4.1 Investment choices
We start our analysis by focusing on the question of whether endowments are more likely to invest in alumni-matched funds compared to other funds. Ideally, we would know the specific fund criteria that endowments were considering before they made a decision to commit capital. As this information is not accessible, we create alternative fund pools for each actual fund investment based on general criteria such as same fund vintage year, strategy type, and size (within a range of 50% to 150% of actual fund size). For example, alternatives to commitments into a USD 1.0 billion buyout fund of vintage year 2010 would include buyout funds with the same vintage year and fund sizes between USD 500 million and USD 1.5 billion. Similar to the approach proposed by Kuhnen (
2009), Siming (
2014) and Bengtsson and Hsu (
2015), the groups of alternative investments determine our counterfactual sample. We delete commitments for which we do not find counterfactual alternatives according to our criteria, so that the number of actual investments used for this identification strategy lowers slightly from 1,590 to 1,523. The number of counterfactual commitments amounts to 15,553 observations. While we match fund managers in the counterfactual sample with potential endowment investors, the number of funds managed by alumni reach approximately 8%, which is notably smaller than the 15% seen in the actual investment sample.
Table 4
Investments and educational ties: actual and counterfactual
Actual | 1,295 | 228 | 91 | 55 | 1,523 |
| 85.1% | 14.9% | 6.0% | 3.6% | |
Counterfactual | 14,322 | 1,231 | 538 | 293 | 15,590 |
| 92.1% | 7.9% | 3.5% | 1.9% | |
Total | 15,617 | 1,459 | 629 | 348 | 17,076 |
| 91.5% | 8.5% | | | |
We recognize that not only more investment criteria may have been used by endowments to decide on an investment but also the presence of networks itself may lead to some investments not necessarily following our strict selection rule. For instance, an endowment could potentially not have been planning to allocate capital to a certain type of fund strategy until it became aware of a specific initiative. However, this would actually mean that we are underestimating the importance of alumni ties, and thus our estimates are rather conservative. While it is possible that our broad set of criteria overestimates the amount of funds that would be considered as close alternatives by endowments, there is also a possibility that our counterfactual approach does not include all potential alternatives. The average and median number of selected fund alternatives for each commitment, counting both actual and counterfactual investments, is at 24 and 17 respectively, and the maximum reaches 104.
14 We do not claim to be able to reproduce the full range of potential fund alternatives, however, we do control for preferences for similar geographies, later fund sequences, existing relationships, and background of fund partners.
15 One can also argue that different finance teams at the endowment level may follow different investment styles, and this heterogeneity among endowments might systematically affect our results. Moreover, investment behavior, or simply the number of investment options available (i.e., competition among investors to access funds), may also change depending on the investment environment of each year and it may be different across fund types. For example, the options to invest into smaller VC funds may be more limited compared to larger buyout funds, which could impact the effect that we see for alumni ties. We address these concerns in our identification strategy by including multi-way fixed effects to control for specific endowment, vintage years, and fund strategy types. The main model specification is as follows:
$$\begin{aligned} \begin{aligned} ln(\frac{p_{i,j}}{1-p_{i,j}})= a+ \beta _1 \textit{Alumni}_{i,j}+\beta _2 \textit{Fund Size}_i+\beta _3 \textit{Fund Sequence}_i \\ + \beta _4 \textit{Same State}_{i,j} +\beta _5 \textit{GP Relationship}_{i,j} +\beta _6 \textit{Experience}_i+ \textit{Fixed Effects}+\epsilon _i. \end{aligned} \end{aligned}$$
(1)
Table 5
The odds of investment
Alumni tie | 0.531*** | | | | |
| (0.194) | | | | |
MBA alumni tie | | 0.687*** | | | |
| | (0.190) | | | |
Undergraduate tie | | | 0.409** | | |
| | | (0.207) | | |
Postgraduate tie | | | | 0.684*** | |
| | | | (0.236) | |
Percentage of alumni | | | | | 0.815*** |
| | | | | (0.312) |
Fund size (log) | 0.600*** | 0.606*** | 0.607*** | 0.605*** | 0.611*** |
| (0.037) | (0.038) | (0.037) | (0.038) | (0.037) |
Fund sequence (log) | -0.223*** | -0.224*** | -0.223*** | -0.224*** | -0.225*** |
| (0.060) | (0.058) | (0.059) | (0.058) | (0.059) |
Same state | 0.559*** | 0.581*** | 0.639*** | 0.580*** | 0.592*** |
| (0.200) | (0.204) | (0.195) | (0.205) | (0.199) |
Previous GP Relationship | 4.124*** | 4.141*** | 4.125*** | 4.134*** | 4.126*** |
| (0.171) | (0.172) | (0.167) | (0.174) | (0.171) |
Consulting experience (%) | 0.416*** | 0.423*** | 0.422*** | 0.426*** | 0.421*** |
| (0.143) | (0.146) | (0.145) | (0.145) | (0.144) |
Banking experience (%) | -0.712*** | -0.708*** | -0.709*** | -0.711*** | -0.709*** |
| (0.118) | (0.118) | (0.118) | (0.118) | (0.118) |
Accounting experience (%) | 0.211 | 0.210 | 0.224 | 0.215 | 0.222 |
| (0.308) | (0.309) | (0.309) | (0.307) | (0.309) |
F.E. Vintage | Yes | Yes | Yes | Yes | Yes |
F.E. Type | Yes | Yes | Yes | Yes | Yes |
F.E. Endowment | Yes | Yes | Yes | Yes | Yes |
Observations | 15,641 | 15,641 | 15,641 | 15,641 | 15,641 |
Pseudo R-squared | 0.3116 | 0.3100 | 0.3092 | 0.3103 | 0.3095 |
Our binary dependent variable
Y\(^{i,j}\) equals one when a commitment in fund
i is made by an endowment
j, and zero when an alternative fund could have been considered as a potential investment according to our criteria but was actually not chosen. We use a logistic regression model, where the left hand-side of the equation represents the log of the odds of
Y\(^{i,j}\), with
p\(^{i,j}\) being the probability of
Y\(^{i,j}\) being equal to one. Our main variable of interest is
Alumni\(^{i,j}\), which takes the value of one for funds where the educational background of managers matches the endowments’ universities and zero where there is no such link. We also show results for variations of our independent variable in Table
5, breaking it down by the degree of commonality (i.e., the number or percentage of individuals with the same background within a fund), degree types (although not available for all alumni ties), and university rankings.
Fund Size\(^{i}\) and
Fund Sequence\(^{i}\) are the natural logarithm of final fund sizes (in USD million) and the sequences of funds managed within fund families (managed by the same GP).
Same State\(^{i,j}\) is a dummy variable that equals to one when endowments and fund headquarters are located within the same U.S. state and controls for a potential home bias, as suggested by Hochberg and Rauh (
2013). Over 11% of endowment investments in our sample are within the same state, which compares to just below 6% in the counterfactual sample.
GP Relationship\(^{i,j}\) is another dummy that equals one when it indicates that an endowment has prior history in investing with a manager and zero otherwise.
16 Table
17 of the Internet Appendix also shows results where we control for previous GP performance in a subsample for which such information is available. The estimates are in line with our main results of Table
5.
Experience\(^{i,j}\) represents a set of three variables related to the percentage of fund managers that have backgrounds in consulting, banking, and finance industry, similarly to the controls applied in Fuchs et al. (
2021).
Table
5 shows the results derived from a logistic regression with coefficients shown in log odds. We confirm our first hypothesis that endowments are more likely to invest into funds with an alumni tie. After exponentiation of the coefficients, we see that such tie increases the odds of an investment by a factor of 1.70, i.e. ceteris paribus, the odds of an endowment investment into an alumni-linked fund are 70% higher than in other funds. By breaking down the ties by degree types, our results remain significant across different degrees, while appearing to be stronger for post-graduate ties and, particularly, for MBA ties.
As previously noted, we observe in our educational background data sample (Table
14 of the Internet Appendix) that certain universities, particularly the higher-ranked institutions with the biggest endowments, have a more abundant alumni presence in PE fund management than others. To test whether the alumni connection matters for different types of institutions, we further categorize our alumni tie variable according to school rankings. We classify American universities according to the QS World University Rankings list for 2010. Therefore, a university is defined as a top-20 school if it is among the top-20 institutions in the worldwide ranking. We also divide MBA ties according to the Financial Times 2010 Global MBA ranking into top-10 (in the United States) and others. As there is a lower number of universities that offer MBA programmes, top universities represent an even larger portion of the sample for this type of degree.
17
To further ensure that our main variable is not influenced by the dominance of alumni from high-ranked universities working in the PE industry, we create a new independent variable, which we refer in the following as “scaled” alumni tie. The introduction of this variable reflects on the idea that there may be situations where an alumni tie with an endowment can be an exclusive feature no other competing fund possesses. Thus, it can be a differential that may impact the corresponding investment odds.
$$\begin{aligned} \begin{aligned} \textit{Scaled tie}_{i,j}=\frac{\textit{Actual tie}_{i,j}}{\sum _{i=1}^n\textit{Alumni tie}_{i,j}}. \end{aligned} \end{aligned}$$
(2)
The “scaled” alumni tie variable in Eq.
2 is defined as the number of alumni ties in actual investments divided by the number of total alumni ties in actual and counterfactual investments within the same criteria group (according to fund strategy, vintage, and size). The value of this variable ranges from zero to one, i.e. Eq.
2 transforms a binary variable into a probability. A value of one represents the situation where, among alternative funds, only the chosen fund had one or more alumni managers from the endowment’s university. It therefore reaches the maximum degree of exclusivity. A value of zero in turn represents the scenario where there are no matches. Accordingly, values between zero and one mean that there were other possible funds to invest that were also managed by alumni. For example, in our data we see that, among 45 possible similar buyout funds with vintage 2000, MIT Investment Management Company selected the only fund where we identify an alumni tie. Therefore, its scaled tie equals to 1. Meanwhile, the scaled tie equals 0.0625 for Harvard Management Company for its investment in 2012 buyout fund since, in addition to the matched alumni in the actual investment, there are 15 other funds among 21 counterfactual opportunities that also have at least one alumna among its managers (e.g., 1/16 = 0.0625). Average scaled tie values by rankings are reported in Table
6.
Table 7
The exclusivity of ties
All Universities | 228 | 0.447 | 0.333 |
QS World rank | | | |
Top 20 | 136 | 0.353 | 0.279 |
Top 21-50 | 27 | 0.316 | 0.231 |
Top 51-100 | 31 | 0.606 | 0.500 |
Others | 34 | 0.781 | 1.000 |
QS US rank | | | |
Top 20 | 163 | 0.346 | 0.273 |
Top 21-50 | 39 | 0.584 | 0.500 |
Top 51-100 | 10 | 0.658 | 0.583 |
Others | 16 | 1.000 | 1.000 |
All Universities (MBA ties only) | 91 | 0.425 | 0.333 |
Global MBA Ranking 2010 | | | |
Top 10 US | 61 | 0.303 | 0.222 |
Others | 30 | 0.672 | 0.500 |
Results of Table
6 highlight that, on average, the higher the ranking position of the university is, the lower the exclusivity ratio. Under the assumption, and as shown in Table
6, that endowments are indeed more likely to invest into funds managed by their own alumni, this finding is not surprising. Graduates of lower ranked universities are underrepresented in the PE industry and are less likely to appear with an alumni match both in the actual and counterfactual sample. Thus, this leads to higher exclusivity ratios. Table
6 represents a first evidence that universities with a smaller footprint in the PE industry tend to rely more on alumni ties when making PE investments. Table
7 further elaborates on this hypothesis within a multivariate setting.
Table 8
The odds of investment according to ranking and exclusivity
Panel A: Regular alumni ties |
Alumni tie | 0.531*** | 0.584*** | | | | |
| (0.194) | (0.218) | | | | |
Redundant alumni tie | | -0.319 | | | | |
| | (0.274) | | | | |
Number of alumni ties | | | 0.152 | | | |
| | | (0.143) | | | |
MBA tie | | | | 0.687*** | | |
| | | | (0.190) | | |
Top 20 alumni tie | | | | | 0.438* | |
| | | | | (0.253) | |
Top 21-50 alumni tie | | | | | 0.436*** | |
| | | | | (0.132) | |
Top 51-100 alumni tie | | | | | 0.422 | |
| | | | | (0.783) | |
Top 100+ alumni tie | | | | | 1.901*** | |
| | | | | (0.537) | |
Top 10 MBA tie | | | | | | 0.584*** |
| | | | | | (0.162) |
Top 10+ MBA tie | | | | | | 0.793*** |
| | | | | | (0.299) |
Panel B: Scaled alumni ties (by number of counterfactual matched funds) |
Alumni tie | 1.377*** | 1.373*** | | | | |
| (0.257) | (0.276) | | | | |
Redundant alumni tie | | 0.029 | | | | |
| | (0.345) | | | | |
Number of alumni ties | | | 1.360*** | | | |
| | | (0.253) | | | |
MBA tie | | | | 1.350*** | | |
| | | | (0.300) | | |
Top 20 alumni tie | | | | | 1.431*** | |
| | | | | (0.385) | |
Top 21-50 alumni tie | | | | | 1.100*** | |
| | | | | (0.300) | |
Top 51-100 alumni tie | | | | | 0.280 | |
| | | | | (1.062) | |
Top 100+ alumni tie | | | | | 2.331*** | |
| | | | | (0.616) | |
Top 10 MBA tie | | | | | | 1.681*** |
| | | | | | (0.432) |
Top 10+ MBA tie | | | | | | 1.192*** |
| | | | | | (0.440) |
Control variables | Yes | Yes | Yes | Yes | Yes | Yes |
F.E. Vintage | Yes | Yes | Yes | Yes | Yes | Yes |
F.E. Type | Yes | Yes | Yes | Yes | Yes | Yes |
F.E. Endowment | Yes | Yes | Yes | Yes | Yes | Yes |
Observations | 15,641 | 15,641 | 15,641 | 15,641 | 15,641 | 15,641 |
Columns 1-4 of Table
7 show results for the regressions on the odds of investment for alumni tie variables that were previously reported and explore the possibility that having more than one tie in a fund might have a greater effect than just one. Column 5 reports the results when we re-run our models based on universities’ ranking positions. Panel B reports results when such variables are scaled as defined in Eq.
2. In Panel A, alumni ties connected to the top-20 universities are significant, however, the effects of ties of universities that do not make it to the top-100 list are not only statistically significant but also economically stronger. Using scaled ties, as displayed in Panel B, our results are overall consistent with our initial analysis in Panel A, with ties from top-20 universities remaining significant. More notably, alumni ties on the level of lower-ranked universities continue to appear as more economically and statistically significant. For scaled ties taking the maximum value of one, top-20 alumni ties lead to an increase in the odds of investment of 318% and that of lower-ranked institutions of 929%. The same pattern holds for MBAs as shown in Column 6. Overall, alumni networks seem to matter in general, but some of them appear to be particularly powerful and alumni ties can be even more important for lower-ranked universities. Following the specification of Eq.
2 a high value for our “scaled” alumni tie variable means that the observed tie is rather exclusive and few fund managers of the counterfactual sample share the same alma mater. With the specification of Panel B we are able to explore these situations in more detail and investigate if the overall presence of a university’s alumni community in the PE industry (e.g., again measured via the counterfactual sample) impacts the odds of an alumni tie. Our results in Panel B display a positive correlation relating to the level of exclusivity. The introduction of a “scaled” alumni tie also allows us to control for the size of the underlying alumni community in the PE industry. As outlined in Table
16 of our Internet Appendix, we observe that higher ranked universities maintain a stronger footprint in the PE industry as lower ranked universities leading to lower values relating to the “scaled” alumni tie variable (e.g., it is more likely that you find another Harvard alumni tie in our counterfactual model as compared to a lower ranked university, which in turn leads to a lower value for the “scaled” alumni tie variable). Our results show that alumni ties do matter for lower ranked universities (e.g., with a lower alumni community in the PE industry) and that the significance of alumni ties is not limited to higher ranked universities but holds also for lower ranked universities and increases with the level of exclusivity.
In a next step, we test whether investments into funds managed by alumni translates into better return performance. Thereby, we regress the PE fund performance of the endowment commitments on our main independent variable, the alumni tie, and control for a comparable set of variables used in prior analyses.
18$$\begin{aligned} \begin{aligned} \textit{Fund Net IRR}_{i,j}=a+\beta _1\textit{Alumni}_{i,j}+\beta _2 \textit{Fund Size}_i+\beta _3 \textit{Fund Sequence}_i \\ +\beta _3 \textit{Same State}_{i,j} + \beta _4 \textit{GP Relationship}_{i,j}+\beta _5\textit{Track Record}_i \\ + \beta _6 \textit{Experience}_i+\textit{Fixed Effects}+\epsilon _i. \end{aligned} \end{aligned}$$
(3)
Compared to Eq.
1, we add a
Track Record\(^{i}\) variable to our performance regressions, which is defined as the average net IRR performance a GP has realized across all previous funds prior to the current fund generation. As our goal is to see whether investments into alumni-managed funds are beneficial or detrimental to endowments, we compare their performance to other endowment commitments to PE funds (without alumni ties). Thus, and in contrast to our odds analysis, we do not need to apply a counterfactual approach. We use ordinary least-squares (OLS) estimates including fixed effects for fund vintage years, fund strategies, and endowments. Standard errors are robust and clustered at the endowment level, similarly to previous studies on performance (e.g., Korteweg and Sorensen (
2017).
The main results of our performance regressions are shown in Table
8 for net IRR measurements, whereas TVPI results are shown in Table
18 of the Internet Appendix. We note that these measurements are popular in the literature but are not risk adjusted, which is a well-known challenge in private markets. Looking at them, we neither observe significant outperformance nor underperformance of fund commitments with alumni ties, which suggests that funds managed by alumni do not tend to perform differently than other funds in endowment portfolios. Thus, we are not able to find empirical evidence supporting our second hypothesis that alumni ties could be advantageous to endowments and translate into higher performance.
Table 9
The performance of investments into alumni funds
Alumni tie | 1.314 | | | | |
| (1.819) | | | | |
MBA alumni tie | | 8.417*** | | | |
| | (3.212) | | | |
Undergraduate tie | | | -3.297 | | |
| | | (2.193) | | |
Postgraduate tie | | | | 4.978 | |
| | | | (3.896) | |
Percentage of alumni | | | | | 13.405* |
| | | | | (7.381) |
Fund size (log) | − 1.304* | − 1.297* | − 1.377** | − 1.290* | − 1.223* |
| (0.681) | (0.662) | (0.668) | (0.679) | (0.671) |
Fund sequence (log) | 0.600 | 0.499 | 0.590 | 0.599 | 0.497 |
| (1.047) | (1.026) | (1.014) | (1.026) | (1.035) |
Same state | 0.025 | − 0.758 | 0.476 | − 0.661 | − 0.817 |
| (2.800) | (2.195) | (2.866) | (2.375) | (2.481) |
Previous GP Relationship | 0.615 | 0.725 | 0.691 | 0.572 | 0.462 |
| (1.365) | (1.368) | (1.343) | (1.357) | (1.362) |
Previous GP IRR | 0.175*** | 0.165*** | 0.174*** | 0.178*** | 0.174*** |
| (0.026) | (0.026) | (0.026) | (0.027) | (0.025) |
Consulting experience (%) | 3.366* | 3.439** | 3.286** | 3.076* | 3.214* |
| (1.625) | (1.570) | (1.597) | (1.759) | (1.669) |
Banking experience (%) | 1.373 | 1.786 | 1.302 | 1.360 | 1.469 |
| (2.833) | (2.543) | (2.952) | (2.734) | (2.642) |
Accounting experience (%) | − 6.471 | − 6.695 | − 6.185 | − 6.728 | − 6.812 |
| (4.967) | (4.848) | (4.987) | (4.951) | (4.965) |
F.E. Vintage | Yes | Yes | Yes | Yes | Yes |
F.E. Type | Yes | Yes | Yes | Yes | Yes |
F.E. Endowment | Yes | Yes | Yes | Yes | Yes |
Observations | 1,054 | 1,054 | 1,054 | 1,054 | 1,054 |
Adjusted R-squared | 0.1050 | 0.1182 | 0.1058 | 0.1108 | 0.1114 |
An interesting exception, however, is MBA ties. As seen in Column 2 of Table
8, they are associated with statistically significant higher performance. Further analyses, shown in Table
19 of the Internet Appendix, suggest that ties for graduates from highly ranked MBA program, which represent over 70% of ties, affect fund performance significantly. A similar pattern was also documented by Wu (
2011), where the performance of non-syndicated leveraged buyout deals is shown to be higher when a team member has an MBA. The author argues that this is evidence for MBAs being better at deal screening and that, when syndication occurs, partnerships involving Harvard MBA social ties seem particularly fruitful. Fund managers with such a background show a strong preference to collaborate and can find a larger number of partners. This highlights the advantages of being part of the alumni network of a highly ranked university. Our findings support such an argumentation. In order to ensure that the positive relationship of MBA ties on performance is not driven by the MBA degrees themselves (see, e.g., Bertrand and Schoar
2003 and Graham and Harvey
2001), we also run regressions as in Eq.
3 with MBA experience reflected by the percentage of fund staff with MBAs as an explanatory variable. Our results, reported in Table
20 of the Internet Appendix, confirm that, although MBA experience is indeed associated with higher performance, MBA alumni ties are still economically and statistically significant.
19 Overall, as we only observe a significant effect in the case of MBA ties, our findings suggest that general alumni ties do not prove to be a systematic factor driving the performance of endowments’ PE investments.
4.3 Robustness tests
We perform a range of different robustness checks to validate our findings. First, we test whether our main finding that endowments seem more likely to invest in alumni-managed funds is not driven by the design of our counterfactual approach. In doing so, we use random draws similarly to Ishii and Xuan (2014) and propensity score matching as alternative selection methods. The results and procedure details are reported in Tables
21 and
22 of the Internet Appendix. In addition, we also use different criteria for the setup of our counterfactual approach. First, we relax size restrictions when selecting counterfactual funds, resulting in an increasing number of potential options for each actual investment. As reported in Table
23 of the Internet Appendix, this adjustment leads to similar conclusions as derived from our main analysis – alumni ties significantly increase the odds of an investment. Second, in contrast to the main analysis, we restrict our sample to investments into “local” funds only, i.e., within the same state or based within a distance of 100km to the location of the endowment fund. We still find positive, but mostly statistically insignificant, effects stemming from alumni ties, as reported in Table
24 of the Internet Appendix. Even though there is a preference for same-state investments in our data, endowments do not only consider local funds. Moreover, such ties could be particularly key for endowments that are not from the same geography due to the absence of local networks and increased information asymmetries.
20 We run a series of subsample analyses according to fund and endowment characteristics and confirm that we can draw similar conclusions for both investment odds and performance regressions as specified in the main models. Results are reported in Tables
9 and
10.
Table 11
Investment odds subsample robustness
Panel A: Fund characteristics | | |
Vintages to 2005 | 0.752*** (0.268) | 0.902*** (0.271) |
Vintages after 2005 | 0.195 (0.244) | 0.234 (0.438) |
Buyout | 0.572*** (0.198) | 0.584*** (0.197) |
VC | − 0.166 (0.475) | 0.981* (0.523) |
Growth | 3.881** (1.689) | 6.761*** (1.373) |
Undersubscribed | 0.501 (0.498) | − 0.035 (0.541) |
Oversubscribed | 0.549*** (0.186) | 0.751*** (0.205) |
Top performers (IRR) | 0.637*** (0.182) | 0.887*** (0.199) |
Low performers (IRR) | 0.452 (0.294) | 0.578* (0.357) |
Top performers (TVPI) | 0.703*** (0.148) | 1.002*** (0.178) |
Low performers (TVPI) | 0.315 (0.344) | 0.278 (0.403) |
Better GP track record (IRR) | 0.593*** (0.189) | 0.886*** (0.246) |
Worse GP track record (IRR) | 0.592** (0.308) | 0.404 (0.442) |
First sequence | 0.070 (0.639) | − 1.100 (0.800) |
Second+ sequence | 0.526*** (0.195) | 0.705*** (0.192) |
Only one endowment investor | 1.137*** (0.263) | 1.163*** (0.344) |
More than one endowment investor | 0.352* (0.211) | 0.566** (0.235) |
Panel B: Endowment | | |
Top 20 | 0.485* (0.262) | 0.494*** (0.128) |
Top 50 | 0.451** (0.198) | 0.682*** (0.200) |
Top 100 | 0.428** (0.192) | 0.652*** (0.195) |
Other endowments (Top 100+) | 2.049*** (0.604) | 1.532 (1.289) |
Top performers (IRR) | 0.437* (0.258) | 0.763*** (0.246) |
Bottom performers (IRR) | 0.665** (0.276) | 0.581* (0.332) |
Endowments with more PE commitments | 0.329 (0.216) | 0.620*** (0.238) |
Endowments with less PE commitments | 1.154*** (0.264) | 1.082*** (0.196) |
Previous GP Relationship | 0.443 (0.445) | 1.387* (0.838) |
No previous GP Relationship | 0.520** (0.231) | 0.611*** (0.194) |
Largest 10 endowments | 0.806*** (0.281) | 0.880*** (0.002) |
Largest 20 endowments | 0.551** (0.231) | 0.582** (0.240) |
Other endowments | 0.496 (0.375) | 0.855*** (0.403) |
Public universities | 0.508 (0.347) | 0.889*** (0.343) |
Private universities | 0.566*** (0.213) | 0.610*** (0.192) |
Top 10 most matched universities | 0.259 (0.316) | 0.509*** (0.139) |
Remaining less matched universities | 0.809*** (0.175) | 1.319*** (0.375) |
Control variables | Yes | Yes |
F.E. Vintage | Yes | Yes |
F.E. Type | Yes | Yes |
F.E. Endowment | Yes | Yes |
Table 12
Performance subsample robustness
Panel A: Fund characteristics | | |
Vintages to 2005 | 1.331 (2.161) | 6.027* (3.497) |
Vintages after 2005 | 0.963 (2.106) | 13.736 (7.372) |
Buyout | 2.259 (2.273) | 11.117*** (3.639) |
VC | 4.896 (4.3664) | 4.805* (2.551) |
Undersubscribed | 3.129 (3.109) | 1.959 (12.864) |
Oversubscribed | 1.311 (2.143) | 9.094*** (3.654) |
Top performers (IRR) | 1.141 (1.740) | 7.599 (4.695) |
Low performers (IRR) | − 0.588 (1.291) | 3.831*** (1.120) |
Top performers (TVPI) | 1.618** (0.798) | 7.695*** (2.899) |
Low performers (TVPI) | 0.047 (1.932) | 4.505*** (1.726) |
Better GP track record (IRR) | − 0.220 (0.961) | 0.662 (1.417) |
Worse GP track record (IRR) | 3.085 (2.184) | 10.234* (5.875) |
Only one endowment investor | 0.430 (8.379) | − 0.696 (8.195) |
More than one endowment investor | 0.277 (1.900) | 5.724* (3.248) |
Panel B: Endowment | | |
Top 20 | 1.840 (2.425) | 9.829*** (2.207) |
Top 50 | 1.736 (1.944) | 7.929** (3.130) |
Top 100 | 1.441 (1.810) | 8.006** (3.012) |
Other endowments (Top 100+) | 3.330 (6.520) | 32.459*** (6.132) |
Top performers (IRR) | 2.008 (2.578) | 10.066** (3.859) |
Bottom performers (IRR) | − 0.713 (1.696) | 3.174 (2.239) |
Endowments with more PE commitments | 1.611 (2.136) | 8.202* (4.055) |
Endowments with less PE commitments | 1.428 (1.687) | 7.811** (3.262) |
Previous GP Relationship | − 0.502 (1.820) | 8.158 (7.608) |
No previous GP Relationship | 3.268 (3.019) | 7.518* (4.235) |
Largest 10 endowments | 0.485 (2910) | 4.437 (5.266) |
Largest 20 endowments | 0.660 (2.138) | 4.155 (3.368) |
Other endowments | 2.392* (1.237) | 14.870*** (2.478) |
Public universities | − 1.537 (1.430) | 9.362 (6.275) |
Private universities | 2.952 (2.070) | 6.832** (2.637) |
Top 10 most matched universities | 5.762* (2.418) | 10.882*** (2.166) |
Remaining less matched universities | − 2.051 (1.767) | 0.615 (5.111) |
Control variables | Yes | Yes |
F.E. Vintage | Yes | Yes |
F.E. Type | Yes | Yes |
F.E. Endowment | Yes | Yes |
Table
9 shows that alumni ties appear to be particularly important for investments into oversubscribed funds, or for funds being raised by fund managers with a track record of high historic investment returns, which supports the hypothesis that alumni ties may facilitate access to highly demanded funds. Investments into growth funds appear to be big outliers with significantly stronger effects, but we take a cautious approach to avoid overinterpreting it since our growth fund sample is very limited (see Table
1). Our results also show that less experienced university endowments in terms of PE investments (e.g., those with less than 20 fund commitments) are more likely to rely on their alumni ties when they invest into PE funds. This is in line with our previous findings as those endowments also tend to represent lower ranked institutions. Similarly, we see that the effect on investment odds is not being driven by the most matched endowments, which again tend to also be the better ranked universities, while those appear to be the ones that show a positive impact on performance, particularly in the MBA case. This also confirms previous findings.
Another key finding, demonstrated in Table
9, is that any impact stemming from alumni ties has weakened in the more recent years as regression coefficients decrease in magnitude and are no longer statistically significant for post-2005 vintage years. This does not come as a surprise given the maturing or professionalization of the PE industry and of endowments as investors. Once endowments establish relationships with private equity firms, fund managers and other industry specialists, the importance of alumni networks for facilitated access to funds and as an information channel weakens. In our robustness checks, we see that alumni ties are particularly important for funds where previous GP Relationships do not exist and that the impact of previous firm relationships seem higher in later periods
21. As endowments became more established in the PE industry over time, the way they approach managers or are approached by them changed. Big endowments now have specialized fund management staff that are often experts in the field of alternative investments, while many smaller endowments are managed by general university financial officers and/or often rely on recommendations given by external investment consultants. Such a higher level of professionalization may have led to an attenuated role of university-related networks over time.
In further regressions, we add an additional category of fixed effects to our main specification to control for variation at the GP level. The rationale for this is that different private equity firms may attract varying levels of endowment investors or show different fundraising strategies. We do not include these fixed effects in our main analysis as many observations would have been dropped in the logistic regressions due to a high number of GPs only being represented with one fund in our data set. This would have resulted in a subsequent selection bias as we would have run our main analysis only for large GPs. However, we still obtain similar results for the odds of investment and performance in Tables
25 and
26 of the Internet Appendix when including GP fixed effects. We also explore using interaction terms and report it in Tables
27 and
28. Table
27 further confirms the relevance of MBA ties and, not surprisingly, the effect of alumni ties differs for endowments representing universities within systems instead of single institutions. In addition to the logit regressions following the main approach of the paper, we report and refer to OLS estimates due to the problems that arise when using interaction terms in non-linear models (see Ai and Norton (
2003)). Table
28 reports the results for performance regressions with interaction terms, where we again see that MBA ties are related to better performing investments, although we do not see any statistically significant interaction for university and endowment characteristics. We do see, however, that the MBA alumni effect itself remains strong and that a better ranking and more experience are linked to lower performance. Our results on the impact of MBA alumni ties remain robust when we also control for outliers by winsorizing performance as reported in Table
29.
Since our access to the fund managers’ biographies is restricted to GPs that manage at least one buyout fund, we note that a key limitation of our study is that our data sample does not capture investments into fund managers who focus exclusively on VC investments. While access to top-performing VC funds can be particularly difficult (compared to larger buyout funds), they are seen as a key driver of the endowments’ investment success (e.g., Sensoy et al. (
2014)). We can therefore expect the results that we derive to be even more pronounced for managers who exclusively follow a VC investment strategy. Thus, our observed estimates may underestimate the effect of alumni ties. However, the fact that we still find significant results, i.e. funds managed by alumni are preferred, is a strong indicator that this effect is non-trivial and must hold for the PE industry as a whole.
Finally, we understand that what we refer to as “alumni ties” is a broad term to classify the connections with individuals that had some sort of experience in or exposure to an institution. We are able to differentiate between types and intensity of these social ties by means of degree types (such as undergraduate or MBA degrees), how extensive or tight an alumni community is, or through university rankings. This allows us to account for different levels of involvement and potential influence of alumni ties and their effect on investment decisions.