Does being focused in private equity pay off?

Amsterdam Business School MSc. Business Economics, track Finance

Master Thesis

Does Being Focused in Private Equity pay off?

Name: Thomas Helder Student number: 0605077 E-mail: tchelder@hotmail.com Supervised by: Dr. Jens Martin, UvA

Date: 15-08-2016

Presented to the Faculty of Economics and Business University of Amsterdam

2 Statement of Originality

This document is written by Student Thomas Helder who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

3 Abstract

In private equity still many questions are left unanswered. The performance of private equity has been the topic of an increasing quantity of research. One of the more prominent questions is what influences the performance of private equity, this research tries to contribute to clarifying this question. The strategies chosen by the funds, either a diversified investment strategy or specialized, either a fund of other funds or a direct fund. And also choosing the stage of the lifecycle of the target companies to invest in. This research shows that the mere fact that a fund is a direct fund has a positive effect on fund performance. In addition this research shows that diversifying over industries may be beneficial to the performance as well.

4 Table of contents

Introduction ... 5

1. Literature review ... 7

1.1 Private Equity funds ... 7

1.2 Factors that influence fund returns ... 10

1.3 This paper ... 11

2. Methodology... 13

2.1 Base analyses - Specialist ... 13

2.2 Deepening analyses - Specialist ... 17

2.3.1 Quantiles ... 17

2.3.2 Time frames ... 18

2.3.3 Fund Value groups ... 19

2.3 Base analyses – type of investor ... 19

2.4 Deepening analyses – type of investor ... 20

3. Data and descriptive statistics ... 21

4. Results ... 24

4.1 Base analyses – Specialist ... 24

4.2 Deepening analyses – Specialist ... 25

4.2.1 Quantiles ... 25

4.2.2 Time frames ... 25

4.2.3 Fund Value groups ... 26

4.3 Base analyses – type of investor ... 28

4.4 Deepening analyses – type of investor ... 28

4.4.1 Quantiles ... 28 4.4.2 Time frames ... 28 5. Conclusion ... 30 6. Discussion ... 34 References ... 36 Appendix ... 38 Appendix A.1 ... 38 Appendix A.2 ... 39 Appendix B ... 41

5 Introduction

Private equity is an investment class that has gained attention in the academic literature. Due to the nature of private equity, being very private about the transactions, the data that is available is limited. The amount of research done on private equity is increasing, one of the more difficult matters on private equity is the performance. More specifically, what are the factors that influence performance in private equity? When a fund is created a strategy needs to be in place. This strategy gives shape to the identity of the fund.

Funds can adopt a diversification strategy, investing in other private equity funds, these are the fund of funds. Or a fund can invest the capital under management itself, in which case it is a direct fund. Direct funds can then choose to focus on specific industries (being a specialist) or decide not to limit themselves to inductees (being a generalist). In addition funds can focus on a specific entry point on the target company’s lifecycle. Investing in startups in an early stage, providing seed capital for example, is a different investment from investing in proven concepts that need capital to grow. Here too a fund can choose to specialize or

generalize their strategies.

The strategy that a fund chooses can affect its profitability, one type of investing can just be more profitable or less risky than another. Research on private equity has given some indications on the implication of strategy choices, although there is only little consensus. This research tries to find an interpretation of the effect of these choices on fund performance, specifically if having a narrowed down focus is profitable in private equity.

To find if adopting one strategy over the other has an impact on performance of private equity funds, and find if specialization contributes to the performance, first a base analysis is made using robust regressions. The independent variables used are two

performance measures, net IRR of the fund and the IRR net of the benchmark’s IRR. In the first base analyses the effect of being a direct fund rather than a fund of funds is tested. This directly contributes to the existing literature, for there is no consensus on this topic yet. Next the effect of choosing a specialist strategy is analyzed in the same way. Then the extent of the specialist strategy is tested, adding an independent variable that indicates on how many industries a particular fund is focused on. A comprehensive analysis is performed testing both the effect of the direct variable and the variable that counts the industries on the performance.

To deepen the analysis the effect of the comprehensive test is used to compare peer groups. Peer groups in this sense are groups that have similar performance. There are many factors influencing the performance of the fund, among which is the experience of the general

6 partner (GP). To compare the impact across funds with similar results the effect can be tested for persistence in this dimension. Next the funds are grouped in two bins depending on the vintage year. The vintage years contain information about, amongst other things, the economic climate when the fund was founded. To see if the effect of the independent variables are consistent across time the two time bins are compared. Lastly, the funds are grouped by fund size (committed capital to the fund). Four groups are made and the effects are tested to be consistent across this third dimension.

To gain more insight in the effect of choosing a focus as a fund, the above procedure is repeated for three different independent variables. The variables used in this section are indicators for which stage of the lifecycle of the invested company a fund focusses on. The three stages are: early stage, growth stage and late stage. The question at hand, in this section, is if it matters for fund performance on which stage they focus on.

The data used in this research is retrieved from Preqin and is a commercially available dataset. The data in the Preqin database are either self-reported by the funds or retrieved from the limited partner (LP). The limited partner is the supplier of capital for the fund, committing capital to the fund.

In part 1 the literature on private equity is reviewed. Research that has been done on specific parts of private equity are related to each other. For example, the literature on the performance of fund of funds. In part 2 the methodology used in this research is explained. Describing the details of the regressions performed. In this section the two base analyses are described in detail, and the subsequent analyses that are derived from that are explained as well. The next part, part 3 is devoted on the data. Here the dataset retrieved from Preqin is described in depth. And the variables used in the analyses are elaborated on. Followed by part 4, where the results of the analyses for both the base analyses and the in depth analyses are discussed. Then, in part 5 the conclusions from this research are drawn and placed in the existing literature. To conclude with part 6, where a discussion is raised on the validity of this research, research on private equity as a whole and suggestions are made for future research.

7 1. Literature review

In recent year Private Equity (PE) firms received gaining attention of researchers. The specific asset class of PE is rather different from public equity or bond investing. Public equity markets have been researched extensively and advanced methods have been developed to gain insight in the risk and returns in these markets. In PE there are still many questions left unanswered.

The data needed to research PE has been one of the most limiting factors. PE firms do not have the same obligations to disclose financial data like public firms do. Also, and this has been a topic of debate, the data from PE does not have the same standard form like public equity market data has. The internal rate of return (IRR) of many PE firms that are known are mostly self-reported or gathered from the LP (which makes it indirect information). Data from publicly traded firms are retrieved either from the exchange on which the company is traded or from specialized firms like Bloomberg or Thomson Reuters. Not only is the standardized data a problem for PE firms but also the frequency, again, equity market data is available by the second where PE data, is per quarter or per year, if available at all.

1.1 Private Equity funds

Fund returns

Investments in PE are different from public equity investments, the returns of PE investments have been topic of much research in recent years. Gompers and Lerner (1999) are among the first to research the revenue based performance of PE firms. They research 419 limited partnerships and find that the compensation of smaller and for newer funds is less sensitive to performance. Perhaps even more striking: they find no evidence for a relation between incentive compensation and performance. This indicates that the GP managing the fund does not perform better when the compensation is higher, in other words, compensation of the GP is not a factor in determining fund performance.

The performance of PE has been a topic of research in the years after Gompers and Lerner published their paper. Kaplan and Schoar (2005), Ljungqvist and Richardson (2003), Groh and Gottschalg (2005) and more recent Harris, Jenkinsin and Kaplan (2014) all find that PE firms outperform the public equity markets. Although the public equity markets is another type of asset class, a comparison is made because it is a valid investment alternative.

A positive Jensen’s alpha is how Groh and Gottschalg (2005) describe the significant excess returns by PE firms that is found in their research. To come to this conclusion they

8 used 199 cash flows from 133 transactions that where completed in the US between 1984 and 2004. The risk adjusted performance measure indicates that investments in PE are indeed more profitable that investments in public equity markets.

Ljungqvist and Richardson (2003) used actual cash flows of 73 funds between 1981 and 1993 for their analysis. They describe that it takes several years for the committed capital to be invested and up to ten years before the capital is generating excess returns. They find that for the capital that is committed and subsequently invested does generate excess returns over the public equity market in the order of five percent per year. Moreover they suggest that this might be a compensation for the illiquidity for this asset class.

Kaplan and Schoar (2005) used a dataset of individual fund returns (from the US) that was collected by Venture Economics. They find that their estimates suggest that the PE firms, both buyout firms and venture capital firms performance exceeds the S&P500, when

considered gross of fees. They do, though, recognize that there might be a selection bias that underestimated the market risk of the S&P500. The results are in line however with other studies and therefore still indicate that indeed PE investments do generate excess returns.

Harris, Jenkinsin and Kaplan (2014) use data of nearly 1400 funds in the US by retrieved from Burgiss. They finds great movements in performance and finds that

macroeconomic factors do play a role in the performance, but overall the researched funds have excess returns in the order of three percent per year.

Direct Fund or Fund of Funds

PE firms can be subdivided into two categories: direct funds and fund of funds. A direct fund invests its capital directly into the companies they want to invest in, a fund of funds invests its capital into direct funds. The fund of funds type of firm acts as an intermediary trying to find those direct funds that are expected to perform best. This

intermediary function is the topic of research, because the limited partner can also choose to invest in direct funds too. This would suggest that fund of funds have either superior

information or superior skill in selecting the direct funds. Because the fund of funds charges fees these costs needs to be compensated by the performance of the direct funds they invest in.

The same kind of division can be made for the more extensively researched public equity markets, where in general the same principals hold. Mutual funds are the public equity

9 markets counterpart of fund of funds in PE. From Carhart (1997) it seems that there is no existence of fund managers with superior skill or information. This conclusion is also

supported by the work of Fama and French (2008), who also conclude that costs for managing a portfolio is not justified by the costs for the management.

PE is another asset class and the market for PE deals does not behave as the market in public equity, which approaches perfect market conditions. For public equity markets the management of a mutual fund does not seem to contribute to the returns on the invested capital. Harris, Jenkinson, Kaplan and Stucke (2015) start with comparing PE funds to the equity markets. They finds that, net of fees, PE firms outperform the public equity market. For the fund of funds type of fund they specifically mention the search and monitoring costs. From an economics perspective these costs would be one of the more important factors why a fund of funds should be better in selecting a direct fund that the limited partner would be. The performance of a fund of funds is lower, on average, than a portfolio of direct funds.

Gresch and Wyss (2011) have researched the performance of both type of funds and made a comparison. This research is based on a dataset of 1641 PE funds that were raised between 1979 and 2005. In this research real fund data is used as opposed to synthetic or simulated fund of funds data. The results of this research is that, when based on IRR, the fund of funds exhibits a more favorable risk-reward profile than direct funds. In addition this research shows that both vintage year and economic conditions influence the performance of a fund, direct or otherwise.

There is no conclusive evidence that indeed a fund of funds performs better, or worse, that a direct fund. The diversification possibilities when investing in a fund of funds could be outweighed by the search and monitor costs associated with investing in direct funds.

Specialist versus generalist

Direct funds can be divided in yet another dimension. A direct fund can be a specialist in a certain industry or a generalists. A generalist does not limit its investments to a certain industry or a limited number of industries. The same diversification question rises with this division of PE firms as with the fund of funds type of PE firms.

Stein (1997) concludes after researching this very topic that investing across industries is an important element of cross industry capital allocation. This is important because the investment opportunities might, at some point, be poor in one industry and not being limited

10 to this industry leaves the possibility to invest in other industries as well. This enables the PE firm to maintain the budget and generate positive returns. Being limited to a narrow array of industries, a PE firm might end up investing in projects with negative net present value (NPV).

On the other hand Rajan, Servaes, and Zingales (2000) and Scharfstein (1998) find that diversified firms have a difficult time redeploying capital into better performing sectors. This suggests that a specialist, as opposed to a generalists, might be better able to spot profitable investment opportunities. Which still gives no convincing argument to the benefit or being a specialist of generalist.

Gomers, Kovner, Lerner (2009) have addressed this very question and found that specialists outperform generalists, at least for venture capital firms. The deterioration of performance when becoming less specialized appears to be the consequence of two factors: inefficient allocation of funds and poor selection of investments within industries.

1.2 Factors that influence fund returns

Performance of private equity is determined by many factors, some general and some specific to the particular fund. There are several factors identified that are of influence on the performance of PE funds, general factors are the geographical distance between the investing firm and the firm in which is being invested, but also economic factors. The experience of the manager is a more specific factor that influences the fund performance.

Geographical differences

The geographical location of the fund matters according to Chen, Gompers, Kovner and Lerner (2010) and to Harris, Jenkinson and Kaplan (2014). The latter study shows that the performance of US based funds and EU based funds do differ but not by much. The fact that there is a difference indicates that it matters if the fund is located in the EU or in the US. The former mentioned study, shows that it not only matters where the fund is located but also where the investments are made.

Monitoring costs are identified as the main cost drivers when it comes to cost drivers associated with geographical distance between the GP and the company that is invested in (Lerner, 1995). In addition to monitoring costs informational flows form the investor to the invested is a relevant factor. Bengtsson and Ravid (2009) finds that venture capital contract are more high-powered as geographical distance increases. Indicating that costs associated with the transfer of information and monitoring costs decreases when distance decreases.

11 Macroeconomic influences

Economic conditions have an effect, for example, on the availability of funds and subsequently on the profitability of the investments (Axelson, Jenkinson, Strömberg and Weisbach, 2013). Berger and Udell (1998) find that the optimal capital structure is different at different points on the growth cycle. And show that the macroeconomic conditions change optimal financing structures. This partially explains the performance of PE funds as their goal is to capitalize any deviation from the optimum.

Funds that stated in boom times are likely to perform worse on average than funds that did not start in boom times and are less likely to raise follow-on funds (Kaplan and Schoar, 2005). Also they indicate that market entry for PE firms is cyclical. Robinson and Sensoy (2011) confirm the findings of the underperformance of funds raised in boom times. Fund specific factors

Gompers, Kovner and Lerner (2009) researched factors that are of influence on fund performance, one of the factors apart from the specialization question is that the experience of the GP is important to the fund performance too. An experienced manager is able to generate more return and is more able to generate follow-on funds.

Kaplan and Schoar (2005), did not only find that the economic conditions are of influence of the fund performance, as discussed above, they also noted that well performing funds are more likely to raise follow-on funds. If a fund, therefore, is a follow-on fund it is likely to perform better than first time funds. Indeed this is what is confirmed in the work of Kaplan and Schoar (2005) as well. In fact one of their conclusions is that heterogeneity in skills and quality of the GP can lead to heterogeneity in performance.

1.3 This paper

This research aims to find an answer to the question if specialization if beneficial in PE. The literature on PE has shown that PE is a valid alternative to investing in public equity markets, it even suggest that it is more profitable in general. The difference in performance among PE funds is a field of study that has still left many questions unanswered. This

research tries to add and contribute to gaining insight in the differences within PE focusing on differences in performance.

First the difference in performance between fund of funds and direct funds is addressed. The literature for mutual funds in public equity markets is fairly clear, the

12 literature on fund of funds is not. The diversification matter is at the heart of this research and underlying the overall question if specialization contributes to fund performance. The first hypothesis that is being tested is: direct funds perform better than fund of funds.

Second, the strategy rather than the identity of a PE fund is researched. A PE fund can adopt a generalist strategy or a specialist identity. The hypothesis tested is: specialists perform better than generalists. In addition the extent of being a specialist is researched. The extent of being a specialist in this research is defined as the number of industries a particular fund invests in. The hypothesis is: focusing on more industries has a negative effect on fund performance. This hypothesis suggest that being focused on a larger number of industries has a negative effect on the performance of the fund.

Lastly a deep dive into the funds strategy is made, researching the effect of adopting a specific strategy based on the stage of life of the company that fund invests in. Much like choosing a particular industry to focus on, a PE fund can choose to focus on certain stages of a company’s life to invest in. In this research three subsequent stages are singled out and tested against each other. The hypothesis is: investing in different parts of the lifecycle of a company has an effect on fund performance. Investing in younger companies is generally regarded as investing in proven concepts or more mature companies.

13 2. Methodology

The literature suggests several characteristics of PE firms that need to be dealt with before being able to test the theory. To research if specialization adds to profitability data is needed on both the profitability, measures of specialization and control variables. The data used is a commercially available set retrieved from Preqin.

The data set indicates both a fund ID and a firm ID. Because this researched is focused on the performance of funds, only fund-level data is used. A PE company can have multiple funds and with each fund a different strategy. Therefore the funds are researched, not the PE companies. As a performance measure of the funds Preqin provided data on the net IRR of the fund, a fund benchmark and in addition an IRR indicator net of the benchmark. Throughout this study the analyses will be performed twice, once for the net IRR as a dependent variable and once for the IRR net of the benchmark performance. This is done so the results of the analysis can be seen both in absolute IRR and relative to the benchmark.

The Preqin dataset provides variables indicating the location of the funds headquarters and in which part of the world the fund is active. Also provided in this set is information about the funds chosen strategy in terms of direct investments or fund of funds, and also in the stage of life of a company the fund invests in. In addition the dataset provides information on whether or not the fund is a generalist or specialist, and if it is a specialist the industries the fund invests in.

2.1 Base analyses - Specialist

Control variables

There are three control variables used to move away influences on the performance caused by other factors than what is being tested. The first control variable is the natural logarithm of the fund’s value. The fund’s value is the size of the fund in terms of committed capital. The natural logarithm of the value is taken for the fund value ranges from several hundreds of thousands of committed capital to tens of billions of committed capital. From the literature on PE it is clear that the fund size is an important factor in the performance of PE and therefore it is controlled for.

The second control variable is if the funds headquarters is located in the same area as where it makes its investments. The literature shows that the performance of a fund is influenced positively if the fund invests in the same region as where the headquarters is located. Because the geographical difference between the investment activities and the

14 headquarters of the fund does not influence the way or the extent that the fund is specialized this effect is controlled for.

A third control variable is the vintage year of the fund is of influence of the

performance of the fund. Funds that stated in boom times might have gotten easy access to capital to stat the fund, and from the literature it is known that in general funds that started in boom years perform worse. In addition the capital flow to more illiquid investments can be due to macroeconomic conditions. The vintage year is used as a proxy to capture the effects of the economic conditions at the time the find was founded.

Direct Fund or Fund of Funds

First the performance of direct funds is researched. Their performance is regressed trough a robust regression on the performance of Fund of Funds, controlled for the four variables mentioned above. The robust regression estimates the standard errors using the Huber-White sandwich estimators. These estimators can deal with minor deviations from the standard assumptions for a regression analysis. Data on PE is commonly less neatly structured than for public equity markets for example, therefore to overcome small deviations from the standard assumptions this method is used. The empirical specification is as follows:

𝐼𝑅𝑅 = 𝛼 + 𝛽₁(𝐷𝑖𝑟𝑒𝑐𝑡) + 𝛽₃₂(𝐿𝑛𝐹𝑢𝑛𝑑𝑉𝑎𝑙𝑢𝑒) + 𝛽₃(𝐺𝑒𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐𝑎𝑙𝐹𝑜𝑐𝑢𝑠) + 𝛽4(𝑉𝑖𝑛𝑡𝑎𝑔𝑒) + 𝜖

The 𝐼𝑅𝑅 variable is the net IRR of the specific fund as reported in the dataset. The 𝐷𝑖𝑟𝑒𝑐𝑡 is a binary variable indicating if a fund is a Direct fund, then it the binary indicates 1 or a Fund of Funds then the binary indicates 0. 𝐿𝑛𝐹𝑢𝑛𝑑𝑉𝑎𝑙𝑢𝑒 is the natural logarithm of the fund value. The 𝐺𝑒𝑜𝑔𝑟𝑎𝑝𝑔𝑖𝑐𝑎𝑙𝐹𝑜𝑐𝑢𝑠 variable indicated if the fund has its headquarters in the same geographical area as where it has its investing activities. Indicating 1 if this is the case and 0 otherwise. 𝑉𝑖𝑛𝑡𝑎𝑔𝑒 is the vintage year, the year the fund was founded.

With this analysis the general performance of the direct funds is measured against the performance of the Fund of Funds. Controlling for the four variables specified the outcome should contribute to interpreting the balance of specialization and diversification. The existing literature is not conclusive as to what the outcome might be. Based on the literature on mutual funds in public equity markets and the little literature on this topic for PE, the outcome of this analysis is expected to be a positive 𝛽1.

15 The same analysis is repeated with the IRR net of the IRR from the benchmark as a dependent variable.

𝐼𝑅𝑅_{𝑛𝑒𝑡 𝑜𝑓 𝐵𝑀} = 𝛼 + 𝛽₁(𝐷𝑖𝑟𝑒𝑐𝑡) + 𝛽₂(𝐿𝑛𝐹𝑢𝑛𝑑𝑉𝑎𝑙𝑢𝑒) + 𝛽₃(𝐺𝑒𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐𝑎𝑙𝐹𝑜𝑐𝑢𝑠) + 𝛽₄₅(𝑉𝑖𝑛𝑡𝑎𝑔𝑒) + 𝜖

The 𝐼𝑅𝑅𝑛𝑒𝑡 𝑜𝑓 𝐵𝑀 is the IRR of the fund minus the IRR of the benchmark specified by Preqin. The control variables are the same as in the previous specification. In this analysis the effect of being a direct fund is expected to be positive too.

Specialist versus generalist

The specialist is the fund that focusses on a limited number of industries and the generalists does not focus at all. The analysis of the performance of a generalist against the performance of the specialist is similar to the previous analysis. Here a robust regression us used , the specifications of the two regressions is as follows:

𝐼𝑅𝑅 = 𝛼 + 𝛽₁(𝑆𝑝𝑒𝑐𝑖𝑎𝑙𝑖𝑠𝑡) + 𝛽₂(𝐿𝑛𝐹𝑢𝑛𝑑𝑉𝑎𝑙𝑢𝑒) + 𝛽₃(𝐺𝑒𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐𝑎𝑙𝐹𝑜𝑐𝑢𝑠) + 𝛽4(𝑉𝑖𝑛𝑡𝑎𝑔𝑒) + 𝜖

𝐼𝑅𝑅_{𝑛𝑒𝑡 𝑜𝑓 𝐵𝑀} = 𝛼 + 𝛽₁(𝑆𝑝𝑒𝑐𝑖𝑎𝑙𝑖𝑠𝑡) + 𝛽₂(𝐿𝑛𝐹𝑢𝑛𝑑𝑉𝑎𝑙𝑢𝑒) + 𝛽₃(𝐺𝑒𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐𝑎𝑙𝐹𝑜𝑐𝑢𝑠) + 𝛽₄(𝑉𝑖𝑛𝑡𝑎𝑔𝑒) + 𝜖

The 𝐼𝑅𝑅 is the net IRR of the specific fund as reported in the dataset. The

𝐼𝑅𝑅_{𝑛𝑒𝑡 𝑜𝑓 𝐵𝑀} is the IRR of the fund minus the IRR of the benchmark specified by Preqin. The 𝑆𝑝𝑒𝑐𝑖𝑎𝑙𝑖𝑠𝑡 is a binary variable indicating 1 if a fund is a specialist, thereby focusing on a limited number of industries in which it invests in. The variable indicates 0 if the fund is a generalist. 𝐿𝑛𝐹𝑢𝑛𝑑𝑉𝑎𝑙𝑢𝑒 is the natural logarithm of the fund value. The

𝐺𝑒𝑜𝑔𝑟𝑎𝑝𝑔𝑖𝑐𝑎𝑙𝐹𝑜𝑐𝑢𝑠 variable indicated if the fund has its headquarters in the same

geographical area as where it has its investing activities. Indicating 1 if this is the case and 0 otherwise. 𝑉𝑖𝑛𝑡𝑎𝑔𝑒 is the vintage year, the year the fund was founded.

With this analysis the impact of adopting the “specialist” strategy is researched. Controlling for the four variables specified the outcome should contribute to interpreting the balance of specialization and diversification. The 𝛽1 form this analysis is expected to be

16 positive. Based on the literature specialists seem to outperform generalists. Also, on this topic there is no consensus, but the little research that has been done suggests that specialists perform better.

Next, an analysis is done on specialist funds. It is indicated on how many industries the fund has indicated it focusses on. Therefore, different from the above analyses, it is possible to give a sense of magnitude to the effect of being specialized. The empirical specification of the two regressions is as follows:

𝐼𝑅𝑅 = 𝛼 + 𝛽1(𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡) + 𝛽2(𝐿𝑛𝐹𝑢𝑛𝑑𝑉𝑎𝑙𝑢𝑒) + 𝛽3(𝐺𝑒𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐𝑎𝑙𝐹𝑜𝑐𝑢𝑠) + 𝛽₄(𝑉𝑖𝑛𝑡𝑎𝑔𝑒) + 𝜖

𝐼𝑅𝑅𝑛𝑒𝑡 𝑜𝑓 𝐵𝑀 = 𝛼 + 𝛽1𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡 + 𝛽2(𝐿𝑛𝐹𝑢𝑛𝑑𝑉𝑎𝑙𝑢𝑒) + 𝛽3(𝐺𝑒𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐𝑎𝑙𝐹𝑜𝑐𝑢𝑠) + 4(𝑉𝑖𝑛𝑡𝑎𝑔𝑒) + 𝜖

The 𝐼𝑅𝑅 is the net IRR of the specific fund as reported in the dataset. The

𝐼𝑅𝑅_{𝑛𝑒𝑡 𝑜𝑓 𝐵𝑀} is the IRR of the fund minus the IRR of the benchmark specified by Preqin. The 𝐼𝑛𝑑𝑢𝑠𝑟𝑦𝐶𝑜𝑢𝑛𝑡 is a variable indicating how many industries the fund has indicated or

reported to be active in. 𝐿𝑛𝐹𝑢𝑛𝑑𝑉𝑎𝑙𝑢𝑒 is the natural logarithm of the fund value. The 𝐺𝑒𝑜𝑔𝑟𝑎𝑝𝑔𝑖𝑐𝑎𝑙𝐹𝑜𝑐𝑢𝑠 variable indicated if the fund has its headquarters in the same

geographical area as where it has its investing activities. Indicating 1 if this is the case and 0 otherwise. 𝑉𝑖𝑛𝑡𝑎𝑔𝑒 is the vintage year, the year the fund was founded.

The 𝛽₁ from this analysis is expected to be negative, being more diversified, and thus less focused, deteriorates performance. The outcome of this analysis gives a more in-depth analysis of the magnitude specialization has on the performance of PE funds.

Lastly the common effects of both 𝐷𝑖𝑟𝑒𝑐𝑡 and 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡 are analyzed. The 𝑆𝑝𝑒𝑐𝑖𝑎𝑙𝑖𝑠𝑡 variable is left out from this analysis because for the observations in the dataset that are indicated that a fund is a specialist also the number of industries are known. Therefore to get a more comprehensive analysis the variable 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡 is used rather than the 𝑆𝑝𝑒𝑐𝑖𝑎𝑙𝑖𝑠𝑡. The empirical analyses are:

𝐼𝑅𝑅 = 𝛼 + 𝛽₁(𝐷𝑖𝑟𝑒𝑐𝑡) + 𝛽₂(𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡) + 𝛽₃(𝐿𝑛𝐹𝑢𝑛𝑑𝑉𝑎𝑙𝑢𝑒) + 𝛽4(𝐺𝑒𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐𝑎𝑙𝐹𝑜𝑐𝑢𝑠) + 𝛽5(𝑉𝑖𝑛𝑡𝑎𝑔𝑒) + 𝜖

17 𝐼𝑅𝑅_{𝑛𝑒𝑡 𝑜𝑓 𝐵𝑀} = 𝛼 + 𝛽₁(𝐷𝑖𝑟𝑒𝑐𝑡) + 𝛽₂𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡 + 𝛽₃(𝐿𝑛𝐹𝑢𝑛𝑑𝑉𝑎𝑙𝑢𝑒)

+ 𝛽₄(𝐺𝑒𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐𝑎𝑙𝐹𝑜𝑐𝑢𝑠) + 𝛽₅(𝑉𝑖𝑛𝑡𝑎𝑔𝑒) + 𝜖

2.2 Deepening analyses - Specialist

The literature on PE performance suggest that performance is influenced by the experience of the GP. Also the literature suggests that funds that started in boom years

perform worse than funs that did not start in boom years. To deepen the analyses specified the regressions are repeated twice more.

2.3.1 Quantiles

First the specified analyses are repeated using quantile regressions. In this way the similarly performing funds are compared filtering out other effects that influence

performance. The literature is clear on the effect the GP has on the performance and that heterogeneity in experience of the GP leads to heterogeneity in performance. By using

quantile regressions the dataset is split into specified quantiles based on performance, both for 𝐼𝑅𝑅 and 𝐼𝑅𝑅𝑛𝑒𝑡 𝑜𝑓 𝐵𝑀. From these analyses the effects of the researched independent

variables on the dependent variables can be compared for different “performance groups”. In this analysis three quantiles are chosen, the 25th, the 50th and the 75th percentile. The standard errors from this analysis are bootstrap standard errors. The empirical specifications are specified below here the 𝜏 indicated the percentile;

For the analysis of the impact of being a direct fund:

𝑄_𝜏(𝐼𝑅𝑅) = 𝛼_𝜏 + 𝛽_𝜏1(𝐷𝑖𝑟𝑒𝑐𝑡) + 𝛽𝜏4(𝐿𝑛𝐹𝑢𝑛𝑑𝑉𝑎𝑙𝑢𝑒) + 𝛽𝜏3(𝐺𝑒𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐𝑎𝑙𝐹𝑜𝑐𝑢𝑠) + 𝛽_𝜏4(𝑉𝑖𝑛𝑡𝑎𝑔𝑒) + 𝜖_𝜏

𝑄_𝜏(𝐼𝑅𝑅_{𝑛𝑒𝑡 𝑜𝑓 𝐵𝑀}) = 𝛼_𝜏 + 𝛽_𝜏1(𝐷𝑖𝑟𝑒𝑐𝑡) + 𝛽𝜏2(𝐿𝑛𝐹𝑢𝑛𝑑𝑉𝑎𝑙𝑢𝑒)

+𝛽_𝜏3(𝐺𝑒𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐𝑎𝑙𝐹𝑜𝑐𝑢𝑠) + 𝛽_𝜏4(𝑉𝑖𝑛𝑡𝑎𝑔𝑒) + 𝜖_𝜏 For the analysis of the impact of being a specialized:

𝑄𝜏(𝐼𝑅𝑅) = 𝛼𝜏 + 𝛽𝜏1(𝑆𝑝𝑒𝑐𝑖𝑎𝑙𝑖𝑠𝑡) + 𝛽𝜏2(𝐿𝑛𝐹𝑢𝑛𝑑𝑉𝑎𝑙𝑢𝑒) + 𝛽𝜏3(𝐺𝑒𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐𝑎𝑙𝐹𝑜𝑐𝑢𝑠) + 𝛽𝜏4(𝑉𝑖𝑛𝑡𝑎𝑔𝑒) + 𝜖𝜏

18 𝑄𝜏(𝐼𝑅𝑅𝑛𝑒𝑡 𝑜𝑓 𝐵𝑀) = 𝛼𝜏 + 𝛽𝜏1(𝑆𝑝𝑒𝑐𝑖𝑎𝑙𝑖𝑠𝑡) + 𝛽𝜏2(𝐿𝑛𝐹𝑢𝑛𝑑𝑉𝑎𝑙𝑢𝑒)

+𝛽𝜏3(𝐺𝑒𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐𝑎𝑙𝐹𝑜𝑐𝑢𝑠) + 𝛽𝜏4(𝑉𝑖𝑛𝑡𝑎𝑔𝑒) + 𝜖𝜏

For the analysis of the impact of the number of industries:

𝑄_𝜏(𝐼𝑅𝑅) = 𝛼𝜏+ 𝛽𝜏1(𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡) + 𝛽𝜏2(𝐿𝑛𝐹𝑢𝑛𝑑𝑉𝑎𝑙𝑢𝑒) + 𝛽_𝜏3(𝐺𝑒𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐𝑎𝑙𝐹𝑜𝑐𝑢𝑠) + 𝛽_𝜏4(𝑉𝑖𝑛𝑡𝑎𝑔𝑒) + 𝜖_𝜏

𝑄_𝜏(𝐼𝑅𝑅_{𝑛𝑒𝑡 𝑜𝑓 𝐵𝑀}) = 𝛼_𝜏 + 𝛽_𝜏1(𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡) + 𝛽_𝜏2(𝐿𝑛𝐹𝑢𝑛𝑑𝑉𝑎𝑙𝑢𝑒) +𝛽_𝜏3(𝐺𝑒𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐𝑎𝑙𝐹𝑜𝑐𝑢𝑠) + 𝛽_𝜏4(𝑉𝑖𝑛𝑡𝑎𝑔𝑒) + 𝜖_𝜏

For the analysis of the impact of the number of being a direct fund and the number of industries: 𝑄_𝜏(𝐼𝑅𝑅) = 𝛼_𝜏+ 𝛽_𝜏1(𝐷𝑖𝑟𝑒𝑐𝑡) + 𝛽𝜏2(𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡) + 𝛽𝜏3(𝐿𝑛𝐹𝑢𝑛𝑑𝑉𝑎𝑙𝑢𝑒) + 𝛽_𝜏4(𝐺𝑒𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐𝑎𝑙𝐹𝑜𝑐𝑢𝑠) + 𝛽𝜏5(𝑉𝑖𝑛𝑡𝑎𝑔𝑒) + 𝜖𝜏 𝑄_𝜏(𝐼𝑅𝑅_{𝑛𝑒𝑡 𝑜𝑓 𝐵𝑀}) = 𝛼_𝜏 + 𝛽_𝜏1(𝐷𝑖𝑟𝑒𝑐𝑡) + + 𝛽𝜏2(𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡) +𝛽_𝜏3(𝐿𝑛𝐹𝑢𝑛𝑑𝑉𝑎𝑙𝑢𝑒) + 𝛽𝜏4(𝐺𝑒𝑜𝑔𝑟𝑎𝑝ℎ𝑖𝑐𝑎𝑙𝐹𝑜𝑐𝑢𝑠) + 𝛽_𝜏5(𝑉𝑖𝑛𝑡𝑎𝑔𝑒) + 𝜖_𝜏 2.3.2 Time frames

In addition to the quantile regressions there the first analysis, the robust regressions, are repeated for two separate time frames, checking for persistence of the results. The dataset is split into two parts based on the vintage years of the funds. The cutoff year is 2005. The exact same analysis as the base robust regression are repeated.

19 2.3.3 Fund Value groups

One last deepening analysis is made for the specialization analyses. The dataset is split into four groups depending on their funds value. The exact same analysis as the base robust regression are repeated. This specific analysis is made to see if the effect of specialization is the same and of the same magnitude for funds differing in fund size. The funds are split into the grouped as follows:

Fund Value in millions of US dollars

Group1 0.4* up to 100

Group2 100 up to 500

Group3 500 up to 2000

Group4 2000 up to 20365*

* The smallest fund in the dataset is 400.000 US dollars, the largest fund is 20.365 million US dollars.

2.3 Base analyses – type of investor

Being a specialists in a particular industry can be one way of specializing, there is one last other type of specializing that can be analyzed to complete this research. A PE fund, or specifically a Venture Capital firm can specialize in investing in a certain stage of

development of the invested company. In this analysis there are three stages distinguished from each other: Early stage, Growth and Late stage. The early stage is an investor that either provides seed capital, which is the first round of financing for a startup. Or the early stage can be series-A investing, which is the first round of issuing shares. The growth investors are series-B or series-C investors getting on board when the concept is proven and capital is needed to let the company (no longer a startup) grow. The late stage investor is getting on board at even a later stage.

The testing of the influence of the stage the PE invests in is similar to the previous analyses on the influence of the 𝐷𝑖𝑟𝑒𝑐𝑡 and 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡 on 𝐼𝑅𝑅 and 𝐼𝑅𝑅𝑛𝑒𝑡 𝑜𝑓 𝐵𝑀, using the same variables as control variables. The three variables independent variables used in this part are binary variables indicating if the particular fund is investing in the particular stage. Three stages are: Early, Growth and Late. Early stage is the starting phase of a company ranging from investing seed capital up to series-A investments. Growth is typical a series-B and series-C investment, contributing working capital to the fund to make investments to grow the company. Late investing is even later in the developing stage when the company

20 operates on a reasonable scale with a proven concept. The empirical specifications can be found in appendix A.

2.4 Deepening analyses – type of investor

The deepening procedures done with the specialization analysis is repeated for the stage analysis in exactly the same manner. Both the quantile regressions and the persistence check are repeated. The empirical specifications of the quantile regression can be found in appendix A. The deepening analysis that is done in for the specialization analysis where the dataset is split into fund size groups is not repeated for the “type” analysis. This analysis is itself a broadening of the research on specialization therefore de deepening is limited.

21 3. Data and descriptive statistics

The dataset used in this research is retrieved from Preqin, and is a commercial data source. The data in the Preqin database are either self-reported or gathered trough LP’s (mostly public pension investors) based on the Freedom of Information Act. Because of this there is no way of knowing if the information gathered is fully correct, it is therefore

considered to be a proxy for the real data. Tables describing the data are included in appendix B.

Full dataset

The full dataset consists of 7948 individual funds that are included in the Preqin database. These funds have vintage years starting from 1969 and ending as recent as 2016. For the years 1969 up to 1980 there are only little funds per vintage year, the same is true for the years 2014, 2015 and 2016. Therefore only the vintage years ranging from 1980 up to 2013 are included.

After dropping all observations for these years 7317 funds are left of which 6336 funds have reported net IRR values. In table 1 the dependent variable is shown, averaged over the funds per vintage year. To start with the number of funds, it can be seen in this table that the number of funds in this database have strongly increased from 12 funds in 1980 up to 540 in 2007, to decrease again to 310 in 2013. For the years 1980 and 1981 there is no benchmark to set off the net IRR of the funds.

The number of funds that are in this dataset increases over the years up to a maximum of 540 funds in 2007, to decrease again to 310 in 2013. When looking at the mean per vintage year net IRR the highest net IRR is accomplished in 1991 and the lowest in 2006, the year before the most funds are reported. In 1991 the lowest net IRR was only -0.5% where -1.19% is average, the highest was 346.6% where 177.93% is the average. The highest net IRR was reported in 1998, at 1015,7%. In 2006 the lowest mean net IRR was reported, -100% is the lowest reported that year and only 45% is the highest that year.

In terms of netting more IRR over the benchmark Preqin has set for the funds (based on their vintage year, where the headquarters of the fund is located and the type of fund), the average of this dataset is 0.017% therefore we might conclude that the benchmark Preqin has set is rather accurate. In 1998 the benchmark has been beaten by 10.08% by one specific fund, where the average of the best performing funds over the years 1980 up to and including 2013 is 1.7%.

22 Table 2 presents the summary statistics of the variables used in all the analyses in this paper. The variables 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡 and 𝐹𝑢𝑛𝑑 𝑉𝑎𝑙𝑢𝑒 are the only two independent

variables that are continuous. The other independent variables are all binary. The mean value of a fund in this dataset is 594 million US dollars, ranging from 400.00 US dollars for the smallest fund up to over 20 billion US dollars for the largest fund. The average of

the𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡, indicating the number of industries a fund is actively investing in (reportedly), is about 3.6. Therefore from this we can conclude that for the specialized funds, not being diversified funds, the average number of industries is less than four. The exact number of industries is not of interest in this study. What is important to extract from this variable is how the particular fund sees itself, as a specialist of generalist. If a fund is

investing in two funds this could be Internet and Mining, two rather different industries, but, it could also be Internet and E-commerce, which is more related. Limiting investments to a number of industries means that there is a focus on these industries and that is the topic of interest in this study. For more details see panel B in table 3.

From the binary variables, the mean gives the centroid of the sample in this dataset. From the 7317 funds observed for the binary 𝐷𝑖𝑟𝑒𝑐𝑡, the mean is 0.87 indicating that 87% of the funds is a direct fund, as opposed to a fund of funds. Of the 7180 funds observed for the binary variable 𝑆𝑝𝑒𝑐𝑖𝑎𝑙𝑖𝑠𝑡 the mean is 0.674, indicating that 67.4% of these funds are reported to be specialists and therefore limiting the number of industries they invest in. Two more binary variables give insight in the configuration of the dataset. The binary 𝐵𝑒𝑎𝑡𝐵𝑒𝑛𝑐ℎ indicates if a fund beats the benchmark Preqin has set for that particular fund. Interestingly 50,2% of the funds beat the

benchmark. From table 1 it was already apparent that the benchmark that Preqin has set was a rather sound benchmark, this result only confirms this thought. Over 90% of the funds invest in the same geographical region in which the headquarters is located, for more detail see table 3.

In panel A from table 3 the funds are split into the region where they are active in, or if a fund does not specialize it is a diversified fund (on this subject). Only a small number of funds is diversified in this particular manner, namely 59 funds. And also only a limited number of funds invest in Africa or in the Middle East, 62 and 65 respectively. By far the most funds invest in the United States of America, 5300 funds. And Europe is second most popular with 1626 funds investing.

In panel B gives insight in how many firms specialize in how many industries. From table 2 it can be seen that the average is 3.6, but by far the most specialized funds focus on just one industry (namely 1840 funds). The number of funds investing in more than 10 industries is 203 funds, the highest number of industries, as reported in table 2, is 67 and this in only one fund.

23 Preqin had subdivided the funds types into 27 categories which are in combined into 17 categories. In table 4 it can be seen that there are six categories affected by this. The category Co-investments in this study is a combination of Preqin’s: Co-investment, Co-investment multi manager and Real estate co-investment. The category Secondaries used in this study is a combination of Preqin’s: Secondaries, Direct secondaries, Infrastructure secondaries and Real estate secondaries. The Early stage category in this study comprises of Early stage, Early stage: Seed and Early stage: Start-up from the Preqin classification. Infrastructure comprises of Infrastructure and Infrastructure fun of funds, Real estate of Real estate and Real estate fund of funds. Lastly the category Venture in this study comprises of Preqin’s Venture general and Venture debt.

These combinations are made to make the type of funds more categorized and less specified than the Preqin database. The total number of funds for which a net IRR is specified is less than the total number of funds in the dataset, 6336 funds with and 981funds without. From the table it can be seen that the Timber type of fund is the least profitable, netting only 5.07% net IRR. The type

Turnaround seems to be the most profitable at 23.75% net IRR. On average the type of funds perform on the level of 13.57 net IRR, which could also be seen in Table 1.

There is some variation in the returns per category, or type of fund. In this analysis the effect of adopting the Early stage, Late stage or Growth identity is researched. From this table is can be seen that these is indeed a difference in performance. Early stage netting 13.68% IRR on average is in the middle of these three types of funds and is about the average over the entire sample. Late stage nets 10.41%, which is the lowest of these three types and is after Timber the lowest performing type of fund. The Growth type is netting 15.5% and is thereby above average although four out of the 17 perform better than the Growth type.

Overall the sample is concentrated mostly in geographical sense in the US and Europe, 66.7% of the funds are active in the US and 21.5% is active in Europe. The funds perform 13.57% on average net IRR, with a dispersion over the type of funds and the vintage years too. From the vintage year it can be seen that performance varies from just over 5% net IRR up to and over 30% in a peak year. Within the vintage years great differences among funds are shows from -100% up to 1015% net IRR, the mean standard deviation is almost 26%.

24 4. Results

The results from the statistical analyses described in the section Methodology are shown in tables in appendix B.

4.1 Base analyses – Specialist

The result of the robust regressions are summarized in table 5a panel A for the net IRR of the funds as a dependent variable and in table 6a panel A for the IRR net of the

benchmark’s IRR. The effect of being a direct fund on net IRR, rather than not being a direct fund (fund of funds) is indeed positive as expected. The result is a significant result on a one percent level. The adaptation of the identity of specialist, focusing on a limited number of industries, appears to have a negative, statistically significant, result. This, as opposed to the effect of being a direct fund is against the expectations. Looking at the contribution of focusing on one more industry, and thereby being less focusses, appears to have a positive contribution on the net IRR. The result of adding one more industry to the focus of a specialist investor was expected to be negative but is positive even on a one percent level.

When looking at the more complete picture, the regression of both the 𝐷𝑖𝑟𝑒𝑐𝑡 and the 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡 variables, it seems the effect of being a direct fund is even bigger, a 5.15% increase in net IRR for adopting this strategy. The effect of adding an additional industry to the focus of the fund contributes, on average, 0.3% to the net IRR of a fund. This regression’s R-squared is 0.0195, though, this indicates that although the results are significant, this only explains a small portion of the IRR generated by PE funds.

When looking at the same analyses for the IRR net of the benchmark’s IRR the first result is similar. Being a direct fund rather than a fund of funds has a positive effect on the net IRR of the fund. The adaptation of the specialist strategy also contributes to the performance of the fund, here the result is significant as opposed to the net IRR counterpart analysis. This is in line with what was expected. Adding another industry to the focus of a fund, making it less focused, has the same, surprising, result as the net IRR counterpart. This too is opposite from what was expected. The regression of both the 𝐷𝑖𝑟𝑒𝑐𝑡 and the 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡

variables indicates that being a direct fund has, on average, a positive result on beating the benchmarks IRR by 0.05% and adding another industry to focus on adds 0.0015%. The R-square of this analyses is only 0.009 and therefore the explanatory power of this analysis is only small.

25

4.2 Deepening analyses – Specialist

4.2.1 Quantiles

In table 5a panel B the results of the analyses of the quantile regression is shown. The regressions show that for all three groups the results have the same sing, indicating a positive effect on both being a direct fund and adding an industry to the focus of the fund. The results differ from each other in magnitude indicating different effects of these variables in different quantiles of the sample. The results are the smallest for the 25th percentile, but are also not significant. The results for the 50th _{percentile is significant and the influence of being a direct} fund on the performance is of approximately the same magnitude as was the base analysis described above. The result of the 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡 variable is about one third of what the base analysis showed. Still this result is differed from what was expected based on the literature. For the 50th percentile the explanatory power is smaller than the base regression, indicating that for this group of funds the impact of the two independent variables is even less. The result of the analysis for the 75th percentile is slightly higher than what was found in the base analysis. These results for both the 𝐷𝑖𝑟𝑒𝑐𝑡 and the 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡 variables are significant on a one percent level. The R-squared for this analysis is slightly higher than the base analysis, this indicates that the two independent variables explain more of the net IRR of these funds.

In table 6a panel B the results of the quantile analysis on the IRR net of the

benchmark’s IRR is shown. The results and the comparison to the base analysis is rather the same. For the 25th percentile the results are not significant. The results of the 50th quantile are significant and slightly smaller than the result of the base analysis. For the 75th quantile the results are slightly higher too and the result for the 75th quantile explanatory power increased slightly when compared to the base analysis.

The results of the quantile regressions show that for the better performing funds the impact of the two indented variables researched are different from the funds that perform worse. The explanatory power of the results is low in all analyses but it is interesting that the impact of the two variables researched is bigger for funds in the 75th _{quantile than for funds in} the 25th quantile.

4.2.2 Time frames

To see if the effect that was found in the base regression was persistent over time the data when it was split into two groups. The division was made based on the vintage year and the cutoff year was 2005. In table 5b the results are shown for the analyses performed. The

26 results show that the effects are not persistent over time. The base analyses are repeated first for the three independent variables separately, 𝐷𝑖𝑟𝑒𝑐𝑡, 𝑆𝑝𝑒𝑐𝑖𝑎𝑙𝑖𝑠𝑡 and 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡. Then the analysis contained both 𝐷𝑖𝑟𝑒𝑐𝑡 and 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡.

The result of being a direct fund on the IRR of a fund is 4.5% for the funds with a vintage year before 2005 but decreases to 0.19% for funds with vintage year 2005 or later. The result found for the younger funds is not significant though. Adopting the strategy of being a specialist if not rewarding for both years, only the result from the analysis with the younger funds is significant. Both result in a negative estimator. Surprisingly the results for adding an industry to the funds focus is not rewarding for funds with the vintage year before 2005 (statistically not significant) but seems to be adding 0.3% to the net IRR of the fund for funds with vintage year 2005 or later (this is significant on a one percent level). Looking at the more complete analysis including both the 𝐷𝑖𝑟𝑒𝑐𝑡 and the 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡 variables the result for 𝐷𝑖𝑟𝑒𝑐𝑡 is significant and large, namely 7.9% for the older funds, but insignificant for the younger funds. The result for 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑜𝑢𝑛𝑡 is not significant for the older funds but is significant for the younger funds, namely 0.3%.

Table 6b shows the results of the same analyses but with the IRR net of the benchmark’s IRR as dependent variable. The results for the analyses with only one of the independent variables are mostly insignificant, only the 𝐼𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡 for the funds with vintage year 2005 or later are significant. The influence of adding an industry to the focus of the fund has a 0.002% positive influence on the IRR net of the benchmark’s IRR. In the more complete regression with both the 𝐷𝑖𝑟𝑒𝑐𝑡 and the 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡 variables the contribution of being a direct fund rather than a fund of funds is positive in both groups, and more than twice as large for the group with funds with vintage years before 2005.

4.2.3 Fund Value groups

In table 7a the results of the analysis of the value groups is shown. The base regression is repeated here for four groups differentiated on the basis of fund size. In panel A of table 7a the results for the net IRR are shown. The impact of both the 𝐷𝑖𝑟𝑒𝑐𝑡 and the 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡 variables are significant for all four groups. The effect of being a direct fund rather than a fund of funds is about 5% for all groups. The effect is largest for the third group, with fund values ranging from 500 up to 2000 million US dollars. Adding another industry to focus on for the specialized funds contributes about 0.3% for all groups. Here the effect is the largest for the second group (fund value ranging from 100 up to 500 million US dollars), and the effect is the smallest for the third group.

27 In panel B of table 7a the results are shown for the same analyses but with the IRR net of the benchmark’s IRR as a dependent variable. Similar results are shown from this analysis, the impact of both the 𝐷𝑖𝑟𝑒𝑐𝑡 and the 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡 variables are roughly the same for all fund size groups. Contributing about 0.04% to the return over the returns from the benchmark for the 𝐷𝑖𝑟𝑒𝑐𝑡 variable and for the 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡 variable the impact is about 0.0015% per additional industry added to the focus of the fund. For the third group the result of the

𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡 variable is not significant.

The split up dataset, based on the vintage year, us used to see if the results that from the overall sample analysis are consistent over time. In table 7b these results are shown for the net IRR, in table 7c these results are shown for the IRR net of the benchmark’s IRR.

In table 7b it can be seen that the fourth group, the group with the larges funds could not be analyzed for there were too little direct funds in this group. The other three fund groups have apparent results. The impact of being a direct fund on the IRR of a fund is largest for the smallest group, even adding 14.6% for the funds in group 1 (400.000 us dollars fund size up to 100 million US dollars). And, comparable with the results from shown in table 7a, about 5% for group 2. The result for group 3 is not significant. The results of the impact of adding another industry to the focus of the fund are not significant for any of the four groups.

Comparing the result of the funds with vintage before 2005 and the results for funds with vintage from 2005 and later, we see that the impact of being a direct fund is much lower, about 2%. This result is not significant, tough, for any of the three groups analyzed. For the 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡 variable we zee similar results for all four groups when compared to their peers in the other vintage year group. Only the results of group 2 and group 3 are significant.

Looking at the same regressions but with the IRR net of the benchmark’s net IRR the impact, when compared to the results in table 7a panel b, are lower for both the 𝐷𝑖𝑟𝑒𝑐𝑡 and the 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡 variable. The impact of being a direct fund is in this analysis the largest for group 1, the funds with the lowest fund values. And the impact decreases as the fund size increases. For the 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝐶𝑜𝑢𝑛𝑡 variable the impact on the IRR net of the benchmark’s IRR is low and only significant for group 3. When, then, compared to the younger

28

4.3 Base analyses – type of investor

In table 8a panel A the results of the base analysis of the impact of focusing on a specific stage of a company’s lifecycle by a PE firm. The first three results are the impact of only the particular stage on the net IRR of the funds. From these three analyses it can be seen that the result for the 𝐸𝑎𝑟𝑙𝑦 variable, indicating investing in the early stage of the lifecycle, is negative but the result is not significant. For the 𝐺𝑟𝑜𝑤𝑡ℎ and the 𝐿𝑎𝑡𝑒 variables the impact is significant on a ten percent level. From this focusing on the growth stage of a company adds 2.8% to the net IRR of the fund, but focusing on the late stage of the company has a negative effect of -4.7%. When analyzed together in a more comprehensive regression only the result for investing in the late stage is significant, and still about -4.8%.

In table 9a panel A the same analyses is performed but with the IRR net of the benchmark’s IRR as a dependent variable. From this analysis the impression is that either of the three stages specified contribute to the IRR over the benchmark’s IRR, the results are for either analyses not significant.

4.4 Deepening analyses – type of investor

4.4.1 Quantiles

In table 8a panel B the result of the quantile regression is shown. The effect of inesting in the early stage has a negative effect on the funds net IRR. The effect is the largest for the funds that perform the worst, and the impact is the smallest for the better performing funds. For the growth stage investors the contribution to the net IRR is positive and is the most positive, about 2.4%, for the funds in the 75th percentile. The effect is the smallest for the funds in the 25th percentile, only about 0.9%.

In table 9a panel B the results of the same analyses but with the IRR net of the benchmark’s IRR are shown. Here only the results for the 25th _{and the 75}th _{percentile are} siginificant for the variable 𝐸𝑎𝑟𝑙𝑦, but the results are interesting. For the funds in the lowest percentile the effect for investing in the early stage is slightly negative, for the funds in the 75th percentile the effect is slightly positive. The results for the 𝐺𝑟𝑜𝑤𝑡ℎ variable are not significant for either quantile. The results for 𝐿𝑎𝑡𝑒 is only significant for the 25th _percentile and is slightly negative.

4.4.2 Time frames

In Tables 8b the results are shown for the base analyses that are repeated but are then compared for the different time frames. The results of the group with vintage year before

29 2005 is only significant for the 𝐿𝑎𝑡𝑒 variable and this indicates that there is a negative effect on investing in this stage by these funds. For the younger counterpart the results for this variable are nog significant so a valid comparison is not possible. For the younger group only the results for the 𝐺𝑟𝑜𝑤𝑡ℎ stage investors is significant. The result suggest that for these younger funds the focus on this stage contributes to the net IRR of the fund by about 4%. The results for the older funds for this variable are not significant and therefore a valid comparison is not possible.

In table 9b the results of the same analyses but with the IRR net of the benchmark’s IRR as dependent variable. The results from none of these analyses seem to be significant.

30 5. Conclusion

Private equity is an investment class that has gained attention in the academic literature. Due to the nature of private equity, being very private about the transactions, the data that is available is limited. The data that is available is not as neatly structured as the data from public equity markets for example and makes it harder to research. Moreover the data that is available on private equity is mostly data retrieved from the limited partner rather than the general partner, making the data less reliable.

The term private equity is a collective name for many forms of investing. One thing they all have in common is that these investments are illiquid. At least when compared to public equity market investments. This illiquidity has been topic of research, for private equity seems to outperform public equity in the long run, the excess returns must be illiquidity premiums. Yet there are great differences within private equity and the performance of the different types of private equity is topic of this research.

In this research, first, the performance of direct funds was measured against the performance of non-direct funds (fund of funds), controlled for several factors that are of influence on the performance in general for private equity. Next the performance within direct funds was researched. The diversification scale was used to discriminate the funds from each other. The first run was to look at the specialist and generalist type of funds, controlling for other factors. Further the extent of being specialized was introduced to gain insight in how this factor influences the performance of the fund.

To deepen the analysis groups where made within the dataset based on several axis to gain insight in the persistence over time and the division amongst performance groups. First a division was made based on performance, for part of the literature suggested that similarly experienced general partners would have similar performance. To control for this, three groups where made based on their performance and the influence of both being a direct fund and having a limited number of industries to focus on was researched.

Next the funds were split into two groups based on their vintage year to check if the effects of the independent variables was persistent over time. The first group contained funds with vintage years up to 2005 and the other groups contained funds with vintage years starting from 2005 up to 2013. Lastly the funds were grouped based on their fund value, literature suggested that the fund value might be the result of the general partner having ran a successful

31 fund in the past. To see if the effects of the independent variables were the same on funds with different fund values.

To gain more insight in the factors that determine private equity performance the above procedure was partially repeated using a different approach. Funds can also specialize in investing in certain periods of the lifecycle of companies. To get a bit of insight in if this type of specializing is of influence on the fund performance three stages where chosen to perform the analysis with. The three variables indicated if the fund was specialized in investing in the early stage of a company, the growth stage or the late stage. These stages fundamentally differ in their risk attribute and therefore the fund performance could be influenced by this. The same procedure was repeated using first base regressions to

subsequently perform the deepening analyses grouping the funds by performance and vintage year.

In this research the performance was measured in two different variables, therefore all the analyses where performed twice. The one measure of performance was the net IRR, the net internal rate of return of the fund, the other measure was the IRR of the fund net of the benchmark’s IRR. The benchmark was set by Preqin based on the funds vintage year, type of fund and location of the general partner’s headquarters.

The results of the analyses performed give insight in the within private equity differences in performance. Looking at the base analyses for both measurements of fund performance the results show that direct funds perform better than non-direct funds (being fund of funds). When looking at the deepening analyses on the performance dimension, the persistence dimension and the fund size dimension, the results show that direct funds perform better. The first hypothesis tested can therefore be confirmed, direct funds do perform better.

The effect of being a specialist rather than a generalist investor is tested next. The effect of being a specialist sec from being generalist is tested trough base analyses and for both measurements of performance. The results show no conclusive result for most analyses the result was not significant. The only significant result of this test was negative, suggesting that the impact of being a specialist is negative on performance. Next the influence of adding another industry to the investment focus of the fund is researched, testing the second

hypothesis. The effect of adding another industry to the focus confirms the suggestion of the result of the specialist test, adding an industry to the focus has a positive effect on the performance of the fund. When looked at over the three dimensions (performance groups,

32 time frames and fund size groups), the results are persistent over all the dimensions. Although there are interesting within group differences. The second hypothesis tested cannot be

confirmed, adding an additional industry to the focus of a fund does not seem to have a

negative effect on performance (on neither performance measure). The evidence even suggest, and confirms, that the effect is in the opposite direction.

Lastly the effect of investing in a certain moment in the lifecycle of a company is researched. The question was if there was a difference and if the choice of entry point matters for the performance of the fund. Similar analysis were performed as was done in the

specialization analyses. The results for the entry point show the most result when only the net IRR performance measure is considered. The IRR net of benchmark gives interesting insight to the within performance groups and suggests (not significant) similarities between the two vintage year groups, but does not contribute to the hypothesis researched.

Looking at the results for net IRR both the base analysis and the performance group analyses show that investing in the early stage of a fund’s lifecycle has a negative effect on the fund’s performance. The effect is larger for the funds that perform worse than average, and the effect is smaller for the funds that perform better than average. Focusing on the late stage of the lifecycle has a negative effect on fund performance too, and also here the effect is larger for the funds that perform worse and the effect is reduced when the funds perform better. Of the three stages of the lifecycle researched the only profitable stage seems to be the growth stage. The effect is larger for the better performing funds than for the worse

performing funds. It is observed that the effect tested is significant for the funds with vintage years starting from 2005. For the group with the older funds the results are not significant.

Overall the third hypothesis tested in this research can be confirmed, the entry point on the lifecycle of a company matters for the fund’s performance. From the three stages tested the growth stage contributes to the performance and both the early and late stage deteriorate performance.

In private equity many questions are still left unanswered, this research has

contributed to gaining insight in what influences the within private equity performance. The most concise finding of this research is that indeed direct funds perform better than fund of funds. This was confirmed along different dimensional axis. This should contribute to finding a definitive answer on this matter. The findings of specialization and being more or less diversified is answered too but caution needs to be taken presuming it to be definitive. There

33 are differences in the results along the different dimensional axis. The findings on the point of entry specialization gave some interesting results, for it seems to matter to the performance of the fund.