS prepares annual data showing PE returns on a one year, three year, five year and ten year horizon compared to their PE Policy Benchmark in their Comprehensive Annual Financial Report. This data is used in our performance analysis section. 20 http://www.CalPERS.ca.gov/index.jsp?bc=/investments/assets/equities/pe/private‐equity‐review/pe‐ perform‐review/home.xml 21 The J‐Curve effect is where in the early years of the PE investment life cycle the actual NAV is often less than 100% as funds have been invested, management fees deducted whilst no distributions will have been received. Then in later years as the investments mature, the market value of the investments should increase, distributions are made on exits and the overall value increases creating a ‘hockey stick’ shape or ‘J‐curve’
4.2 SBA 4.2.1 Background: SBA PE Investments SBA is also one of the largest pension funds in the United States and is based in Florida with over $122 billion assets under management22. SBA has been investing in PE even longer than CalPERS, starting in 1989 and has invested a total of $14,7 billion up to the end of 2012. As of June 30th 2012, SBA had $6,7 billion invested in the asset class representing just over 5% of the asset allocation in the SBA fund, significantly less than CalPERS both in terms of asset allocation and nominal amounts invested. 4.2.2 Cash Flow Data ‐ SBA The data set supplied by SBA is in a different format from CalPERS and contains a column with a record for each Capital Call and a separate column identifying fee payments by date, which is presumed to include the management fees but not performance fees (i.e. carried interest). Distributions, however, are recorded as a single payment on a full‐year basis at the end of the year (SBA year ends on 30th June). Therefore, the distribution payments were adjusted to be paid in the middle of the year (so 31st December) except for the first distribution payment recorded per fund, which was kept at the last date in the full year ensuring that there may not be cash inflows prior to cash investments having been made (for some funds with very early distribution payments). As per CalPERS, the Distributions are net‐of‐performance fees As per CalPERS the distributions may include shares of an underlying portfolio company that underwent an IPO or was acquired by an existing stock listed company but this cannot be seen from the data set and is treated in the same way as CalPERS (i.e. treated as a cash distribution). 4.2.3 Net‐Asset Values (NAVs) Unlike CalPERS, SBA provided an up‐to‐date Net‐Asset‐Value report with the cash flow data set as of 31st December 2012. Matching the funds to the cash flow data we could record the NAV for each fund as a cash distribution as of 31st December 2012. 4.2.4 Time Periods 22 SBA 2012 Investment Report
Whilst the first PE investment of SBA was in 1988, they only invested in a total of 3 funds prior to 1995 although these funds did represent 4,6% of the total capital invested over the entire period (1988‐2012). As already explained in the CalPERS description, due to some benchmarks not having data prior to 1995, the large majority of the performance analysis will only use vintage years from 1995 onwards. For robustness of the sample, an analysis on the full data period f is also undertaken and shows no effect on the results (Table 11). As with the CalPERS data set we will also we run a number of robustness checks over different time periods to only include funds that are running for some period of time for example, that are at least ten years old and so should be either fully liquidated or close to realizing all their investments. This is explained in more detail later on. 4.2.5 Fund Types SBA splits its PE portfolio into two separate asset classes. The first asset class is called “Private Equity” and presumably contains VC and Buy Out investments – but this is not specified. The second asset class is called “Strategic Investments” which included ‘special situations’, ‘debt related’ and ‘hedge funds’. Although it appears hedge funds and other non‐PE type of investments were excluded from the Strategic Investments file, nonetheless it contained investments that were largely made from 2008 onwards (and none before 2005). Therefore, given the early nature of this data set it was decided to only analyse the Private Equity file which started in 1988. The file contained 57 funds with Capital Calls of $5,8 billion approximately. 4.3 Benchmarks In the performance analysis section we will use a variety of equity indices to benchmark the relative performance of PE. These benchmarks have been listed in Appendix 4 along with a brief description of the benchmark/index and the source. The returns data for each benchmark index comes from one of three sources. Primarily we have used Thompson Datastream to obtain the majority of indices but the other two sources include Ken French’s website and the Wharton Research Data Service (WRDS). All benchmark indices are based on Total Returns 23 and are on a monthly basis. 23 Total Returns means that the index return assumes that all dividends are reinvested in the index as opposed to the Price Return of an index which excludes dividends.
5. DESCRIPTIVE STATISTICS AND DATA ISSUES 5.1 CalPERS Data Set Table 1, Panel A shows the total amount of capital invested (Capital Calls) by CalPERS, comparing the full data set with vintages from 1991 until June 30th 2012 versus the slightly reduced sample from 1995 until 2012. As can be seen, the 1995 data set only contains 0,5% less capital and as will be shown later, these excluded funds do not affect the results. CalPERS invested just over $50 billion of capital into PE over this time period, comprising 285 funds with an average of $175 million per fund, a median of $105 million and a standard deviation of just under $196 million. Capital invested (calls) included payments for capital committed to the fund for use by the fund in its investing activities, as well as the management fees which are paid to the General Partners (GPs) who manage the investing activities of the funds. Panel B of Table 1 shows how CalPERS invested by PE fund type with almost 60% of the total capital invested going to ‘Corporate Restructuring’ (Buy Out) funds. The next largest categories are Credit Related (13%), VC (10%) and Expansion Capital (9%). Then there are a smaller amount of Opportunistic funds (4%) and a number of funds (11 funds representing 5% of capital invested) where it was not possible match the fund with the CalPERS website to determine the classification. These proportions do not change with the 1995 data set. One key point to note is that whilst we look at the overall performance of the whole PE portfolio, we will carry out a robustness check to test performance without Venture Capital (VC) funds included as most of the academic research to which this thesis is comparing benchmark performance is based around ‘Buy Out’ investments in particular. It has been documented that Venture Capital performed well in the early to mid‐1990s but that it has not outperformed most public equity benchmarks in the 2000s. The focus of the current debate is really around later stage PE, generally termed “Buy Out” to refer to the fact that the PE fund will buy a controlling stake in the company, which will allow it to install its own management team and restructure the company in line with a strategic plan to create more value from the business. The restructured and more valuable company is then usually sold in a trade sale or sometimes through a public offering for a profit if successful. Buy Outs typically involve much larger transactions, as they involve buying established companies, and therefore account for a much larger share of the total value of PE as an asset class. A second key reason to also exclude VCs is that benchmarking is really dependent on matching the PE portfolio companies with a public market index that best represents the invested companies. In this way relative risk, return and thus performance can be compared. However, VC has a very different profile from later stage PE, such as Buy Outs. VC typically involves smaller, not‐yet‐profitable
companies, traditionally in the information technology or bio‐tech sectors with a much higher risk profile. Buy Outs involve larger, often already profitable companies across a variety of sectors. Therefore, by controlling for the effect of VCs in our sample we will get a fairer view of CalPERS PE performance relative to recent academic papers. This will be discussed in more detail in the Data Analysis section. Finally, Panel C shows how much capital was invested per vintage year over the whole time period. As can be seen, approximately 9% of the total capital invested occurred between 1991 and 1999, covering 34 funds, whilst 89% of the capital was invested between 2000 and 2009 across 240 funds, with the large majority of that occurring between 2006 and 2008 (over 60% of the total capital invested). It is quite remarkable that a disproportionately large amount of funds were committed to these vintage years which represented the peak of the last PE cycle. The fact that such a large amount of capital was invested in these relatively recent times means that the performance analysis will be heavily weighted by the NAV values provided. Indeed, of the total distributions amount shown in Panel A of Table 1, approximately 50% is made up of NAV data which means that NAV reporting bias is an important issue. Similarly, any effort to look at performance of largely liquidated funds (i.e. from vintages of 2003 and less), will result in a greatly reduced sample. All of these issues will be discussed further in the data analysis section. 5.2 SBA Data Set Table 2 gives an overview of the PE investments made by SBA. Panel A compares the full data sample running from 1988 to 31st December 2012 versus the adjusted data set running from 1995 to 31st December 2012. The full data set shows that SBA invested in 145 PE funds during this period, totaling to an amount of $14,28 billion in capital calls (being the capital transferred to the PE funds for use in investing activities) and an additional $446 million in management fees, which represents 3,1% of capital calls. This 3% seems slightly on the higher side of industry norms of 2% of committed capital perhaps due to the fact that capital called is usually slightly less than the capital committed as most PE funds don’t call up exactly the full amount of capital committed. Comparing with the adjusted data set, there are now 142 funds which cover $13.6 billion of investments ‐ a reduction of almost 5% of the capital called in the full data set. Furthermore, using the 1995 data set reduced Distributions including NAVs from $19,07 billion to $17,76 billion – a reduction of almost 7%. Therefore, despite there being only three funds less between the data sets, they represent a relatively high amount of the total capital invested and so we shall perform some robustness checks to see the effect on the performance benchmarking analysis.
We can also see that the average and median levels of capital calls in the 1988 data set, $100,2 million and $59,5 million respectively , are considerably lower than CalPERS. The standard deviation is $123 million and these figures don’t change significantly in the 1995 data set. Finally, we can also see that NAVs account for less of the total Distributions plus NAVs amount than CalPERS, as NAVs represent 35% (37,9% ‐ 1995) of the total, suggesting that more of the value comes from realized gains in the SBA sample. Finally, Panel B supports the above conclusion as it can be seen that the distribution of investments by funds and capital invested is more spread‐out across the time period and is also invested over an earlier time period. SBA has invested 50% of its total capital by 2004 whereas CalPERS takes until 2006 to do so. Unlike CalPERS, where the majority of investments take place between 2006‐2008, SBA has invested more evenly across the vintage periods (the standard deviation of capital invested by vintage years is 3,7% with SBA versus 6,9% with CalPERS). Nonetheless, SBA has three investment peaks, 1998 (12%), 2002 (10,6%) and 2005‐2008 (35,4%). Indeed, the decade of 2000 to 2009 accounted for 64% of the total capital invested, so like CalPERS the majority of PE capital invested is relatively new meaning that performance analysis may be largely affected by the NAVs and an analysis of only liquidated funds would lead to a greatly reduced sample. 5.3 Data Discrepancies It is important to note that there are some discrepancies between our data sets and the published data from both sources which are discussed in more detail further on. For both data sets we find a significant increase of both total capital calls and distributions in our data sets versus what is published on the LP websites or their annual financial and investment reports. For example, the CalPERS data set has $49,3 billion of capital calls whilst the website reports cash‐in of only $43,8 billion – a difference of $5,5 billion. Distributions are $33,2 versus $30,6 billion (cash‐out) – a difference of $2,6 billion. At the time of writing it is not clear what the reason for this discrepancy is. It could be that the data set provided to us includes co‐investments or includes performance fees such as carried interest that are not included on the website data although neither of these seem very likely. Certainly, some of the discrepancy is due to offsetting double entries and where possible we tried to eliminate these from the data set. However, there is still a surplus of $2,9 billion of capital calls for CalPERS (deducting the difference in capital calls from the difference in distributions) and so one may expect that the results of CalPERS are biased slightly downwards. Indeed, this may explain the slightly lower value‐weighted net‐IRRs of our data set versus the website. However, as the difference is relatively small we assume that the discrepancy does not significantly alter the key findings of this thesis.
Indeed, when running the PME analysis against the policy benchmarks of CalPERS, we find very similar performance results as reported in CalPERS annual reports. CalPERS reports an average of 2,4% annual over‐performance of the PE investments on a ten year time frame when averaging the last eleven annual reports which compares to a 2,1% annual over‐performance as found through our PME analysis of the cash flow data set. Although these calculation methodologies are not directly comparable, this supports the possibility of a slight under‐performance bias in our analysis but that results are still very close (the average of 2,4% is not significantly different from 2,1% at the 5% level). There are two other issues related to the cash flow data set regarding funds not included in our cash flow data set. Firstly, seven funds appear on the website but are not in the cash flow data set. According to the CalPERS website $148m has been invested in them so far of a total capital commitment of $180m. Secondly, there are 72 funds which are shown in the CalPERS 2011‐2012 Investment Report which are not in the cash flow data set which had a market value of $4,3 billion as of June 30th 2012. Many of these funds are also not included on the CalPERS website. The missing funds could lead to a bias if they was some systematic reason why they were omitted (if they were randomly omitted then there shouldn’t really be a bias – it just reduces our sample size). For SBA the issue is smaller and goes in the other direction. Our data set has $14,7 billion capital calls versus $13,4 in the investment report and $19,1 billion versus $17,4 billion distributions respectively. This leaves a surplus of $0,3 billion in distributions in our data set assuming that the most of the additional capital calls and distributions directly offset each other. Therefore, it may be that our results are biased slightly upwards for SBA which may explain the slightly higher net‐IRR of our data set. However, as this thesis will show, SBAs’ PE data set largely underperforms and so a correction of this upward bias would further lower their performance. Ideally it would be best to have a clear explanation for these discrepancies but we believe any potential bias is relatively small and somewhat random, and that the general results of our analysis will still hold.
6. DATA ANALYSIS – PE PERFORMANCE AND BENCHMARKING In this section, we will begin by examining the performance data provided by CalPERS and SBA on their websites and/or through their annual financial and investment reports. In particular, we will compare the publicly reported net IRRs and multiples with the cash flow data sets provided and we will discuss the reported PE performance history as found in the respective financial investment reports. It is interesting to note the different calculation methodologies used by each LP as well as their choices of policy benchmarks, which serves to highlight some of the key issues of PE performance measures and which benchmark to select. Of particular interest to us in this analysis is the relative performance of PE as an asset class; thus whether PE investments have outperformed against an appropriate risk adjusted benchmark. To begin this exercise we will turn to examining the relative performance of the two PE investors by using the Public Market Equivalent (PME) method. The PME method discounts capital calls and distributions to determine the net present value of the PE investments and the resulting present value of distributions is divided by the present value of the capital calls to give a ratio. The discount rate used is the return of a public equities market index or benchmark such as the S&P 500 where a number above 1 suggests that the investment return has outperformed the benchmark (index) and less than one indicates underperformance. Under the PME measure, the risk adjustment is taken into account largely through selecting an appropriate benchmark, ensuring the investment and risk profile of the benchmark matches that of the average PE investment. Indeed, this issue is at the heart of the current debate around PE performance where recent academic papers have shown strong outperformance against the S&P 500 benchmark whereas Phalippou (2012) has argued that PE has underperformed when using more appropriate benchmarks such as the Fama‐French small‐value index. Thus, we will largely follow the path of Phalippou’s paper in examining performance against different benchmarks that account for size, style and leverage premiums as well as comparing and analyzing the policy benchmarks that CalPERS and SBA themselves use to measure their own PE performance. These results will then be discussed and compared in view of the larger discussion around PE performance. 6.1 CalPERS Absolute Performance Analysis – Net IRRs and Multiples We begin our performance analysis by comparing the net IRR and multiple data provided by both CalPERS and SBA with the cash flow data file obtained from them under the Freedom of Information Act. We can also use these results to compare with the stated performance data found in the financial reports and investment reports of the two PE investors as well as how their PE portfolio has
performed against their own policy benchmarks for the asset class. Indeed, comparing the benchmarks and reporting methodology of the two investors also gives a clear illustration of the different approaches and challenges in PE performance reporting. 6.1.1 CalPERS Website IRR and Multiples vs Data Set Table 3 Panel A compares the value weighted IRR as calculated from the individual fund net‐IRRs on the website versus the value weighted IRR which was found by determining the net‐IRR of each fund from the data set and then value‐weighting these to give an overall IRR for the data set. The weighting applied is the percentage of the value of capital calls of an individual fund against the total value of capital calls of all the funds. As can be seen the website has a slightly higher IRR of 9,93% versus 9,65% for the full data sample. This reduces slightly when using the 1995‐2012 data set. The TVPI figure is also slightly lower with our data set compared to the website, 1,32 versus 1,41. <Table 3 ‐ Panel A> Panel C of Table 3 also shows the IRR when we pool the cash flows from our data sample and treat the portfolio as a single fund. Calculated in this way the IRR is 10,22%, somewhat higher than the value‐weighted IRR of 9,65% for the full data set whilst our reduced 1995 data set shows an IRR of 9,95%. 6.1.2 CalPERS Data Set IRRs and Multiples by Vintage Years and Periods Table 3 Panel C shows the net IRRs based on the vintage years and periods for the cash flow data set as provided by CalPERS. As can be seen, the returns appear to follow cycles with periods of stronger performance followed weaker performance and so on. <Table 3 ‐ Panel C> The early 1990’s until mid‐1990s showed stronger performance (1991‐1994 IRR is 20,4%) although with very limited capital invested. Funds that started in 1998 and 1999 show weaker performance (9,60%), presumably as they started to invest at the peak of the dot‐com bubble and then saw their investments perform poorly in the subsequent crash. However, performance starts to pick up again in 2000 as funds raised in the early 2000’s will benefit from investing in cheap assets and the subsequent uptick in the economy (14,13% from 2000‐2004 vintages). However, by 2005 one can start to see the lower performance effect of vintages that we just starting to invest at the peak of the overheated economy which led to the major financial credit crisis of 2007‐2008. The IRR for vintages between 2005 and 2009 was 7,89%. Finally, we can see that funds raised during and shortly after the
crisis could benefit from cheap assets and the economic recovery that was to follow. The most recent funds from 2010 onwards are too young to have meaningful returns. 6.1.3 CalPERS Annual Financial Reports – Historical PE Performance as Reported CalPERS comprehensive annual financial reports provide data on their PE performance over 1, 3, 5 and 10 year periods alongside the performance of their PE policy benchmark and whether the asset class has over‐ or under‐performed. CalPERS uses a time‐weighted methodology calculating the monthly returns in the PE portfolio NAV and determining the annualized geometric average of those monthly returns for the time period in question (i.e. 1 year, 3 year etc). Unfortunately, this means our return calculations are not directly comparable as we only have the data that allows us to calculate the money (dollar) weighted returns (effectively the IRR method)24. Nonetheless, it gives us an indication of how the LP (as asset manager) has performed over the time period (as it disregards cash inflows and outflows which are not in the control of the LP) and is a useful comparison with the money weighted or IRR that we can calculate. Table 4a lists the reported performance data for CalPERS last 11 annual reports to give a view on how their PE portfolio has been performing over time. If we begin with 2012 we can see that CalPERS outperformed their policy benchmark over a three year period, returning 17,1% versus 16,3% (annualized) and also outperformed their benchmark over a ten year period, returning 9,8% versus 8,3%. However, over the five year the benchmark outperformed CalPERS, with a 7,3% versus 4,7% annualized return. <Table 4a> In general, we can see that the returns are more extreme and volatile over shorter time periods, as illustrated in Appendix 1a which graphs CalPERS PE returns data between 2002 and 2012. As can be seen from these graphs, CalPERS largely outperforms their policy benchmark, particularly over the ten year time period, achieving an arithmetic average of the nine reports available for this time period of 9,5% which is not too far away from the since inception IRR calculated from our cash flow data set of 10,22%. The Benchmark returned on average 7% per year for each of the ten year periods preceding the year of the report and as of 2012, CalPERS PE portfolio had outperformed their policy benchmark by 1,5% annually over a ten year period. <Appendix 1a> 24 To calculate the time weighted returns we would need to know the monthly NAVs, changes in portfolio values and net cash flows.
This brings us to the question of which benchmark CalPERS uses to determine the relative performance of its portfolio. In 2012 CalPERS Investment Policy benchmark for PE was: 67% FTSE US Total Market Index (TMI) + 33% FTSE All‐World excluding US Index + 3%, with the indices lagged one quarter. However, as can be seen in Appendix 1a, CalPERS has used a number of different benchmarks over the last 11 years. However, because these benchmark indices are value‐weighted they are largely driven by the performance of large stocks and so we should expect performance to be fairly similar to the S&P 500 (which shall be seen in the analysis results further below). The 3% additional basis points is added to cover additional risks associated with investing in PE as opposed to the respective benchmarks such as illiquidity, long investment time horizon etc. 6.2 SBA Absolute Performance Analysis – Net IRRs and Multiples 6.2.1 SBA Reported IRRs and Multiples vs.Data Set Whilst SBA does not openly provide data on its PE funds such as the net‐IRRs on a fund‐by‐fund basis on its website, it did provide this information in our data set in a file containing the NAVs as of 31.12.2012. As with CalPERS this allows us to compare the value weighted average IRR of this file with the value weighted average IRR of the funds in our cash flow data set. Table 3 ‐ Panel B, the data set shows a higher return of 7,82% versus the investment report, 6,64%. The multiples are very close being 1,295 versus 1,305 respectively since inception fund IRR as calculated from the cash flow data set provided by SBA. <Table 3 ‐ Panel B> Table 3 Panel B, also shows the since inception IRR if we pool all cash flows from the underlying PE funds as if it were a single fund. Meanwhile, Appendix 2 is an extract from the SBA website which shows the since inception returns for SBA’s PE assets as of 31.12.2012. SBA shows a return of 7,42% which is lower than the since inception return we calculate from our data set of 8,10%. .Part of this difference may be explained by the fact that Distributions in the cash flow data set have been included as a single aggregate yearly distribution made in the middle of a given year. If in reality, more distributions were made at the end of the year this could slightly upward bias our data set IRR although one wouldn’t expect the difference to be as large as it is. With respect to the cash flow data set 1995‐2012 we can see that the PE IRR is 1% lower at 7,08% given that some of the early PE funds had very strong IRRs.
