Performance of Private Equity Funds: Another Puzzle? - CiteSeerX

43 downloads 42 Views 203KB Size Report
private equity (PE) funds and their underlying investments, we find that the ..... compute a value-weighted average of factor X over the life of the fund. ...... Fernandez, Pablo, 2004, The value of tax shields is not equal to the present value of tax.
Performance of Private Equity Funds: Another Puzzle? Oliver Gottschalg, Ludovic Phalippou, and Maurizio Zollo INSEAD

October 2003 Using a novel and comprehensive database on the performance of US and EU private equity (PE) funds and their underlying investments, we find that the performance of PE funds is comparable to public market performance. We show how sensitive this result is to various assumptions and thereby reconcile existing divergent estimates. We find evidence consistent with the hypothesis derived by Jones and Rhodes-Kropf (2003), who show that idiosyncratic risk should be priced for PE funds. We also document that the performance of venture funds is more sensitive to business cycles while the performance of buyout funds is more sensitive to the state of both the bond and the stock market. Nonetheless, PE funds appear to offer some attractive hedging properties.

Please address all correspondence to: Ludovic Phalippou, INSEAD, Boulevard de Constance, 77305 Fontainebleau, France. Tel : 33 (0) 1 60 72 42 94 or (33) (0) 6 22 12 41 29. Financial support from the R&D Department at INSEAD, the Wharton-INSEAD Alliance and the INSEAD Gesellschaft Scholarship (Oliver) is gratefully acknowledged. The authors would like to thank Jesse Reyes and Thomson Venture Economics for making this project possible through generous access to their databases. We also thank seminar participants at INSEAD and Robert Kosowski for their helpful comments.

The activities of private equity (PE) funds have received inc reasing attention from both the investor and the academic community, primarily because of two factors. First, the amount of capital committed to this asset class grew from US$2 billion in 1980 to US$134 billion in 2000, totaling over US$780 billion over the last 25 years (see Graph 1, as well as Kaplan and Schoar (2003)). 1 Second, it has been argued that fund managers (general partners, GPs) play an active and important strategic role in the companies financed. This role has been extensively studied both theoretically (e.g., Admati and Pfleiderer, 1994; Hellmann, 1998) and empirically (e.g., Gompers, 1995; Lerner, 1995; Hellmann and Puri, 2000 and 2002). Given the importance of private equity both as an investment vehicle and a catalyst for economic growth, the need for a comprehensive assessment of the performance and risk profile of this industry is apparent. In a recent review paper, Gompers and Lerner (2001) classify the understanding of risk and return as “what we don’t know about venture capital,” a statement which is equally true for buyout investments, the other main sub-category of private equity. 2 Graph 1 Using a unique and comprehensive dataset containing information on the cash flows to investors (limited partners, LPs) and investments of about 500 mature private equity funds, this paper provides three key contributions. First, we test whether idiosyncratic risk is priced as argued by Jones and Rhodes-Kropf (2003), hereafter referred to as JRK. JRK model the GP-LP relationship and derive several hypotheses. In particular, their theorem 4 states: “All else equal, the return received by the investors is increasing in the amount of realized idiosyncratic risk, even though they face competitive market conditions.” To test this implication they argue tha t, for buyout (BO) funds, a reasonable proxy for idiosyncratic risk is the number of investments that a fund makes. In the absence of this information they use the number of take-downs and find a negative relationship with returns. Our dataset, in contrast, enables us to construct the number of investments as

1

The figures reported in Graph 1 are a lower estimate as it is based on our dataset, which includes only funds for which we have cash flow data available. 2 Sometimes, other non-public investment vehicles such as real estate and entrepreneurial investments in nonpublic companies are called private equity. In this paper, only buyout and venture investments are labeled private equity.

1

well as several other proxies that capture more closely the spirit of the model. We find that the number of investments for both venture and buyout funds is indeed negatively related to returns. In addition, concentration indices for both industry and investment type are strongly positively related to returns. Such evidence is consistent with the prediction of JRK’s model. Second, we use two methodologies to assess the extent to which systematic risk affects fund performance. For the first, we assume that the CAPM holds and that betas on assets are the same within each industry. 3 Using information on the industry and the type of each investment, we estimate the beta of buyout funds to be on average 1.7 and of venture funds to be 1.2. However, we do not find that performance is linearly related to betas. For the second methodology, we attempt to “mark-to- market” investments, that is, each investment is assumed to return what its corresponding industry or market returns over the same time period. In this instance, we find a strong and positive relationship between this measure of public market performance and fund performance suggesting that systematic risk influences returns, especially for buyout funds. We also evaluate the influence of business cycles and corporate bond yields on PE returns. We find that bond yields strongly influence buyout performance and that unexpected GDP growth strongly influences venture returns. Third, we estimate the overall retur n from investing in Private Equity (PE) funds in the US and Europe over the last 25 years using the most comprehensive dataset to date. As data has become available, several papers have recently documented both the gross performance and systematic risk at the deal level for venture investments (Cochrane, 2003; Quigley and Woodward, 2003; Peng, 2003; and Hege, Palomino, and Schwienbacher, 2003), and the net performance and risk at the fund level. For both venture and buyout funds, estimates of the overall (net of fees) performance of PE funds vary greatly from poor (Jones and Rhodes-Kropf, 2003) to high (Ljungqvist and Richardson, 2003), via intermediate (Kaplan and Schoar, 2003). 4 The evaluation of risk is not more consensual:

3

As assets are not traded continuously, the CAPM does not hold, in theory, in the context of PE investments. Throughout the paper, we often refer to Jones and Rhodes -Kropf (2003) as JRK, Ljungqvist and Richardson (2003) as LR, and Kaplan and Schoar (2003) as KS. 4

2

CAPM-betas for venture (buyout) funds are estimated to be 1.8 (0.7) by JRK, 1.1 (1.1) by LR, and 1.7 (0.5) by KS. We show that the performance of PE funds is similar to the public stock market performance and that the divergent estimates proposed in the literature can be reconciled as being due to differences in sample selection and the treatment of unrealized investments. Using a sample of largely liquidated funds only introduce a selection bias, as the decision to liquidate is endogenous and is probably influenced by the success of the investments. When we include funds with investments that are not yet liquidated in the sample, the question becomes how to value these on-going investments (residual values, RV). We estimate the present value of RVs based on an analysis of how RVs historically converted into cash flows for now liquidated funds. The resulting conversion matrix shows that writing-off accounting values leads to a systematic underestimation of fund returns. Furthermore, we find several characteristics to be highly related to performance. GPs that are often the main investor, are experienced, manage larger funds, and sell investments rapidly, significantly outperform This paper complements the pioneering work of Moskowitz and Vissing-Jorgensen (2002) on the return of entrepreneurial investment in non-public companies - another private equity class. They also find that private equity funds’ performance is very close to public market returns, despite their lack of diversification. This paper is also related to the work of KS, who focus on the persistence of fund performance and the flow-return relationship; and to that of LR, who focus on the draw-down and capital return schedule, and JRK, who focus on the pricing of idiosyncratic risk. 5 The paper continues as follows: section 1 describes our data and methodology, section 2 discusses the evaluation and pricing of risk, section 3 evaluates the overall performance of funds, section 4 presents various sensitivity tests, section 5 presents additional considerations regarding the liquidity of PE fund investments and section 6 concludes our discussion.

5

As mentioned, these three papers also assess the risk and return of PE funds with different sub-sets of our dataset and make different assumptions about accounting values and performance measurements.

3

I. Performance Measurement

In this section, after detailing our data sources and our sample selection, we offer descriptive statistics of the characteristics of both the funds and their investments. We then describe our approach to account for the residual value of ongoing investments. Finally, we discuss how to measure performance.

A. Data sources In this study we use several sources of data. Data on both Treasury bill rates and stock performance are from CRSP (via WRDS). Data on accounting are from Compustat (via WRDS). Data on real GDP and corporate bond yields are from the Federal Reserve Bank of Saint Louis. Data on private equity funds have been obtained from two datasets maintained by Thomson Venture Economics. These datasets cover funds raised from 1980 to 2002. Venture Economics (VE) records the amount and date of all cash flows as well as the aggregate quarterly book value of all unrealized investments for each fund until June 2003. Cash flows recorded in this dataset reflect net returns to LPs, as cash flows from LPs to GPs (“take-downs”) include all fee payments, and cash flows from GPs to LPs (“distributions”) are already reduced by the carried interest or other charges. VE also collects information on the private equity investments that each fund undertakes through its VentureXpert database. The basic information we use from this dataset is the company’s industry, amount invested, and entry/exit dates of each investment. Details of these databases as well as certain corrections that we undertake are provided in Appendix A.II.

B. Sample selection For each fund we observe a series of cash flows to investors and the “residual value” (RV) of ongoing investments. Until a fund is entirely liquidated, the existence of a positive RV prevents a proper estimation of performance as neither the fund nor its underlying investments are publicly traded. The unique assessment of RV that we observe is the accounting value that is reported quarterly by funds. However, these accounting

4

valuations are relatively uncertain and unreliable. 6 This observation led KS, for example, to select funds with no cash flows over the last six quarters and for which the ratio of residual value to capital invested (RV/CI) was below 10%, in addition to funds officially liquidated. 7 In such a sample, the treatment of accounting values has a minimal impact but it will be biased in that it is likely to be a sample of “winners”. Indeed, funds that are not fully liquidated (and hence excluded from the sample) may be finding it difficult to sell their current investments or may simply be waiting before realizing, and officially acknowledging, a poor performance. 8 The liquidation decision, then, may be endogenous and partly influenced by the success of investments. Our concerns in this case are twofold. First, the average performance of funds may be upwardly biased. Second, funds in this sample may have different characteristics from the universe which would influence our findings about the pricing of risk. As a consequence, we primarily consider a sample of all funds raised before 1993. These funds are said to be “mature” as they have reached typical liquidation age (see Appendix A.I.). We further restrict our sample to funds on whose investments we have enough data. We require that the total amount of investments made (on which we have data) be above 50% of the capital committed. This sample, denoted as Sample 1, contains 393 venture funds and 98 buyout funds. We report the characteristics of this sample in Tables I and II. We also offer descriptive statistics of other samples to assess the various differences. Table I Our sample of interest (Sample 3) over-samples VC funds: it contains 80% of VC funds compared with 66% in the universe of Venture Economics (Sample 0). In terms of 6

The US National Venture Capital Association proposed certain mark-to-market guidelines for the valuation of PE fund investment in 1989 which have become a quasi-standard for the industry since then. Nevertheless, the discussion in the PE industry about appropriate rules for the valuation of unrealized investments is ongoing and accounting practices vary to the point that LPs sometimes receive significantly different valuations from different GPs who jointly invest in the same company. Interested readers may refer to Blaydon and Horvath (2002, 2003) for a detailed discussion of accounting practices. 7 Note that liquidation is no guarantee that all investments are realized. For example, our sample contains 487 funds (22% of the universe) that are officially liquidated of which 152 funds still have some latent investments in the book. Among them, 63 funds have more than 30% of their capital invested in the book. 8 There is also a bias that can arise if GPs start by investing in their best ideas. Early evaluations would then be optimistic.

5

fund size, funds in our sample are smaller because those raised after 1993 are larger. Nonetheless, our investment requirement implies that the selected funds are slightly larger than the non-selected funds raised before 1993. Overall, our sample does not appear to be very different, the main difference being that it has fewer European funds, in particular European VC funds. Generally, the descriptive statistics are similar to those reported in the literature (see, for example, KS). Venture funds are significantly smaller than buyout funds with an average committed capital of $97 million for VC funds compared to $410 million for BO funds. 9 The number of take-downs is similar although slightly lower than that reported by JRK. We have about eight take-downs for VC funds and 15 for BO funds in Sample 3. In addition, it is important to note the significance of accounting values even in a sample of funds raised before 1993. RV/CI is on average 26% for BO funds and 16% for VC funds. Moreover, reflecting the fact that the PE industry is young, over one third of the funds are first-time funds. Nonetheless, certain firms are quite successful and for funds raised before 1997 a quarter of the funds are fourth-timers or more, that is, the parent firm has already raised more than four funds. Table II We now turn to the description of investments as reported in Table III. We find that VC funds have almost twice as many investments as BO funds. VC funds invest in 32 (28) companies on average (median) and BOs in 15 (12) companies. These figures are consistent with the findings of Gompers and Lerner (1999) for a different sample: they note that VC funds typically invest in two dozen firms over about three years. We can also compare these figures with what LR report for their sample: that on average BO funds invest in 16 companies versus 37 for venture funds. Several important elements are worth mentioning. First, we find that venture funds have 11% of their investments in buyouts and that as much as 30% of the investments of buyout funds are in venture deals. Nonetheless, the medians are smaller. This indicates that a few funds in each category (more so in BO funds) are diversified across deal types. Consequently, care sho uld be taken when interpreting differences in the behavior of funds

6

according to their objectives. Second, the average holding period and industry focus is similar for both VC funds and BO funds. Finally, deals last on average about five years and over 50% of the deals undertaken by VC funds are in high-technology industries.

C. Valuing non-liquidated investments To avoid sample selection bias we include all funds raised up to 1993. This implies that certain funds still have some ongoing investments. As already mentioned, neither funds nor their underlying investments are publicly traded. The unique assessment of residual values (RVs) that we observe is the accounting value that is reported quarterly by the funds. The problem is then to decide how to translate these residual values into performance measurement. Given a general distrust of accounting valuations we may want to write off RV. 10 This is the option chosen by LR, for example, but is extremely conservative in that performance is systematically understated and may bias regression results when the dependent variable is performance. KS and JRK, in contrast, use Venture Economics’ calculation which involves applying a certain growth rate to RV. In this paper, we analyze the historic pattern whereby resid ual values translate into cash flows. Using a set of liquidated funds we compute the net present value (NPV) of a dollar of residual value as a function of the age of the fund and the proportion of unrealized investments (see Appendix A.III. for details of these calculations). The resulting matrix is reported in Table III. We implicitly assume that the historic pattern continues to hold in the future. We then convert any accounting value reported in June 2003 in the corresponding immediate cash inflow according to the matrix in Table III. Table III We observe that, historically, RV significantly underestimates future cash flows except for “overdue” funds with a high relative residual value. Such conservatism is actually encouraged by the US National Venture Capital Association which provides 9

Note that all figures are in 2002 US dollars. Certain practitioners also distrust accounting valuations and look at performance analyses based on realized returns only, such as the Investment Benchmarks index of fund returns published by Venture Economics, which includes several performance measures that do not consider residual values. 10

7

guidelines that in most cases result in carrying investments “at cost” in the books, especially in the early years of an investment and in the absence of any alternative unambiguous valuation procedure. This implies that residual values understate expected cash flows if investments earn a positive return. Furthermore, these figures are audited and thus may give GPs an incentive to remain conservative in order to avoid lawsuits. Another interesting fact to emerge from Table III is the relationship between accounting conservatism and fund characteristics such as RV/CI and age. Within each age category conservatism increases as RV/CI decreases. It is likely that funds with many successfully realized investments are even more conservative in reporting the value of “ongoing” investments to signal their success more aggressively. This prompts the observation that accounting practices vary considerably across funds and that using averages may lead to biased results. First-time funds may be more aggressive in order to raise subsequent funds; loser funds may be less conservative than winner funds.11 Nonetheless, we do not make any further attempt to evaluate the link between RV and subsequent cash flows. The above matrix is indicative that RV has a positive value and we use this conversion matrix in our performance estimation.

D. Performance Measures Having selected a sample and found a way to treat accounting values, we need to decide on an appropriate measure of performance. The principal performance measurement in the industry is the internal rate of return (IRR) which has also been used (in addition to alternative performance measures) by KS, LR, and JRK. Here, three important remarks are in order. First, IRR implicitly assumes that capital distribution occurring before liquidation can be reinvested at the fund’s IRR. This is problematic as even if another PE fund raises capital when the distribution occurs, investment is not immediate. Second, the profile of investments in the industry allows GPs to manipulate their IRR by strategically reporting

11

Moreover, the above matrix may not be appropriate for recent years. Indeed, when investments are reported at cost and markets go up, RVs appear conservative. Symmetrically, when markets go down, RVs may appear optimistic.

8

their residual values and timing their cash flows. 12 Third, inflows and outflows are treated as two flows with the same risk, an assumption that seems unwarranted as discussed below. Due to the shortcomings of the IRR, we compute a profitability index (PI). PI equals the present value of cash inflows divided by the present value of cash outflows.13 The main issue is then the choice of appropriate discount rates. 14 LR, for example, propose the use of Treasury-bond yields (of corresponding maturities) to discount outflows (from LP’s perspective) and either the NASDAQ or S&P 500 index returns for inflows. Discounting outflows this way is appropriate if investors know the amount and timing of the take-downs in advance, the argument being that when they commit to the capital they can simultaneously buy bonds with corresponding maturities so as to match each takedown. The problem is that, in practice, the take-down schedule is unknown, rendering this option inadequate. In addition, when a call is made, cash has to be delivered in days (see appendix) and it is therefore likely that investors will have to rely on short-term bonds as an alternative. Consequently, we use the 30-day Treasury-bill yield. Note that a premium over the risk- free rate could be justified, however, to discount outflows. Indeed, investment opportunities and thus outflows may be more likely to appear in good times and hence command a positive beta (although it is arguable that beta should be less than one). Finally, inflows are discounted using the market portfolio as proxied by the CRSP value-weighted index.

12

When an investment is performing poorly (i.e., has a negative IRR), GPs have an incentive to keep a small positive value in the book as then the IRR increases toward zero as time (deal length) increases (toward infinity). 13 For example, if one invests $1 and receives $2 after one period and the opportunity cost of capital is 10% (over this period) then PI is 182%. In this simple case, PI is how much more gross return one obtains compared to the gross opportunity cost of capital. More generally, it is the (gross) return on one’s investment, when both total investments and total payoffs are expressed in present value terms. 14 When discounting cash flows, we consider June of the vintage year as being the present time.

9

II. Is Risk Priced?

Many institutions invest heavily in PE based on the belief that PE returns are largely uncorrelated with market returns. Unfortunately, there is little (if any) documentation about how certain sources of systematic risk influence returns. This issue is discussed in sub-section A. In sub-section B we test whether idiosyncratic risk is priced. Finally, in sub-section C we report additional important determinants of performance.

A. Systematic risk A.1 CAPM Beta The fact that a secondary market does not exist for partnership stakes in PE funds (see Appendix A.I.) implies that the classical assumptions on which the CAPM is based do not hold for this asset class. In the absence of an alternative theoretical model, evaluating the systematic risk of PE investment becomes challenging. Nonetheless, given our dataset we can assess how market-wide risk affects the performance of funds. In this sub-section we assume that the CAPM holds and that the betas on asset are the same inside a given industry. 15 By making further assumptions regarding the leverage of the deals we can then assess the beta of PE funds that would be observed if each company in which GPs invest were publicly traded, and if the secondary market for PE funds were frictionless. We mainly follow the methodology proposed by Kaplan and Ruback (1995) and present the construction details in the appendix. The basic idea is to evaluate the beta on asset that prevails in each industry, assign it to each investment that the fund has made, and lever it up as a function of the deal type (VC or BO). Each beta is then aggregated at the fund level. To compute betas we assume that the leverage of BOs decreases linearly over time from a debt-to-equity ratio of 3 down to the leverage that prevails in the industry. The beta on debt is assumed to be 0.25 and the beta for VC deals is assumed to be the same as the beta of the corresponding industry. The choice of the beta on debt is based on Cornell and Green (1991), who evaluate the beta for high- grade debt (from 1977 to 1989). The choice 15

Note that here, and throughout this paper, we do not separate US and EU funds. We use US only benchmarks: market portfolio and industry.

10

of equity-to-debt ratio is based on Cotter and Peck (2001), who document the debt structure of several LBOs. 16 Table IV Descriptive statistics about estimated betas are reported in Table IV. We report the median and average beta for various sub-samples. We also document how betas differ among first-time funds, venture vs. buyout funds, large vs. small funds, and how leverage assumptions change estimates. 17 The average beta of BO funds is 1.7 compared to 1.2 for VC funds. Second, small VC funds are slightly riskier (in the beta sense) than large VC funds but small BO funds are slightly less risky than large BO funds. To the extent that beta can be interpreted as a measure of aggressiveness, it suggests that first-time funds and small funds may be more aggressive in order to become top performers so as to be able to raise larger subsequent funds. 18 Note also that the median beta is higher than the average beta, showing that most funds have a high beta and relatively few funds have a low beta. Finally and importantly, moving from the industry leverage level to our leverage assumption increases beta estimates significantly, which explains why our beta estimates are higher than LR (2003) reports, especially for BOs. 19 We then test whether fund betas are related to performance. We do not find any such evidence. 20 For example, in Panel B of Table V, we report that betas are negatively (positively) related to performance for VC (BO) funds, but coefficients are not statistically significant. This result is, moreover, robust (unreported tests) to changes in performance 16

An alternative solution, that gives excessive credit to accounting valuations, is to assume that the beta of all BO funds (VC funds) is identical and then estimate the beta of the industry based on co-movements of the aggregate value of funds with the market portfolio. This solution is explored by JRK (2003) who further assume that the value of a fund is given quarterly by the accounting residual value. They then aggregate all funds into a single value-weighted portfolio and project the quarterly returns on contemporaneous and lagged market returns. Summing up all the betas, they estimate the long-run beta of VC funds to be 1.80 and the long-run beta of BO funds to be 0.65. Interestingly, their estimate for VC funds is higher than ours (we report 1.28 on average). However, the beta of BOs is diametrically different. 17 In order to assess the impact of the assumption on leverage we compare it to two other betas. For beta2, we assume that the beta on debt is 0.25, that the debt to equity ratio of BO deals is 3 throughout the investment life, and that VC deals have the same equity beta as their industry peers. For beta3, we assume that the debt to equity ratio of BO deals is the same as that in the industry. Finally, for beta4, we assume that VCs have the same beta as the 25%-smallest firms in the industry in which they operate. 18 Another possibility is that larger funds are more diversified, hence have lower portfolio betas. 19 LR (2003) assume that the leverage of all deals is as the industry. They report an average beta of 1.08 for BOs (we report 1.15 for beta3) and of 1.12 for VCs (we report 1.15 for beta3).

11

measures, control variables, and assumptions about leverage. Probably, this poor explanatory power relates to the three principal assumptions we made. First, that the CAPM holds; even for the public market and despite disputes the CAPM does not successfully explain cross-sectional differences in performance. Second, that returns on assets are the same within industries. This is a particularly strong assumption for venture deals which, by definition, concern companies that are not yet publicly “tradeable” and therefore do not have a public market equivalent. Third, we assume implicitly that the risk is constant over the life of a deal. Despite having partially controlled for this by assuming a decreasing leverage, we have not captured all the sources of a decrease. In particular, for venture investments it is arguable that the main uncertainty (final success of the product) decreases over time. This uncertainty is, in addition, likely to be higher for early venture investing. 21 However, we do not find evidence that funds that invest more in early stage VC deals outperform. 22

A.2 Fund exposure to non-diversifiable risks Using available data about the underlying investments of funds, we can compute the exposure of each fund to certain macroeconomic factors. A cross-sectional analysis will then shed light on the underlying time-series dyna mics. For each fund we compute the fraction of the portfolio that was invested in every quarter. Using these weights we compute a value-weighted average of factor X over the life of the fund. If X influence investment returns, then cross-sectionally a positive relationship should exist between the value-weighted average of X of each fund and its performance. We report the exposure to the following factors: unexpected real GDP growth rate, interest rates, return of the market portfolio, and aggregate level of earning-to-price ratio. The unexpected GDP growth rate is obtained by modeling real GDP growth as an AR(1) process. 23 Interest rates are the yields of Corporate-BAA bonds as reported by Moody’s. Exposures are computed both over the 20

We introduce and detail our model (control variables etc.) in the next sub-section. Note that this is the case in the model of Berk et al. (2003). The risk premium decreases over the life of the investment. 22 One could also argue that bankruptcy/distress risk should be priced for LBOs. Nonetheless, whether distress risk should be priced is still being debated in the finance literature. 21

12

life of the investment, as described above, and at entry. For entry we use as weights the portion of the portfolio invested in deals that started over a given quarter. 24 In Panel A of Table V, we report how these exposures are related to performance. In this and all subsequent regressions we also control for certain characteristics. We control for fund size and the square of fund size. Based on the discussion in section I.C., we include among the independent variables the average length of investments and a dummy variable for liquidated funds. This is to capture whether successful funds exit deals and liquidate funds earlier on. As Hege, Palomino and Schwienbacher (2003) argue that European funds underperform US funds for various institutional reasons, we also include the proportion of European deals undertaken by a fund. 25 Finally, we control for the total capital committed during the vintage year of the fund (see LR and Gompers and Lerner, 2000). In addition to this set of “pure” characteristics, we add a set of characteristics that may be related to risk: the proportion of investments that are made in early stage funds (as they are likely to be riskier), and the proportion invested in the two principal industries (health and high- technology). 26 These three variables can also be seen as proxies for idiosyncratic and systematic risk. Indeed, uncertainty in general, and information asymmetries in particular, often characterize young firms in the early stages of venture financing and firms in the high- technology industry. 27 Table V Panel A of Table V shows several interesting interactions between macroeconomic conditions and performance. At entry, being in a period of high yield or high GDP growth is generally bad, especially for VC funds. VC funds that invest during an economic boom significantly underperform (t-stat is -2.1) unlike BO funds. This suggests that venture deals 23

Hamilton (1994, p112) shows that the real GDP growth rate follows an AR(1) process. The proportion is in terms of length times amount. For example, suppose there are two deals of an amount X: if deal 1 lasts twice as long as deal 2, then the weight assigned to deal 1 is 2/3. Results are, however, insensitive to this assumption as deals are of similar length. Such a weighting scheme is also used for proportion of investments (e.g., proportion invested in the health industry). 25 We actually find that it is more the proportion of European deals that is related to performance rather than the fund being European. 26 Industry groups are defined as follows. Health comprises three Fama French (1997) industries: healthcare, medical equipment, and pharmaceutical products. High-tech comprises four Fama French (1997) industries: electrical equipment, telecommunications, computers, and electronic equipment. 27 Cochrane (2003) finds that, at the deal level, early stage VC deals are riskier than late stage VC deals. 24

13

are too costly in boom times, potentially because a lot of money is chasing too few deals during these periods (see Gompers and Lerner, 2000). Interestingly, this is not the case for BO investments. For buyouts, the main determinant of performance is the interest rate paid for their very high debt. In boom times interest rates are high, making it costly to undertake BO investments. This is confirmed by the results in columns 7 and 8 (Panel A of Table V). Once GDP levels are controlled for, BO funds that mostly invest in periods of high interest rates for corporate debt significantly underperform (t-stat is -3.3). This is not so for VC funds. The key determinant of the performance of VC funds is the GDP growth during the holding period. VC funds that invest during periods when GDP growth rates are unexpectedly high outperform (t-stat is 3.5). This is not so for BO funds. If anything, BO funds do better in an economic slump, but coefficients are not significant. The correlation of returns to the stock market portfolio is also different for VC and BO funds. VC funds appear insensitive to stock market fluctuations unlike BO funds, although results are only weakly significant. The overall picture from this table is therefore as follows. The performance of VC funds is sensitive to business cycles but not to stock market fluctuations. What matters for venture investments is that the economy is booming. The performance of BO funds, in contrast, is unrelated to business cycles except at entry via the impact of corporate bond yield, but is related to stock market fluctuations. At exit, high levels of earning- to-price ratios on public stock markets are, as expected, negatively related to the performance of both BO and VC funds. Once other macroeconomic factors are controlled for, the loading is still negative but non-significant for VC funds. In contrast, the loading remains strongly negative for BO funds (t-stat is -2.5).

A.3 Marking to market the investments In this sub-section we estimate implicit betas. For each investment that a fund undertakes we assume that the same amount is invested on public stock markets. When the investment is exited, the corresponding stock market position is closed. For each fund we then have two time series of cash flows with the same timing and same inflows but

14

different outflows. If the return on each investment (which we do not observe) is related to stock market fluctuations then the two series should be related. For this exercise we use IRR as a measure of performance. The reason for this is that we do not want the discount rates used to calculate PI measures to bias our findings and the coefficients will have a “beta-type” interpretation. Indeed, when we regress the IRR of funds on their market equivalent IRRs, the loading can be interpreted as a beta. It is as if we had a time-series of portfolio returns regressed on a time-series of market returns. For the market equivalent investments we use three time-series of returns. First, we use the industry return, that is, the return on a value-weighted portfolio of publicly traded companies in the same industry as the company in which the fund has invested. Second, we adjust the industry returns for leverage for each BO deal. Using the CAPM (see Appendix A.IV), we compute the return on asset in the industry. Then, assuming a debt-toequity ratio of 3 for BO deals, we obtain a time-series of levered industry returns. Finally, we use the market portfolio (CRSP value weighted index). Results are reported in Panel B of Table V. We find a positive and significant relationship between the actual IRR and the market equivalent IRR for both VC and BO funds and for both the industry and the market portfolio (though not for the industry levered one). It is interesting to note that the industry portfolio leads to stronger results than the market portfolio. In addition, the finding that VC funds are less sensitive to market conditions than BO funds is confirmed here, but the difference is small. The loading on the market IRR (in column 6) is 0.84 for VC funds (t-stat is 2.27) and 0.97 for BO funds (t-stat is 2.70). The implicit market beta is thus less than 1 but relatively close. The practitioner’s assertion that the performance offered by PE funds is unrelated to public market performance is therefore clearly exaggerated, but PE funds would seem to offer certain hedging properties that are rather attractive. Finally, the loading on the industry IRR is statistically more significant but lower in economic magnitude (this is primarily explained by the higher volatility of industry IRRs).

15

B. Idiosyncratic risk JRK (2003) point out that GPs are forced to hold considerable idiosyncratic risk. Indeed, primarily due to incentive considerations they have a substantial amount of wealth in the form of contingent stakes. JRK (2003) propose a model whereby more idiosyncratic risk should be associated with higher returns, in equilibrium. In particular, their theorem 4 states: “All else equal, the return received by the investors is increasing in the amount of realized idiosyncratic risk, even though they face competitive market conditions.” To test this implication they argue that, for BO funds, a reasonable proxy for idiosyncratic risk is the number of investments that a fund makes. In the absence of this information they use the number of take-downs. Our dataset, in contrast, enables us to construct several proxies that capture more closely the spir it of the model. In addition to the number of take-downs and of investments, we construct the proportion invested in the dominant industry and use Herfindahl indices for type and industry concentration based on the portfolio weight of each investment. The weights are computed either in terms of duration multiplied by capital invested, as mentioned above and in the appendix. 28 These indices capture the dispersion of investments across type (VC deals vs. BO deals) and across 48 industries.29 Being focused increases the amount of idiosyncratic risk to both GPs and LPs. Nonetheless, there are advantages to being focused - indeed, it enables GPs to work with smaller and more specialized teams that may learn faster via a high number of similar deals. An additional benefit is that GPs build tight links with the industry which may improve performance. 30 It may be that poor managers diversify because they have failed and thus try something different. Furthermore, one could argue that it is not so much the number of investments as the correlation between them that should be related to returns in their model. In order to capture this we construct a variance-covariance matrix of industry

28

For example, if there are two investments in two different industries for an amount i1 and i2 and their length is repectively t1 and t2, then 1 index will be equal to 0.5 and the other to {(t1*i1)/ 2 2 (t1*i1+t2*i2)} +{(t2*i2)/ (t1*i1+t2*i2) } . 29 The 48 industries are as defined in Fama and French (1997). 30 If seen as a specialist, targeted firms may bargain less and the GP may get cheaper prices. In addition, bankers trust you more, which reduces debt cost, (see Cotter and Peck (2001) for such evidence). Buyout specialists are likely to be repeat players in the BO debt market and their reputation may be at stake. Lenders are then likely to lend to them at easier terms.

16

returns. Matching each deal to an industry we assess the volatility of the “marked-tomarket” portfolio of investments. 31 This variable is referred to as volatility. Results are reported in Panel C of Table V. We confirm the findings of JRK on the relationship between take-downs and performance, although ours is not statistically significant. Nonetheless, their intuition and claim cannot be rejected as, when we use the number of investments as an independent variable, there is a strong negative relationship between performance and number of investments. This relation, moreover, holds for both VC and BO funds and is strongly significant for VC funds (t-stat. is -2). Furthermore, for all concentration proxies a positive relationship exists between focus and performance. Overall, then, funds that do fewer investments, or invest in a single type of deal (venture vs. buyout) or a single industry, strongly outperform. Note, however, that volatility is not significant which suggests that industries may be more different than their market correlation would indicate. As a conclusion, we cannot reject the hypothesis that idiosyncratic risk is priced and plays an important role in the heterogeneity of fund performance. 32

C. Other determinants of performance In panel D of Table V we focus on the characteristics that are related to performance. A number of characteristics appear to be strongly related to performance in a robust fashion. Most notable is the increasing and concave relationship with size. If a fund is too large, it is unlikely to find enough good investment opportunities and may have to diversify (invest in new industries, for example) or invest in projects with poor perspectives. It is interesting to compute where the turning point, i.e., the decrease in returns, is. The answer is typically about US$20 billion as of 2002. In Table I, we report 31

Construction details are provided in the appendix. There are potentially other sources of risk faced by LPs. For example, investors face the risk that fund managers will not find an investment opportunity. If such opportunities are related, one way or another, to business cycles and market conditions, then cash flows will also be related to business cycles and to the return of the market portfolio. In other words, finding an investment opportunity is conditional on the resolution of systematic uncertainty. It can then be expected that in equilibrium, investors require a premium for this feature. This type of risk is similar to what Berk, Green, and Naik (2002) call “technical” uncertainty in the R&D investments context. They present a model in which the uncertainty regarding the success or failure at each stage is an idiosyncratic risk that is priced. 32

17

that the 75th percentile BO fund has a size of 0.7 billion. The turning point is therefore distant but note that our largest fund in the sample is close to this threshold. In addition, in Panels A, B, and C the concavity is often significant only at a 10% el vel test. In our sample, then, larger funds do better. In most specifications of Panels A, B, and C investment length is one of the strongest explanatory variables. The longer investments are held, the poorer the (net of fees) performance of the fund. This observation is in line with our suspicion that GPs may have an incentive to hold onto poorly performing investments, as mentioned in Section I.D. When fixed-year effects are included and for funds raised before 1993, the explanatory power is reduced. In addition, in Panel A when macroeconomic variables are included, the explanatory power of length is also reduced. This indicates that investment length is related to macroeconomic conditions. In bad times, funds are forced to hold onto investments longer and will subsequently underperform. This further stresses the importance of business and market cycles for private equity investments. As documented in Hege et al. (2003), the European private equity industry and investments are different from those in the US and are found to underperform their US cousins. We report a difference of some magnitude: investing 100% in EU deals compared with 100% in US deals results in a profitability index that is 0.6 lower. This is a significant amount and means that EU deals return only half as much as their US counterparts. We find that first-time and second-time funds underperform while experienced funds strongly outperform. Being experienced increases PI by as much as 0.38. Such effects, however, are not present in the early years. The sample of funds raised before 1993 exhibits such an effect but neither coefficient is statistically significant. These are of considerable economic magnitude and in line with Gompers and Lerner (1999) and KS. Moreover, KS find that first-time funds are raised after periods of high returns for the PE industry. It is possible that unskilled GPs enter the market at such times so as to attract naive investors. The significance that appears mainly for funds that were raised before the Internet ‘hype’ is very much consistent with these findings. For funds raised between 1993 and 1997, being experienced (as opposed to inexperienced) made an immense difference

18

while for funds raised before 1993 it did not. So it could be said that experience counts in times of hype! Investments in high-technology industries paid off for funds raised before 1997 but not significantly so for those raised before 1993, thus capturing the Internet boom. Also of interest is the fact that when a GP is the main investor, its fund outperforms. This may explain part of the size and experience effect: if a GP is large and experienced it can only leverage this advantage as the lead investor. Similarly, only dominant investors can “time” the exit decision according to their objectives. Note also that we do not find a negative effect between the capital committed to all PE funds in the inception year of a given fund and its performance. At first sight this observation may seem to contradict the rationale of “money chasing deals” (Gompers and Lerner, 2000) which suggests a negative impact of capital committed to PE funds and fund performance. We offer two possible explanations for this finding. First, LPs may be able to accurately predict future fund returns and heavily commit to PE funds in years during which there are good investment prospects for PE funds. Alternatively, we can stay within the Gompers and Lerner (2000) framework of “money chasing deals” and argue that the money committed to PE funds in the entry year of a given investment will push up valuations of PE investments in subsequent years. This will increase the valuation of the investment for subsequent rounds of funding and thus the (accounting) valuation of the initial investments.

III. Overall Performance of PE Funds

A. Performance before 1997 Evaluating global performance is important. It gives the expected return of this investment class. As mentioned above, sample selection and accounting issues render the problem rather tricky. In Table VI we show aggregate performa nce statistics for two samples: a sample of funds raised before 1993 and another before 1997. In Panel A we account for residual values (RV) as proposed above, that is, we use the matrix evaluated in

19

Table I to convert accounting values as of June 2003 into cash flow equivalents. In Panel B, to compare, we write off these accounting values. Table VI While we can say that the 983 funds raised before 1993 underperformed the market severely, to make a definitive estimate we have to wait until they liquidate their final positions. In Table II, we report that 25% of what they have invested is not yet liquidated even though they have attained their typical ‘lifespan’ (10 years - see appendix). Nonetheless, we can assess the final profitability index (PI) to be about 0.95 on a valueweighted basis and 0.83 on an equally-weighted basis. Without accounting for RVs, these figures are respectively 0.76 and 0.87. The latter figure means that about 90% of the money invested has been returned to investors in present value terms (that is assuming a beta of 1 with the market portfolio). The final profitability index is likely to be about 0.95 but this is speculative. However, loosely speaking it means that PE funds have a 4.5% annual return when the market has a 10% annual return, which is close to the risk-free rate. In other words, their performance appears very low. 33 The analysis in the previous sub-section indicates that inexperienced GPs underperform. It is possible that this young industry has learned over time and thus performed better in more recent periods. In equilibrium, it may mean that LPs and GPs may be tied by some implicit agreement stating that the performance of the first funds will be poor but will improve and that the GP will continue to work with the same LP. Another possibility is that since a large number of funds first entered the market, many of them unskilled, screening has separated the good from bad. In this instance, LPs must be guaranteed one way or another to be able to invest in subsequent funds, otherwise they would not have invested in the early stages of the industry and a (rational) equilibrium would not exist. Consistent with the above assertion, the performance of funds raised before 1997 seems to improve. Nonetheless, a fair amount derives from accounting values. If these funds manage to convert current investments according to Table I then they will strongly 33

From the example in footnote 12. If the opportunity cost of capital is 10% yearly, then PI=0.95, means that a one year investment X has returned 0.95*1.1*X=1.045X.

20

outperform the market. In particular, on a value-weighted basis their expected PI is 1.09. On the other hand, if all their current investments are worthless then their PI is about 0.8 meaning that they have not yet returned the opportunity cost of capital (far from it). Evaluating overall performance is, then, very difficult and further robustness and sensitivity tests are carried over in section IV. Interestingly, no matter what happens to current investments of VC funds, the performance of VC funds raised before 1997 is very high on a value-weighted basis. Even if RVs are written off, VC funds post a PI of 1.12 (VW). Loosely speaking it means that they have returned about twice as much as the market over the same period. This is not the case for VC funds raised before 1993. This also shows the sensitivity of expected performance estimates. Finally, performance is heterogeneous in every sample. The median funds severely underperform. This further stresses the point that a few large funds have posted strong performances but many smaller funds have returned close to nothing. The heterogeneity is larger for VC funds. The VC fund at the 25th percentile has a PI of 0.40 (in Sample 2, 0.46 in Sample 5). This basically means that half the value of the capital invested is returned to investors. If we write-off RVs this figure decreases to one third.

B. Performance After 1997 Since 1997, the last fund vintage year considered in our main sample, the economy overall and stock markets in particular have reached a high and then faced a serious recession. This raises the question of how these events influenced the performance of young funds. In addition, if their performance is poor the overall performance figures reported above may be quite optimistic. As pointed out in sub-section I.C and in the appendix, a performance assessment of these younger funds is difficult as most of them are still in the investing phase and have not yet had the time to realize more than a fraction of

21

their investments. We compare these young funds to mature funds at the same age and compute several indicators related to their capital distributed and capital called. 34 Table VII The first striking observation from Table VII is that the overall amount of money raised by PE funds during these five years is immense - roughly equivalent to the amount raised over the previous 20 years (see also Graph 1). If we compare the most recent vintage years with their mature peers, we can see clear differences in terms of both investments and divestments. Younger funds are significantly behind the ‘historic’ investment schedule. For all vintage years they invested less than half the percentage of committed capital that their mature peers had done at the same point. For example, the mature funds had on average 81% of their committed capital invested by the end of their fourth year, whereas funds from vintage year 1999 had only invested 30% of committed capital so far. At the same time, the younger funds are also behind schedule in terms of the realization of their investments, especially for less recent vintage years. We note that, traditionally, funds break even (without discounting) in their sixth year as they have then returned 91% of called capital to their investors. In contrast, 1997 funds have so far returned no more than 30% of called capital on average. The present difficult economic environment is also reflected in the accounting values of the youngest funds which are also consistently below the comparable values in the sample of mature funds. Historically, the book value of unrealized investments for funds aged three to five years was roughly equal to the capital committed. For younger funds of comparable age the ratio is between 58% and 69% - a finding that is even more striking if we remember that the residual values of the mature funds have been reduced by their greater number of realizations. Overall, this suggests that the younger funds may have difficulty identifying enough suitable targets to place the enormous amounts of capital they have raised and finding profitable exit opportunities for all their investments prior to the end of the fund’s life.

34

For a meaningful comparison, we match young funds from recent vintage years with mature funds from our main sample on a same -age bases. For example, the situation of (six-year-old) funds raised in 1997 is compared to the situation of mature funds in their sixth year.

22

Having described the difficulties of assessing the performance of very recent vintage years, we still need to calculate two PI values for them that make a comparison with the mature funds possible. In the first case we treat accounting values of the young funds as immediate cash flows of the same value. If we then use the CRSP- index and the 30-day Treasury-bill rate to discount inflows and outflows respectively, we find that the young funds outperform the market as well as their mature peers in all vintage years. This finding is driven by the negative returns on the CRSP index in recent years applied in the discounting of cash flows. Given the flexibility of GPs in the choice of accounting values, one may wonder whether accounting values have been sufficiently adapted to changed expectations about the economic environment (which are reflected in the CRSP decline). To find a lower bound of fund performance we also calculate the PI using the same discount rates but for completely written-off accounting values. Not surprisingly given our earlier discussion, we find a dramatic underperformance of younger funds vis-à-vis the market as well as their mature peers. Overall, our findings suggest that the most recent funds are likely to have difficulty generating for investors the same return achieved by the pre-1997 vintage funds in our sample.

C. Performance, discount rates, and systematic risk If PE funds have a beta different from 1 then the global performance figures that we reported in Table V should be adjusted accordingly. In fact, even the figures in Table I should be changed as the NPV of cash flows should likewise be adjusted for systematic risk. In this sub-section we abstract from the second point and focus on how the profitability index reported in Table V is modified when each fund’s cash flow is discounted by the modified discount rate. 35 Table VIII reports the resulting estimation of expected performance for the funds raised before 1997 (Sample 5). This performance is, moreover, displayed for all funds (Panel A), VC funds (Panel B), and BO funds (Panel C). Five scenarios are compared in each case: when beta is 1 for all funds (base) and when it is one of the four betas we computed and reported in Table VIII. 35

To adjust discount rates, we apply the CAPM formula and only inflows are corrected for systematic risk.

23

Table VIII For all funds the adjustment for systematic risk does not decrease performance dramatically but contributes to bring it closer to market performance. For example, on an equally- weighted basis, the outperformance we found in Table VI (1.08) is now 1, that is, the expected return is exactly equal to the public market retur n. On a value-weighted basis, we still find outperformance. PI is 1.08, down from 1.18 with the base scenario. Note that the very high performance of venture funds that we pointed out in sub-section III.A is significantly reduced by systematic risk adjustment even though most of this adjustment concerns buyout investments (via our leverage assumption). PI is still a high 1.27 on a value-weighted basis (down from 1.45) and on an equally- weighted basis is it is close to 1. As a conclusion, adjustment for systematic risk brings the performance of PE funds raised after 1997 closer to the market performance, with certain size discrepancies and despite accounting for RVs.

IV. Sensitivity analysis: Accounting, sample selection, and overall performance In this section we assess the sensitivity of the average fund performance (reported in the previous section) to the various assumptions made (Section I). Results are reported in Table IX. Table IX In Panel A we change the selected sample. In particular, we document how average performance changes when we focus on funds that are close to liquidation, as do KS. The evidence is mixed. For funds raised before 1993, focusing on liquidated funds raises average performance. Basically, funds go from underperforming to market performance. Indeed, on a value-weighted basis, the performance of liquidated funds and funds raised before 1993 that are close to liquidation exactly matches the performance of the market portfolio as the average PI is 1. In contrast, for those raised before 1997, funds close to liquidation also perform in line with the market portfolio while the whole sample outperforms. Potentially, these funds may have invested during the Internet boom and may not have written down their investments since the bubble burst. To investigate this issue, and by the same token the influence of our accounting value treatment, we turn to Panel B.

24

In Panel B, for the sample of funds raised before 1997, we report seven different performance measures and their correlation. We make three assumptions about RV. The first is that residual values are written off (denoted by ‘wo’). Obviously, this is very conservative. In this worst-case scenario, only past cash flows (as of June 2003) are accounted for. In other words, this measure assumes that the world ends in June 2003 and thus the corresponding performance measure represents the lower bound for fund performance. The second assumption is that there is a one-to-one correspondence between net present value of future cash flows and RV. This is still conservative but less so than writing off RV. This alternative is denoted by ‘a0’. Finally, the third assumption is the one made throughout the paper, that is, residual values reported in June 2003 are converted into an immediate cash inflow according to the matrix in Table III. It is denoted by ‘a1’. In addition, two sets of discount factors are used: the one used throughout the paper and one that uses the S&P 500 to discount both inflows and outflows (as in KS). Finally, we report the Venture Economics IRR. We first note that performance measures are all very highly correlated. The IRR has a correlation of about 80% with each PI measure. In addition, PI measures consistently have coefficients of correlation above 90% with one another and there are several instances of 99%. These high coefficients, coupled with the observation that overall performance varies substantially across measures, imply that changing assumptions about accounting values or discount rates may translate the returns but do not change the ordering, that is, using any of the alternative PIs as a dependent variable in the above regressions should not change results. Changing assumptions affects only the overall performance estimate. Indeed, performance (on a value-weighted basis) ranges from a PI of 0.77 to a PI of 1.43. Our accounting treatment is only partly responsible for this as most of the variation comes from the choice of discount rates. Nonetheless, as we expected, most of the outperformance of the funds raised before 1997 comes from their ongoing investments and so should be treated with caution. If we write off their residual values, we move from a strong outperformance (1.1) to a severe underperformance (0.8). These figures concern most of the US and European ve nture capital and buyout universe, and, as such, offer some of the first and up-to-date measures of the overall

25

performance of this asset class. Our results also bring some clarity into the controversy about the overall performance of PE funds. JRK report a gross underperformance (net performance is about zero as they estimate beta for BO to be less than 1); KS find that PE returns are similar to public market returns, and LR find that PE funds strongly outperform. To reconcile this controversy we need to consider the differences among these studies regarding sample selection, discounting and treatment of accounting values, in light of the above results. LR use a sample of PE funds in which a large and sophisticated LP has chosen to invest. They report that this LP may have reasons other than maximizing its PE fund portfolio returns to select PE funds for investment. Still, one is inclined to think that the “picking ability” of this LP may be the origin of the strong performance they report, despite the fact that they write off any residual accounting values. Moreover, their sample over-represents large and experienced funds. Results in Panel D of Table V show clearly that these types of funds strongly outperform. KS report a PI of 0.96 (if equallyweighted) and 1.05 (if value-weighted). 36 Their estimate is then similar to ours as the effect of their sample choice (which is likely to over-sample winners) and discount rates (S&P returns are lower than those of the market portfolio) is counter-balanced by the effect of the writing off of residual values (decrease performance). Finally, JRK include very young funds that have not yet returned much capital but have called sizeable amounts. Consequently, it is not surprising that the overall performance in their sample is low.

VI. Further considerations: Liquidity Both the investments made by a GP and the partnership stake of LPs are ‘illiquid’ in the sense that selling any of them before maturity would be extremely difficult and expensive. This illiquidity is expected to command a return premium for LPs. In order to reduce this premium, we would expect GPs to take action to facilitate the transferability of partnership stakes. In practice, however, the exact opposite seems to prevail: GPs try hard to prevent a secondary market coming into existence. 37

36 37

PI s-s-w0 corresponds to what KS report and call Public Market Equivalent. See Lerner and Schoar (2003).

26

When the identity of an LP changes, GPs are not forced to sell any of the illiquid assets held in their portfolio, so bank runs as described by Diamond and Dybvig (1983) cannot occur. Conversely, information-based bank runs as described by Jacklin and Bhattacharya (1988) can occur. This may be one reason why GPs want to avoid a secondary market. There are other explanations for the non-existence of a secondary market that go beyond the will of GPs. For example, the asymmetry of information between incumbent LPs and outside investors may be so flagrant that a secondary market fails to come into existence. 38 Lerner and Schoar (2003) investigate this illiquidity puzzle and propose the following explanation. They argue that when a GP raises a new fund, outside investors tend to suspect that incumbent investors do not reinvest because of a liquidity shock or because the fund is a ‘lemon’. This implies a higher cost of capital for follow-on funds. An optimal response for GPs is to make sure that the fund is as illiquid as possible so that only LPs with a low probability of facing a liquidity shock invest in the first fund. GPs then trade off the current cost of equity against the future cost of equity. Lerner and Schoar (2003) present a model with such features and empirical evidence consistent with it. The question yet to be answered is how much compensation illiquidity requires. If LPs hold enough liquid assets, having a small stake in an illiquid security may not be a major handicap when faced with liquidity needs. Indeed, in such a situation, LPs are likely to rely first on more liquid assets such as (safe) bonds, then maybe large stocks, etc. and only in very extreme situations will PE positions be liquidated. This would still require a premium but it may not be sizeable as LPs are large, diversified, institutional investors. In addition, the Lerner and Schoar (2003) model implies that LPs are those with the lowest probability of a liquidity shock, which suggests an even lower compensation. The above findings of an average performance similar to the public market performance may then be consistent with a rational equilibrium in which there is a low illiquidity premium and where private equity returns are not highly correlated with public equity returns.

38

See Akerlof (1970) and the subsequent literature.

27

VII. Conclusion This paper sheds light on the risk and return of the private equity industry. We show that the average performance of private equity funds is similar to the performance of the market portfolio. Nonetheless, a definitive assessment of expected returns is premature as it is unclear how to value ongoing investments. Furthermore, our sample enables us to test the hypothesis of Jones and Rhode-Kropf (2003) that idiosyncratic risk is priced in this industry. We find substantial evidence consistent with their predictions. In addition, we find that venture fund performance is very sensitive to business cycles while buyout fund performance is sensitive to the level of corporate bond yields and, to a lesser extent, to public market performance. This evidence, together with the claim of Lerner and Schoar (2003) that the illiquidity premium is likely to be low, makes it possible that this relatively low performance is consistent with a rational pricing of this class of asset.

28

References Admati, A., and P. Pfleiderer, 1994, Robust financial contracting and the role of venture capitalists, Journal of Finance 49, 371-402. Akerlof, George A., 1970, The market for ‘lemons’: Quality uncertainty and the market mechanism, Quarterly Journal of Economics 84, 488-500. Berk, Jonathan B., Richard C. Green, and Vasant Naik, 2002, Valuation and return dynamics of new ventures, Review of Financial Studies, forthcoming. Blaydon, Colin, and Michael Horvath, 2002, What’s a company worth? It depends on which GP you ask, Venture Capital Journal, May. Blaydon, Colin, and Michael Horvath, 2003, LPs need to trust General Partners in setting valuations, Venture Capital Journal, March. Cochrane, John, 2003, The risk and return of venture capital, Working paper, University of Chicago. Cornell, Bradford, and Kevin Green, 1991, The investment performance of low-grade bond funds, Journal of Finance 46, 29-48. Cotter, James F., and Sarah W. Peck, 2001, The structure of debt and active equity investors: The case of the buyout specialist, Journal of Financial Economics 59, 101147. Diamond, Douglas W., and and Philip Dybvig, 1983, Bank runs, deposit insurance, and liquidity, Journal of Political Economics 91, 401-419. Fama, Eugene F., and Kenneth R. French, 1997, Industry cost of equity, Journal of Financial Economics 43, 153-193. Fernandez, Pablo, 2004, The value of tax shields is not equal to the present value of tax shields, Journal of Financial Economics, forthcoming. Gompers, Paul, 1996, Grandstanding in the venture capital industry, Journal of Financial Economics 42, 133-156. Gompers, Paul, and Josh Lerner, 1999, An analysis of compensation in the U.S. venture capital partnership, Journal of Financial Economics 51, 3-44.

29

Gompers, Paul, and Josh Lerner, 2000, Money chasing deals? The impact of fund inflows on private equity valuations, Journal of Financial Economics 55, 281-325. Gompers, Paul, and Josh Lerner, 2001, The venture capital revolution, Journal of Economic Perspectives 15, 145-168. Hamilton, James D., 1994, Time-series analysis, Princeton University Press. Hege, Ulrich, Frederic Palomino, and Armin Schwienbacher, 2003, Determinants of venture capital performance: Europe and the United States, Mimeo. Hellmann, Thomas, 1998, The allocation of control rights in venture capital contracts, Rand Journal of Economics 29, 57-76. Hellmann, Thomas, and Manju Puri, 2000, The interaction between product market and financing strategy: The role of venture capital, Review of Financial Studies 13, 959984. Hellmann, Thomas, and Manju Puri, 2002, Venture capital and the professionalization of Start- up firms: Empirical evidence, Journal of Finance 57, 169-197. Jacklin, Charles J., and Sudipto Bhattacharya, 1988, Distinguishing panics and information-based bank runs: Welfare and policy implications, Journal of Political Economy 96, 568-592. Jones, Charles and Matthew Rhodes-Kropf, 2003, The price of diversifiable risk in venture capital and private equity, Working paper, Columbia University. Kaplan, Steven and Antoinette Schoar, 2003, Private equity performance: Returns, persistence, and capital flows, Working paper, University of Chicago. Kaplan, Steven and Jeremy Stein, 1993, The evolution of buyout pricing and financial structure in the 1980s, Quarterly Journal of Economics 108, 313-358. Kaplan, Steven and Richard Ruback, 1995, The valuation of cash flow forecasts: An empirical analysis, Journal of Finance 50, 1059-1093. Lerner, Josh, 1994, Venture capitalists and the decision to go public, Journal of Financial Economics 35, 293-316. Lerner, Josh, 1995, Venture capitalists and the oversight of private firms, Journal of Finance 50, 301-318.

30

Lerner, Josh and Antoinette Schoar, 2003, The illiquidity puzzle: Theory and evidence from private equity, Journal of Financial Economics, forthcoming. Ljungqvist, Alexander and Matthew Richardson, 2003, The cash flow, return and risk characteristics of private equity, NBER working paper 9454. Moskowitz, Tobias and Annette Vissing-Jorgensen, 2002, The private equity premium puzzle, American Economic Review 92, 745-778. Peng Liang, 2003, Building a venture capital index, Working paper, Yale University. Quigley, John and Susan Woodward, 2003, An index for venture capital, Working paper, University of California at Berkeley.

31

Appendix A.I. Industry description Private Equity funds are typically structured as limited liability partnerships in which a specialized Private Equity firm serves as the general partner (GP) and institutional investors or high-net-worth individuals provide the majorit y of capital as limited partners (LP). Most Private Equity funds are closed-end funds with a finite life of about 10 years, which may be extended by up to two years with the consent of the majority of the shareholders (Gompers and Lerner, 1999). During this period, the GP undertakes Private Equity investments of various types (venture capital, bridge financing, expansion capital, leveraged buyouts, mezzanine debt) on behalf of the fund, with the obligation to liquidate all investments and return the proceeds to the investors at the end of the fund’s life. A minority of funds, so-called “evergreen” funds, have, in contrast, an infinite life and no obligation to liquidate their positions. At the time of the fund’s inception, LPs commit to a percentage of tota l fund size. In the first years of the fund life (typically 6 first years), the GP makes capital calls (or takedowns) to LPs whenever it finds an investment opportunity. Typically, within two weeks, LPs have to provide the corresponding cash. The total amount of such “capital calls” can exceed the capital committed at the fund’s birth, but this is relatively rare. In fact, it is more common for a fund to liquidate without having invested all the capital committed. Whenever a fund receives returns on its investments, proceeds are proportionally distributed to the LPs – net of fees and the so-called “carried-interest”. These distributions can be in form of cash or shares, often preferred shares or convertibles. For its services the GP receives compensation in varying forms. As a fixed component, a yearly management fee (between 1% and 3%) of the total committed capital (i.e. the size of the fund) is charged to LPs. In addition, the GP can receive fees for each transaction performed (fixed or as a percentage of deal value) and participates in the fund returns through the so-called “carriedinterest” which often specifies that 20% of all net gains (or gains beyond a certain “hurdle rate”) accrue to the GP whilst the rest is distributed among LPs.39 PE firms often manage several finite limited liability partnership funds concurrently, raising a new fund three to five years after the closing (“inception”) of the fundraising process

32

for the previous fund. Note also that some PE funds are structured as non-partnership captive or semi-captive vehicles with one dominant (or exclusive) LP. It is mainly the case with Private Equity funds that are managed by subsidiaries of large insurance companies or banks that primarily or exclusively invest the parent company’s money and can then have a different fee structure. These funds are relatively more common in Europe but overall the PE industry is dominated by non-captive (or limited partnership) funds.

A.II. VentureXpert content and corrections Venture Economics’s Private Equit y Performance Database is the most comprehensive source of financial performance of both US and European Private Equity funds in existence and has been used in previous studies of PE returns (e.g., KS and JRK). Venture Economics builds and maintains this dataset based on voluntarily reported information about cash flows between GP and LPs in Private Equity funds. Venture Economics obtains and cross-checks information from both GPs and LPs, which increases the reliability of this data. The aggregate residual book values of unrealized investments (i.e., non exited investments, also sometimes called net asset values) are obtained by Venture Economics from audited financial reports of the partnership. Venture Economics makes some simplifying assumptions about the cash flows. First, cash flows are assumed to take place at the end of the month. Second, stock distributions are valued based on the closing market price (as of) the day of distribution to LPs. In the case of an IPO, GPs have to hold on to the stock until the end of the lock-up period. After this date, however, they have some flexibility regarding when to distribute the stock to the LPs. This means that GPs can “time” distributions to some extent. This valuation at the time of stock distribution to LPs ma y be different from the value and timing of actual realizations by LPs, hence this treatment of stock distribution leads to a certain degree of inaccuracy in the cash flow data. For each fund, Venture Economics also collects information on underlying Private Equity investments through its VentureXpert database, starting from 1980. This database contains information on Private Equity investments in 29,739 companies. Several of these

39

See also Gompers and Lerner (1999) for a detailed study of compensations in the US venture capital industry at the partnership level

33

investments have received funding at different points in time, possibly as different types of Private Equity (e.g. subsequent rounds in VC investments) and by different Private Equity funds, so that the total number of investments (different in company, fund or date) amounts to 134,641. About half of this sample concern funds that are not in the cash flow dataset of Venture Economics. This sub-sample is denoted by ‘VX-B’. Data on PE fund investments obtained from VentureXpert include information about the target company (location, industry description, age), the investment (time of investment, stage, group of co-investors, equity amount provided by each fund, exit date and exit mode for liquidated investments), the fund (fund size, investment focus, year of inception or “vintage”) and the GP (age, size, location). Due to the confidential character of Private Equity investments, the composition of this dataset was based on information Venture Economics could obtain through its relationships with the GP and LP community and its market research activities in the Private Equity industry over the past decades. However, despite all these efforts a complete coverage of all investments by all funds remains difficult to achieve. Consequently, missing information on some investments is accommodated in the following way: VentureXpert includes a number of investments with a zero value. These correspond to confidential investments with an undisclosed equity amount and account for 5% of the entry in sample VX-B. We assigned an equity value to these deals according to the following logic. If we had information about at least three other investments of the same fund in the same stage (four stages are defined: early, intermediate, and late VC, and buyout), we assigned the average amount of these investments to the focal investment (71% of the missing cases). Whenever there were too few investments of the same category by the same fund, we turned to the firm level (i.e., consider all investments made by the same GP) and applied the same procedure. Whenever there were too few investments made even by the firm, we relied on the average per stage across the entire sample. Similarly, VentureXpert provides information on many investments but relatively few divestments. This can be explained by the confidential character of many divestitures. We then have to correct for certain missing holding periods. First, certain investments are still in the database as “active investments” with a holding period of more than seven years (i.e., that

34

started before 1996). 40 Second, some investments are reported as terminated but lack an exit date (28% of the cases in sample VX-B). The same logic as above is then applied. We estimate the average length for each type of deal and deduce the exit date. For 82% of the cases there were enough investments in the same stage operated by the same fund to use the stage-fund average length.

A.III. Conversion of accounting values In order to evaluate the relationship between accounting measures and subsequent cash flows, we compare reported RVs to the net present value of subsequent cash flows that accrued to investors. We proceed by selecting a sample of sufficiently liquidated funds, that is, we select funds with vintage year less or equal to 1997 that have a Residual Value (RV) of less than 1% of total capital invested (CI) as of June 2003. There are 431 such funds. For each of these funds, we divide their life into three periods (whenever it applies). The first period is from 5 to 9 years, the second from age 10 to 11, and the third from 12 years onward. 41 For each of these periods we consider five different categories of the RV/CI ratio (0 to 10%, 10% to 25%, 25% to 50%, 50% to 75%, and over 75%). For each of these 15 categories, we average all the reported residual values and the corresponding net present value (NPV) of subsequent cash flows. This way we obtain what each fund, classified by age category and RV/CI category, had on average as a residual value and the corresponding NPV of cash flows. Repeating this computation for all funds, we obtain two 5-by-3 matrices, one for average NPV and one for average RV. Finally, we sum up the averages in each category over all funds (for which we have valid observations) and divide the cumulative RV by its corresponding cumulative NPV. If RV is a good proxy for the value of ongoing investments, then the ratio should be close to one.

40

We use this cut-off as it is three standard deviations above the average length of a deal. Results are insensitive to this choice. 41 As funds typically liquidate in years 10 or 11, the cut is such that we have funds before, during, and after, their “normal” lifespans.

35

A.IV Systematic risk of PE portfolios We evaluate the systematic risk of PE portfolios in three steps. We first estimate industry-wide unlevered beta. Second, we assign each company (deal) to an industry and assume that the unlevered beta of the company is the same as the industryaggregate. We lever it up for LBOs and VCs in a different fashion. Finally, we aggregate each deal beta at the fund level. I. Estimation of industry-wide unlevered betas First, we assume that the CAPM holds, hence the risk-return trade-off for the equity of firm h is: (1) Rhe = Rf + βhe (Rm − Rf ) where Rf is the risk-free rate (proxied by T-bill 30 days), βhe is the stock’s systematic risk, and Rm is the market portfolio (proxied by the CRSP value-weighted index). Using (1) at a monthly frequency, and rolling over the past 60 months, we ese timate the time series of βh,t . These betas are then unlevered using the following relationship: u = (2) βh,t

e ME d βh,t h,t +(1−τh,t )Dh,t βh,t MEh,t +(1−τh,t )Dh,t

where MEh,t is the capitalization of firm h at the beginning of month t, Dh,t is the book value (from Compustat) of the long-term debt, updated yearly. We further assume that the corporate tax rate (τ ) is constant across both firms and time, and u equals 35%. Finally, we denote βi,t the systematic risk of the unlevered firm, and d we denote βi,t the systematic risk of debt. We make two alternative assumptions d about βi,t . It is assumed to be a constant that equals to either 0.25 or 0.50. These figures correspond to the estimation range of the systematic risk of high-grade debt documented in Cornell and Green (1991). The relationship (2) makes some simplifying assumptions. In particular, it values interest tax shields at the full corporate tax rate and assumes that the value of tax shields has the same systematic risk as the unlevered firm. More details and comments on these issues can be found in Kaplan and Ruback (1995), Brealey and Myers (2000), and Fernandez (2003). Next, each stock is assigned to one of the 48 industries defined in Fama and French (1997). Then, individual betas are aggregated in each industry to obtain a time-series of industry betas: (3) βitu =

Ni X

h=1

MEh,t

u wh,t βh,t , with wh,t = P Ni

MEh,t

h=1

36

, i = 1, ..., 48.

Then, for each of the 48 industries, we compute two “summary” equity betas. ae First, is the aggregated equity beta: βi,t =

Ni X

e wh,t βh,t

h=1

e Second, is the value-weigthed average of the betas (βi,t ) of the 20%-smallest se stocks in the industry. We denote it βit .

II. Estimation of equity betas for venture and LBO deals: Each deal j (or investment), carried out by a given fund, has a VEIC number assigned by VentureXpert. Using their industry description, we assign each VEIC number to one of the 48 industries mentioned above. Using information about the type of the deal, we separate venture and LBO investments. Each venture investment is assigned an equity beta for each month between the date of entry (dentry) and the date of exit (dexit) of the deal. The equity beta is ae se either βi,t or βi,t . That is each venture investment operating in industry i is assigned .e .e a time-series of equity betas: (βi,dentry , ..., βi,dexit ). For LBO deals, we need to lever-up the unlevered beta. To begin with, we need to make some assumptions about the debt level of the deal, as we do not have this information. We investigate two scenari: S1: LBO deals start with a debt-to-equity ratio of 3, i.e. a debt-to-asset ratio of 75%. This ratio then linearly decreases down to the industry (value-weighted) average debt-to-equity ratio. That is Dj,dentry = 3, ..., Dj,dexit = Di,dexit . S2: LBO deals start with a debt-to-equity ratio of 3 throughout the investment life. S1 is the most realistic scenario while S2 serves as a sensitivity check of our assumption about the decrease in leverage. A debt-to-asset ratio of 75% is common in practice, as document by Cotter and Peck (2002). The market value of the investment throughout its life is assumed to be the total value of this deal at entry. That is, we search for all the funds that invested in the given company and add up each deal value (capital invested, Ij ). The market value of equity is then a constant denoted V Di . Finally, using equation (2), and assuming that the systematic risk of the unlevered investment (in industry i at time t) is βitu , we obtain the systematic risk of the given LBO deal for each month (of its life).

37

III. Aggregation at the fund level Now, we need to aggregate the information about each deal at the fund level. This convention is adopted for all the characteristics of the deals that we aggregate at the fund level. The systematic risk (or characteristic X) of each deal, each month, is multiplied by the amount invested in this deal (Ij ). We take the sum of all these observations and divide it by the sum of the weights. That is, the characteristic Xj of each investment of fund f is summarized by Xf at the fund level as follows: Xf =

1 T N P Pf

Ijt

t=0 j=1

Nf T X X

Ij,t Xj,t ,

t=0 j=1

with Ij,t = Ij if the deal j is active at time t, 0 otherwise. Then, let us denote Lj the length of deal j, we have: Xf =

1 Nf

P

1 Nf

P

Ij,t Xj,t . When the characteristic is not time varying,

Lj Ij t=0 j=1

j=1

Xf =

Nf T X X Nf X

Lj Ij Xj , this is so when, for example, X is a dummy for deal j is

Lj Ij j=1

j=1

in the health industry. And when the characteristic is not varying across investment, Xf =

1 Nf

P

T X

Xt

Lj Ij t=0

j=1

Nf X

Ijt this is so when, for example, X is the amount of M&A

j=1

in the year of entry of deal j.

A.V Volatility of PE portfolios Here, we abstract from leverage and CAPM considerations. We estimate what would be the expected volatility of the portfolio if the same investments would have been made in the public market in the matching industry. To begin with, we estimate a rolling (past 60 months) variance-covariance matrix of the 48 industry returns. That is, at date t, we have σia,ib,t , which denotes the covariance, from date t-60 to date t-1, between the returns of industry ia = 1, ..., 48 with the returns of industry ib = 1, ..., 48. As above, each deal j operated by fund f is assigned to one of the 48 industries. The volatility of fund f is then defined as: Vf =

1 T

T X t=0

V(

Nf X

a=1

wjt Rjt ) =

1 T

Nf Nf T X X X

wat wbt σa,b,t , where wat =

t=0 a=1 b=1

38

Iat Nf

P

j=1

Ijt

Table I: Descriptive statistics – Sample characteristics This table displays descriptive statistics of five different samples, as of June 2003. Statistics for venture and buyout funds within each sample are reported separately. We report, respectively and for each sample, the average (equal weights) capital committed to a fund in 2002-US-dollars along with the 25th and 75th percentile figures (Size), the average of the ratio of residual accounting value to total capital invested (RV/CI), of the ratio of capital invested to capital committed (Invested), of the number of take-downs, of the proportion of first-time funds, second-time funds, and experienced funds (fourth fund or more raised by the same firm), of European funds, and of liquidated funds, as well as the average (value-weighted) profitability index (see text for definition). Finally, we report the number of observations for each sub-sample. The five samples consist of: the universe of funds with cash flow data in Venture Economics (Sample0), funds that are officially liquidated and with less than 5% RV/CI (Sample4), funds with vintage years of 1993 or less (Sample2), and funds raised before 1993 (1997) for which the total value of investments is more than 50% of the capital committed, labeled Sample1 (Sample3). Sample0: All

Sample1: 93- ,I

Sample2: 93-

Sample3: 97- ,I

Sample4: liquidated

VC

BO

VC

BO

VC

BO

VC

BO

VC

BO

Mean

163.04

615.27

97.26

410.47

86.75

359.63

113.13

450.32

75.79

291.57

25%-pctile

30.21

83.82

33.39

70.26

25.21

57.80

37.68

81.81

22.01

51.34

75%-pctile

159.38

664.87

110.64

341.66

91.24

341.66

127.61

441.30

88.04

297.44

RV/CI

0.52

0.53

0.16

0.26

0.27

0.25

0.29

0.42

0.00

0.00

Invested

0.85

0.85

1.03

1.07

0.99

0.98

1.01

1.05

0.97

0.95

Take downs

9.30

13.77

7.75

15.05

8.37

12.44

9.03

16.96

6.13

10.23

First time

0.38

0.41

0.38

0.50

0.44

0.54

0.35

0.44

0.51

0.51

Second time

0.21

0.20

0.27

0.22

0.25

0.21

0.24

0.22

0.25

0.23

Experienced

0.28

0.27

0.18

0.16

0.16

0.14

0.24

0.24

0.13

0.13

European

0.28

0.37

0.05

0.29

0.20

0.43

0.07

0.34

0.13

0.42

Liquidated

0.22

0.16

0.52

0.38

0.45

0.40

0.40

0.21

1.00

1.00

Profitability index

1.30

1.09

1.04

1.04

0.97

0.97

1.45

0.99

0.91

1.06

Nber of obs.

1451

767

393

98

701

282

510

178

260

105

Size

39

Table II: Descriptive statistics – Investment behavior This table displays descriptive statistics about the investment characteristics of funds for Sample1 and Sample3 (see definition in the previous table). We report, for all funds in this sample as well as for venture funds and buyout funds separately, the average (equal weights) and median (in italics) of the number and length (in months) of investments, of the proportion (in terms of amount invested) invested in venture deals, buyout deals, in the health industry (healthcare, medical equipment, and pharmaceutical products), in the high-technology industry (electrical equipment, telecommunications, and computers), in Europe, and in the dominant industry, and, finally, of four Herfindahl indices, based on the proportion invested in each of the 48 industries and invested in venture vs. buyout deals (type) respectively, in terms of both number and value of deals. Sample3: 93- ,I

Nber deals Deal length % VCs % BOs % Health % Hi-tech % Europe % Dominant industry Herfindahl type value Herfindahl ind. value Nber of observations

Sample4: 97- ,I

All

Venture

Buyout

All

Venture

Buyout

28.58

31.86

15.43

26.40

30.51

14.62

24.00

28.00

12.00

22.00

27.00

11.00

58.40

59.80

52.78

54.22

55.59

50.30

57.14

58.22

51.32

51.56

53.31

48.33

0.77

0.88

0.30

0.72

0.89

0.26

0.92

0.96

0.22

0.91

0.96

0.12

0.23

0.11

0.69

0.27

0.11

0.74

0.07

0.04

0.77

0.08

0.03

0.87

0.20

0.23

0.08

0.19

0.23

0.08

0.15

0.18

0.00

0.13

0.18

0.00

0.45

0.51

0.18

0.43

0.51

0.20

0.43

0.53

0.12

0.42

0.53

0.16

0.11

0.06

0.29

0.16

0.09

0.38

0.00

0.00

0.01

0.00

0.00

0.05

0.42

0.40

0.46

0.41

0.41

0.42

0.37

0.37

0.38

0.38

0.38

0.35

0.84

0.86

0.75

0.84

0.87

0.77

0.91

0.92

0.74

0.92

0.93

0.79

0.30

0.28

0.37

0.30

0.28

0.33

0.24

0.24

0.25

0.24

0.25

0.24

491

393

98

688

510

178

40

Table III: Net present value of future cash flows versus accounting residual values This table displays the ratio of the net present value of future cash flows to the accounting value for various age and RV/CI categories. Age is taken from January 1st of the vintage year to the date of the accounting observation. RV/CI is the ratio of the accounting residual value (RV) reported by the fund to the sum of investments made up to that date. We use the CRSP Value-Weighted index to discount cash inflows (flows to investors) and the Treasury bill 30-days interest rate (Rf) for cash outflows. The number of observations for each age -RV/CI category is reported below. These statistics are derived from a sample of 431 funds with vintage years less or equal to 1997 and an RV/CI below 1% as of June 2003.

RV/CI

Fund age (as of December 2002)

categories

5 to 9 years

10 to 11 years

over 12 years

[0.75, 8 [

0.99

0.95

0.89

Discounting:

[0.50,0.75[

1.24

1.12

0.94

CRSP-VW for inflows

[0.25,0.50[

1.26

1.16

1.13

Rf for outflows

[0.10,0.25[

1.66

1.34

1.23

[0.00,0.10[

2.95

2.00

2.26

[0.75, 8 [

387

99

50

[0.50,0.75[

270

98

43

[0.25,0.50[

244

176

97

[0.10,0.25[

150

176

145

[0.00,0.10[

114

223

415

Number of observations

41

Table IV: Systematic risk This table displays the (fund) CAPM-Beta for various assumptions regarding the estimation of the (deal) CAPM-Beta. The sample of funds is Sample3. Construction details are given in the appendix and in the text. Beta1 is the base case. It assumes, for buyout deals, a beta on debt of 0.25 and a debt-to-equity ratio that decreases linearly from 3 (at entry) to the industry average (at exit); and, for venture deals, it assumes that Beta is the same as in the industry (value-weighted). Beta2 (beta3) is like Beta1 but assumes that the debt-to-equity ratio for buyout deals is always 3 (the industry average). Finally, Beta4 is like Beta1 but the beta of venture deals is the average of the smallest (25%) public firms in the same industry.

Beta1

Beta2

Beta3

Beta4

Median

1.25

1.34

1.16

1.06

EW-Mean

1.22

1.28

1.15

1.01

VW-Mean

1.19

1.21

1.16

0.95

Median

1.72

2.33

1.11

2.28

EW-Mean

1.65

2.17

1.15

2.05

VW-Mean

1.68

2.27

1.14

2.27

Median

1.51

1.89

1.13

1.73

EW-Mean

1.32

1.49

1.15

1.26

VW-Mean

1.22

1.25

1.15

1.01

Largest 25%

VW-Mean

1.26

1.38

1.16

1.13

Smallest 25%

VW-Mean

1.60

2.08

1.12

1.96

First time funds

VW-Mean

1.55

1.96

1.15

1.82

VCs

Bos

All

42

Table V: Risk and performance of Private Equity funds This table presents the results of OLS regressions. The dependent variable is the profitability index (see text) in Panel A, Panel C and Panel D, and it is the IRR reported by Venture Economics in panel B. The common set of control variables includes a dummy variable for liquidated funds (1 if so, 0 otherwise), the natural logarithm of size and the square of it, the natural logarithm of the amount committed to the industry during the year of the fund creation, and the natural logarithm of the average length of the deals. It also includes the proportion of the money invested in deals that are European, early VC stage, health industries, high-technology industries, and in deals in which the GP is the main investor. The additional explanatory variables are different in each panel. In Panel A, we include macroeconomic variables: average corporate bond yield (BAA) when entering deals, average unexpected real GDP growth when entering deals, average corporate bond yield during deals’ life, average unexpected real GDP growth during deals’ life, average return of the market portfolio during deals’ life, and the average earning-to-price ratio (for all stocks listed on the NYSE/AMEX/NASDAQ) when exiting deals. In Panel B, additional explanatory variables are beta (estimated using the CAPM and assuming that beta on asset is the same within industries; see text) and the IRR that the fund would have posted if they had invested in the industry X portfolio (publicly traded) whenever they made an investment in industry X (labeled ‘Marked Ind’.). We also compute the same IRR but we lever up buyout investments (labeled ‘Marked Ind.L’; see text). Finally we compute a similar IRR by replacing the industry portfolio with the market portfolio. In Panel C, we add the number of take-downs, the (natural logarithm of the) number of deals, the Herfindahl index based on type (venture vs. buyout) and based on industries, the proportion invested in the dominant industry, and the volatilit y of the portfolio (using the variance-covariance matrix of industry returns). In Panel D, we add a dummy variable for funds that are first-time, second-time, and experienced (fourth time and more); we also use vintage dummies in certain specifications. Furthermore, additional explanatory variables in Panel A, Panel B, and (certain) in Panel C are crossed with a dummy variable that takes a value of 1 if the fund’s objective is Venture, 0 if Buyout. There are 476 observations in Panel A, Panel C and Panel D (col. 1 to 4), 467 in Panel B, and 671 in Panel D (col. 5 to8). The sample is Sample1 in panel A, B, C, and D (col. 1 to 4) and Sample3 in Panel D (col. 5 to 8). The t-statistics (corrected for serial correlation and heteroskedasticity) are reported next to the corresponding coefficients, in parentheses.

43

Panel A: Macroeconomic factors and performance (PI) 1

2

3

4

5

6

7

8

Liquidated

0.06 (0.63)

0.02 (0.19)

0.07 (0.68)

0.05 (0.52)

0.02 (0.23)

0.09 (0.93)

0.09 (0.93)

0.13 (1.34)

Size

0.43 (2.23)

0.41 (2.12)

0.43 (2.20)

0.40 (2.03)

0.40 (2.08)

0.42 (2.14)

0.40 (2.16)

0.41 (2.37)

-0.04 (-1.92)

-0.04 (-1.70)

-0.04 (-1.89)

-0.03 (-1.65)

-0.04 (-1.74)

-0.04 (-1.90)

-0.03 (-1.80)

-0.03 (-1.86)

Size-squared Committed year

0.11 (1.32)

Deal length

-0.38 (-1.80)

-0.58 (-3.53)

-0.41 (-2.10)

-0.52 (-2.99)

-0.62 (-3.82)

-0.43 (-2.52)

-0.28 (-1.59)

-0.23 (-1.06)

% European

-0.64 (-3.54)

-0.48 (-2.71)

-0.64 (-3.57)

-0.59 (-3.31)

-0.45 (-2.42)

-0.70 (-3.70)

-0.33 (-2.36)

-0.32 (-2.33)

0.49 (2.38)

0.60 (3.00)

0.52 (2.53)

0.64 (3.11)

0.64 (3.10)

0.61 (2.89)

0.34 (1.94)

0.41 (1.99)

% Dominant inv. % Early stage

0.24 (0.93)

% Health

-0.21 (-0.82)

% Hi-tech

0.30 (1.46)

Yield entry – VC

-0.08 (-1.77)

0.10 (0.61)

0.04 (0.24)

Yield entry – BO

-0.05 (-1.14)

-0.75 (-3.25)

-0.76 (-3.34)

GDP entry - VC

-0.38 (-2.13)

-0.51 (-2.11)

-0.47 (-2.01)

GDP entry – BO

-0.27 (-1.35)

-0.19 (-1.19)

-0.26 (-0.93)

Yield during – VC

-0.13 (-1.82)

-0.25 (-0.95)

-0.12 (-0.45)

Yield dur ing – BO

-0.10 (-1.42)

1.47 (4.89)

1.53 (5.05)

GDP during – VC

1.98 (1.52)

6.32 (3.51)

7.02 (3.50)

GDP during – BO

2.09 (1.59)

-3.77 (-2.07)

2.66 (-1.18)

Mkt during – VC

0.10 (0.83)

-0.04 (-0.42)

-0.05 (-0.33)

Mkt during – BO

0.20 (1.63)

0.21 (1.41)

0.19 (1.23)

E/P exit – VC

-0.19 (-2.96)

-0.09 (-0.66)

-0.09 (-0.69)

E/P exit – BO

-0.16 (-2.36)

-0.42 (-2.47)

-0.44 (-2.56)

11.3%

20.8%

23.8%

Adj. R-square

10.3%

9.9%

9.6%

10.2%

44

9.6%

Panel B: Systematic risk and performance (IRR) 1

2

3

4

5

6

Liquidated

1.39 (0.68)

0.06 (0.03)

0.16 (0.53)

0.14 (0.20)

0.07 (0.04)

0.18 (0.09)

Size

12.82 (2.88)

10.91 (2.67)

12.11 (2.69)

11.25 (2.74)

11.21 (2.70)

10.84 (2.59)

Size-squared

-1.07 (-2.15)

-0.84 (-1.89)

-0.91 (-1.89)

-0.87 (-1.97)

-0.85 (-1.89)

-0.81 (-1.78)

0.50 (0.35)

-0.74 (-0.47)

Committed year Deal length

-14.66 (-4.00)

-10.64 (-2.99)

-12.54 (-4.09)

-10.46 (-2.73)

-11.10 (-2.89)

-11.12 (-2.83)

% European

-9.47 (-2.43)

-9.62 (-2.48)

-9.69 (-2.29)

-8.74 (-2.23)

-9.81 (-2.51)

-9.04 (-2.30)

9.20 (1.91)

13.63 (3.16)

11.03 (2.46)

11.88 (2.68)

14.97 (3.04)

14.56 (2.92)

% Early stage

-1.45 (-0.26)

-0.33 (-0.06)

% Health

-1.53 (-0.27)

-2.05 (-0.35)

% Hi-tech

7.19 (1.63)

8.52 (1.90)

% Dominant inv.

Beta – VC

-3.41 (-0.60)

Beta – BO

0.61 (0.13)

Marked Ind. – VC

0.50 (3.05)

0.48 (2.88)

Marked Ind. – BO

0.58 (2.97)

0.62 (3.07)

Marked Ind.L - VC

-0.10 (-1.15)

Marked Ind.L - BO

0.44 (1.61)

Marked mkt – VC

0.61 (1.88)

0.84 (2.27)

Marked mkt – BO

0.70 (2.18)

0.97 (2.70)

Adj. R-square

13.6%

15.1%

13.2%

45

13.6%

15.6%

14.9%

Panel C: Idiosyncratic risk and performance (PI) 1

2

3

4

5

6

7

Liquidated

0.03 (0.37)

0.03 (0.27)

0.02 (0.19)

-0.01 (-0.07)

-0.01 (-0.02)

0.02 (0.20)

0.06 (0.59)

Size

0.35 (1.76)

0.49 (2.46)

0.35 (1.79)

0.48 (2.44)

0.47 (2.40)

0.40 (2.05)

0.49 (2.47)

-0.03 (-1.27)

-0.04 (-1.81)

-0.03 (-1.27)

-0.04 (-1.91)

-0.04 (-1.89)

-0.03 (-1.65)

-0.04 (-1.94)

Size-squared Committed year

0.15 (2.28)

Deal length

-0.56 (-3.35)

-0.56 (-3.35)

-0.65 (-3.96)

-0.57 (-3.50)

-0.60 (-3.69)

-0.61 (-3.74)

-0.49 (-2.75)

% European

-0.49 (-2.80)

-0.56 (-3.18)

-0.49 (-2.79)

-0.55 (-3.16)

-0.54 (-3.09)

-0.53 (-3.03)

-0.52 (-2.99)

0.74 (3.85)

0.50 (2.32)

0.76 (4.06)

0.57 (3.04)

0.61 (3.28)

0.67 (3.66)

0.60 (2.68)

% Dominant inv. % Early stage

0.21 (0.81)

% Health

-0.28 (-1.07)

% Hi-tech

0.15 (0.72)

Nber take - VC

0.01 (1.10)

Nber take – BO

-0.00 (-0.23)

Nber deals - VC

-0.16 (-2.00)

Nber deals – BO

-0.18 (-1.82)

Herfindahl stage

0.61 (2.06)

Herfindahl ind.

0.61 (2.46)

% dominant ind.

0.55 (2.16) 0.51 (2.15)

Volatility Adj. R-square

-0.27 (-0.48) 9.4%

9.9%

9.8%

10.2%

46

10.0%

10.5%

11.7%

Panel D: Fund characteristics and performance (PI) 1 – 93-

2 – 93-

3 – 93-

4 – 93-

5 – 97-

6 – 97-

7 – 97-

8 – 97-

Liquidated

0.13 (1.38)

0.14 (1.40)

0.13 (1.37)

0.04 (0.49)

0.12 (1.01)

0.12 (1.02)

0.11 (0.93)

0.02 (0.20)

Size

0.52 (2.63)

0.53 (2.67)

0.53 (2.64)

0.53 (2.66)

0.44 (2.14)

0.45 (2.20)

0.44 (2.15)

0.45 (2.18)

-0.05 (-2.21)

-0.05 (-2.24)

-0.05 (-2.21)

-0.05 (-2.16)

-0.04 (-1.79)

-0.04 (-1.81)

-0.04 (-1.87)

-0.04 (-1.82)

Size-squared Committed year

0.15 (2.28)

0.22 (3.25)

Deal length

-0.22 (-1.11)

-0.22 (-1.10)

-0.22 (-1.11)

-0.47 (-2.61)

-0.44 (-2.05)

-0.42 (-1.98)

-0.47 (-2.21)

-0.63 (-3.26)

% Venture

-0.16 (-0.67)

-0.14 (-0.60)

-0.15 (-0.66)

-0.13 (-0.53)

0.02 (0.08)

0.04 (0.20)

0.02 (0.11)

0.04 (0.18)

% European

-0.64 (-3.49)

-0.65 (-3.54)

-0.64 (-3.48)

-0.55 (-3.11)

-0.51 (-3.09)

-0.51 (-3.09)

-0.56 (-3.39)

-0.47 (-2.93)

Herfindahl ind.

0.44 (1.72)

0.45 (1.73)

0.44 (1.71)

0.50 (1.95)

0.37 (1.38)

0.40 (1.47)

0.41 (1.50)

0.49 (1.82)

% Health

0.18 (-0.68)

-0.20 (-0.75)

-0.19 (-0.70)

-0.16 (-0.68)

-0.06 (-0.21)

-0.07 (-0.26)

-0.14 (-0.51)

-0.16 (-0.60)

% Hi-tech

0.31 (1.39)

0.31 (1.43)

0.31 (1.38)

0.23 (1.04)

0.67 (2.93)

0.67 (2.96)

0.59 (2.60)

0.58 (2.61)

% Dominant inv.

0.47 (1.96)

0.46 (1.96)

0.47 (1.96)

0.53 (2.26)

0.32 (1.33)

0.30 (1.25)

0.34 (1.45)

0.39 (1.69)

0.33 (3.04)

0.38 (3.47)

st

1 time

-0.02 (-0.17)

2nd time

-0.07 (-0.70) -0.11 (-1.05)

Experienced Vintage dummies Adj. R-square

-0.21 (-2.00) 0.02 (0.17)

0.08 (0.75)

YES

YES

YES

NO

YES

YES

YES

NO

14.3%

14.5%

14.3%

11.8%

16.4%

16.9%

17.5%

14.4%

47

Table VI: Fund performance This table reports aggregate performance statistics of PE funds. In Panel A we account for residual values according to Table III whereas in Panel B we write them off. We report performance for funds raised before 1993 (Sample1; 983 observations) and before 1997 (Sample5; 1456 observations). The performance measure we use is the profitability index based on actual cash flows and accounting values as of June 2003. We discount inflows with the CRSP value weighted index, and outflows with the risk-free rate (T-bill 30-days). We report performance percentiles (25%, 50%, and 75%) as well as the value-weighted (VW) and equally-weighted (EW) profitability index in each case.

Panel A: With Residual Values Sample1: 93-

Sample5: 97-

All

Venture

Buyout

All

Venture

Buyout

25%-Percentile

0.42

0.40

0.55

0.51

0.46

0.62

50%-Percentile

0.68

0.63

0.82

0.78

0.73

0.85

75%-Percentile

1.00

0.95

1.09

1.11

1.10

1.12

EW-Average

0.83

0.78

0.95

1.02

1.05

0.96

VW-Average

0.95

0.97

0.97

1.09

1.35

0.98

Nber of observations

983

701

282

1456

963

493

Panel B: Without Residual Values Sample1: 93-

Sample5: 97-

All

Venture

Buyout

All

Venture

Buyout

25%-Percentile

0.35

0.33

0.42

0.31

0.31

0.31

50%-Percentile

0.62

0.58

0.75

0.59

0.59

0.59

75%-Percentile

0.94

0.90

1.03

0.96

0.97

0.94

EW-Average

0.76

0.72

0.86

0.83

0.88

0.72

VW-Average

0.87

0.87

0.87

0.77

1.12

0.63

48

Table VII: Performance of young funds This table compares the young funds, i.e. those with vintage years from 1997 to 2001, to mature funds (vintage years before 1997) at the corresponding stage of their life. For example, the situation of funds raised in 1997 (six years old as of June 2003) is compared to the situation of mature funds in June of their sixth year. We report, for young funds, the number of funds and the sum of the committed capital for each vintage year. Then the following ratios are displayed for both young and mature funds: total capital called (called) to total capital committed (com.), total cash distributed (dist) to called, and total residual value (book) to called. Finally, two profitability indices are reported: one that writes off accounting values (PIc-r-wo) and one that converts them on a one-to-one basis (PIc-r-a0). Discount rates are the same as in Table VI.

Young funds of these vintage years Vintage year

Nber

(age)

Mature funds at this age

Capital

Called

Dist.

Book

PI

PI

Called

Dist.

Book

PI

PI

com.

to

to

to

c-r-a0

c-r-wo

to

to

to

c-r-a0

c-r-wo

(billion)

com.

called

called

com.

called

called

1997 (6 y)

177

60

0.30

0.65

0.70

138

69

0.92

0.91

0.85

93

54

1998 (5 y)

202

91

0.42

0.38

0.58

136

50

0.88

0.68

0.95

90

44

1999 (4 y)

229

92

0.30

0.21

0.64

122

19

0.81

0.45

1.02

86

31

2000 (3 y)

248

134

0.24

0.09

0.69

105

10

0.69

0.26

1.00

82

19

2001 (2 y)

114

73

0.20

0.04

0.85

109

5

0.51

0.15

0.96

82

12

49

Table VIII: Fund performance - Net of systematic risk: Venture, buyout, and all funds This table displays profitability indices of funds when discount rates are adjusted by the CAPM-Beta reported in table IV. We present the results for all funds in Panel A, for venture funds in Panel B, and buyout funds in Panel C. We discount inflows (to the investor) with the CRSP value-weighted index adjusted for systematic risk (according to the CAPM) and outflows with the Treasury bill (30-days) rate. We also assume that all residual values as of June 2003 translate to immediate inflows according to the matrix displayed in Table III. We report results for fove adjustments for systematic risk. The base scenario is when beta for each fund is assumed to be 1. The other four cases correspond to the four betas reported in Table IV. We report five statistics for each profitability index. The 25th , 50th , and 75th percentiles, and the equally-weighted (EW) and value-weighted (VW) averages. The sample is Sample3 (see Table VI).

Panel A: All funds base

beta1

beta2

beta3

beta4

th

0.46

0.51

0.43

0.49

0.47

th

50

0.70

0.80

0.66

0.77

0.71

75th

1.09

1.19

1.08

1.14

1.19

EW-Mean

1.09

1.00

0.98

1.02

1.06

VW-Mean

1.18

1.09

1.06

1.12

1.11

Profitability Index

base

beta1

beta2

beta3

beta4

25th

0.42

0.48

0.42

0.44

0.46

50th

0.65

0.75

0.65

0.70

0.70

th

1.04

1.14

1.04

1.08

1.16

EW-Mean

1.07

0.97

0.97

1.01

1.09

VW-Mean

1.45

1.27

1.27

1.43

1.41

percentiles

Profitability Index 25

percentiles

Panel B: Venture funds

75

Panel C: Buyout funds base

beta1

beta2

beta3

beta4

th

0.55

0.69

0.47

0.63

0.47

th

50

0.78

0.93

0.69

0.88

0.72

75th

1.12

1.19

1.12

1.17

1.22

EW-Mean

1.11

0.99

0.97

1.07

1.00

VW-Mean

0.99

0.83

0.83

0.96

0.91

percentiles

Profitability Index 25

50

Table IX: Fund performance – Sensitivity analysis This table reports the sensitivity of overall performance to sample selection and performance measurement. Panel A shows how the overall profitability index (same PI as in Table VII) changes as we change samples. We consider a sample of fully liquidated funds (Sample4), a sample of funds raised before 1993 and that have a residual value to capital invested (RV/CI) ratio below 10%, a sample of funds raised before 1997 and that have an RV/CI ratio below 10%, funds raised before 1993 (Sample2), funds raised before 1997 (Sample5), funds raised before 1993 for whic h we have more than 50% of their investments (Sample1), and funds raised before 1997 for which we have more than 50% of their investments (Sample3). Panel B reports the coefficients of correlation between different measures of fund performance and aggregate performance statistics for each of them. The performance measures we report include the internal rate of returns reported by Venture Economics (IRRve) and several profitability indices (PI) based on actual cash flows and accounting values. To compute PIc-r-wo, PIc-r-a0, and PIc-r-c1, we discount inflows with the CRSP value-weighted index and outflows with the risk-free rate (T-bill 30 days). In contrast, to compute PIs-s-wo, PIs-s-a0, and PIs-s-c1, we discount both inflows and outflows with the S&P 500 index. In addition, PIc-r-wo and PIs-s-wo assume that accounting values are written off (wo), PIc-r-a0 and PIs-s-a0 treats the accounting values reported in June 2003 as immediate cash inflows of the same amount (a0), and PIc-r-a1 and PIss-a1 assume that the accounting values reported in June 2003 translate to immediate inflows according to the results shown in Table III (a1). Panel B includes all funds raised before 1997. For both Panel A and B, we report performance percentiles (25%, 50%, and 75%) as well as the value-weighted (VW) and equally-weighted (EW) average performance.

51

Panel A: Sensitivity to sample selection Liquidated

93-, low RV

97-, low RV

93-

97-

93-,I

97-,I

25%-Percentile

0.43

0.44

0.46

0.42

0.51

0.47

0.54

50%-Percentile

0.70

0.71

0.72

0.68

0.78

0.74

0.81

75%-Percentile

1.03

1.04

1.05

1.00

1.11

1.05

1.20

EW- Mean

0.85

0.85

0.88

0.83

1.02

0.93

1.10

VW- Mean

1.01

0.99

1.01

0.95

1.09

1.04

1.20

Nber obs.

365

568

602

983

1456

491

688

Panel B: Sensitivity to performance measure IRRve 1.00

PIc-r-wo 0.81

PI c-r-a0 0.83

PI c-r-a1 0.83

PI s-s-wo 0.81

PI s-s-a0 0.83

PI s-s-a1 0.83

PIc-r-wo

0.81

1.00

0.96

0.96

0.99

0.93

0.93

PI c-r-a0

0.83

0.96

1.00

1.00

0.95

0.99

0.99

PI c-r-a1

0.83

0.96

1.00

1.00

0.95

0.99

0.99

PI s-s-wo

0.81

0.99

0.95

0.95

1.00

0.94

0.94

PI s-s-a0

0.83

0.93

0.99

0.99

0.94

1.00

1.00

PI s-s-a1

0.83

0.93

0.99

0.99

0.94

1.00

1.00

25%-Percentile

1.80

0.31

0.50

0.51

0.38

0.64

0.67

50%-Percentile

9.85

0.59

0.77

0.80

0.73

1.01

1.04

75%-Percentile

19.79

0.96

1.11

1.14

1.16

1.42

1.45

EW- Mean

16.89

0.82

1.01

1.04

1.01

1.29

1.32

VW- Mean

17.78

0.77

1.06

1.11

0.95

1.37

1.43

IRRve

52

Graph 1

25 years of capital raising by US and European Private Equity Funds (Total of 783 billions, all figures are in 2002 million US dollars) 160000 140000 120000 100000 80000 60000 40000 20000 0 1978

1980

1982

1984

1986

1988

1990

53

1992

1994

1996

1998

2000

2002