The effect of performance on Private Equity firm survival

The effect of performance on Private

Equity firm survival

Msc. Finance, Corporate Finance Thesis

Derk Majoewskij

10012079

July 2017

Supervisor:

Abstract

This thesis studies the relationship between performance and outperformance of private equity firms and their survival rate. More specific, this research focuses on the fact if returns and outperformance of average expected returns of previous funds have a significant effect on the probability of raising new funds by using a logit model. In addition, the effect of performance on the survival rate of private

equity firms is analyzed by using a COX survival model. The results show that the net Internal Rates of Return (IRR) and returns above expectations of the current fund both have a significant positive

effect on the probability of raising new funds and this effect is the largest for buyout funds. Furthermore, in line with the probability of raising new funds, the net IRR and the abnormal returns

both have a positive and significant effect on the number of funds a PE firm raises before it stops surviving when controlling for the survival bias. Again, this effect is the highest for buyout funds. This thesis shows that no matter the market conditions, fundraising and survival of PE firms are also

Table of Contents

Abstract... 2

Introduction

... 4

Literature review

... 7

Data

... 9

Empirical Strategy

... 13

Net IRR and fundraising ... 14

Methodology

... 17

Logit Model ... 17

Effect of returns on fundraising ... 17

Effect of abnormal returns of PE firm on fundraising ... 18

Effect of fund sequence on fundraising ... 18

Survival Analysis ... 19

Kaplan Meier survival estimate ... 19

Effect of returns on survival rate ... 19

Effect of abnormal returns of PE firms on survival rate ... 20

Results ... 21

Probability of raising new funds ... 21

Survival analysis ... 24

Discussion

... 27

Conclusion

... 28

Reference List ... 30

Introduction

For investors, it is important where to put their money. One can make an investment through a bank, broker, insurance company or a Private Equity (PE) firm. To achieve certain financial goals, the investors must make important decisions where and how to invest. It is commonly accepted PE will outperform public equity and other investments over the long-term and that is why investors are so keen to invest in it. Private equity consists of funds and investors that invest in private companies that are not noted on a public exchange or engage in buyouts of public companies. The most common types of PE are :

- Buyouts (BO), the purchase of a company’s shares to gain controlling interest in the targeted company, and

- Venture Capital (VC), a form of capital that is provided to small and early stage firms (Start Ups) that are forecasted to have high growth potential.

Most PE investments are done by a PE firm, the general partner (GP), that manages PE funds. To raise these funds, firms try to acquire capital from institutional investors also known as the limited partners (LP’s) by doing roadshows and sell their funds. Institutional investors can be banks, insurance companies, pension funds or even wealthy individuals. By using the raised funds, several investments are made by the PE firms and they earn their profits by receiving an annual management fee and/or a percentage of the generated returns.

To survive by making money through investments, fundraising is obviously a major objective for PE firms. Investors will look at the market conditions as well as the track record and reputation of PE firms when they decide where to invest. There is a clear distinction between the performance of various PE firms when looking at the number of raised funds and the survival rate of these companies. It can be argued investors are likely to invest when market conditions are favorable and that survival of a PE firm is due to luck but it can also be argued the performance and skill of a PE firm has a significant effect on the chance of raising new funds and the survival rate of PE firms.

This thesis focuses on the performance and outperformance of PE firms’ previous results of funds they managed to try to find an explanation why several PE firms raise more funds than others and are able to survive longer. It will try to prove fundraising is not only affected by favorable market conditions but also by the skills and expertise of the PE firm. Therefore, the research question will be: What effect does the performance of PE firms have on their success to raise additional funds and the eventual survival of PE firms?

There are two reasons why it is interesting to answer this question. First, PE firms want to raise as many funds and are looking for going concern. Therefore, it is important to know what drives the probability to raise additional funds on an ongoing basis. Second, investors want to invest in the best performing PE firm. If an investor is able to understand how a PE firm has performed and what skills it possesses by looking at the historic performance, this may be a deciding argument for them.

To answer the research question, an analysis should be performed to measure the effect of several performance variables on the chance of raising new funds and the survival rate of PE firms. By using data on 238 PE firms obtained by Preqin, first a graphical analysis should make clear if there is a relationship between a PE firms’ survival rate and the returns generated in their funds. It can be expected that investors are likely to invest in firms that have a better track record than others, so firms with high fund sequences are likely to have generated high returns in the past.

If there is such a relationship, the magnitude of the effects should be measured. To get a clear view of the effects of performance on fundraising, a model is used to see if performance relates to the probability of raising new funds. To do this, a logit model is used. It can be expected that returns in historic funds are likely to have a causal positive correlation to the probability of raising a new fund and therefore, the first hypothesis is:

H1: The net Internal Rate of Return (IRR) has a positive causal relationship with the probability of raising a new fund.

When a PE firm is able to outperform its expectations, the returns can be seen as abnormal and when these are high, it will probably attract more investors. So, to measure not only the effect of returns on fundraising but also outperformance, the abnormal returns will be calculated aside from the net IRR. If a firm is able to generate high abnormal returns, it means that the skills of the managers are higher. To test the effect of the outperformance of the firm on the chance of raising a new fund, the same logit model as for the IRR is used. It can be expected that higher abnormal returns will lead to a higher probability of raising a new fund and therefore, the second hypothesis is:

H2: Abnormal returns have a positive causal relationship with the probability of raising a new fund.

When holding performance constant, investors are likely to look at the experience of a PE firm before they invest in one. A way to look at this is to use the number of funds a PE firm already raised. A firm that has been able to raise a significant amount of funds probably performed very well in the past, so investors are more likely to invest in their funds.

To test this, the logit model is used where the number of the last raised fund of a PE firm is measured to have effect on fundraising. It can be expected that the higher the last fund number, the higher the probability to raise a new fund when controlling for performance. Therefore, the third hypothesis is:

H3: When holding performance constant, higher sequences lead to a higher probability of raising new funds.

The chance of raising a new fund does not necessarily say anything about the survival rate of PE firms, because there is a survival bias. The survival bias implies that the logit model takes into account the entire sample and does not control for the firms that did not survive past a certain fund number. Therefore, the results may skew higher. So, to measure the survival rate of PE firms on their own, a survival analysis is performed. With the use of the COX model to estimate the hazard rates, it can be estimated how many funds PE firms have under their control before they stop raising funds and how this is affected by returns and performance of the managers. It can be expected that firms that generate higher net IRR’s and abnormal returns are most likely to survive longer, so hypothesis 4 is: H4: The net IRR and the abnormal returns have a positive effect on the survival rate of PE firms when controlling for the survival bias.

The results from the graphical analysis show that there is a nonlinear relationship between firm survival and fund returns. This relationship is further examined in the models and the results from the logistic regression show that returns of the current fund have a significant and positive relationship with the probability of raising a new fund. This effect however, does only account for the final fund of a PE firm but not for earlier funds, indicating that only the most recent performance has its effects on fundraising.

In addition to the returns, also the abnormal returns show a positive significant relationship to fundraising. This means investors do not only look at the current returns but also at the fact if a firm outperformed its expectations due to managerial skill. In contrast to the net IRR, the abnormal returns do show a nonlinear relationship with fundraising. These effects are the highest for BO funds. When controlling for performance, the sequence number does not show a significant result, indicating that investors do not look at the number of the current fund when deciding to invest in a PE firm.

The results from the survival analysis show that most firms stop fundraising after their second and third fund. In line with the logit model, the results show that returns and the abnormal returns have a positive and significant effect on the survival rate of PE firms and these effects are also the highest for BO funds. The coefficients in the survival analysis are lower than the ones in the logit model because it takes into account the firms that quickly stopped raising new funds.

This thesis proves performance and skills of PE firms do have a positive causal effect on their ability to raise new funds and their long-term survival. This could either be due to the fact these firms are in a better position to convince potential investors to invest in their funds or investors look at previous performance when they decide to invest. PE firms will therefore raise more funds when they perform well and outperform their expectations. When looking at the survival rate of PE firms, it can

be concluded that firms with higher performances and abnormal positive returns do not only have a higher chance of raising a new fund, but are also able to survive longer.

The thesis is organized as follows: First, the existing literature will be examined to show how previous researchers studied this subject of PE and where there are still gaps open for further research. Second, a data analysis will describe which data is used, where it comes from and how variables are constructed. Hereafter, a strategy section will introduce the relationship between returns and the survival rates by several plotted graphs. Following this section, a methodology section will describe how different models will be used to measure the relationship between performance, fundraising and survival. Finally, the results, discussion and conclusion sections will shed light on the obtained results and place them in a proper perspective.

Literature review

This thesis focuses on performance of PE funds, outperformance in PE and the effects of this on capital inflows. The most cited and important of many researchers to study PE performance are Kaplan and Schoar (2005), who study PE performance on a fund level basis. By using the Public Market Equivalent (PME) they try to find out if private equity funds outperform the S&P 500 and conclude that, on average, BO and VC fund returns exceed those of the S&P 500 gross of fees. Several studies follow this research of which Hochberg, Lunqvist and Lu (2007) conclude that better networked VC firms experience significant better fund performance. Phalippou and Gottschalg (2009) conclude that PE funds outperform the S&P 500 index gross-of-fees. And finally, Harris, Jenkinson & Kaplan (2014) also conclude in their research that PE funds outperform the S&P 500 averages by using a new dataset from Burgiss and by comparing different databases.

Aside from performance of one fund, performance persistence is often used to find out if performance is driven by luck or skill and Kaplan and Schoar (2005), Phalippou and Gotschallg (2009) and Harris, Jenkinson & Kaplan (2014) all find significant performance persistence in their results.

Additional articles proving performance persistence are by Phalippou (2010) who shows that when a fund is being raised, its performance is only weakly related to that of the previously fund, Braun, Jenkinson and Stoff (2013) who conclude persistence has declined for BO firms after 2000 and by Buchner, Mohamed, & Schwienbacher (2016), who show that risk is an important driver of

performance persistence. A reaction to the findings of Kaplan and Schoar (2005) is the article by Korteweg and Sorensen (2017) who argue that the AR(1) model introduced by Kaplan and Schoar (2005) is a model that does not allow for long term performance differences. They come up with a variance decomposition model of PE performance and they find substantial long-term persistence in net-of-fee returns across all types of PE funds.

The effects of performance and in particular on the probability of raising new funds is a topic that has been examined before by Chung, Sensoy, Stern & Weisbach (2012). They conclude both the likelihood of raising a new fund as well as the size of that fund if it is raised are strongly positively related to current performance. This relation, between future fund-raising and performance, is stronger for BO funds than for VC funds and it declines in the sequence of a partnership’s fund. They provide a rational learning model, that is based on the learning model first introduced by Berk and Green (2004.) It predicts both the likelihood of raising a new fund and the size of the fund if it is raised depends on current performance. Gompers and Lerner (1999) study determinants that influence fundraising in VC by looking at supply and demand effects as well as individual firm performance and reputation. They also conclude fund performance is an important determinant of the ability for VC firms to raise new funds.

In this thesis, outperformance is measured as the abnormal returns above average expected returns but many prior studies use the net Alpha to investors as abnormal returns to measure managers skills in mutual funds. In one of the most important studies with this measure, Carhart (1997)

concludes there is no evidence of skilled or informed mutual fund managers. In addition, Fama and French (2010) test skill by using alpha measures and simulations and conclude that there are no managers who are significantly skilled enough to outperform their benchmark. Some articles that provide evidence that managers do have skill are by Kosowski, Timmermann, Wermer and White (2006) who show performance of funds is not solely due to luck and that this skill is persistence and by Cremers and Petajisto (2009) who use a new measure of active portfolio management and

conclude that funds are able to outperform their benchmarks and this persists. One article that argues the use of the net alpha as a measure for skill is by Berk and Binsbergen (2012). They conclude the net alpha is not the skill of the manager, but the value added to the fund should also be included for a precise measurement.

Previous literature confirms the hypotheses stated in the introduction by showing the drivers behind performance, that this performance persists and the main positive effects of this on subsequent capital inflows. From the literature, it can be concluded that performance and performance persistence have a positive relationship to capital inflows, but this thesis goes deeper into firm related

performance effects.

Where Chung et al (2012) only use the net IRR to predict the probability of raising a new fund, this thesis also shows outperformance of expected returns has its effect. This way, the results are better able to show that fundraising is affected by the skill and performance of a PE firm. In contrast to previous literature, this research also examines the survival rate of PE firms by controlling for the survival bias. Where Chung et al (2012) and Gompers and Lerner (1999) show a relationship between performance and fundraising for their entire sample, this study also examines the effect of

made between the fact if performance has its effect on fundraising for the entire sample alone and the fact if performance does really affect the survival rate of PE firms.

Data

Descriptive statistics

To examine PE firms and funds and to answer the research question, fund-level data for BO and VC funds provided by Preqin are used. Preqin is an online platform that provides information on deals, funds and firms in PE. Preqin has summary performance data for a large number of funds and firms, even compared to other databases. To have a detailed sample and a broad picture, the time period in the sample starts in 1980 and to allow for sufficient time to ascertain whether a fund raises a

subsequent fund, the time period ends in 2007. Examining the effect of historic funds on new funds is the main part in this study, so only firms with more than one fund under their management are included. In addition, only fully liquidated funds are included. To obtain estimates of the relationship between future fund-raising and previous performance, the sample consists only of funds for which the net IRR is available. Furthermore, funds with values of less than 5 million dollars are excluded to reduce the influence of potential high growth rates of smaller funds in the sample.

The final sample of funds and PE firms is shown in table 1. The exclusion of funds leads to a final sample of 781 funds that are raised by 238 PE firms. There are 238 final - sequence and 498 preceding funds in the sample.

Table 1.

Sample statistics of PE funds in the sample by type.

1 2 3 Total VC BO Number of funds 781 375 406 Number of firms 238 105 133 number of funds/firms mean 3.28 3.49 2.9 Std. dev. 2.05 2.1 1.46 median 3 3 2 min 2 1 1 max 18 16 10

Column 1 of table 1 shows the sample statistics for the total amount of funds, column 2 for the VC funds and column 3 for the BO funds, from which is clear to see that there are more BO funds in the sample. Preqin also classifies other types of funds, for example, venture debt and real estate funds, but these are excluded from the dataset. The average number of funds per firm is 3.28 and the distribution

of this average is shown in Figure 1. The majority the firms in the dataset manage 2 to 3 funds and this amount declines in the number of funds.

Figure 1.

Number of funds per firm frequency.

Figure 2 shows the frequency of the number of years between the vintage year of the first fund of a firm and the last fund of a firm. What can be seen is that most firms have on average 3 to 6 years between raising their first and last fund. There are very few firms that only have 1 year between this and there are several firms with a higher number of years between their first and last fund. This indicates that there are firms that have a higher survival rate than others.

Figure 2.

The frequency of time spans between the vintage year of the first fund and the last fund of a PE firm.

Figure A in the appendix shows the frequency of funds by vintage year in the dataset. For the late 2000s and the eighties, there is only a small amount of funds available on Preqin with known

0 50 1 00 F req ue nc y 2 3 4 5 6 7 8 9 12 13 14 18 Number of funds per firm

0 10 20 30 40 F req ue nc y 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 22 Years between first and last fund

performance measures. The largest number of funds with known IRR’s have vintage years between 1990 and 2000. After 2000 a steep decline in funds is clearly visible from the graph and this can be explained by either the dot-com bubble and/or the financial crisis. PE firms could have decided to raise fewer funds during these tumultuous financial periods or there is simply no information available about the funds yet.

The main limitation to the data obtained by Preqin, is that Preqin lacks cash flow data needed for the calculation of the Public Market Equivalent (PME), a performance measure that is needed to compare the funds’ performance with a certain index.

Variables

This section describes the several variables that are used in this research and how they are obtained and constructed. To examine if historic performance has its effect on fundraising and survival of a PE firm, the net IRR is used. Several studies use the PME measure to examine performance of PE, but due to lack of information to calculate the PME in this study, this is not possible. The net IRR is the annualized return to LP’s after management fees and carried interest have been accounted for and is the most used measure in PE performance analysis. Other main variables used in this study are the:

- Vintage year - Sequence

- Abnormal returns

The net IRR, vintage year and sequence of the funds are all provided by Preqin. The vintage year is the year when the first capital is contributed to a certain fund and has first values of 1980 and final values of 2007. The variable sequence is the number of the fund from the PE firm and ranges from 2 for the lowest amount of funds under control to 18 for the highest amount of funds under control.

Abnormal Returns

Next to the net IRR, to further measure the effect of performance on fundraising and the survival rate of a firm, the outperformance of a PE firm is calculated. The net IRR could be higher due to several factors, so the abnormal returns in comparison to the average expected returns show the

outperformance of managers. The most common measure of manager’s skill in previous studies is the net alpha or the abnormal returns in comparison with its benchmark, but in this study, it is not

possible to compare returns to their benchmark so a regression is run on the following equation: 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼,𝑁𝑁 = 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝐼𝐼,𝑁𝑁+ 𝐹𝐹𝑉𝑉𝑉𝑉𝐹𝐹𝐼𝐼,𝑁𝑁+ 𝜀𝜀𝐼𝐼,𝑁𝑁 (1)

In this regression, 𝐼𝐼𝐼𝐼𝐼𝐼_{𝐼𝐼,𝑁𝑁}is the net internal rate of return of fund I of firm N, 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉_{𝐼𝐼,𝑁𝑁}is the vintage year of fund I of firm N and 𝐹𝐹𝑉𝑉𝑉𝑉𝐹𝐹_{𝐼𝐼,𝑁𝑁}is the firm of fund N. Whenever the returns of a certain

fund are controlled for the vintage year and the firm that has the fund under its management, the residuals are assumed to be the abnormal returns of a certain fund of a particular firm. They resemble the returns made that are above or below the average expected returns of the fund from a certain vintage year. In other words, they measure the outperformance of the PE firms’ manager.

Summary Statistics

Table 2 shows the summary statistics of the variables used in this study. The mean fund returns (IRR) of 25.28% for VC funds is significantly higher than for BO funds. In addition, the mean of the sequence number is the highest for VC funds as well. The median sequence of the funds together and by type is 2, which confirms figure 1 that shows most firms manage 2 to 3 funds. The mean and median of the abnormal returns are lower than the returns and are 0% for the entire sample. The VC funds have mean abnormal returns that are positive, while they are negative for BO funds.

Table 2.

Summary statistics for the variables. The table shows the mean, standard deviation, median, min and max for every variable used in this study.

1 2 3 Total VC BO Net IRR (%) mean 22.27 25.28 19.5 std dev. 39.2 49.53 23 median 14.9 12.6 17.95 min -100 -46.1 -100 max 514.3 514.3 279.7 Sequence mean 2.78 2.93 2.63 std dev. 2.3 2.35 2.31 median 2 2 2 min 1 1 1 max 18 17 18 Vintage year median 1999 1994 1997 min 1980 1980 1980 max 2007 2006 2007 Abnormal returns mean 0 2.5 -2.31 std dev. 38.11 49.6 22.69 median -7.59 -11.52 -4.81 min -119.57 -66.7 -119.57 max 493.82 493.82 254.16

Empirical Strategy

To answer the research question using the data described in the previous section, it is first interesting to see how returns relate to PE firm survival rates. In this section, a graphical analysis tries to show if PE firms that survive longer also experienced higher returns in previous funds and if there is a relationship between this.

Survival bias

Before it is possible to analyze the relationship between returns and survival of PE firms, it is important to show the implication of the sample used in this study. This implication is visible when plotting the average net IRR per fund number for all PE firms. This is shown in Figure 3.

From the graph can be concluded that for funds 2 to 8, the net IRR tends to move normal, that is, without any large outliers, but after 8 funds, more volatility arises in the line. The figure clearly shows that firms that do not survive after a certain fund number have an impact on the results. For every fund number value, the return is a combination between PE firms that create new funds and the ones that stop at that number. The line shows very large fluctuations after the 8th sequence and this is because a significant amount of funds included in the dataset already stopped existing. This effect, the error of concentrating on funds that made it past a certain selection point and overlooking those that did not, is called the survival bias. The survival bias causes results of financial studies to skew higher Figure 3

because only firms that were successful enough to survive until a certain fund number are taken into account past this point.

Net IRR and fundraising

To control for the survival bias and to get a clear view of the relationship between previous returns and fundraising, the average net IRR per number of raised funds is plotted. Figure 4 shows the average net returns for firms that raised a maximum of 2 to 5 funds separately.

The firms that stopped raising funds the earliest, the ones with 2 funds under their management, did on average experience low returns in their first fund that even declined in their second fund. This decline after an already low return in the first fund was probably noticed by investors and made it hard for the firms to continue raising funds.

Firms that raised 3 funds had almost the same return pattern like the ones that raised only 2, however these firms generated higher returns in their first. These higher returns probably convinced investors to invest in these firms, but the declining returns after the first fund made investors more skeptical.

Firms that raised 4 funds also experienced relatively higher returns in their first fund and performed well until their 3rd fund when a decline in returns eventually led to the end of fundraising. Firms that raised 5 funds generated high returns in their first fund, but this average quickly declined in their second fund. Hereafter, the returns increased steeply and then declined very sharp.

0 20 40 60 80 A ve rag e ne t I R R 1 2 3 4 5 Sequence numbers 2 funds 3 funds 4 funds 5 funds Figure 4

The average net IRR and fund sequence per firm. The graph shows the average net IRR for firms separated by fund sequence. The graph shows the average net IRR for firms with sequences from 2 to 5.

As a general conclusion, the graph shows firms had difficulties raising funds after they had a decline in returns from one fund to another. In addition, when the returns eventually declined to a certain low point, the firms stopped raising funds.

Figure 5 shows the average net IRR for firms that survived longer and raised 6 to 9 funds. Figure 5

The average net IRR and fund sequence per firm. The graph shows the average net IRR for firms separated by fund sequence. The graph shows the average net IRR for firms with sequences from 6 to 9.

Firms that raised a number of 6 funds started with relatively low returns but experienced a solid increase in returns after their second fund. This solid increase probably convinced investors to invest their money in the firms which led to a higher survival rate. After the 5th fund however, the returns were lower what made fundraising harder. This was quickly followed by the end the of the firms.

Similar to firms with 6 funds, firms with a total of 7 funds also experienced a solid increase in returns in their first funds, a positive peak in their second to last fund and a steep decline after this peak.

Firms that were able to raise 8 funds generated on average relatively high stable returns in their first funds which was then followed by a small decline. Probably due to the stable returns in their first funds, these firms were able to raise funds after this decline and had solid positive returns

hereafter. In contrast to other fund sequences, the firms with 8 funds stopped raising funds after an increase in returns.

Firms with 9 funds experienced high and increasing returns in their first funds, but a strong decline in returns in their last funds. Although these firms experienced negative returns in their last funds, they were still able to raise new funds. The overall conclusion from this graph is firms that were able to raise more funds had on average stable or increasing returns in their first funds, but

-2 0 0 20 40 60 A ve rag e ne t I R R 0 2 4 6 8 10 Sequence number 6 funds 7 funds 8 funds 9 funds

eventually reached a peak after which the returns were lower. Quickly after this, the firms were not able to raise any new funds.

Figure 6 shows the returns for the firms that were able to raise the highest number of funds and survived the longest.

Figure 6

The average net IRR and fund sequence per firm. The graph shows the average net IRR for firms separated by fund sequence. The graph shows the average net IRR for firms with sequences from 12 to 18.

Due to a lower amount of firms in this part of the sample, the line shows strong volatility. These firms show very large differences between the returns of their funds. In some funds, the returns are very high but this is then followed by very low returns. The most important fact from this figure is that investors are likely to invest in these firms because they are able to generate very high returns although these are not stable. The fact that they are able to do so makes investors probably keen to invest in them.

Figures 4, 5 and 6 show a nonlinear relationship between the net IRR, fundraising and the survival of PE firms. Firms that stopped raising funds quickly had declining returns in their first funds which eventually reached a certain low point. Firms with a higher number of funds under their

management and a higher survival rate experienced a solid increase in returns in their first funds that reached a positive peak in time. After this peak however, returns quickly decreased and fundraising was stopped. Firms that raised a high number of funds and had the highest survival rate show very volatile returns. Negative peaks are followed by positive peaks and vice versa. This effect could either be explained by regression to the mean or simply by managers who become more conservative. Reversion to the mean in finance is the theory suggesting that returns and prices eventually move back to their average. However, in figure 6, the lines show no decreasing peaks or IRR’s that come closer to their mean as the sequences increase, so regression to the mean will most likely not be the

0 20 40 60 80 A ve rag e ne t I R R 0 5 10 15 20 Sequence number 12 funds 13 funds 14 funds 18 funds

case. Other factors play a role in the great volatility of the net IRR’s over fund sequence, like the skill of the managers or market factors.

Methodology

Looking at figures 4,5 and 6, it is clear to see that there is a nonlinear relationship between returns and the number of funds raised. It is therefore interesting to find out if there is a significant causal effect of performance on fundraising and survival. This section covers the models to measure the effects of returns and abnormal returns on the probability of raising new funds and the survival rates of PE firms.

Logit Model

Effect of returns on fundraising

PE firms raise funds to survive, so it is interesting to find out if performance measures have a relationship with the probability of raising new funds when holding market conditions constant. To test this, a logit model is used. By using a logit model, one is able to measure the effect of several independent variables on the probability of having a certain dependent binary variable. In this case, the binary variable is described as a fund that has a following fund or, in other words, when there is another fund raised after the current fund. Therefore, the binary dependent variable is 1 if a fund has a following fund and 0 if it is the last fund of the firm. Equation 2 is used to measure the effect of returns on this probability. Vintage year effects of the current fund are added to control for any market- and time-varying factors that could affect the probability to raise a new fund. Furthermore, fund type fixed effects are added to control for any type-varying factors that could affect this probability.

𝐿𝐿𝐿𝐿𝑉𝑉𝑉𝑉𝑉𝑉 (𝑝𝑝) = 𝛽𝛽0+ 𝛽𝛽1𝑁𝑁𝑉𝑉𝑉𝑉𝐼𝐼𝐼𝐼𝐼𝐼𝑁𝑁+ 𝛽𝛽2𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑁𝑁+ 𝛽𝛽3𝑇𝑇𝑇𝑇𝑝𝑝𝑉𝑉𝑁𝑁+ 𝜀𝜀N (2)

In this equation, 𝐿𝐿𝐿𝐿𝑉𝑉𝑉𝑉𝑉𝑉 (𝑝𝑝) is the fact if a fund is followed by another fund, 𝑁𝑁𝑉𝑉𝑉𝑉𝐼𝐼𝐼𝐼𝐼𝐼_𝑁𝑁 the net IRR of current fund N, 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉_𝑁𝑁 the vintage year fund N was raised and 𝑇𝑇𝑇𝑇𝑝𝑝𝑉𝑉_𝑁𝑁 the type of fund N. Not only current returns could affect the chance of raising a new fund, but also previous returns. Figure 4 and 5 show sequence patterns where the second to last fund is a peak that is followed by lower returns in the next fund which is also the last one of the firm. It is therefore interesting to see what effect the second to last funds’ net IRR has on the chance of raising a new fund. To do this, the lagged net IRR is added to equation 2. Again, the Vintage Year effects and Fund type effects of the lagged fund are added as control variables. In addition, the current IRR is also used as control variable.

As mentioned before, Figures 4, 5 and 6 show a nonlinear relationship between the net IRR and the number of funds raised by a PE firm. To test if there is such a relationship, the squared net IRR is added to the regression. This quadratic term implies that the higher the net IRR will become, the effect on fundraising will become less and eventually become negative. In other words, if there is a nonlinear relationship, it means that as the returns get higher, they eventually have a negative effect on the chance of raising a new fund. Again, there is control for vintage year and fund type.

Effect of abnormal returns of PE firm on fundraising

Fundraising could be affected by returns in previous funds, but PE firms are probably also able to raise more funds when investors know that outperformance is high. It can be expected the chance of raising a new fund increases when investors are aware of managerial skills or abnormal returns. When a fund generates high abnormal returns, it means that it performs above average expected returns. Investors are likely very keen to invest in firms that outperform expectations. To see if these skills contribute to fundraising and to measure the magnitude of this, another logistic regression is run similar to equation 2. Instead of using the net IRR as independent variable, the abnormal returns as calculated in the data section are used. Again, the Vintage Year and Type variables are added for control. The equation used is as follows:

𝐿𝐿𝐿𝐿𝑉𝑉𝑉𝑉𝑉𝑉 (𝑝𝑝) = 𝛽𝛽0+ 𝛽𝛽1𝐼𝐼𝑉𝑉𝑅𝑅𝑉𝑉𝑅𝑅𝑁𝑁+ 𝛽𝛽2𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑁𝑁+ 𝛽𝛽3𝑇𝑇𝑇𝑇𝑝𝑝𝑉𝑉𝑁𝑁+ 𝜀𝜀𝑁𝑁 (3)

Where 𝐼𝐼𝑉𝑉𝑅𝑅𝑉𝑉𝑅𝑅_𝑁𝑁 are the residuals, or abnormal returns generated in fund N.

Abnormal returns of previous funds are also expected to influence fundraising. If a manager is able to generate high abnormal returns in one or more funds prior to its last, it can be expected to have a positive and significant effect on the chance of raising a new fund. To test this, the abnormal returns of preceding funds are added as lagged abnormal returns to equation 3. Furthermore, to test for a nonlinear relationship between abnormal returns and the chance of raising a new fund, the squared abnormal returns are added to the regression. A nonlinear relationship would imply that if the

abnormal returns get higher, the probability of raising a new fund would eventually become negative. Effect of fund sequence on fundraising

Performance can also be measured by the number of funds a PE firm already raised. It can be argued that if a firm has a high number of funds already raised, investors may see this firm as high

performing and experienced. A good way to measure the effect of this experience is to look at the number of funds a firm already raised. If a firm has a high current fund - sequence number, it can be expected that investors see this firm as successful even when not looking at performance. To test the

effect of the current sequence number on the chance of raising a new fund, another logistic regression is run. In this regression, there is control for performance. Equation 4 shows the regression used.

𝐿𝐿𝐿𝐿𝑉𝑉𝑉𝑉𝑉𝑉 (𝑝𝑝) = 𝛽𝛽0+ 𝛽𝛽1𝑁𝑁𝑉𝑉𝑉𝑉𝐼𝐼𝐼𝐼𝐼𝐼𝑁𝑁+ 𝛽𝛽2𝑆𝑆𝑉𝑉𝑆𝑆𝑆𝑆𝑉𝑉𝑉𝑉𝑆𝑆𝑉𝑉𝑁𝑁+ 𝛽𝛽3𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑁𝑁+ 𝛽𝛽4𝑇𝑇𝑇𝑇𝑝𝑝𝑉𝑉𝑁𝑁+ 𝜀𝜀𝑁𝑁 (4)

Here, 𝑆𝑆𝑉𝑉𝑆𝑆𝑆𝑆𝑉𝑉𝑉𝑉𝑆𝑆𝑉𝑉_𝑁𝑁 is the sequence number of fund N.

Fundraising could be affected more or less in certain sectors of PE. To see if there is a difference between PE fund types, above regressions are also run separately for the BO and VC funds. In these regressions, the PE firms with only 1 BO or VC fund are excluded and there is no control for fund type.

Survival Analysis

In the previous section, the probability of raising a new fund is examined for the entire sample. Following this, it is interesting to see how performance relates to the number of funds a firm is able to raise before it stops surviving. The logit model does not control for the survival bias because it estimates the odds ratios which are cumulative over the entire study. Therefore, after a certain fund number, they do not include the firms that already stopped existing. In this section, to test the survival rate of PE firms and to control for the survival bias, a survival analysis is performed. A survival analysis can be used to analyze the expected duration of time until one or more events happen per firm. In other words, the survival models control for the survival bias by testing all firms separately until an event. In this research, the event is the last fund of a firm, or the fact that a firm is no longer able to raise a new fund and stops existing. The time measured is the number of funds a firm raised, so the survival model tests how performance measures affect the number of funds a firm raises until they stop raising new funds.

Kaplan Meier survival estimate

To test the survival rates of PE firms, first the relationship between the survival rate and the number of funds raised until the end of the PE firms is plotted in a Kaplan – Meier survival graph. The Kaplan Meier estimate is a statistic that is used to measure the time, or number of funds, until firms quit raising funds. With this plot, a clear image can be made to show firm survival.

Effect of returns on survival rate

To find out what drives these survival rates of PE firms and the magnitude of performance on this, the COX proportional hazard function is used. By using the COX model, the hazard rate can be estimated. Hazard ratios differ from odds ratios, like the ones in the logit model, in the fact that the latter are cumulative over an entire study as previously described. Therefore, hazard ratios suffer less from

survival bias with respect to the endpoint chosen. Another advantage of the COX model is that next to the fact that one can estimate if there is an effect of a certain explanatory variable on the survival rate, also the size of the effect can be estimated. Furthermore, several explanatory variables can be

included in the model.

To test the magnitude of the effect of the returns on the survival rate, equation 5 is used for the COX model. In this equation, the effect of the net IRR on the number of funds until firms stop existing is estimated. As in the logistic regressions, there is control for Vintage Year and Fund Type effects.

ln(_𝐻𝐻𝐻𝐻(𝑡𝑡)

0(𝑡𝑡)) = 𝛽𝛽1𝑁𝑁𝑉𝑉𝑉𝑉𝐼𝐼𝐼𝐼𝐼𝐼𝑁𝑁+ 𝛽𝛽2𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑁𝑁+ 𝛽𝛽3𝑇𝑇𝑇𝑇𝑝𝑝𝑉𝑉𝑁𝑁 (5)

Where ln(𝐻𝐻(𝑡𝑡)

𝐻𝐻0(𝑡𝑡)) is the hazard ratio at sequence t, 𝑁𝑁𝑉𝑉𝑉𝑉𝐼𝐼𝐼𝐼𝐼𝐼𝑁𝑁 is the net IRR of fund N and

𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑁𝑁 and 𝑇𝑇𝑇𝑇𝑝𝑝𝑉𝑉𝑁𝑁 are the vintage year and type of fund N respectively.

Because figures 4,5 and 6 show a nonlinear relationship between returns and survival, a quadratic term of the net IRR is added to test for this relationship. If this relationship is indeed significant, it means that as the returns get higher, the survival rate eventually becomes negative.

Effect of abnormal returns of PE firms on survival rate

The fact that a PE firm performs above expectations most likely has its effect on the survival rate of the firm. Therefore, the abnormal returns are used in the COX model to estimate the relationship between this performance and the survival rate of the PE firm. Equation 6 shows the equation for this.

ln(_𝐻𝐻𝐻𝐻(𝑡𝑡)

0(𝑡𝑡)) = 𝛽𝛽1𝐼𝐼𝑉𝑉𝑅𝑅𝑉𝑉𝑅𝑅𝑁𝑁+ 𝛽𝛽2𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑁𝑁+ 𝛽𝛽3𝑇𝑇𝑇𝑇𝑝𝑝𝑉𝑉𝑁𝑁 (6)

Here, 𝐼𝐼𝑉𝑉𝑅𝑅𝑉𝑉𝑅𝑅_𝑁𝑁 are the abnormal returns for fund N.

Again, to find out if there is a nonlinear relationship between the abnormal returns and the survival rate, a quadratic term of the abnormal returns is added to equation 6. A positive significant coefficient would imply that as the abnormal returns eventually get very high, the survival rate will go down. Like the logit model, the COX regressions are also run separately for the BO and VC funds to measure the effect of the net IRR and the abnormal returns on the survival rate per fund type.

The COX model is valid when proportional hazard rates for the variables are constant over time. To test this, a proportional hazard rates test is performed. Appendix C shows the results for this. It can be concluded that the results are non-significant so there is no evidence to contradict the proportionality assumption.

Results

In this section, the results of the described models and regressions are shown. First, the results of the logistic regressions are presented for the entire sample as well as for BO and VC funds separately. Second, the results of the survival analysis are shown for the entire sample and the BO and VC funds separately.

Probability of raising new funds Logit results for all funds Table 3 shows the results of the logistic regressions for all PE firms. Table 3

Regression results from the logit model for all funds together. The dependent variable is the probability of being a fund that has a following fund. Net IRR’s are obtained by Preqin, Lagged Net IRR is the IRR of the previous fund, the abnormal returns are obtained from regression 1, lagged abnormal returns are the abnormal returns from the previous fund and Sequence is the sequence number of the current (preceding) fund. Vintage year and fund type firm effects are added for control and heteroskedastic – robust standard errors are clustered by firm level.

(1) (2) (3) (4) (5) (6) (7) VARIABLES Having a following fund Having a following fund Having a following fund Having a following fund Having a following fund Having a following fund Having a following fund Net IRR 0.00775** 0.0107** 0.0120*** 0.00791** (0.00393) (0.00471) (0.00456) (0.00400)

Net IRR2 _-1.56e-05

(9.76e-06)

Net IRRN-1 0.000701

(0.00205)

Abnormal returns 0.00793** 0.0103** 0.00846**

(0.00397) (0.00438) (0.00381)

Abnormal returns2 _-1.62e-05*

(9.84e-06) Abnormal returnsN-1 0.000864 (0.00206) Sequence -0.0585 (0.0866) Constant 703.7*** 698.8*** 46.90* 710.6*** 707.9*** 54.13** 698.4*** (62.69) (62.23) (26.94) (62.47) (62.03) (27.08) (62.95)

Vintage Year fixed effects

Fund Type fixed effects

Observations Yes Yes 781 Yes Yes 781 Yes Yes 543 Yes Yes 781 Yes Yes 781 Yes Yes 543 Yes Yes 781 Robust standard errors in parentheses

Column 1 shows the relationship between the net IRR of the current fund on the probability of raising a new fund. The significant coefficient of 0.0075 means that a 1% increase of the net IRR of the current (preceding) fund leads to an increase in the probability of raising a new fund of +0.75%. This means that, when holding market and fund type conditions constant, returns do indeed positively affect the probability of raising a new fund. According to this coefficient, higher performing PE firms are able to raise more funds.

Column 2 shows the results from the regression when adding the quadratic term to test for a non-linear relationship. The coefficient is negative, which indicates that as the returns get higher, the chance of raising a fund eventually gets lower. The coefficient however, is very small and

insignificant. Column 3 shows the chance of raising a new fund when a previous fund is added. The coefficient is small and positive, but it is insignificant. Therefore, no conclusion can be drawn regarding the effect of this previous fund.

Like the net IRR, column 4 shows the relationship between the abnormal returns of the current fund and the probability of raising a new fund. The significant coefficient of 0.00793 indicates that a 1% increase in the abnormal returns leads to a 0.793% increase in the probability of raising a new fund. This coefficient is higher than the one from the net IRR, indicating that when controlling for market and fund type conditions, firms that are able to outperform the expectations are able to raise new funds more easily.

Unlike the net IRR, the squared abnormal returns have a significant effect as shown in column 4. Although the effect is very small, as the abnormal returns get higher, the probability of raising a new fund will eventually drop. This could be caused by investors who expect that successful firms are likely to generate lower losses after a high performing fund. Column 6 shows that similar to the net IRR, the abnormal returns of a previous fund do not have a significant effect on the chance of raising a new fund.

Finally, column 7 shows that, when controlling for performance, there is no significant relationship between the sequence number of the current fund and the probability of raising a new fund. This indicates that investors do not look at the fund number of a PE firm solely when investing.

Logit results for Buyout and Venture Capital funds Table 4 shows the results of the logistic regressions for BO and VC funds separately. Table 4

Regression results from the logit model for BO and VC funds separately. The dependent variable is the probability of being a fund that has a subsequent fund. Net IRR’s are obtained by Preqin and the abnormal returns are obtained from regression 1. Vintage year effects are added for control and heteroskedastic – robust standard errors are clustered by firm level.

(1) (2) (3) (4) (5) (6) (7) (8)

VARIABLES BO BO BO BO VC VC VC VC

Net IRR 0.0240*** 0.0264*** 0.00717* 0.00929

(0.00583) (0.00656) (0.00427) (0.00584)

Net IRR2 -7.10e-05 -1.05e-05

(9.20e-05) (1.19e-05)

Abnormal returns 0.0246*** 0.0240*** 0.00723* 0.00895*

(0.00588) (0.00603) (0.00428) (0.00537)

Abnormal returns2 -8.56e-05 -1.08e-05

(8.27e-05) (1.19e-05)

Constant 770.4*** 768.6*** 794.3*** 791.6*** 538.3*** 532.0*** 544.7*** 540.0***

(101.9) (101.9) (103.8) (103.5) (67.01) (66.77) (66.63) (66.02)

Vintage Year effects Yes Yes Yes Yes Yes Yes Yes Yes

Observations 401 401 401 401 368 368 368 368

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

It is clear to see from column 1 that the effect of the net IRR on the probability of raising a new fund is higher for BO funds than for VC funds. The coefficient of 0.0240 means that an increase of 1% in the net IRR of BO funds leads to a 2.4% increase in the probability of raising a new fund. This is high in comparison to the full sample. Column 2 shows that there is no nonlinear relationship between the net IRR and fundraising in BO funds. Furthermore, the abnormal returns have significant and positive effects on the probability of raising subsequent funds and these effects are higher than the net IRR. The abnormal returns do not show a nonlinear relationship. For VC funds, it is clear to see from column 5 that there is lower significance and a lower effect than for BO funds regarding the net IRR. This is also seen from column 7, where the effect of abnormal returns is lower for VC funds than for BO funds. Both the net IRR and the abnormal returns show no significant nonlinear relationship to fundraising.

Survival analysis

Figure 7 shows the Kaplan-Meier survival estimate for all the firms in the dataset.

From the steep decline after the second fund number, it is clear to see that most companies stop raising funds after their second fund. From there on, an almost linear decrease is shown in the survival rate until the 8th fund. After 8 funds, the survival rate becomes flatter, implying that during these fund numbers, little to none firms reached their last fund. This could be due to less firms in the dataset that manage more funds of the fact if a firm is able to raise this number of funds, the survival rate gets higher. The line reaches 0 at the 18th fund, meaning that after 18 funds, there were no more funds in the sample. The overall conclusion from this graph is that the survival rate is linear until the 8th fund after which less firms stopped existing. This indicates that when a firm is able to reach a certain fund number, the survival rate will go up. This however is not solely due to the number of funds as is seen from the results of equation 4.

COX results for all funds

The effect of performance on the survival rates plotted in the Kaplan – Meier graph are tested by the COX model and the results are shown in Table 5.

0. 00 0. 25 0. 50 0. 75 1. 00 Pe rce n ta g e o f su rvi va l 0 5 10 15 20 analysis time

Kaplan-Meier survival estimate Figure 7

Table 5

Regression results from the COX model regressions for all funds together. The dependent variable is the chance of hazard, or being the last fund of the fund. Vintage year and fund type fixed effects included for control and heteroskedastic – robust standard errors are clustered at PE firm level.

(1) (2) (3) (4) VARIABLES _t _t _t _t Net IRR -0.00479* -0.00607** (0.00264) (0.00310) IRR2 8.41e-06 (7.25e-06) Abnormal returns -0.00489* -0.00589** (0.00265) (0.00286)

Abnormal returns2 _8.92e-06

(7.19e-06) Vintage year fixed effects

Fund type fixed effects

Yes Yes Yes Yes Yes Yes Yes Yes Observations 781 781 781 781

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

The hazard rates in the table are converted to coefficients to more easily compare them with the results from the logit regressions.

Column 1 shows the relationship between the net IRR and the number of funds to reach the firm’s last fund. The significant coefficient implies that an increase of 1% in the net IRR results in a -0,479% chance of hazard, or raising the last fund of the firm. Looking at the results from column 1 in the logit model and this coefficient, it is clear to see that the net IRR has a positive effect not only on the chance of raising a new fund, but also on the survival rate. The effect on the survival rate is lower than the logit model, because the effect is examined per entity and there is control for the survival bias. As mentioned earlier, the logit model skews the results higher due to the fact that successful funds are taken into account when estimating the probabilities.

Column 2 shows that there is no significant effect of the squared net IRR on the chance of hazard, implying that there is no nonlinear relationship between these. Column 3 and 4 show the same effects as the first and second columns, but for the abnormal returns. Similar to the net IRR, the abnormal returns have a negative coefficient, indicating that an increase of 1 in the abnormal returns, leads to a -0,489% chance of hazard. Again, this is in line to the findings of the logit model implying that an increase in abnormal returns not only leads to a higher chance of raising a new fund but also on the survival rate of a PE firm. The coefficient is also slightly lower than the one from the logit model due to controlling for the survival bias.

Cox results for Buyout and Venture Capital funds

The results of the COX regression for VC and BO firms separately are shown in table 6.

Table 6

Regression results from the COX model regressions for VC and BO funds separately. The dependent variable is the chance of hazard, or being the last fund of the fund. Vintage year and fund type fixed effects included for control and heteroskedastic – robust standard errors are clustered at PE firm level.

(1) (2) (3) (4) (5) (6) (7) (8)

VARIABLES BO BO BO BO VC VC VC VC

Net IRR -0.00952** -0.00939*** -0.00533 -0.00682

(0.00372) (0.00357) (0.00325) (0.00431)

Net IRR2 -7.33e-06 8.05e-06

(5.16e-05) (9.30e-06)

Abnormal returns -0.00978*** -0.00984** -0.00532* -0.00653*

(0.00374) (0.00430) (0.00323) (0.00394)

Abnormal returns2 -2.68e-06 8.29e-06

(4.90e-05) (9.18e-06)

Vintage fixed effects Yes Yes Yes Yes Yes Yes Yes Yes

Observations 401 401 401 401 368 368 368 368

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Column 1 shows that the net IRR has a negative and significant effect on the hazard rate for BO funds, implying that a 1% increase in the net IRR leads to a -0,95% decrease in the chance of having the last fund in the sequence. Like the funds pooled together, this is in line with the results from the logit model, however the effect from column 1 is lower than the one in the logit model. Again, this is due to the control for the survival bias. It is very clear to see that the effects are the highest for BO funds. The squared net IRR shows no significant nonlinear relationship to the chance of hazard in column 2.

Similar to the net IRR, the abnormal returns also show a negative significant effect on the chance of having the last fund, implying that a 1% increase in the abnormal returns leads to a -0,98% chance of hazard. This effect is in line with the one in the logit model, but this coefficient is lower. In contrast to the BO funds, the VC funds show no significant effect of the net IRR on the chance of reaching the last fund. The effect of this in the logistic regression also had low significance, but this disappeared when controlling for the survivor bias. This however, is not the case for the abnormal returns, which show a negative significant effect, indicating a 1% increase in the abnormal returns will lead to a -0.5% chance of having the last fund. The abnormal returns also had a significant effect in the logit model, so this is in line with each other.

Discussion

By using a logit model and a survival analysis it is possible to support or reject the hypotheses proposed in the introduction section. The plotted graphs in the strategy section show that there is a nonlinear relationship between fund returns and survival of PE firms so a further research is performed to find out what the effects are of performance on fundraising. First, the results from the logit model show that fundraising is positively affected by the net IRR and the abnormal returns of current funds. It is clear to see that firms that perform well do have a significant higher chance of raising a new fund and this chance is even higher for firms that have high abnormal returns. This indicates that firms with more skill also have higher chances of raising new funds. These effects account for current funds but do not account for previous funds. These results are in line with

previous literature like Kaplan and Schoar (2005) and Chung et al (2012) who show that performance is positively related to future capital inflows. Further literature shows that performance of previous funds is also related to capital inflows, but the results in this thesis do not support this because the coefficients of the lagged variables are not significant. Furthermore, outperformance of expectations has a nonlinear relationship with the chance of raising a fund. This implies that as the abnormal returns get higher, the probability of raising a new fund eventually decreases. The significant

coefficients for current funds support hypothesis 1 and 2, but this only accounts for the current fund. The rational learning model presented by Berk and Green (2004) and Chung et al (2012) supports the hypothesis that investors learn from previous performance and that firms with higher fund sequences are able to easier raise new funds. Against expectations, the results show that investors do not look solely at the number of funds that a firm already managed when they decide where to invest. Therefore, these results contrast with hypothesis 3. The insignificant coefficient of the current fund number proves that when holding performance constant, investors do not use the current fund number. The Kaplan – Meier graph shows that firms with higher fund numbers have a higher survival rate, but this does not solely come from the fund numbers. Finally, the results from the logistic regression show that the effects of the net IRR and the abnormal returns do have a

significantly larger effect for BO funds than for VC funds.

Like the logit model, the results from the survival analysis show that the net IRR and the abnormal returns also have a positive and significant effect on the survival rate of PE firms. When controlling for the survival bias, it is clear to see that firms that perform well in terms of net IRR and abnormal returns have a lower chance of raising their final fund. The results are in line with

hypothesis 4 and the results from the logit model, although the effect is lower which is because there is control for firms that do not survive past a certain point. Finally, the results show greater effects for BO funds than for VC funds. The positive and significant results from the survival analysis are in line with the expectations drawn from previous literature. A positive relationship between performance and fundraising was already found, but the effect on survival of PE firms was not yet examined this

way.

In addition to the logit model, to make a distinction between direct and indirect/spurious effects of the variables in the logit model, the Karlson, Holm, Breen (KHB) method is used. With the KHB method, a distinction can be made between the direct effect of variables on the probability of raising a new fund and the indirect effect. The indirect effect is the effect that goes through the control variables. The results of the KHB test are presented in appendix C. What can be seen from the table is that the effect of the net IRR decreases when adding control variables. This means that other factors also play a significant role in the probability of raising new funds but there is good control for these factors. Furthermore, the fund number does have a significant relationship with fundraising but when there is control for market and fund type factor, this significance disappears.

Conclusion

Fundraising is essential for the existence and survival of PE firms. Whether the market conditions are favorable or bad, PE firms do want to raise a significant amount of money to invest in order to generate their profits. This paper tries to find out if there is a relationship between the performance of PE firms and the probability of raising new funds and their survival rates. When holding the market conditions constant, it is interesting to see if investors consider performance and skill of a PE firm when they decide to invest.

A graphical analysis shows that there is a nonlinear relationship between the net IRR and the probability of raising a new fund. Funds that are able to raise more funds clearly show a solid increase in their returns over the span of their funds which eventually lead to a peak. After the peaks, there is a steep decline in returns after which the firms are no longer able to raise new funds. By using a logit model this thesis shows, when holding market conditions constant, returns and abnormal returns of the PE firms do have a positive significant effect on the chance of raising a new fund. More specific, by using a logit model the effects of returns and outperformance on the probability of fundraising are estimated and both show a positive relationship. The effect of the abnormal returns is even higher which indicates that firms that are able to outperform the average expected returns have a higher chance of raising new funds. These abnormal returns do however have a nonlinear relationship with the chance of raising a new fund, which indicates that as they get higher, the chance of raising a new fund may eventually become negative. Furthermore, the effects are highest for buyout funds. Unlike the significant effects of current funds’ performance measures, the effects of performance of previous funds are not significant.

Furthermore, the results from a survival analysis show that performance and skill do not only have a positive effect on the probability of raising new funds, but also on the chance of survival when controlling for the survival bias. When the survival is examined for all firms separately, it can be concluded that firms that have generated higher returns and higher abnormal returns are able to raise

more funds and survive longer no matter what the market conditions are. The main conclusion from this thesis is that PE firms that are able to raise a lot of funds and survive longer are not only able to do so because the market conditions are prosperous or by luck, but also because they generate high returns and outperform the average expected returns.

The main limitation to this research is that Preqin only provides information on a fund-level basis. With fund-level base data, this research does not consider aggregate performance and skill and it does not investigate whether certain funds do outperform the market or not. Because the sample lacks cash-flow level data, the PME cannot be calculated and this is necessary for this type of research. It would be very interesting for future research to see if the outperformance of the market leads to lower or higher chances of hazard. In addition, other measures of skill as used in previous literature are also interesting to use in a survival analysis.

This study shows in several graphs that the net IRR for firms with high fund sequences is very volatile. A following research on this volatility, if one is able to predict net IRR peaks and the

manager’s skilled reaction on this would be a very interesting study. The net IRR in itself was not the focus of this study and therefore, this subject is still open for debate.

Finally, performance persistence and survival rates are very interesting to examine. This study only focuses on performance measures for funds on themselves, however it is most likely that

performance persistence over several funds influences survival too. This would be a very interesting topic to examine in further research.

Reference List

Berk, J. B., & Green, R. C. (2004). Mutual fund flows and performance in rational markets. Journal of political

economy, 112(6), 1269-1295.

Berk, J. B., & Van Binsbergen, J. H. (2015). Measuring skill in the mutual fund industry. Journal of Financial

Economics, 118(1), 1-20.

Braun, R., Jenkinson, T., & Stoff, I. (2017). How persistent is private equity performance? Evidence from deal-level data. Journal of Financial Economics, 123(2), 273-291.

Buchner, A., Mohamed, A., & Schwienbacher, A. (2016). Does risk explain persistence in private equity performance?. Journal of Corporate Finance, 39, 18-35.

Carhart, M. M. (1997). On persistence in mutual fund performance. The Journal of finance, 52(1), 57-82. Chung, J. W., Sensoy, B. A., Stern, L., & Weisbach, M. S. (2012). Pay for performance from future fund flows: the case of private equity. Review of Financial Studies, 25(11), 3259-3304.

Cremers, K. M., & Petajisto, A. (2009). How active is your fund manager? A new measure that predicts performance. Review of Financial Studies, 22(9), 3329-3365.

Fama, E. F., & French, K. R. (2010). Luck versus skill in the cross‐ section of mutual fund returns. The journal

of finance, 65(5), 1915-1947.

Gompers, P. A., & Lerner, J. (1999). What drives venture capital fundraising? (No. w6906). National bureau of economic research.

Harris, R. S., Jenkinson, T., Kaplan, S. N., & Stucke, R. (2014). Has persistence persisted in private equity? Evidence from buyout and venture capital funds.

Harris, R. S., Jenkinson, T., & Kaplan, S. N. (2014). Private equity performance: What do we know?. The

Journal of Finance, 69(5), 1851-1882.

Kaplan, S. N., & Schoar, A. (2005). Private equity performance: Returns, persistence, and capital flows. The

Journal of Finance, 60(4), 1791-1823.

Korteweg, A., & Sorensen, M. (2017). Skill and luck in private equity performance. Journal of Financial

Economics.

Kosowski, R., Timmermann, A., Wermers, R., & White, H. (2006). Can mutual fund “stars” really pick stocks? New evidence from a bootstrap analysis. The Journal of finance, 61(6), 2551-2595.

Ljungqvist, A., Hochberg, Y., & Vissing-Jorgensen, A. (2009). Informational hold-up and performance persistence in venture capital.

Phalippou, L., & Gottschalg, O. (2009). The performance of private equity funds. Review of Financial

Studies, 22(4), 1747-1776.

Phalippou, L. (2010). Venture capital funds: Flow-performance relationship and performance persistence. Journal of Banking & Finance, 34(3), 568-577.

Appendix

A graphical view of the funds per vintage year in the sample.

B.

Results of the proportional hazard rates assumption tests.

CHI2 df Prob>CHI2 0,25 3 0,9687 CHI2 df Prob>CHI2 0,23 4 0,9939 CHI2 df Prob>CHI2 0,25 3 0,9687 CHI2 df Prob>CHI2 0,23 4 0,9938

C.

Results from the KHB test. The reduced effect is the effect when no controls are added and the full effect when controls are added.

(1) (2) (3) (4) (5) (6) (7)

VARIABLES Net IRR Net IRR2 Net IRRN-1 Abnormal

returns Abnormal returns 2 Abnormal returns N-1 Sequence

Reduced 0.0107*** 7.31e-06 0.000443 0.00747* 4.78e-06 -0.00138 -0.237*** (0.00388) (5.44e-06) (0.00225) (0.00395) (5.68e-06) (0.00213) (0.0907) Full 0.00775** -1.56e-05 -0.000436 0.00793** -1.62e-05* -0.000305 -0.0585 (0.00393) (9.76e-06) (0.00233) (0.00397) (9.84e-06) (0.00228) (0.0866) Diff 0.00298 2.29e-05* 0.000879 -0.000463 2.10e-05* -0.00107 -0.178***

(0.00191) (1.27e-05) (0.00168) (0.00191) (1.27e-05) (0.00167) (0.0347)