Private Equity Fund of Funds vs. Funds – APerformance Comparison

Nathalie Gresch and Rico von Wyssa

October 13, 2010

Based on a comprehensive sample of 1641 funds this article investigates

the performance of private equity fund of funds versus direct fund in-

vestments. On a risk adjusted basis, fund of funds outperform the ag-

gregated direct funds. When separated into categories such as buyout,

venture and fund of funds, buyout funds exhibit the most attractive risk

return profile.

Analyzing how fund performance depends on macroeconomic variables,

direct funds generate procyclical returns: returns increase with high

public market performance and economic growth as well as declining

corporate bond yields. For fund of funds we cannot observe such a

pattern.

JEL classification: G11, G12

Keywords: Private Equity Fund, Fund of Funds, Performance

a Swiss Institute of Banking and Finance, University of St. Gallen, Switzerland.

We kindly thank SAM for providing the data.

Address for correspondence: Swiss Institute of Banking and Finance, University of St. Gallen, Rosen-

bergstrasse 52, CH-9000 St. Gallen, Switzerland. Email : heinrich.vonwyss@unisg.ch.

1 Introduction

In the last decade fund of funds have become one of the most important investor groups in

private equity funds. According to Preqin (2010), their share of total capital contributions

to direct private equity funds amounted to as much as 22% in 2009.1 Despite the increasing

importance of private equity fund of funds the risk and return characteristics of this fund

category are not yet fully understood. Due to the scarcity of data, the academic literature

on this topic, especially empirical research based on real fund of fund data, is rare. The

article closest to ours is Weidig and Mathonet (2004), who emphasize the positive return

characteristics of fund of funds. They use a sample of 1027 direct funds and a simulation

approach to obtain fund of fund returns and find that fund of funds, as a managed portfolio

of twenty funds, offer significant diversification effects relative to direct fund investments.

Furthermore, they report a similar risk profile of fund of funds with respect to a public

market index, with no total losses and a symmetric distribution of returns.

Most articles study risk and return characteristics as well as performance drivers of

direct fund investments. Ljungqvist and Richardson (2003) provide the first analysis of

private equity returns based on actual cash flows by looking at returns of large institutional

investors that invested into 73 funds between 1981 and 1993. They report that it takes

over eight years for the internal rates of return (IRRs) to turn positive and more than

ten years for private equity returns to exceed public equity returns. In addition, they

document an out performance of private equity relative to public market returns of 5%

or more per annum. Superior performance of private equity investments, expressed by

significant positive Jensen’s alphas, is also reported by Groh and Gottschalg (2006), who

use a sample of 199 cash flows from 133 transactions completed in the US between 1984

and 2004. Ick (2006) uses a data set containing information on 243 funds spanning from

1975 to 2003. He finds marginal out performance of private equity investments on a gross

of fee basis. Kaplan and Schoar (2005), who study returns of more than 746 funds of

vintage years 1980 to 2001, find that besides large heterogeneity across funds, private

equity investments earn returns (net of fees) approximately equal to the S&P 500. In

addition they document that the main drivers of performance are past fund performance

and fund size. Furthermore, they provide evidence for pro cyclical fund performance.

Phalippou and Gottschalg (2009) use a later version of the data set examined by Kaplan

and Schoar (2005), which includes 852 funds raised between 1980 and 2003. They report

1Compare The 2010 Preqin Private Equity Fund of Funds Review, p. 1.

1

average gross performance of 3% per annum above the S&P 500, but a net of fees under

performance of 3% per annum. Furthermore, they find that adjusting for risk decreases

performance by about 3% per year, resulting in a net of fees alpha of -6% per year. This

shows the importance of fees when measuring fund performance. Driessen et al. (2008),

by analyzing 958 mature funds of vintage 1980 to 1993, find a negative alpha and a high

CAPM beta for venture funds as well as a lower beta and a slightly positive but statistically

insignificant alpha for buyout funds.

While the authors above mainly compare private equity fund returns to public market

returns, the following articles concentrate on determinants of fund returns. Diller and

Kaserer (2007) use a dataset of 200 mature European private equity funds of vintage 1980

to 2003 and show that fund flows and general partners’ skills have a significant impact

on performance. In addition, they find a negative impact of the stock market return in

the vintage year of the fund on its final return. Moreover, fund returns are found to

be negatively correlated with the growth rates of the economy as a whole and unrelated

to public market returns. In contrast, Aigner et al. (2008), by using data on 358 funds

raised between 1974 and 2007, find that strong stock market and GDP growth during the

fund’s lifetime positively affect fund returns. Furthermore, they support the case for out

performance of experienced private equity funds and document that high interest rates

during the fund’s lifetime and high commitment volumes have an adverse impact on fund

returns. The approach in our article to explain fund returns is very similar to that of

Phallipou and Zollo (2005), who use a dataset of 705 funds raised between 1980 and 2003.

They provide evidence for positive correlation between fund performance with both the

business cycles and public stock markets. In addition they attribute low average fund

performance to weak returns of small and inexperienced funds. In summary, the findings

in previous literature, both concerning relative private equity performance and drivers of

returns, are rather divergent.

Our article extends the existing literature for the following reasons. First, the analysis

is based on real fund of fund data as opposed to simulated fund of fund returns. Second,

the data set includes detailed cash flow information which allows for the calculation of a

number of different performance measures and net of fees figures. This has the positive

effect that modeling of the missing values (e.g. fees) is redundant. Finally, as most of the

current research focuses on final IRR values, in this report the investigation is extended

to the full IRR structure, which allows for a more detailed performance comparison.

By using a data set provided by Preqin, we examine the following issues. First, we

2

investigate the performance of private equity fund of funds and relate it to the performance

of direct fund investments. Performance measures used for the evaluation besides total

value to paid-in (TVPI) and final IRRs are duration and some characteristics of the IRR

structure, such as the time it takes until the IRR turns positive or the maximum IRR

during the fund’s life. In addition, we compare the performance over different vintage

years. On an aggregate level, fund of funds’ risk and return characteristics are found

to be superior to direct funds, which is mainly due to less dispersion in fund of funds

returns. However, when split up by categories, buyout funds exhibit the most attractive

risk return features. Second, we analyze how macroeconomic conditions at the time of

investments as well as during the fund’s life influence fund performance. Regarding the

latter, we document for direct fund investments pro cyclical behavior with increasing fund

returns in line with rising stock market returns, growing real GDP and decreasing returns

on corporate BAA bonds. Moreover, when corporate bond yields and credit spreads are

high at the time investments are made, fund performance is higher too. Fund of funds

seem to be less prone to the general market development during the fund’s life. The most

significant explanatory variable for those is the real GDP growth rate prior to the start

date, with 52.6% of explained variation in returns.

The remainder of the article is structured as follows. In Section 2 we describe the

sample. Section 3 provides the results of the performance comparison of fund of funds

and direct fund investments. Section 4 investigates the exposure of fund performance to

macroeconomic factors and Section 5 briefly concludes.

2 Data

2.1 Full Sample

We use two main sources of data. Private equity fund cash flow data stems from a database

maintained by Preqin. Additionally, we obtain data on stock performance (MSCI World

Index), real GDP (US, chain weighted gross domestic product), corporate bond yields

(corporate BAA bonds as reported by Moody’s) and treasury bill rates (10 year treasury

bonds) from Bloomberg. The sample from Preqin provides the following data for each

fund: the amount and date of all cash flows (to/from investors), the quarterly net asset

values (NAVs) from 1979 to February 2010 and fund characteristics, such as fund type,

status, fund focus and size. Cash flows are net of fees and include all fee payments to

general partners as well as carried interest. The dataset consists of 1641 funds of which

3

204 are officially liquidated. Figure 1 shows the split-up of the different fund categories as

defined by Preqin.

Figure 1. Fund Categories

This figure illustrates the split-up of the different fund categories. Preqin distinguishes between

24 different fund types, of which Buyout, Venture, Early Stage and Fund of Funds are the largest

groups. Other fund types, representing more than 3% of total number of funds, include Real

Estate, Mezzanine and Distressed Debt. Finally, fund types with less than 3% of the funds are

Expansion (2.3%), Natural Resources (2.3%), Secondaries (2.2%), Balanced (2.1%), Late Stage

(2%), Early Stage: Start-up (1.2%), Early Stage: Seed (1.1%), Infrastructure (0.8%), Special Sit-

uation (0.7%), Co-investment (0.5%), Venture Debt (0.5%), Timber (0.3%), Co-Investment Multi-

Manager (0.2%), Real Estate Co-Investment (0.2%), Real Estate Secondaries (0.2%), Turnaround

(0.2%) and Direct Secondaries (0.1%).

2.3%2.3%

2.2%2.1%

2.0%1.2%

Mezzanine3.6%

Distressed Debt3.5%

Real Estate4.9%

Early Stage8.2%

Fund of Funds8.3% Venture (General)

22.0%

Buyout32.6%

1.1%

As for the geographic distribution, while the largest share of funds (>80%) invest in

the US, 11% of the funds focus on Europe, which leaves 9% concentrating on the rest of

the world. Table 1 summarizes the full sample. Due to the rapid industry growth in the

90s, the earlier years contain relatively fewer fund observations. We refer to “Combined”

as for all the fund types excluding fund of funds.

2.2 Samples for Performance Comparison

Since there is no market value for ongoing investments, accurate performance calculations

are only possible for sufficiently mature funds. Throughout Section 3, two samples of the

data are used. In sample A, we follow the approach of Phallippou and Gottschalg (2009)

4

Table 1. Sample Overview

The sample consists of combined funds (Combined), buyout funds (Buyouts), venture funds (Ven-

tures) and fund of funds (FoFs) raised between 1979 and 2001 (Vintage). All funds (All) are a

combination of 24 different fund types as classified by Preqin, of which Buyout (32.6%), Venture

(22%), Early Stage (10.5%) and Fund of Funds (8.3%) represent the largest share. Combined

stands for all the fund types excluding fund of funds. Out of the 1641 funds, 204 are officially

liquidated. Over 80% of the funds have a geographic focus on the US, 11% are investing in Europe

and 9% concentrate on the rest of the world.

Vintage Combined Buyouts Ventures FoFs All

1979 1 - 1 - 1

1980 3 2 1 - 3

1981 1 - 1 - 1

1982 3 - 2 - 3

1983 3 - 2 - 3

1984 6 3 3 - 6

1985 10 3 5 - 10

1986 11 3 4 1 12

1987 9 4 4 - 9

1988 10 5 3 1 11

1989 11 5 4 - 11

1990 19 6 5 - 19

1991 10 1 4 - 10

1992 24 8 7 - 24

1993 24 11 9 - 24

1994 38 18 9 1 39

1995 36 12 14 2 38

1996 50 21 8 - 50

1997 63 25 20 1 64

1998 89 36 21 4 93

1999 96 33 29 4 100

2000 138 34 53 8 146

2001 90 17 30 14 104

2002 75 27 15 13 88

2003 58 20 12 12 70

2004 98 30 14 11 109

2005 140 54 22 19 159

2006 155 54 27 16 171

2007 136 56 26 20 156

2008 88 32 11 6 94

2009 6 2 3 1 7

2010 6 1 1 - 6

Total 1507 523 370 134 1641

Liquidated 200 74 64 4 204

5

as well as Driessen and Phallipou (2008), which means that funds that have reached their

normal liquidation date, i.e., funds that are older than ten years, are included. As this

only leaves 15 FoFs, we use a second sample B that consists of funds that have either

been officially liquidated or were started before 2004 (i.e., funds that are older than six

years). We eliminate funds with no cash flow activities over the last six quarters of the

observation period as in Kaplan and Schoar (2005). This leaves 41 FoFs that we consider

to effectively be liquidated. In order to compare direct fund and FoFs performance, we

include in both samples only those direct funds, that are of the same vintage year as the

corresponding FoFs. We display the two subsamples A and B in Table 2.

Table 2. Sample for Performance Comparison

Sample A consists of funds that were either officially liquidated or are older than ten years (normal

fund life). Only funds of vintage years for which fund of funds data is available are included. There

are 343 combined funds (Combined), 132 buyout funds (Buyouts), 100 venture funds (Ventures)

and 15 fund of funds (FoFs) of vintage years 1986-1999 and 2005 which fulfill the criteria. Sample

B includes funds that were either officially liquidated or that are of vintage younger than 2004 and

that did not have any cash flow activities over the last six quarters. Again, funds are only included

when there are fund of funds in the same vintage year. 481 combined funds, 144 buyout funds,

153 venture funds and 41 fund of funds raised between 1986 and 2005 meet the requirements.

Sample A Sample B

Vintage Combined Buyouts Ventures FoFs All Combined Buyouts Ventures FoFs All

1986 11 3 4 1 12 11 3 4 1 12

1988 10 5 3 1 11 10 5 3 1 11

1994 38 18 9 1 39 33 15 7 1 34

1995 36 12 14 2 38 28 9 11 2 30

1997 63 25 20 1 64 54 21 18 1 55

1998 89 36 21 4 93 67 23 19 2 69

1999 96 33 29 4 100 72 22 24 2 74

2000 - - - - - 87 16 39 6 93

2001 - - - - - 50 7 16 10 60

2002 - - - - - 33 12 6 8 41

2003 - - - - - 36 11 6 6 42

2005 - - - 1 1 - - - 1 1

Total 343 132 100 15 358 481 144 153 41 522

6

2.3 Samples for Explaining Internal Rates of Return

The samples for the regression analysis and IRR patterns include the same direct funds

and FoFs as A and B. However, since that section aims at explaining the IRRs rather

than comparing performance measures across different fund types, we also include direct

funds of vintage years for which no data on FoFs in available. Table 3 gives an overview

of these two enlarged samples A+ and B+.

2.4 Possible Biases

Due to the nature of the sample, a number of biases could arise. First, FoFs returns

possibly include low performance funds in which they have invested, whereas some of

those funds might choose not to report to Preqin. Therefore the Preqin return data for

direct funds might be slightly upwards biased as compared to data on FoFs. Second,

the FoFs in sample B are relatively younger than Combined, Buyout or Venture funds.

While over 50% of Combined, Buyouts and Ventures are of vintage earlier than 2000, the

percentage of FoFs is only 24%. As net asset values usually grow over time, this could

lead to poorer results for FoFs. Third, as the data are self reported they are potentially

subject to selection biases. Finally, calculations are based on realized as well as unrealized

investments, which due to subjective accounting treatment of net asset values possibly

introduces further biases. Driessen et al. (2008) document that net asset values reported

by inactive funds are highly upward biased. This could affect performance measures for

both samples.

3 Performance Comparison

Rather than providing an analysis of absolute returns or relating private equity to public

market returns, we want to compare the performance of different types of private equity

funds. The following section describes the three different return measurements, total value

to paid-in, internal rate of return and fund duration, used in the performance comparison

between direct funds and FoFs.

3.1 Methodology

The Total Value to Paid-In (TVPI), also called “multiple”, is calculated as the ratio

between the funds’ distributions and its contributions. Grabenwarter and Weidig (2005)

stress the fact that this definition can only be applied to funds that have been liquidated

7

Table 3. Sample for Regression Analysis and IRR Patterns

Sample A+ consists of funds that were either officially liquidated or are older than ten years (normal

fund life). There are 524 combined funds (Combined), 196 buyout funds (Buyouts), 157 venture

funds (Ventures) and 15 fund of funds (FoFs) of vintage years 1979-2005 which fulfill the criteria.

Sample B+ includes funds that were either officially liquidated or that are of vintage younger than

2004 and that did not have any cash flow activities over the last six quarters. 641 combined funds,

201 buyout funds, 209 venture funds and 41 fund of funds raised between 1979 and 2005 meet the

requirements.

Sample A+ Sample B+

Vintage Combined Buyouts Ventures FoFs All Combined Buyouts Ventures FoFs All

1979 1 - 1 - 1 1 - 1 - 1

1980 3 2 1 - 3 3 2 1 - 3

1981 1 - 1 - 1 1 - 1 - 1

1982 3 - 2 - 3 3 - 2 - 3

1983 3 - 2 - 3 3 - 2 - 3

1984 6 3 3 - 6 6 3 3 - 6

1985 10 3 5 - 10 10 3 5 - 10

1986 11 3 4 1 12 11 3 4 1 12

1987 9 4 4 - 9 9 4 4 - 9

1988 10 5 3 1 11 10 5 3 1 11

1989 11 5 4 - 11 11 5 4 - 11

1990 19 6 5 - 19 19 6 5 - 19

1991 10 1 4 - 10 10 1 4 - 10

1992 24 8 7 - 24 24 8 7 - 24

1993 24 11 9 - 24 22 9 9 - 22

1994 38 18 9 1 39 33 15 7 1 34

1995 36 12 14 2 38 28 9 11 2 30

1996 50 21 8 - 50 36 16 8 - 36

1997 63 25 20 1 64 54 21 18 1 55

1998 89 36 21 4 93 67 23 19 2 69

1999 96 33 29 4 100 72 22 24 2 74

2000 - - - - - 87 16 39 6 93

2001 4 - 1 - 4 50 7 16 10 60

2002 - - - - - 33 12 6 8 41

2003 1 - - - 1 36 11 6 6 42

2004 2 - - - 2 2 - - - 2

2005 - - - 1 1 - - - 1 1

Total 524 196 157 15 539 641 201 209 41 682

8

at the time of calculation. To include also non liquidated funds, the funds net asset value

serves as a proxy for future cash flows. This means that for non liquidated funds, the

TVPI is the ratio of the funds’ distributions plus the funds net asset value to its total

contributions. While the TVPI reflects the effectiveness of the investment with regard to

returning money to the investor, it completely disregards the time dimension as nominal

cash flow are used. This means that a TVPI of 2 does not declare whether the investor

doubles his money within two or within ten years. In this paper, we calculate TVPI for

all funds for which we could obtain an IRR. Since TVPI neglects the time dimension,

explanatory power increases when it is known over what time span the return is achieved.

Thus, average fund life for each of the categories as well as per vintage year is calculated

as well.

The Internal Rate of Return (IRR) is according to Grabenwarter and Weidig (2005)

the most widely used performance measure in private equity. It is the annualized effective

compounded rate of return that can be earned on the invested capital. As opposed to the

TVPI, the IRR describes how time efficient the fund has invested by considering discounted

instead of nominal cash flow. This means that the shorter the investment period for a

profitable investment, the higher the IRR. In addition to the dependence on time, the

IRR is money weighted, meaning that larger amounts account for a bigger part of the IRR

value.

Mathematically, the IRR is the discount rate that makes the net present value of all

cash flows equal to zero,

0 =∑t

CFt

(1 + IRR)t, (1)

where CF is the cash flow at time t and includes all capital contributions, capital distri-

butions as well as the last reported net present value of the respective fund.

Besides calculating IRRs by using all cash flows from the first contribution date until

the evaluation date, also IRRs where only part of the investment’s life is considered can

be computed. In this article, for each fund (if possible) the complete term structure of

IRRs is obtained by calculating an IRR for every cash flow date. Consider n cash flows

CFt at time t = 0, . . . , T where t = 0 is the base date. Calculating the IRR at time k

requires all cash flows with t ≤ k. I.e., for each cash flow, a new IRR is calculated by

using all precedent cash flows and including the new cash flow. Following this procedure

for all cash flows up to t = T results in the IRR term structure of a fund.

As the different fund’s cash flows do not occur at equal times from the base date,

9

and therefore the calculated IRRs correspond to different times, it is difficult to compare

results. To overcome this, the IRR term structures are normalized by interpolating values

at equal time steps from the base date. The time steps chosen in this report are quarterly

dates, meaning that for each fund, t = 1 corresponds to 90 days after the base date, t = 2

to 180 days after the base date and so forth. Given the different fund life, the IRR term

structure is interpolated up to 13,000 days (longest fund life). Outside the interval of the

first and last calculated IRR, the nearest calculated IRR is used as the interpolated value.

We apply the normalization to all funds for which more than one distinct IRR exists. In

cases of only one IRR, all points in time are assigned this value.

The Macaulay Duration measures the sensitivity of an asset’s price to interest rate

movements. Analogously to Phallipou and Gottschalg (2009) we calculate the duration

for a fund as the difference between the duration of distributions and contributions i.e.,

distributions and contributions are treated separately with opposite sign.

The distributions are the positive cash flows CF+ to the fund, whereas contributions

are the negative cash flows CF− from the fund. The calculation steps are as follows: First,

we calculate the present value PV ± of all the cash flows:

PV ± =∑t

CF±t

(1 + IRRt)t, (2)

where t is the time (in years) of the respective cash flow since the base date. We obtain

the time weighted present value by multiplying each cash flow with its t:

TWPV ± =∑t

t · CF±t

(1 + IRRt)t. (3)

Subsequently, the duration of the distributions Dur+ and the duration of the contributions

Dur− are computed as:

Dur± =TWPV ±

PV ± . (4)

Following Phallipou and Gottschalg (2009), the funds’ Duration is given by the difference

Duration = Dur+ −Dur−. (5)

The higher Dur+, the longer it takes until invested money is returned, and the higher

Dur−, the later capital has to be paid in. An investor favors a small Duration (or even

negative) which means a moderate Dur+ and a long Dur−.

The Zero Year represents the number of years it takes for the fund’s IRR to turn

positive. Given an interpolated IRR term structure defined for every time t, this is math-

ematically defined as the first root of the term structure. This date does not necessarily

10

correspond to a time of a cash flow. For some funds, i.e., when the first distribution is very

large, the interpolated IRR is positive from the beginning. In these cases the Zero Year

is the time difference between the first distribution date and the base date. Never Zero

are the number of funds for which the IRR term structure does never turn positive (and

therefore also has no first root). The Final IRR is the last IRR, i.e., the IRR obtained

when considering all capital flows and the last NPV. The Maximum IRR is the largest

IRR observed in the IRR term structure and often coincides with the Final IRR. The Fund

Life is the difference between the date of the last cash flow (usually a distribution) and

the base date. Finally, the Risk Return Ratio of the TVPI is given as either the average

TVPI minus one divided by the standard deviation of the TVPIs and the Risk Return

Ratio of the IRR is calculated as the average IRR divided by the standard deviation of

the IRRs.

3.2 Results

3.2.1 Fund of Fund vs. Combined Fund Performance

Table 4 resumes summary figures both for sample A and sample B. Average TVPI for

funds is with 1.61 and 1.48 for both samples larger than for FoFs with 1.32 and 1.25,

respectively. However, on a risk adjusted basis FoFs show an out performance of roughly

50% with risk return ratios of 0.53 vs. 0.34 for sample A and 0.60 vs. 0.30 for sample B.

Furthermore, while one out of three direct funds ends in a loss (i.e., TVPI smaller than

one), this is true only for one out of four FoFs. In addition, the size of the loss exceeds

the one for FoFs (+4% sample A and +18% sample B). The median TVPI is smaller than

the average TVPI for both samples and fund categories, indicating a negative skew in the

TVPI distribution. The difference between the maximum and minimum TVPI is much

larger for funds than for FoFs, which can partly be explained by the different sample size.

Findings for sample B can also be retraced in the probability distribution (Figure 2) and

the cumulative distribution (Figure 3) of the TVPI. Average fund lives for sample A and

sample B are 10.3 years and 8.48 years for Combined and 10.8 years and 7.23 years for

FoFs. This means that Combined has a longer fund life for sample A (+5%) and a shorter

life for sample B (-15%).

As for the IRR, results for sample A and B are mixed. For the former, the risk return

ratio for Combined is slightly better than for FoFs, which is mainly due to the much higher

average IRR of 7.28% vs. 4.23% for FoFs. However, with a minimum IRR of -73.72% and a

11

Tab

le4.

TV

PI

an

dIR

RS

tati

stic

sfo

rFu

nd

of

Fu

nd

san

dC

om

bin

ed

Fu

nd

s

Th

ista

ble

rep

orts

TV

PI

and

IRR

stat

isti

csfo

r343

dir

ect

fun

ds

(Com

bin

ed)

an

d15

fun

dof

fun

ds

(FoF

s)fo

rsa

mp

leA

as

wel

las

for

481

dir

ect

fun

ds

and

41fu

nd

offu

nd

sfo

rsa

mp

leB

,ra

ised

bet

wee

n1986

an

d2005.

TV

PI

isca

lcu

late

das

the

rati

oof

dis

trib

uti

ons

and

rem

ain

ing

net

asse

tva

lue

toin

vest

edca

pit

al.

Th

ep

rob

ab

ilit

yof

alo

ssis

the

per

centa

ge

of

fun

ds

exh

ibit

ing

TV

PI/

IRR

small

er

than

one/

zero

and

the

aver

age

loss

give

na

loss

isob

tain

edby

sub

tract

ing

the

mea

nof

all

TV

PIs

small

erth

an

on

e/IR

Rs

small

er

than

zero

from

one.

Th

eri

skre

turn

rati

ois

the

aver

age

TV

PI

min

us

on

e/av

erage

IRR

div

ided

by

the

stan

dard

dev

iati

on

(SD

).

Sam

ple

AS

am

ple

B

TV

PI

IRR

[%]

TV

PI

IRR

[%]

FoF

Com

bin

edF

oF

Com

bin

edF

oF

Com

bin

edF

oF

Com

bin

ed

Ave

rage

1.32

1.6

14.3

27.2

81.2

51.4

85.8

46.2

2

Med

ian

1.12

1.3

37.9

37.3

31.1

81.2

58.5

15.7

0

Max

2.67

20.3

823.8

496.8

32.6

720.3

824.0

096.8

3

Min

0.37

0.0

6-2

3.3

8-7

3.7

20.3

70.0

6-2

7.7

1-8

9.3

1

SD

0.60

1.8

112.9

720.7

90.4

21.5

811.3

620.5

5

Pro

b.

ofa

Los

s[%

]26

.67

32.9

426.6

733.2

426.8

036.3

826.8

336.5

9

Ave

rage

Los

s0.

340.3

8-1

2.1

4-1

2.1

40.1

80.3

6-8

.60

-12.4

5

Ris

kR

etu

rnR

atio

0.53

0.3

40.3

30.3

50.6

00.3

00.5

10.3

0

12

probability of loss of 33.24% as well as a negatively skewed distribution, it is not conclusive

whether Combined show a superior return profile. For sample B, the return distribution

for FoFs with less standard deviation, lower probability of loss, lower average loss and

higher risk return ratio looks more attractive, as documented in Table 4 and Figure 2 and

3. Due to the biases explained in Section 2.4, FoFs’ returns are downward biased, which

suggests that the out performance of FoFs in reality is even more pronounced.

Duration for FoFs is 0.62 for sample A and 0.99 for sample B. This is smaller than for

Combined with 1.50 and 1.49 respectively. The lower Duration for FoFs can be explained

by the higher Dur−, meaning that FoFs investors on average need to pay in capital at a

later stage than investors of direct funds. In summary, based on TVPI, IRR and Duration,

on an aggregate level, FoFs clearly outperform direct fund investments.

3.2.2 Fund of Fund Performance vs. Venture Fund and Buyout Fund Perfor-

mance

In this section we describe the findings for sample B, however, unless otherwise stated,

results for sample A are qualitatively the same. Table 5 sheds light on the return charac-

teristics for FoFs, Buyouts and Ventures.

Table 5. TVPI and IRR Statistics for Fund of Funds, Buyout Funds and Venture

Funds

This table reports TVPI and IRR statistics for 481 Combined, 41 FoFs, 153 Ventures and 144

Buyouts raised between 1986 and 2005. TVPI is calculated as the ratio of distributions and

remaining net asset value to invested capital. The probability of a loss is the percentage of funds

exhibiting TVPI/IRR smaller than one/zero and the average loss given a loss is obtained by

subtracting the mean of all TVPIs smaller than one/IRRs smaller than zero from one. The risk

return ratio is the average TVPI minus one/average IRR divided by the standard deviation (SD).

TVPI IRR [%]

FoF Buyout Venture FoF Buyout Venture

Average 1.25 1.62 1.43 5.84 11.80 1.11

Median 1.18 1.52 0.94 8.51 11.56 -2.13

Max 2.67 5.45 20.38 24.00 68.71 96.83

Min 0.37 0.18 0.13 -27.71 -28.77 -73.72

SD 0.42 0.74 2.43 11.36 15.05 22.58

Prob. of a Loss [%] 26.80 19.44 52.90 26.83 19.44 53.59

Average Loss 0.18 0.25 0.41 -8.60 -8.63 -13.03

Risk Return Ratio 0.60 0.84 0.18 0.51 0.78 0.05

13

Buyouts exhibit the most attractive risk return characteristics with a mean TVPI of

1.62, a standard deviation of 0.74 and only 19.4% probability of a loss. While the largest

TVPIs (up to 20.38) can be achieved with Ventures, the general risk profile of those

funds, due to the large dispersion appears unfavorable (risk return ratio of 0.18). These

findings suggest that investors seeking exposure to venture capital should consider a FoFs’

approach.

Analogue to TVPI statistics, with an average IRR of 11.8%, probability of loss of

19.44% and a risk return ratio of more than twice the risk return ratio of FoFs, Buyouts

outperform other fund categories with respect to IRR characteristics.

Results for TVPI are also demonstrated in Figure 2 and 3, while IRR returns are

depicted in Figure 4 and 5.

One might argue that since Buyouts are relatively older than FoFs, with 68% vs. 24%

of funds of vintage older than 2000, due to the J-curve effect of private equity returns, this

could explain part of the Buyouts out performance. However, as results are very similar

for sample A, this argument is questionable.

Figure 2. TVPI Distribution for Combined Funds, Fund of Funds, Buyout Funds and

Venture Funds

This figure illustrates the TVPI distribution for 481 Combined Funds, 41 Fund of Funds, 144

Buyout- and 153 Venture Funds raised between 1986 and 2005. Combined Funds compasses

Buyout- and Venture Funds and 184 funds of other types. TVPIs larger than 10 are added up and

displayed in the bin where TVPI is 10.

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10

0%5%

10%

15%

20%

25%

30%

35%

40%

45%

50%

TVPI

Pro

babi

lity

Combined FundsFund of FundsBuyout FundsVenture Funds

empty line

14

Figure 3. Cumulative TVPI Distribution for Combined Funds, Fund of Funds, Buyout

Funds and Venture Funds

This figure illustrates the cumulative TVPI distribution for 481 Combined Funds, 41 Fund of Funds,

144 Buyout- and 153 Venture Funds raised between 1986 and 2005. Combined Funds compasses

Buyout- and Venture Funds and 184 funds of other types.

●

●

●

●

●

●

●

●

●

●

●

●●

●

●

0%10

%20

%30

%40

%50

%60

%70

%80

%90

%

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8

TVPI

Cum

ulat

ive

Pro

babi

lity

● Combined FundsFund of FundsBuyout FundsVenture Funds

Figure 4. IRR Distribution for Combined Funds, Fund of Funds, Buyout Funds and

Venture Funds

This figure illustrates the IRR distribution for 481 Combined Funds, 41 Fund of Funds, 144 Buyout-

and 153 Venture Funds raised between 1986 and 2005. Combined Funds compasses Buyout- and

Venture Funds and 184 funds of other types.

−100 −90 −80 −70 −60 −50 −40 −30 −20 −10 0 10 20 30 40 50 60 70 80 90 100

0%5%

10%

15%

20%

25%

30%

35%

IRR [%]

Pro

babi

lity

Combined FundsFund of FundsBuyout FundsVenture Funds

15

Figure 5. Cumulative IRR Distribution for Combined Funds, Fund of Funds, Buyout

Funds and Venture Funds

This figure illustrates the cumulative IRR distribution for 481 Combined Funds, 41 Fund of Funds,

144 Buyout- and 153 Venture Funds raised between 1986 and 2005. Combined Funds compasses

Buyout- and Venture Funds and 184 funds of other types.

● ● ● ● ● ●●

●

●

●

●

●

●

●●

●● ● ● ●

0%10

%20

%30

%40

%50

%60

%70

%80

%90

%

−100 −90 −80 −70 −60 −50 −40 −30 −20 −10 0 10 20 30 40 50 60 70 80 90

IRR [%]

Cum

ulat

ive

Pro

babi

lity

● Combined FundsFund of FundsBuyout FundsVenture Funds

Regarding the cash flow structure, Dur+ is highest for FoFs, which in combination

with an average Dur− results in the lowest Duration of 1 as compared to 1.4 (+40%) for

Buyouts, 1.5 for Combined (+50%) and 1.7 (+70%) for Ventures. This indicates that

FoFs are leading in terms of cash flow timing.

FoFs and Ventures TVPI figures are in contrast to findings of Weidig and Math-

onet (2004). While they report similar risk return profiles for Buyouts, with numbers for

all TVPI measures lying between numbers of sample A and sample B in Table 5, figures

for Ventures and FoFs are more favorable in their paper. Their Ventures show lower stan-

dard deviations of TVPIs (2.97 in sample A and 2.43 in sample B compared to 1.9), less

probability of a loss (46% in sample A and 52.9% in sample B compared to 30%) and

smaller average loss (over 40% for both samples compared to 29%). As for FoFs, Weidig

and Mathonet (2004) calculate numbers for two different types of FoFs: average numbers

of 50,000 simulated FoFs once on the basis of 300 venture funds (FoF Venture) and once

on the basis of 200 European buyout funds (FoF Buyout). While mean TVPIs (1.7 and

1.8) are higher than for direct funds, standard deviations are much lower (0.2 and 0.5),

resulting in larger risk return ratios of 3.1 for FoF Buyouts and 1.7 for FoF Ventures.

In addition, they report a probability of any loss close to zero for both FoF types. The

16

discrepancy of results could be due to several reasons. First and most important, their

approach is based on simulated FoF returns whereas data in our article is real. Second, the

data on direct funds differ. The sample period of Weidig and Mathonet (2004) goes from

1980 to 1998, while direct funds for sample A are of vintage 1986-1999 and direct funds

for sample B of vintage 1986-2003. Thus, their sample includes the period of 1980-1985

which on average generated high returns for both fund categories and neglects more recent

vintages for which Ventures performed rather poorly. In addition, they only consider Eu-

ropean Buyouts whereas the Preqin sample contains more than 80% American Buyouts.

This should, however, lead to poorer results, as European Buyouts on average perform

worse than US Buyouts.2 Third, Weidig and Mathonet (2004) base their calculation on

gross of fees returns and then deduct management fees, set up costs and carry according

to a simplified calculation. The high FoF performance reported by them could therefore

also be due to an underestimation of fees.

In summary, TVPI and IRR characteristics of FoFs seem to be superior to Combined

and Ventures, but less attractive than Buyouts. In terms of cash flow timing, FoFs exhibit

the most favorable features.

3.2.3 Vintage Year Analysis

While the previous two sections concentrated on summary statistics over the whole sam-

ple period, the emphasis of this section lies on a performance comparison over different

vintage years. However, drawing conclusions for FoFs based on a single vintage year seems

somewhat arbitrary, as for vintages before the year 2000 the maximum amount of FoFs

per vintage year is two. Again, as outcomes for sample A and sample B are very similar,

we discuss only results for sample B.

Table 6 reveals results for different performance measurements for individual vintage

years as well as for the total range. In addition to Final IRR and TVPI, further measures

such as the mean ZeroYear, NeverZero, the maximum IRR and more detailed Duration

figures are shown.

Funds of vintage 1986 performed well by means of TVPI, with Combined, FoFs and

Buyouts generating very high returns and Ventures showing the most attractive risk return

ratios of TVPIs. With the exception of FoFs of vintage 1994, the trend continued for

vintage years 1988 and 1994-1995. Probably due to the effect of the dot com bubble,

performance declined in the subsequent years, with Combined and Ventures reaching the

2Compare Phalippou and Gottschalg (2009), p. 1752.

17

Tab

le6.

Perf

orm

an

ce

Com

pari

son

for

Diff

ere

nt

Vin

tage

Years

Th

ista

ble

rep

orts

TV

PI

and

IRR

stat

isti

csfo

r41

FoF

s,153

Ven

ture

san

d144

Bu

you

tsra

ised

bet

wee

n1986

an

d2005.

Com

bin

edco

mp

ass

esB

uyo

uts

,V

entu

res

and

184

fun

ds

ofot

her

typ

es.

TV

PI

isca

lcu

late

das

the

rati

oof

dis

trib

uti

on

san

dre

main

ing

net

ass

etva

lue

toin

vest

edca

pit

al.

RR

stan

ds

for

risk

retu

rn

rati

oan

dis

given

asth

eav

erag

eT

VP

Im

inu

son

e/av

erage

IRR

div

ided

by

the

stan

dard

dev

iati

on

.T

he

Fu

nd

Lif

eis

the

diff

eren

ceb

etw

een

the

date

of

the

last

dis

trib

uti

onan

dth

ed

ate

ofth

efirs

tco

ntr

ibu

tion

.T

he

mea

nze

roye

ar

isth

eti

me

as

mea

sure

din

years

itta

kes

unti

lth

eIR

Rtu

rns

posi

tive

.F

pV

is

the

nu

mb

erof

fun

ds

per

vin

tage

and

%N

ever

Zer

osh

ows

the

per

centa

ge

of

fun

ds

per

vin

tage

for

wh

ich

the

IRR

nev

ertu

rns

zero

.M

ax

IRR

isth

eh

igh

est

IRR

per

vin

tage

.W

hil

eD

ur+

isth

eD

ura

tion

ofdis

trib

uti

on

s,D

ur−

isth

eD

ura

tion

of

contr

ibu

tion

san

dD

ura

tion

isD

ur+

min

us

Du

r−.

Th

ela

stro

wsh

ows

aver

age

nu

mb

ers

for

all

fun

ds

(not

eth

atth

isis

not

equ

al

toth

eav

erage

of

vin

tage

years

figu

res,

as

that

wou

ldin

trod

uce

aw

eighti

ng).

Mea

nT

VP

IR

RT

VP

IF

un

dL

ife

Mea

nZ

ero

Yea

rN

ever

Zer

o[%

]#

Fp

V

Vin

tage

CF

oFB

VC

FoF

BV

CF

oF

BV

CF

oF

BV

CF

oF

BV

CF

oF

BV

1986

2.10

1.93

2.92

1.35

0.87

NA

1.26

3.2

615.9

121.1

215.6

815.8

28.7

99.8

28.3

69.9

29

00

011

13

419

882.

062.

671.

812.

541.

58N

A1.

841.4

714.7

217.5

114.5

813.7

67.3

15.8

27.4

37.1

60

00

010

15

319

942.

230.

481.

932.

100.

60N

A1.

080.3

811.2

113.1

410.1

911.8

95.0

1N

aN

5.1

55.2

033

100

13

71

33

115

719

952.

531.

501.

463.

570.

405.

930.

670.4

511.4

812.2

611.9

111.4

95.6

58.9

55.0

64.9

318

022

18

28

29

11

1997

1.65

0.37

1.50

1.90

0.55

NA

0.84

0.5

29.9

710.4

010.2

29.5

66.3

8N

aN

7.7

44.3

030

100

24

33

54

121

18

1998

1.56

1.04

1.39

2.06

0.23

0.60

0.80

0.2

49.3

39.0

09.3

89.3

66.9

29.5

67.6

57.1

037

50

26

63

67

223

19

1999

1.02

1.81

1.40

0.68

0.03

23.8

30.

60-0

.77

8.3

17.7

58.7

38.0

27.1

56.8

66.9

.7.9

353

032

79

72

222

24

2000

1.08

1.35

1.82

0.83

0.15

0.84

1.75

-0.7

77.4

37.8

27.8

57.3

46.9

17.2

86.3

78.0

053

33

669

87

616

39

2001

1.31

1.10

1.81

0.99

0.48

0.49

1.00

-0.4

46.5

56.8

46.8

56.5

45.5

46.9

24.9

46.7

536

40

14

56

50

10

716

2002

1.39

1.22

1.59

1.10

0.66

1.12

1.35

0.3

75.3

84.5

05.3

75.6

64.9

74.9

04.8

05.9

421

13

17

17

33

812

620

031.

351.

181.

691.

180.

460.

710.

520.8

84.2

53.7

64.2

23.8

14.1

74.4

14.2

43.8

425

17

18

17

36

611

620

051.

12N

A2.1

31.7

50

1A

ll1.

481.

251.

621.

430.

300.

600.

840.1

88.4

87.2

38.9

48.5

36.1

66.1

96.3

36.3

237

27

19

54

481

41

144

153

Max

IRR

[%]

Mea

nF

inal

IRR

[%]

RR

Fin

al

IRR

Du

rati

on

Du

r+D

ur−

Vin

tage

CF

oFB

VC

FoF

BV

CF

oF

BV

CF

oF

BV

CF

oF

BV

CF

oF

BV

1986

258

256

8.92

7.93

15.2

14.7

71.0

6N

A1.3

94.1

02.0

83.4

41.2

02.4

15.2

85.1

54.9

96.1

73.2

01.7

13.7

93.7

619

8824

2418

2413

.59

23.8

412

.51

17.4

82.3

8N

A3.5

51.9

01.6

11.0

82.1

90.1

85.2

04.9

25.5

54.5

73.5

93.8

43.3

64.3

919

9469

-20

6955

17.5

3-1

9.83

21.5

44.9

60.7

4N

A1.1

50.1

81.2

4-1

.23

1.0

90.4

74.6

27.0

53.9

66.4

03.3

88.2

82.8

75.9

.19

9597

1240

9718

.28

10.4

911

.03

22.9

50.6

94.9

10.6

00.6

71.6

91.0

72.9

90.9

54.5

85.2

25.6

43.8

42.8

94.1

52.6

42.9

019

9796

-23

2496

12.7

9-2

3.38

7.98

20.6

50.5

4N

A0.8

10.6

21.5

00.0

91.1

61.6

44.6

46.7

95.2

54.0

23.1

46.7

04.0

92.3

819

9877

217

442.

520.

665.

11-5

.04

0.1

20.5

50.5

4-0

.20

1.6

1-0

.34

1.6

62.2

94.9

86.4

75.1

95.6

53.3

76.8

13.5

43.3

619

9943

1923

13-2

.86

16.9

94.

63-1

0.5

6-0

.17

5.2

00.3

2-0

.88

1.5

61.3

61.1

12.2

55.5

25.5

85.2

26.3

93.9

64.2

24.1

14.1

420

0029

1929

18-0

.65

7.27

16.1

8-5

.87

-0.0

40.8

41.7

1-0

.52

1.6

60.7

31.1

51.9

35.9

75.3

84.9

36.6

14.3

24.6

53.7

84.6

820

0143

1343

266.

853.

1818

.84

-2.8

20.4

00.4

41.0

9-0

.23

1.1

10.9

71.0

61.4

74.8

85.8

94.2

25.8

73.7

74.9

23.1

54.3

920

0258

2431

1210

.95

9.95

17.5

93.2

70.6

11.2

51.4

20.3

31.4

01.4

21.2

60.9

94.3

54.3

44.1

04.8

52.9

52.9

22.8

43.8

620

0365

2065

1810

.20

5.01

16.3

97.0

90.6

10.3

00.6

60.9

81.1

51.1

41.3

21.0

23.5

03.6

63.7

63.6

42.3

52.5

32.4

42.6

120

059

9.31

NA

0.9

61.6

70.7

2A

ll97

2469

976.

225.

8411

.80

1.1

10.3

00.5

10.7

80.0

51.4

90.9

91.3

91.7

15.0

05.0

74.8

45.6

33.5

14.0

83.4

53.9

2

18

lowest mean TVPIs of 1.02 and 0.68 in 1999, while the through for FoFs and Buyouts was

in 1997 and 1998 respectively. Not surprisingly Buyouts of vintage 2000 performed rather

well, presumably taking advantage of low valuations in the aftermath of the crisis. As can

be observed, eight out of ten times when TVPI for Ventures and FoFs increased/decreased

from the previous vintage year, TVPI for Buyouts moved towards the opposite direction.

This tendency also applies to Final IRR. It appears that changes in the returns for Ventures

and FoFs are inversely correlated to the changes in Buyout returns. From an investor’s

or asset management’s perspective it could thus be prudent to include both fund types

(Buyouts as well as either Ventures or FoFs) to better diversify the investment portfolio.

Column three shows that fund life strongly decreased for all fund categories. While in

1986 average fund life was longer than 15 years, in 1999 it was almost halved to around

eight years. As fund life from vintage year 1995 onwards on average decreases by one with

each vintage year increase, this effect is obviously due to the fact that cash flow data is

only available up to the first quarter of 2010. Surprisingly, FoFs with 7.23 years exhibit

the shortest average fund life.

The mean ZeroYear is on average over all vintages with 6.16 years lowest for Combined,

followed by 6.19 years for FoFs. Mean ZeroYear is generally large for vintages 1986, 1988

and 1998-2000 and decreases substantially over the vintage years 2001-2004.

The average share of NeverZero for all vintages coincides with the probability of a

negative IRR (see Table 5, prob. of a loss), which means that no fund that once crossed

the zero IRR line exhibits a negative final IRR. Obviously, when there are many funds

that never achieve positive IRRs in a vintage year, the mean Final IRR for that year is

low. As one could expect, maximum IRRs were achieved with Ventures of vintages 1995

and 1997.

Most of the time, high TVPIs coincide with high Final IRRs. This is not entirely true

for vintage year 1986, which shows some of the largest TVPIs for Combined, FoFs and

Buyouts but not topmost Final IRRs, which is probably a result of the long fund life for

funds of vintage 1986.

Regarding Duration, we note that since vintage 2000, both Dur+ and Dur− constantly

decreased, indicating that cash flows are being exchanged at a faster rate. Generally,

vintage years 1994 as well as 2001-2003 show the most favorable Durations. FoFs, in most

years have among the largest Dur+. However, as they exhibit very large Dur− as well,

average Duration over the whole sample period is lowest.

All in all, funds of vintage years 2002-2003 seem to perform well, with short mean

19

ZeroYear, low percentage of funds that exhibited a negative Final IRR, decent level of Final

IRRs, low Durations and effectual TVPIs. Regarding mean TVPI, mean Final IRR and

the percentage of NeverZero, except for FoFs in 1988, overall performance for vintage years

1986, 1988 and 1994-1995 is also considerable. As opposed to that, lowest performance is

for all fund categories shown for funds of one of the vintage years between 1997 and 1999,

which is probably due to the effects of the dot com bubble. Since return changes of FoFs

and Ventures seem to be negatively correlated to return changes of Buyouts, it could be

prudent to include both Buyouts and either FoFs or Ventures to the portfolio in order to

better diversify investments.

4 Regression Analysis

4.1 Methodology

In this section, we aim to explain private equity fund returns. Similar to Phalippou and

Zollo (2005), we include the stock market return, real GDP growth rate, corporate BAA

bond yields (CBAA yields) as reported by Moody’s and credit spreads as explanatory

variables. The CBAA yields and the credit spreads, which capture the probability of

default and the expected recovery in case of default in the economy, reflect the cost of

financing buyout investments. Both variables also correspond to different stages in the

business cycle and are thus relevant for all fund categories. High credit spreads and high

CBAA yields are usually found in difficult market environments (i.e., economic recession).

Finally the GDP growth rate and stock market return assess the impact of the market

sentiment in the year the fund first invested (start year) on final fund returns (Phalippou

and Zollo, 2005). In this study, the dependent variables are the average Final IRRs per

Start Year.3 To exploit the macroeconomic impact on different fund categories, the average

IRRs of All, Combined, FoFs, Buyouts and Ventures are explained separately.

We run two different types of regressions, corresponding to the different time periods,

which we refer to as the Fund Life Regression and the Start Year Regression. For the Fund

Life Regression, we calculate returns of the different explanatory variables over the fund’s

life to measure how macroeconomic factors during the funds life influence its returns. For

the Start Year Regression, we calculate factor returns for the year prior to the investment

start. This allows for an evaluation of the impact of the economic environment at start of

3Note that the Start Year is not always the same as the vintage year, i.e. for the total sample, the Start

Year is on average eight months later than the Vintage Year.

20

the fund life on fund returns. First, the average Final IRR as well as the average fund life

for each Start Year are computed. The fund life is thereby set as the difference between

the date of the last transaction (usually a distribution) and the first contribution. This

is calculated for each fund and then averages for all funds of the same Start Year are

computed. Then, for the Fund Life Regression, the real GDP growth rate, MSCI return

and CBAA spread over the average fund life per Start Year are calculated,4 i.e. when

a fund has Start Year 1980 and average fund life for this Start Year is 8.9, returns for

all variables are calculated for the period 1980-1989. For the Start Year Regression, we

take the MSCI return and real GDP growth rate for the year before the Start Year into

account. In addition, we consider the absolute value of the CBAA yield and the credit

spread (calculated as the difference between the CBAA yield and treasury bond yield) at

the end of the year before the Start Year. We repeat this for each Start Year, i.e., when a

fund has Start Year 1980, we consider the MSCI return and real GDP growth rate between

1979 and 1980 and the CBAA yield as well as the credit spread as of December 31, 1979.

In order to evaluate the economic importance of the variables, we standardize them by

subtracting their sample mean and dividing by their sample standard deviation. Finally,

we regress the three Fund Life regression variables and the four Start Year regression

variables against the mean Final IRRs per Start Year of the six different fund categories.

Since the variables exhibit high correlations among each other, we regress them against

the IRRs one at a time. Carrying out the analysis for sample A+ as well as for sample B+

results in 84 single regressions.

4.2 Results

We present the results of the analysis in Table 7. While Panel A reports figures for sample

A+, Panel B shows regression outputs for sample B+.

For Combined, Buyouts and Ventures signs of the coefficients for the Fund Life Re-

gression are with one minor exception all pointing towards the same direction. Funds that

invest during times with high public market performance and economic growth as well

as declining corporate bond yields show higher IRRs. This suggests that private equity

funds generate pro cyclical returns, which is in line with the findings of Phallipou and

Zollo (2005). Coefficients are statistically significant for all variables for Combined and

All in sample B+, as well as for CBAA Spread (Buyouts sample A+) and GDP growth

(Ventures sample A+ and B+). For Ventures, the real GDP growth rate during the time

4The CBAA spread is the CBAA yield at End Date minus the CBAA yield at Start Date.

21

Table 7. Explaining IRRs

This table reports the results of OLS regressions to explain the drivers of the IRR for different fund

categories. The dependent variables are the mean Final IRRs per Start Year of combined funds

(Combined), fund of funds (FoFs), buyout funds (Buyouts), venture funds (Ventures) and all funds

(All). Explanatory variables are per Start Year (year in which first contribution took place): 1)

MSCI return and real GDP growth rate during the average fund life as well as the difference of the

average corporate bond yield (CBAA yield as reported by Moody’s) at the end of the investment

and the Start Year, 2) MSCI return and real GDP growth rate for the year before the Start Year,

as well as average corporate bond yield (CBAA yield as reported by Moody’s) and credit spread

at the end of the year before the Start Year. ∗ denotes significance of the coefficient at a level of

10%, ∗∗ at a level of 5%. Standard deviation of the coefficients are given in parentheses.

Panel A: Explaining IRRs for Sample A+

Fund Life Regression Start Year Regression

IRR MSCI GDP CBAA MSCI GDP CBAA Credit Spread

FoFs -0.017 0.006 0.025 -0.051 -0.115∗∗ 0.035 0.063

(0.060) (0.060) (0.059) (0.057) (0.041) (0.058) (0.055)

Combined 0.010 -0.009 -0.011 -0.016 -0.009 -0.002 0.033∗

(0.018) (0.018) (0.018) (0.018) (0.018) (0.018) (0.017)

Buyouts 0.017 0.020 -0.037∗∗ -0.003 -0.040∗∗ 0.025 0.032∗

(0.018) (0.018) (0.017) (0.019) (0.016) (0.018) (0.017)

Ventures 0.027 0.060∗∗ -0.034 0.023 0.000 0.025 -0.034

(0.026) (0.023) (0.026) (0.026) (0.027) (0.026) (0.026)

All 0.012 -0.006 -0.013 -0.017 -0.010 0.000 0.034

(0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.016)

Panel B: Explaining IRRs for Sample B+

Fund Life Regression Start Year Regression

IRR MSCI GDP CBAA MSCI GDP CBAA Credit Spread

FoFs -0.017 -0.003 0.022 -0.011 -0.061 0.013 0.022

(0.039) (0.039) (0.038) (0.039) (0.034) (0.039) (0.038)

Combined 0.025∗ 0.030∗∗ -0.028∗∗ 0.001 -0.004 0.024∗ 0.005

(0.013) (0.012) (0.012) (0.014) (0.013) (0.013) (0.013)

Buyouts 0.005 0.003 -0.012 -0.005 -0.031∗∗ 0.010 0.027∗

(0.016) (0.016) (0.016) (0.016) (0.015) (0.016) (0.015)

Ventures 0.024 0.048∗∗ -0.031 0.006 -0.010 0.026 -0.026

(0.020) (0.018) (0.019) (0.020) (0.020) (0.020) (0.020)

All 0.026∗∗ 0.031∗∗ -0.029∗∗ 0.000 -0.005 0.025∗∗ 0.005

(0.012) (0.011) (0.011) (0.013) (0.013) (0.012) (0.013)

22

of investment is the only significant factor. Even though none of the Fund Life Regres-

sion variables are statistically significant for FoFs, the signs of coefficients indicate that as

opposed to the other fund categories, FoF returns do not follow public stock markets or

economic growth and can be said to be more robust. This robustness towards the economic

environment could be due to the following reasons. First, a FoF has more diversification

of the end investment, i.e., of portfolio companies which are heavily dependent on the

macroeconomic environment. Second, a FoF has more vintage year diversification than a

direct fund. Both these effects could lead to more stable returns.

In case of the Start Year Regression, the MSCI return one year prior to the first

investment is not statistically significant for any dependent variable and generally only

explains very little of the variation in fund returns. At times with high real GDP growth

previous to the Start Year, IRRs were lower for all categories. For both samples, with

the exception of Ventures, funds that started investing in periods of high corporate bond

yields and high credit spreads, i.e., in a difficult market environment, outperformed. As

opposed to the Fund Life Regression loadings, this contradicts the results of Phallipou and

Zollo (2005). For FoFs, the most significant explanatory variable is the real GDP growth

rate prior to the Start Date, with significant coefficient and 52.6% of explained variation

(R square) in sample A+. As opposed to that for Combined, for sample A+ the credit

spread and for sample B+ the CBAA yield at the Start Date are significant. As expected

for Buyout returns, the growth rate of the economy one year prior to the Start Year is a

very important signal, as can be observed from the significance level and with the negative

coefficient being rather large as compared to other explanatory variables. When there’s

an economic downturn, Buyouts can take advantage of low valuations. On the other hand,

a positive coefficient for credit spreads is surprising. It would be anticipated that when

financing is expensive, Buyouts under perform.

To summarize, while signs of coefficients for All, Combined, Buyouts and Ventures

suggest a pro cyclical behavior of fund returns, fund of funds seem to be less prone to the

general market development during the fund’s life. Generally, R squares are rather low for

most of the regressions, indicating that there are more relevant factors than the chosen

macroeconomic conditions explaining the IRR structure.

23

5 Conclusion

Based on a comprehensive dataset of 1641 private equity funds raised between 1979 and

2005, we study risk and return characteristics of different fund categories. Unlike previ-

ous studies, real FoF data is used, allowing for an authentic performance comparison of

FoFs and direct fund investments. We show that based on TVPI and IRR, FoFs exhibit

a more attractive risk return profile than aggregate direct fund investments. For both

performance measures, FoFs exhibit a higher risk return ratio, less probability of loss and

lower average loss. When split up by fund category, Buyouts feature the most favorable

risk return characteristics, followed by FoFs, Combined and Ventures. Examining the cash

flow structure, we document that the duration of the contributions is highest for FoFs,

which in combination with an average duration of distributions, results in the lowest total

duration of 1 as compared to 1.4 (+40%) for Buyouts, 1.5 for Combined (+50%) and 1.7

(+70%) for Ventures. This indicates that FoFs are leading in terms of cash flow timing

as investors on average need to pay in capital later and receive distributions earlier than

investors of other fund categories.

We also assess the performance by vintage years and find that overall, funds of vintage

years 2002-2003 seem to perform well, with low percentages of funds that exhibit a negative

Final IRR, decent levels of Final IRRs, low durations and effectual TVPIs. Regarding

mean TVPI and mean Final IRR overall performance for vintage years 1986, 1988 and

1994-1995 is also considerable. As opposed to that, lowest returns are for all fund categories

shown for funds of one of the vintage years between 1997 and 1999, which is probably due

to the effects of the dot com bubble. It appears that changes in the returns for Ventures

and FoFs are inversely correlated to the changes in Buyout returns. From an investor’s or

asset management’s perspective it could thus be prudent to include both Buyouts as well

as either Ventures or FoFs to better diversify the investment portfolio.

Finally, analyzing how macroeconomic conditions at the time of investments as well

as during the fund’s life influence fund performance, we show that direct fund returns

increase with MSCI and real GDP growth. This indicates pro cyclical return behavior.

As opposed to that, fund of funds seem to be less prone to the general market develop-

ment during the funds life. Generally, there is for all fund categories only little variation

explained by the used factors, suggesting that there are more relevant factors than the

chosen macroeconomic conditions explaining the IRR structure.

24

6 References

AIGNER, P., S. ALBRECHT, G. BEYSCHLAG, T. FRIEDRICH, M. KALEPKY, AND

R. ZAGST (2008): “What drives PE? Analyses of Success Factors for Private Equity

Funds,” The Journal of Private Equity, 11(4), 63-85.

DILLER, C., AND C. KASERER (2007): “What Drives Private Equity Returns? Fund

Inflows, Skilled GPs, and/or Risk,” Working Paper.

DRIESSEN, J., L. TSE-CHUN, AND L. PHALIPPOU (2008): “A New Method to Esti-

mate Risk and Return of Non-Traded Assets from Cash Flows: The Case of Private

Equity Funds,” Working Paper.

GRABENWARTER, U., AND T. WEIDIG (2005): Exposed to the J-Curve: Understand-

ing and Managing Private Equity Fund Investments. Euromoney Books, London.

GROH, A., AND O. GOTTSCHALG (2006): “Risk-Adjusted Returns of Private Equity

Investments,” Working Paper.

ICK, M. M. (2006): “Performance Measurement and Appraisal of Private Equity Invest-

ments relative to Public Equity Markets,” Working Paper.

KAPLAN, S. N., AND A. SCHOAR (2005): “Private Equity Performance: Returns,

Persistence, and Capital Flows,” Journal of Finance, 60 (4), 1791-823.

LJUNGQVIST, A., AND M. RICHARDSON (2003): “The Cash Flow, Return and Risk

Characteristics of Private Equity,” Working Paper.

PHALIPPOU L., AND O. GOTTSCHALG (2009): “The Performance of Private Equity

Funds,” Review of Financial Studies, 22(4), 1747-1776.

PHALIPPOU, L., AND M. ZOLLO (2005): “What Drives Private Equity Fund Perfor-

mance?,” Working Paper.

PREQIN LTD. (2010): The 2010 Preqin Private Equity Fund of Funds Review, London,

Preqin Ltd.

WEIDIG, T., AND P. MATHONET (2004): “The Risk Profiles of Private Equity,” Work-

ing Paper.

25