Social capital, syndication, and investment performance: Evidence from PE investing in LBOs

This study examines the influence of social capital on leveraged buyout (LBO) investments. Exploiting proprietary global private equity data at the investment-level for leveraged buyouts, we find that alumni of Harvard ’ s MBA program are more inclined to co-invest and form syndicates in LBO with each another. The phenomenon of Crimson pairing manifests in deals that involve uneven investments in co-investor capital, necessitating trust to alleviate agency costs and enabling investors to diversify their portfolios. The outcome of Crimson pairing is an increase in value and investment returns relative to all other typical LBO syndication partnerships.


Introduction
Since the 1970s, syndicated leveraged buyouts (LBOs) spearheaded by private equity (PE) have exceeded $2.5 trillion globally (Degeorge,  Martin, & Phalippou, 2016).The investor base for these LBOs comprises a small, sophisticated group of investment experts whose social ties can significantly impact syndicate creation, investment yields, and the LBO market (Acharya, Gottschalg, Hahn, & Kehoe, 2013; Lopez-de-Silanes,  Phalippou, & Gottschalg, 2015).This study investigates whether social capital, embodied by shared college alumni connections among institutional investors, affects LBO syndicate structure and performance by alleviating information and agency problems typical in syndicated deals (Stacchini and Degasperi (2015) and Javakhadze, Ferris, and French  (2016)).Utilizing a novel dataset of worldwide LBO fundraising prospectuses provided by financial institutions, we find that shared connections among Harvard Business School (HBS) alumni within the PE investor community enhances the likelihood of LBO syndication, resulting in superior transaction-level performance in Crimson syndicated deals. 2ocial capital can be categorized into: (i) personal relationships, (ii) support from social networks, (iii) participation in civic activities, and (iv) trust and cooperative norms.Prior research on social capital in financial contracting stresses the importance of civic participation and trust in promoting financial growth and economic stability at the firm level (Guiso, Sapienza, & Zingales, 2004, 2008; Gurun, Stoffman, &  Yonker, 2018; Hasan & Habib, 2019; Lins, Servaes, & Tamayo, 2017).Only a handful of studies, all utilizing venture capital (VC) data, investigate the impact of personal relationships and trust on investment outcomes at the transaction level (Gompers, Mukharlyamov, & Xuan,  2016; Bottazzi, Da Rin, & Hellmann, 2016; Bubna, Das, & Prabhala,  2020).
Our research builds upon previous studies on social capital and financial contracting in two notable ways.First, we explore the role of social capital in LBO syndication, with a specific emphasis on the selection of PE co-investors in LBO syndications.Second, we examine whether social capital influences the performance of investment transactions executed by PEs in syndicated LBOs.Fundamentally, our goal is to bridge a gap in the existing literature by exploring shared college connections as a symbol of social capital and trust in the choice of LBO syndication partners by PE co-investors, and their transaction-level investment performance.As a result, our study can be considered groundbreaking in scrutinizing the factors influencing LBO syndication and carries significant implications, especially concerning the elements that propel the performance of PE syndicated transactions that differ from VC investments.
In a nutshell, we hypothesize that social connections influence the selection of syndication partners and PE investment performance.Regarding the selection, we expect that PE investors will syndicate with those who share social capital via their college affiliations and the trust inherent in these relationships.However, we predict this effect will vary among colleges, potentially being more pronounced for those tasked with fostering and preserving strong alumni networks. 3As for the outcome, the expected impact of such connections among PE investors in a syndicated LBO is not immediately clear and is likely a matter for empirical investigation.On one side, the management team of an LBO co-investor and syndicate member, who has personal obligations to another co-investor team due to college alumni ties might overlook flaws in a deal.This could potentially jeopardize their own investors' capital (Fang, Ivashina, & Lerner, 2015; Lopez-de-Silanes et al., 2015).On the other side, a personal relationship rooted in strong college affiliations among PE management teams could encourage a reliable flow of information strengthened by alumni ethos.This could lead to decreased monitoring costs due to trust established by college alumni credos, giving the alumni-connected deal an edge over all other coinvestor deals (Lins et al., 2017; Malenko & Malenko, 2016).
Considering this conflict, we assemble a dataset that extracts proprietary data from private placement memorandum (PPM) on 941 global LBOs from 1991 through 2004. 4In line with Degeorge et al.  (2016), we utilize proprietary data in PPMs to accurately capture the characteristics and returns of PE investments at the deal level, rather than at the firm-level where the bulk of prior work has concentrated (see Arcot, Fluck, Gaspar, & Hege, 2015; Boone & Mulherin, 2011;  Humphery-Jenner, 2013 and Officer, Ozbas, & Sensoy, 2010; among  others).The performance metrics available in the PPMs include valuation multiple and gross internal rate of return (GIRR). 5To enhance these data, we trawl Capital IQ, VentureXpert and SDC to find any public information on sample LBO syndications, and verify whether an LBO was ultimately sold or underwent an IPO. 6ext, we identify the college affiliations of all senior management in PE firms investing in all LBOs, using data from the PPMs and various databases.This information highlights shared college connections based on Masters of Business Administration (MBA) degrees, which have been used in related studies on the influence of peer effects on financial performance (Ahern, Duchin, & Shumway, 2014; Cohen, Frazzini, &  Malloy, 2008, 2010; Lerner & Malmendier, 2013; Shue, 2013).We place particular emphasis on Harvard Business School (HBS) due to its strong alumni ties (e.g., Lee, 2017; Lerner & Malmendier, 2013) and its historical role as a significant source of company executives and finance professionals (Lee, 2017; Lerner & Malmendier, 2013).It may not come as a surprise that 51% of the management teams of PE investors from the 941 sample LBOs hold an MBA degree from HBS.
The analyses of these data reveal that deal value, capital invested, geographic distance, and management buyout (MBO) are drivers of LBO syndication. 7Team characteristics, such as diversity of skills or MBA cohorts, do not appear to play a role in a PE's decision to syndicate an LBO.It's worth noting that the drivers we identify could be incidental due to a PE investor's preference in syndicating an LBO (Boone &  Mulherin, 2011; Degeorge et al., 2016; Gorbenko & Malenko, 2014 and  Officer et al., 2010).To address this concern and potential bias from omitted variables, we employ a modified selection model in pooled cross-sectional regressions based on whether IRRs are disclosed for an investment in an LBO.The primary findings remain consistent to this adjustment.
While we find no clear evidence that team characteristics motivate PEs to co-syndicate, it is still possible that shared alumni connections could potentially influence PE partner selection in LBO syndications.To reduce the effect of selection on observable variables in the pooled regressions, we estimate McFadden conditional logit regressions on a subset of 153 LBO investments involving only two PE co-investors.This analysis reveals that MBA graduates from Harvard and Chicago are more inclined to syndicate with their fellow alumni.Furthermore, the existence of at least one pre-existing HBS alumni connection among the management teams of co-investors, which predates the LBO, significantly boosts deal value and capital invested.For example, a single overlap of HBS alumni results in an average increase of 80% in capital invested by PE co-investors.In comparison, when there is an overlap of other Ivy League alumni or no overlap at all, both the capital invested, and deal value are generally much lower.
One possible interpretation of the results so far could be a home bias among the management teams of PE co-investors.If PE investor teams are geographically close, then the observed results could be attributed to local labor markets and/or information networks (Engelberg, Gao, &  Parsons, 2012; Knyazeva, Knyazeva, & Masulis, 2013).However, we find no evidence to suggest that geographic proximity among the management teams of PE co-investors is linked to the likelihood of HBS-HBS syndication, which dismisses the home bias explanation.
Another possibility is that the Crimson alumni connections among the management teams of PE co-investors are merely capturing deals that 3 There is evidence that the social connections developed at HBS have a profound and enduring impact, influencing entrepreneurial and managerial choices even years after graduation.A study by Lerner and Malmendier (2013)  on HBS MBA students revealed a lower level of unsuccessful ventures.Lee  (2017) discovered that having social ties in venture capital results in more deal opportunities and larger funds under management for the initial funds of HBS executive-led venture capital firms.This suggests that HBS graduates gain an advantage by accessing top-tier deals through their network. 4The size and duration of our sample are not uncommon in private equity (PE) research.For instance, the study by Braun, Engel, Hieber, and Zagst (2011)  used a sample of 460 buyouts from 1990 to 2005, while Kim and Palia (2014)  analyzed 526 PE transactions from 1980 to 2009.We recognize that our sample size and time frame are somewhat smaller compared to the research conducted by Braun et al. (2017), and Braun, Jenkinson and Schemmerl (2020), which focus on PE co-investments among limited partners (LPs) and institutional investors.Therefore, a potential direction for future research could be to explore the influence of college alumni networks on PE co-investments involving LPs and other institutional investors, in addition to other PE firms.
recognized that investors experience substantial wealth gains when a leveraged buyout (LBO) firm re-enters the public market. 7Braun, Jenkinson, & Schemmerl, 2020 report that larger LBOs are more likely to be offered as co-investments for the purpose of diversification, which in the scope of their study refers to co-investing with limited partners.We believe a similar rationale applies to our work, where co-investments club PE investors together in larger syndicated LBOs.Kim and Palia (2014) examine reasons for alliance formation between PE bidders when compared to solesponsored PE deals and find that private bidders are more likely to form an alliance in a diversifying acquisition.We moderate both these effects in our study.
might not have otherwise been syndicated.For example, differences in capital investment among PE investors might deter some PE investors from participating in LBO syndication.This hurdle might be overcome by a bond formed through an alumni connection that fosters mutual trust.Indeed, differences in capital invested among PE co-investors are positively related with syndication for Crimson management team pairings.This suggests that social capital among management teams of PE investors facilitates LBO syndication reducing agency costs.2F.
In the second part of the study, we delve into the normative implications of our findings.Having established that affiliation based on alumni relationships influence co-investing by PEs, we explore if transactions linked through alumni networks are superior in terms of overall returns for LBO investors.The overall relationship between the performance of PE co-investments and syndication is negative, although not consistently statistically significant.In line with Kaplan and Lerner  (2010), Harris, Jenkinson, and Kaplan (2014), Fang et al. (2015), Buchner, Mohamed, and Schwienbacher (2016), and Robinson and  Sensoy (2016), among others, the lowest valuation multiples and IRRs are observed in syndicated LBOs.However, this trend reverses when we focus on matched pairs of PE co-investor management teams that are HBS graduates.Within this subset, we find that PE co-investors connected through HBS alumni consistently outperform all other alumni connected PE co-investor teams, as well as pairings where there are no alumni-connected PE co-investor teams.This finding is not solely driven by an HBS effect per se and remains valid after accounting for key dealspecific variables and fixed effects. 8e realize a need to address endogeneity issues before we can infer causality from our findings.One concern is the possibility of reverse causality, where PE co-investments result in the formation of social relationships (Cohen et al., 2010 and Engelberg et al., 2012).On average, an 11-year time lag exists between attending HBS and forming relationships and making LBO investment decisions.This helps mitigate concerns about reverse causality. 9Another concern is that our findings could be driven by unobserved attributes of PE firms or syndicated LBOs that correlate with the personal connections of PE co-investor management teams based on an alumni network.PE co-investors with greater assets under management (AUM) are likely to share more common alumni connections with other PE co-investors, not just those they coinvest with (Lopez-de-Silanes et al., 2015).Similarly, more established PE co-investing firms may have more alumni ties with other PE investors, potentially mechanically, leading to better scaling and screening capabilities (Robinson & Sensoy, 2016).Finally, larger PE investors may share more alumni connections with other PE investors if they encounter fewer financial constraints, enabling them to partake in more and larger syndicated LBOs (Demiroglu & James, 2010).
Our study adds to earlier work on trust and social capital in financial contracting, LBO syndication and co-investing, and buyout performance in several ways.First, we investigate shared college connections in PE co-investing management teams across various countries where information access and agency cost concerns are likely more acute.This supplements studies on (Harvard) alumni networks among security analysts and directors in U.S. publicly traded firms (Brochet, Miller, &  Srinivasan, 2013; Cohen et al., 2010 and Ishii & Xuan, 2014).Second, our work on social capital in PE investments builds upon the work of Gompers, Mukharlyamov, and Xuan (2016), Bottazzi et al. (2016) and Bubna, Das and Prabhala (2020) who focus on VC investments.Third, we highlight shared connections as a key driver in the formation of LBO syndicates (Boone & Mulherin, 2011; Degeorge et al., 2016; Giovannetti  & Pipic, 2023; Officer et al., 2010).Fourth, the detailed nature of PPM data enables us to measure PE investment performance more accurately.Fifth, we add to the work of Acharya et al. (2013) and Gompers, Kaplan,  and Mukharlyamov (2016), among others, on the experience of PE management teams and fund performance.Finally, we offer preliminary evidence on shared investor connections as a determinant of PE performance persistence and scalability of equity returns (see Braun, Jenkinson, & Stoff, 2017; Buchner et al., 2016; Giovannetti & Pipic, 2023;  Humphery-Jenner, 2012; Lopez-de-Silanes et al., 2015). 11ompers, Mukharlyamov, and Xuan ( 2016), a study closely related to ours, uses a sample of VC co-investments from 1973 to 2003 to study how personal traits (homophily, or affinity) influence the selection of syndication partners and the success of investments (IPO or not).Our paper focuses on syndication decisions at the senior management team level, both within and across firms, which is more suitable to PE investments.We connect these decisions to investment success using performance metrics.Importantly, our emphasis is not on homophily, or similarity, but on trust, or cooperative norms among people that can be cultivated during education, specifically an MBA, a prevalent qualification in the PE industry uniquely tailored for each top school.Our findings indicate that not all such connections are significant and some work much better than others.For example, Harvard affiliations among PE co-investors decrease agency costs in LBO syndications by undertaking deals that might not have been executed otherwise, while overcoming financial constraints through diversification.Crimson PE pairings also uniquely generate superior investment returns in LBO coinvesting.Our findings underscore the role of social capital in deal formation and value creation, offering practical insights for investors contemplating a PE co-investment.
The structure of the paper is as follows.The next section details the collection of data on sample LBOs, PEs, and their senior management teams, and outlines characteristics of the sample.In Section 3, we conduct univariate and multivariate analyses to identify the factors influencing LBO syndication, including the role of college alumni connections among PE co-investor management teams.Section 4 is dedicated to exploring PE co-investment returns and shared college 8 Our findings regarding Crimson connections also add to the discussion on the scalability of private equity returns, as explored by Humphery-Jenner  (2012), Lopez-de-Silanes et al. (2015), Malenko and Malenko (2016), and Robinson and Sensoy (2016).A comprehensive examination of this topic is reserved for future research. 9Another less significant concern is the issue of self-selection, or the endogenous grouping among graduate batches.In this scenario, school affiliations simply mirror the shared abilities or skills within a cohort.According to Lerner and Malmendier (2013) and Shue (2013), the Harvard MBA student cohorts are grouped in a balanced and random manner, supervised by the HBS administration, which tempers this concern.

Data and sample
In this section, we assemble a database on LBOs using PPMs and investor data from various sources.PPMs are fund raising prospectuses produced by PE firms and distributed to select clients.They include information on securities offered in a placement, as well as details and performance data on all prior investments made by the PE firm, whether they are currently held in a portfolio or have been divested.The key benefit of using PPMs is that they provide performance metrics and characteristics for each PE firm's LBOs at the investment level. 12he proprietary PPMs provided by various financial institutions, cover 6611 PE investments dating back to as early as 1971. 13Our primary interest in PPMs pertains to LBO investments.Therefore, we initially screen PPMs to identify all references to buyouts between 1991 and 2004. 14We then scrutinize each reference to narrow down the sample to investments that specify an exit date and report information about the investment fund and portfolio company (PC).This process reduces the initial sample to 1317 LBO investments.For each of these, we rescan PPMs to record: the initiation and exit (if realized) date of each buyout; the status of each buyout classified as realized, partially realized, or unrealized; the exit route if relevant; the industry classification (SIC code) and country of incorporation of the LBO firm; the total equity value invested by each PE firm; the total distributions (if any) to each PE firm; and the current value of any remaining unsold equity stake by a PE firm at the time of PPM issuance.We also gather performance metrics from PPMs on each PE firm investment: multiple is the value of total equity scaled by investment size and the gross internal rate of return (GIRR). 15We then cross-reference each investment, whenever possible, with data from Capital IQ, VentureXpert, and Securities Data 14 Ending the sample in 2004 aligns with the time required for all our investments to be exited, with the latest exit year in our sample being 2008.Ideally, we would like the sample to extend beyond 2004, but due to the confidential nature of the data and the sensitivity of the providers, this is not feasible.We have no reason to suspect that the results would vary over time. 15The term 'Multiple' refers to the current equity valuation plus cash received, scaled by the total cash invested.'GIRR' is calculated by interpolating multiple^ (1/duration)-1.Both Multiple and GIRR are at the fund level and are gross of fees, including cash transfers and loans from the fund to the portfolio company.This table provides summary statistics on management team characteristics involved in sample LBO investments.Panel A shows the size distribution of investment firms, where size is measured by the number of senior professionals in the firm at the time of investment.Panel B reports the nationality on investment professionals.
Panel C provides other managerial characteristics.Panel D reports summary statistics on MBA graduates and other educational attainments and disciplines (engineering, law, and regular master's degree).
Corporation (SDC) to confirm a buyout is syndicated and identify PE firms.2.PPMs provide biographical details about the senior management in each PE firm participating in the LBO, including data on senior managers who have left the firm. 16We gather biographical background information for all senior managers in each PE firm from sample PPMs, as available. 17The depth of employment history and educational background information in PPMs can vary.We fill any obvious gaps using resources like Galante's Venture Capital and Private Equity Directory, ZoomInfo, LinkedIn, and company websites to compile the most comprehensive education and employment profile on each PE firm senior management team. 18The key bio variable that we need for each senior manager includes academic degree(s) along with the name of the granting institution(s) and graduation date(s), as well as professional certifications with respective dates.
We drop an LBO investment by a PE firm from the sample when there is incomplete or conflicting information across various data sources.This requirement reduces the sample to 941 LBO investments, of which 406 are syndicated with two or more PE firms across 825 unique PCs.The data we assemble is unique in three ways.First, we have comprehensive historical biographies of college alumni and employment records for senior PE management.Second, we know the identity and senior management of PE firms involved in a large sample of both syndicated and non-syndicated LBOs. 19Last, we have performance data at the investment level for each PE firm.
Table 1 presents summary statistics for each sample LBO.Panels A, B, and C report the deal value, capital invested, and transaction year, respectively.Columns D and E enumerate the locations of portfolio companies and their industry affiliations.An inspection of Table 1 reveals six key observations.First, 43% of sample LBOs are syndicated between two or more buyout firms, aligning with the findings of Officer  et al. (2010) who studied U.S. LBOs from 1984 to 2007.Second, syndicated LBOs have deal values that are more than triple those of nonsyndicated LBOs.Third, over 90% of both syndicated and nonsyndicated LBOs have less than $100 million in capital invested.Fourth, 59% of LBOs took place between 1995 and 1999, coinciding with the buyout boom of the 1990's.Fifth, 53% of portfolio companies are based in the U.S., with another 20% domiciled in the U.K. The share of portfolio companies by country is similar for both syndicated and nonsyndicated deals.Finally, over half of the sample portfolio companies operate in the manufacturing or services sectors, with the trend evenly This table provides coefficient estimates (marginal effects) on LBO syndication using binomial probit models.The dependent variable equals one for syndicated LBO conducted by multiple co-investors and 0 otherwise.For the explanatory variables, investment size is measured by two proxies: deal value and capital invested in U.S. million dollars.Geographic distance is defined as the distance between the capital city of the portfolio company and the investment firm.Firm Experience is the difference between the founding year of the investment firm and the acquiring year of the portfolio company.Geographic distance and Firm Experience are transformed by logarithm.Management buyout is an indicator variable which equals the value one for management buyouts and zero otherwise.Investment team characteristics are proxied using density variables (i.e., number of investment professionals who have specific characteristics scaled by size of the investment team within a PE firm).Skill diversity is the standard deviation on three variables of skills, i.e., Law, Business, and Engineering.Standard deviations are clustered at the industry level.Z-values are reported in parentheses, and the symbols *, **, and *** denote statistical significance at the 0.10, 0.05, and 0.01 levels, respectively. 16Commercial PE databases such as CEPRES do not provide current or historical bios on PE management. 17The titles that are considered as senior management include: managing director, partner (but not operating, administrative, advising, recruiting, technology, venture, or special partner), principal (excluding finance principal), director, executive director (excluding independent), senior director, controller, senior manager, investment director, chief executive, chairman (excluding vice chairman), chief financial officer, and founder.Titles such as vice president, analyst, investment manager, investor relations, associate director, marketing, associate, assistant, account, and advisor are not considered as senior titles. 18The Directory, which began in 1996, is an annual publication by Dow Jones & Company.It includes information on over 3000 professionals in private equity (PE), venture capital (VC), and mezzanine investments from 1995 onwards.However, the depth of coverage can differ greatly.For instance, some firms list all their employees in the directory, while others only report their top professionals.ZoomInfo is an online resource that offers biographical and employment data for 65 million business professionals.LinkedIn is a social networking website for professionals, boasting a user base of 225 million individuals spread across >200 countries and territories.
spread between syndicated and non-syndicated deals. 20fter detailing the LBO investments, we shift our focus to their management teams.Table 2 provides summary statistics on the size of the management team, nationality, background, and academic qualifications in Panels A, B, C and D respectively.Panels A and B show that the size of LBO management teams varies from 1 to 20 members, with a third of firms having between 5 and 10 members.The nationality of team members aligns with the location of portfolio companies.
As shown in Panel C, there are 1304 PE senior management that participated in our sample deals.We focus on the College alumni connections among these individuals and classify each manager's education by discipline and degree type.The disciplines include business, legal studies, and engineering, while the degree types are MBA and other nonbusiness-master's degree.Interestingly, about 70% of the senior management from PE firms involved in LBOs have business education, and half hold an MBA.Panel D provides the distribution of MBAs by academic institution and functional background.It reveals that 30% of MBAs graduated from HBS, which surpasses the combined total of MBA graduates from Wharton, Columbia, Stanford, and the University of Chicago. 21Moreover, it's worth noting that an average of 13% of all MBA graduates also have an Engineering degree.
In the next section, our first task is to scrutinize the factors influencing LBO syndication.Then, we delve into the relationship between the alumni networks of top PE managers and the formation of LBO syndicates.That is, we estimate how overlapping college connections among PE management teams affect the choice of partners in LBO syndication.To alleviate the influence of selection based on observable factors, we run McFadden's conditional pooled logit regressions.

LBO syndication and shared college connections
This section begins with a simple question: Given all other factors being equal, do the senior management of PE firms who share college connections tend to form syndicates? To answer this question, we use multivariate regressions to pinpoint the factors influencing LBO syndication. 22The results of these analyses are displayed in Table 3. Across all the specifications in this table, we estimate binomial probit models to regress LBO syndication, represented by an indicator variable that is set to one, against a variety of deal and PE investor attributes.This approach facilitates a 'horse race' among the principal independent variables to discern the primary catalysts of LBO syndication.The selection of independent variables in this table is guided by prior studies on LBOs, PE, and PE syndication (Acharya et al., 2013 and Lopez-de-Silanes et al., 2015), as well as survey data (Gompers, Kaplan, &   Mukharlyamov, 2016).
The base regressions in the first two columns of Table 3 consider deal value, capital invested, geographic distance, PE experience, and MBO as main independent variables.Acknowledging that the sample is not random, we address selection issues by using IRR reporting in PPMs as an instrument within our two-stage framework.For brevity, we only present the results of the second stage of the regression models in Column 3. The subsequent columns in Table 3 augment the base regression in Column 2 with attributes of the PE firm and its senior management team, including skill diversity and the representation of MBA alumni in the top team. 23n inspection of Table 3 reveals that both the size of the investment and geographic distance play a significant role in the decision to syndicate LBOs.For example, an increase in capital invested by $10 million raises the probability of syndication by 3.3% (p-value = 0.06).Similarly, a 10-km increase in the geographic distance between the PE investor and the investee raises the likelihood of syndication by 1% (p-value ≤0.05).LBO syndication is also more common in MBOs by 18% (p-value ≤0.01), potentially to alleviate financial constraints.The proxy for team attributes by MBA subgroups and the experience of the PE investor is not statistically significant at the 0.10 level or less in the regressions.After accounting for deal-specific characteristics, holding an MBA or PE experience does not impact LBO syndication. 24Similarly, skill diversity does not correlate with the likelihood of LBO syndication, implying that homogeneity in discipline-specific skills across PE management teams does not determine whether an LBO will be syndicated. 25he findings indicate that deal-specific characteristics such as deal value, capital invested, the distance between the portfolio firm and the PE investor, and MBO are key determinants of the likelihood of LBO syndication.Attributes of the PE management team do not seem to influence the decision to syndicate an LBO.However, this analysis falls short as it ignores the impact of college alumni networks on LBO syndication.Cohen et al. (2010) argue that a major benefit of MBA alumni membership is the capacity to exchange information and secure resources, leading to improved deal performance.To rectify this oversight, we investigate whether PE senior managers with MBAs have a greater propensity to syndicate LBOs with other PE managers who also hold MBAs, and subsequently check if an alumni connection increases the likelihood of syndication.
In order to model the selection of LBO syndication partners and tease out whether MBAs tend to syndicate more with specific partners due to alumni connections, we screen syndicated LBOs that involve a pair of PE co-investors.This filtering process identifies 153 syndicated LBOs with 20 In general, the trends over time for both syndicated and non-syndicated leveraged buyouts (LBOs) are alike and align with the description of buyout and venture capital (VC) activity as outlined in Stromberg (2008). 21In our sample of syndicated investments from 1992 to 2004, approximately 30% of executives are Harvard graduates.This is slightly lower than the 38% reported by Gompers, Kaplan, and Mukharlyamov (2016a) in their 2011 survey of large firms in the private equity (PE) industry.The discrepancy could be attributed to the different time periods considered and/or our focus on PE firms involved in syndicated LBOs. 22Table 1 in the Appendix presents the characteristics of deals and private equity (PE) investors for both syndicated and non-syndicated LBOs.The main observations from this table are that the deal value, capital invested, size of the PE fund, age, experience, and the distance between the portfolio firm and the PE investment firm are all greater in syndicated LBOs.The team size in syndicated LBOs is also larger.LBO syndication is less likely in transactions involving company founders, but more likely when senior management hold MBA degrees.This increased likelihood could be a function of MBA specific skills needed in syndication or the ability to network among the PE investment community possibly thought a college alumni connection.None of the other team attributes presented in this table are different between syndicated and non-syndicated LBOs at the 0.10 level or less. 23The term 'skill diversity' refers to the standard deviation of three density variables that represent three functional degree disciplines (law, business, and engineering) pursued by the senior management in each sample PE firm (at the investment level).Each density variable is calculated as the number of team members who have earned related degrees, normalized by the team size (with values ranging from 0 to 1).A higher value of the skill diversity variable indicates a greater uniformity in educational skills and functional background.Similarly, each MBA subgroup variable is a density measure that calculates the ratio of the number of PE team professionals with specific characteristics to the team size.The primary benefit of using density is that it minimizes any potential influence of team size on other variables. 24This result remains when we employ a different metric to gauge PE experience, which is the number of deals in the 3 years preceding the PE firm's key investment, normalized by the total number of deals in the same 3-year period.It's important to note that using measures based on past deal numbers leads to a decrease in the sample size.We consider a 3-year window to be a prudent choice that doesn't result in the loss of too many investments for analysis. 25Table 2 in the Appendix presents the pairwise correlation coefficients between deal characteristics and PE team attributes in relation to the probability of LBO syndication.The main takeaway from this table is that the size of the investment and invested capital, along with numerous PE team attributes, are highly correlated.paired PE co-investors.The rationale behind this approach is grounded in the assumption that PE investors seek a single 'best' syndication partner, where each PE investment firm (f) at a given time (t) can choose from the pool of all other PE investing firms (i) in the sample that have available data on team attributes at time t., 2627 The McFadden conditional logit model is employed to handle the selection of one alternative among many.The dependent variable in this model is an indicator that takes the value one for PE co-investing-candidate pairs at time t and zero otherwise. 28For each PE firm ʄ denotes the utility derived from selecting a syndication partner i ∈ {0, …, I} as y * fi = β ʹ x fi + ϵ fi .Here x fi is a vector of observable PE co-investor attributes, while ϵ fi captures unobservable characteristics that might affect their utility.We then let i be the choice of PE firm ʄ that maximizes its utility y f = argumax Following McFadden (1974), if Since our primary focus is on the relations among senior management teams of PE co-investors.The most significant interactions are the team characteristics within PE firms, specifically between firmcandidate pairs.In our regressions we prioritize the attributes of PE co-investor pairs over those of individual PE investors (f).Hence, we initially investigate if PE co-investor management teams, which include MBA graduates, syndicate LBOs among themselves.Following this, we incorporate college alumni membership as an incremental factor to determine its influence on LBO syndication.The base regression in this model states Matching Propensity = ʄ (MBA(f)*MBA(i), MBA(f)*Engineer(i), MBA(f)*Law(i), MBA(f)*Other Master(i)), where "f" refers to the PE co-investor top management team of the focal investment and "i" refers to each potential syndication (or co-investing) partner.The density variable measures certain attributes at the PE level, where MBA(f) represents the number of MBA graduates, adjusted by the team size in the PE co-investor.By using interaction terms in individual density variables of potential pairs, we can accommodate non-linear relationships of team characteristics between pairs of firms.
Table 4 reports the coefficients of these regressions, with the independent variables being interaction terms on density variables associated with team attributes as described above.Column 1 is considered as the base regression specification.Columns 2 to 6 provide estimates on PE co-investor management teams with MBAs categorized by college alumni.Despite the low fraction of PE investment professionals with post-graduate degrees from non-business disciplines, the results in Column 1 of Table 4 indicate a preference among MBAs to collaborate with each other rather than with investment professionals with graduate degrees from non-business disciplines.The key coefficient of 1.81 on MBA(f)*MBA(i) is positive and significant at the 0.01 level.Column 6 of the same table shows that the Harvard-Harvard syndication coefficient is 7.23 and statistically significant at the 0.01 level, suggesting that Harvard MBAs prefer to co-invest with each other and that Harvard affiliations play a significant role when it comes to syndicating LBOs.It is well-known that a strong network, like the HBS alumni, not only offers valuable business connections but also promotes mutual trust among its alumni members (Cohen et al., 2008).MBAs from other colleges do not express any alumni preference for co-investing in LBOs, with the exception of Chicago alumni who also appear to favor LBO syndication among themselves, as indicated by a statistically significant syndication coefficient at the 0.05 level in Column 5.
Nevertheless, we have not yet dismissed omitted variables that could potentially moderate agency costs among co-investors, which might explain the relationship among Crimson alumni in LBO syndication.One such variable could be co-investing in deals where there are significant This table reports coefficient estimates on the conditional logit model of syndication partner selection for MBA graduates from leading universities on the subsample of matched PE co-investments.Each investment firm(f) at time t can choose among all other investment firms(i) in the sample with available team attributes data at time t.The dependent variable is an indicator that equals one for the investment firm-candidate pairs that co-invest with each other at the time when the deal is initiated and 0 otherwise.Independent variables are measured in density, which is defined as the number of the professionals who have specific characteristics, scaled by the number of the whole investment team within each firm.Model ( 1) is the baseline model for MBA graduates.Models ( 2)-( 6) provide coefficient estimates for different subgroups by University.Models ( 7)-( 8) extend Model (1) to include two proxies on agency costs: differences in capital invested and geographic difference between co-investors.The proxies on agency costs are transformed by logarithm.Z-values are reported in parentheses, and the symbols *, **, and *** denote statistical significance at the 0.1, 0.05, and 0.01 levels, respectively. 26The findings outlined in this section remain consistent even when the method is adjusted to include more than two co-investors.A discussion on alternative strategies for pairing PE firms in LBO syndication is provided in Section 5. 27 Counterfactuals are drawn from the initial dataset, which includes 6611 investments dating back to 1971.Depending on the year, the count of possible pairs fluctuates between 132 and 177. 28The selection process is in line with Kuhnen (2009) and follows the random utility model of McFadden (1974).
differences in the capital committed to a syndicated transaction, especially if such partnerships impact future deal flow generation.To address this issue, we examine the difference in capital invested among PEs in Column 7 of Table 4.The coefficient on the interaction between HBS-HBS co-investors and the difference in capital invested is 2.69 with a p-value of 0.01.This reaffirms the view that agency costs stemming from asymmetric investments in syndicated LBOs can be alleviated when PE firm managers, who are Harvard alumni, invest with each other.Home bias could be yet another omitted variable.If co-investors are located nearby, this could offer advantages in monitoring, and our results could reflect local labor markets and information networks.In Column 8 of Table 4, we demonstrate that the marginal effect of geographic distance on the HBS-to-HBS co-investor connection is not statistically significant.This implies that home bias among Crimson PE co-investors does not drive LBO syndication. 29n sum, investment size, geographic proximity, and MBOs are key considerations for PE investment teams when syndicating an LBO.Regarding syndication partners, PE investment teams with MBAs tend to syndicate with each another, a trend largely driven by the Harvard MBA alumni network.Moreover, a preference for co-investment among Harvard alumni is found in deals where there are significant differences in capital invested among PE co-investors.This aligns with the view that Harvard ties not only facilitate information sharing and resource acquisition but also foster trust in deals to mitigate agency costs related to differences in invested capital and financial constraints that arise in syndicated transactions.An obvious question this analysis raises is whether deals connected through Crimson alumni are beneficial to LBO co-investors.30

LBO syndication, shared college connections and performance
To assess if Harvard-to-Harvard co-investing pays off, and whether matching MBAs by their college alumni creates value for co-investors in syndicated LBOs, we propose a straightforward game that is illustrated in Fig. 1.
Undoubtedly, syndication is favored when it results in better deal selection and/or provides additional value (NPV) to PE co-investors.If this is true, then syndicated deals should, all else being equal, offer lower returns to PE co-investors.This is because PEs wouldn't need to copartner LBOs that they consider promising and more certain, assuming a PE possesses the necessary scale, expertise, and resources to carry out such deals.However, it's reasonable to assume that PEs wouldn't resort  This table reports summary statistics on LBO investment performance using value multiple and gross internal rate of return (GIRR).The value multiple is defined as the ratio of total cash received from the investment plus its current valuation (if not fully liquidated) to the total cash invested (that is, total value divided by investment size, gross of fees) and GIRR is defined as the internal rate of return, gross of fees, of the investment in the PPM.some noise.For instance, it's likely that most members of a graduating class won't know all their peers, much less keep in touch after graduation, which could dilute any effect we might detect.We exercise caution to balance the potential for information exchange among PE co-investor teams syndicating LBOs to mitigate error in variables bias.
to syndication unless necessary. 31n the first stage of the game, firms assess deals and, if required, they proceed to the second stage to seek a co-partner.At this stage, a typical deal can result in three outcomes: NPV 1 = A > 0, NPV 1 < 0, and NPV 1 = 0. Firms ignore investments with negative NPV and only invest in positive NPV projects.For all other deals, including those with uncertain NPVs, or those that involve value-added services to/from a co-investor, syndication is more likely.In the second stage of the game, a syndicated deal can yield three possible NPVs: NPV 12 = A > 0, NPV 12 < 0, and NPV 12 = 0.This assumes that any payoff will be equally shared among two PE co-investors. 32As shown in Fig. 1, LBO investments can either be syndicated or non-syndicated, hence the best performers will satisfy the following conditions: Identifying superior LBO investments, whether syndicated or nonsyndicated is quaint.However, the key question is whether the expected outcomes are realized in the data.To address this, we collect data on two investment performance measures --value multiple and gross internal rate of return (GIRR) -from PPMs for each LBO investment.We present these data at the investment level for both syndicated and non-Fig.2. Histogram of LBO Performance. 31Looking ahead to Tables 5 and 6, we find a negative correlation between LBO syndication and performance, although this is not consistently statistically significant. 32Relaxing this assumption does not change the primary conclusions.This table shows the relationship between syndication and LBO investment performance.The dependent variable is investment performance, proxied by multiple (Panel A) and gross internal rate of return (Panel B), both winsorized at the 5% level.The independent variables include an indicator variable that equals one for syndicated investments and 0 otherwise.Deal value and capital invested are in $U.S. million.Geographic distance is measured in miles between the capital city of the portfolio company and that of the investment firm.Firm experience is the difference between the founding year of the PE firm and the acquiring year of the portfolio company.All variables are transformed by logarithm.The skill diversity variable is the standard deviation of three variables of skill --Law, Business, and Engineering.Team attributes are measured by density, which is defined as the number of the professionals who have specific characteristics, scaled by the size of the investment team in a firm.Models (1)-( 4) report OLS coefficient estimates with standard deviations clustered at the industry level.Models ( 5)-( 6) report coefficient estimates on the two-stage treatment effect model.Z-values (T-values) are reported in parentheses, and the symbols *, **, and *** denote statistical significance at the 0.1, 0.05, and 0.01 levels, respectively.syndicated LBOs in Table 5. 33 Table 5 (and Fig. 2) confirm the statistics from the two-stage game, indicating that syndicated LBO investments typically earn lower average returns for PE co-investors.The value multiple for syndicated LBOs stands at 3.74, which is one-fifth the size compared to non-syndicated deals.To build on this, we run multivariate regressions with the performance at the investment level for each PE firm as the dependent variable and deal size, capital invested, geographic distance, PE experience, skill diversity, and syndication as independent variables.Panel A of Table 6 reports regression coefficients when the dependent variable is the value multiple, while Panel B shows coefficients when the dependent variable is gross IRR. 34The independent variables are borrowed from the base specifications in Columns 1 and 2 of Table 3.The key coefficient in Columns 1 through 4 of Table 6 is syndication, which is negative and statistically significant at the 0.01 level or less.This negative correlation between syndication and performance remains robust and statistically significant when deal value is the independent variable but becomes less pronounced when gross IRR substitutes deal value.This implies that the relationship between PE investment performance and LBO syndication is not clear cut.
It's expected that syndicated investments in the game are more likely to be inferior compared to non-syndicated deals, making syndication a self-selected 'treatment' instead of a random variable.To better understand this relationship, we employ a treatment effect model that yields two-stage consistent coefficient estimates.We present the coefficients for value multiple and GIRR in Columns 5 and 6 of Panels A and B in Table 6, respectively.In this setup, the correlation between investment returns (GIRR and value multiple) and LBO syndication disappears.Thus, the main takeaway from Table 6 is that the relationship between LBO syndication and investment performance is complex, even when we account for selection bias.
Next, we examine whether the MBA qualification by college alumni drives LBO investment performance for PEs when syndicating LBOs.We report performance data for deals involving Harvard alumni in Panel A of Table 7.A quick look at Panel A shows higher average and median value multiple and GIRR for deals involving Harvard alumni.We then run average and median tests on investment performance for syndicated and non-syndicated deals that involve Harvard MBA alumni compared to those that do not.The results in Panel B of Table 7 indicate that LBOs managed by Crimson alumni are linked with higher investment returns in non-syndicated deals, but this is not the case for syndicated deals.
To elaborate on the impact of MBA alumni on syndication performance, we run multivariate OLS regressions as shown in Table 8.The primary independent variable is a density variable that represents the fraction of the PE team with MBAs in Column 1, and the college affiliations of MBAs in Columns 2 to 6.We include key deal-specific variables and fixed effects across all specifications.The first six columns of this table pertain to non-syndicated LBOs, and the last six to syndicated deals.An inspection of this table reveals three main insights.First, only HBS MBA graduates show a positive correlation with investment performance in both syndicated and non-syndicated LBOs.The coefficient for HBS alumni membership is positive and statistically significant at the 0.10 level or less for both GIRR and value multiple in non-syndicated deals, and is significant only for the value multiple in syndicated transactions.Second, we see that other MBA college alumni, notably Wharton, Stanford, and Chicago are associated with positive performance, but these results are only evident in non-syndicated transactions and are not consistent for both value multiple and GIRR.This leads us to conclude that there is little consistency in the relationship between This table reports LBO investment performance using value multiple and gross internal rate of return (GIRR) of deals made by investors involving Harvard MBA graduates.The value multiple is defined as the ratio of total cash received from the investment plus its current valuation (if not fully liquidated) to the total cash invested (that is, total value divided by investment size, gross of fees) and GIRR is defined as the internal rate of return, gross of fees, of the investment in the PPM.Panel A shows summary statistics of these deals.Panel B shows comparisons in investment performance (winsorized at the 5% level) of deals with versus without Harvard MBA graduates. 33Figure 2 presents corresponding histograms for syndicated and nonsyndicated leveraged buyouts (LBOs), with data trimmed at the 0.05 level.This chart shows that the value multiple and GIRR are concentrated around the mean with smaller standard deviations for syndicated LBO investments, while they are more dispersed in non-syndicated deals. 34The standard errors on the coefficient estimates in Table 6 are clustered at the industry level.These results are robust when clustered at the PE firm level.Moreover, a small number of investments by each sample PE firm does not appear to explain the result as these typically range anywhere from 1 to 48, with a mean and median of 7 and 4, respectively.college alumni, syndicated LBOs, and investment performance.Therefore, alumni connection alone does not dictate the relationship between syndicated LBOs and investment performance.
After establishing that PE alumni connections alone do not influence the relationship between LBO syndication and performance, we proceed to check if matching college alumni across different PEs in syndicated transactions impacts performance.In this context, we switch the dependent variable in Table 4, which is LBO syndication, with a performance measure and report the results in Table 9. Panels A and B of this table, report regression results when the dependent variable is value multiple and GIRR, respectively.Columns 1 to 12 in both panels indicate that the key coefficient MBA(i)*MBA(f) representing the pairing of college alumni, is statistically significant with p-values below 0.01 for the HBS-HBS pairing, regardless of the performance metric used.In Column 12 of Panel A, the coefficient on value multiple is 6.49 with a pvalue of 0.01, and GIRR in the same column of Panel B is 2.12 with a pvalue of 0.05.No other alumni pairings report superior performance (in both metrics) in syndicated LBOs.This result supports the view that Harvard connections produce above-average performance when Crimson alumni co-invest to syndicate LBOs.The outperformance is not attributed to any performance advantage gained from an MBA-educated team, a large presence of Harvard MBAs in the team, or deal and/or investor characteristics.Performance is simply higher when Harvard MBA alumni from different PEs opt to syndicate LBOs together.
Collectively, the evidence supports a simple conclusion: Harvard MBAs collaborate with each other because such cooperation can reduce agency costs that might arise in large asymmetric co-investments requiring trust to create value, such as in syndicated LBOs.That said, it's challenging to comment on the potential benefits of syndication for other MBA alumni networks, as syndication could be pursued for reasons beyond those we have explored.

Robustness checks
Having already addressed concerns related to reverse causality, selection, and endogeneity, we now conduct two additional robustness tests to eliminate spurious correlations.The first test deals with data and modeling challenges, while the second test investigates whether known agency costs linked to PE investments can explain the observed Crimson alumni effect in LBO investment performance.5F25 This table shows how MBA team attributes affect investment performance of non-syndicated and syndicated investments, respectively.The dependent variable is investment performance captured by multiple (Panel A) and gross internal rate of return (Panel B), both winsorized at the 5% level.The key independent variable is MBA(f), which is a density variable defined as the number of the professionals who have (specific) MBA degrees, scaled by the size of the investment team in the firm.
The remaining independent variables are size which is the capital invested in US million dollars, distance as measured by miles between the capital city of the portfolio company and investment firm, and experience is the difference between the founding year of the investment firm and the acquiring year of the portfolio company.
Aforementioned independent variables are transformed by logarithm.Each specification provides coefficient estimates for different subgroups of MBA graduates.Standard deviations are clustered at the industry level.T-values are reported in parentheses, and the symbols *, **, and *** denote statistical significance at the 0.1, 0.05, and 0.01 levels, respectively.

Data and modeling
(i) IPOan alternative performance metric: To further verify the impact of Harvard alumni connections on the performance of syndicated LBOs, we examine the proportion of syndicated LBOs that undergo an IPO for a second time.This analysis reveals that 22% of the co-invested LBOs reenter the public market, and among these, 41% involve an HBS-HBS pairing.This percentage surpasses all others in the sample and could explain the superior investment performance in HBS matched deals.35(ii) Country-effect: Our sample comprises LBO investments worldwide.As a test for sample stability, we exclude a sub-sample of 23 PCs from smaller-developing countries.This exclusion has no meaningful impact on the key regression coefficients in Tables 6   through 9. 36 We repeat this exercise, removing all European LBOs from the sample, and once again observe no change in the main results.(iii) Outliers: In Section 4, we present histograms that trim both tails of the performance data at 5%.To verify if this truncation affects the results, we re-run the regressions in Tables 6 through 9 using performance data trimmed at 1%.The results remain unchanged when trimmed at 1%, indicating that they are not influenced by how we treat outliers.(iv) Indicator on HBS-HBS overlap: To quantify the degree of overlap of Crimson alumni across PE firms, we employ a density variable.
An alternative approach could be to use an indicator variable when PE firms co-syndicating an LBO have Harvard alumni on their investment teams.The switch from a density variable to an indicator does not significantly impact the key findings.This table shows how MBA team attributes affect investment performance of the subsample of investments co-invested by only two investors.The dependent variable is investment performance, proxied by multiple (Panel A) and gross internal rate of return (Panel B), both winsorized at the 5% level.The main explanatory variable, MBA(f), is a density variable, defined as the number of the professionals who have (specific) MBA degrees, scaled by the number of the whole investment team members within the firm.The control variables include size, proxied by the capital invested in US million dollars.Distance is measured by the distance between the capital city of the portfolio company and that of the investment firm.Experience is the difference between the founding year of the investment firm and the acquiring year of the portfolio company.All these control variables are transformed by logarithm.Each specification provides coefficient estimates for different subgroups of MBA graduates.Standard deviations are clustered at the industry level.T-values are reported in parentheses, and the symbols *, **, and *** represent statistical significance at the 0.1, 0.05, and 0.01 levels, respectively.
the superior performance in Harvard syndicated LBOs. 37Another potential criticism concerns the consistency of the sample selection process, since we use a restricted sample in Tables 4 and 9 and the full sample in other tables.To address this issue, we repeat the analysis using the restricted sample and discover that the main findings still apply. 38We also rerun the regressions after removing repeated deals, and the results are nearly identical to those reported in the tables.39(vi) Higher bar: From Tables 7 through 9, we understand that LBO coinvestors connected to Crimson exhibit superior ex-postperformance.However, this result doesn't necessarily align with the idea that HBS connections boost syndicated returns, as they could potentially harm deal efficiency while enhancing performance for other reasons.In our work, we argue that information about high performing LBOs is available to all potential co-investors, but co-investors connected by college alumni simply use the information differently.Proponents of a higher bar story might argue that co-investors connected to alumni receive a different set of information, resulting in friction with no offsetting advantage.If there's a cost associated with co-investing related to alumni, then the uncertainty of meeting the higher bar implies that co-investors without connections must steer clear of those with connections.The results in Tables 4 and 9 show that HBS participation generally increases the likelihood of a deal being syndicated, and these deals perform better (with the most significant gains involving Crimson connected co-investors), which appears to contradict the higher bar narrative.

Other known agency costs
Prior research on LBO performance, PE scalability, and investment returns has found that factors such as buyout leverage, macroeconomic effects and endogenous timing, fund reputation, overfunding, and premature fund exits influence fund performance.Evidently, agency-based alternatives could also account for the relationship between HBS connections among co-investors and syndicated LBO performance.To manage this possibility, we adjust the regression in Column 12 of Table by sequentially adding a proxy for each of the variables and present the results in Table 10.
Consistent with Axelson et al. (2013), we employ two different indicator variables in the first two columns of Panel A in Table 10 to capture high leverage in buyouts: (a) the amount of debt; and (b) the ratio of debt to deal value.The inclusion of these proxies does not alter the significance of the key coefficient, MBA(f)*MBA(i).In Columns and 4 of Table 10, we adopt definitions of PE fund reputation from Demiroglu and James (2010), which are based on completed deals.In Column 3, the reputation variable is an indicator that is set to one for PEs  Robinson and Sensoy  (2013).They suggest that larger PE funds have a smaller proportion of managing partners responsible for allocating the fund's capital, leading to potential issues of incentive misalignment and diseconomies of scale.The proxy we employ is an indicator variable that equals one when the PE fund size exceeds the top 25% fund value for the entire sample (greater than $1 billion), and zero otherwise.The overfunding proxy in Column 5 is not significant at the 0.10 level or lower.
In Column 6 of Table 10, we use investment duration as a proxy for early fund exit, due to the lack of data on waterfall threshold or early IPOs (Gompers (1996), Robinson and Sensoy (2013), and Lopez-de-Silanes et al. ( 2015)).The early fund exit proxy is set to one when the investment duration is 2 years or less, and zero otherwise.The coefficient on duration in Column 6 has a p-value of 0.10, suggesting that short-term investments do yield marginally higher value multiples, while the MBA(f)*MBA(i) coefficient remains significant at the 0.05 level.In Panel B of Table 10, we replicate the regressions in Panel A with GIRR as the dependent variable.The result on the coefficient MBA(f) *MBA(i) is still significant at the 0.05 level or less in all columns of this panel, as are the coefficients on fund reputation and early exit in Columns 3, 4, and 6.Even after accounting for other agency cost factors in Table 10, we find that PE co-investment teams with overlapping Harvard MBA alumni produce superior investment performance in syndicated LBOs.
In our final examination of agency-based explanations, we scrutinize deals on management buyouts (MBO) and notice that Crimson syndicated LBOs are less frequent in MBO transactions.Out of 68 MBOs in the co-investment sample, we find that only 22% are syndicated by HBS alumni.Intriguingly, the performance in the HBS-HBS subsample is significantly superior to that in all other MBOs.This suggests that the enhanced performance in Crimson deals may be driven by MBOs, although we don't heavily rely on this result due to the small sample sizes involved. 40n sum, the process through which Crimson connections add value involves syndicating LBO transactions with significant disparities in capital invested among co-investors.This encourages PEs to undertake deals that otherwise would not be possible, leading to increased fund diversification and improved performance for PE co-investors. 41

Conclusion
We investigate if social capital among institutional investors can alleviate agency issues in syndicated transactions.Utilizing confidential investment-level data from financial institutions on global LBOs, we explore whether college connections among PE co-investors yield enhanced investment performance.The anticipated effect of shared college connections on LBO syndication is not clear-cut, as social capital could stimulate information exchange and decrease transaction expenses, thereby boosting co-investor returns.Conversely, deal concerns might be overlooked by trusting co-investors, leading to diminished returns.
Given this tension, we conduct an empirical analysis to determine if MBA alumni networks among PE co-investors influence LBO syndication decisions and investment performance.Our findings indicate that PE coinvestors with shared HBS alumni connections are more likely to engage in LBO syndication, and these co-investments generate superior performance.In contrast, co-investments involving other college alumni, or no alumni connections are less likely to be syndicated and realize significantly lower performance for PE co-investors.
As one might expect, tying 'alumni network' to LBO syndication comes with selection and endogeneity issues.We perform a battery of checks to alleviate these concerns, ensuring that identification arises from performance changes rather than an omitted variable at the coinvestor level.We show that differences in capital invested among PE co-investors are positively associated with LBO syndication in deals involving Crimson alumni.A question prompted by this result is whether these investments yield better results.We find that performance (value multiple and GIRR) is indeed higher when Harvard alumni are part of co-investor teams.These findings align with PE co-investors relying on mutual trust nurtured by a strong alumni connection like HBS, which facilitates the completion of LBOs with large asymmetric information (i.e., co-investments) that might be overlooked by other PE co-investors.Consequently, this subset of PE co-investors can reduce syndication agency costs and diversify their investments to achieve higher returns.
We also conduct several robustness tests and eliminate other recognized agency costs associated with PE investments in syndicated LBO using proxies derived from previous studies.These tests reaffirm that HBS alumni have a higher propensity to co-invest in syndicated LBO with each other, resulting in increased investment returns and valuation.
We acknowledge that our sample has limitations, and our analyses leave some questions unresolved.First, our conclusions are drawn from data sourced from a subset of financial institutions, which may not be representative of a wider population.Second, previous research on sources of PE outperformance categorizes them into three broad areas: operational improvements; governance; and financial engineering (Cumming, Siegel, & Wright, 2007; Gompers, Kaplan, & Mukharlyamov,  2016).Our data does not allow us to identify the category contributing to PE co-investment gains, nor does it allow us to determine whether information flow or trust is the primary driver of the relationship we observe.Third, we only offer preliminary evidence on the scalability of PE performance (Lopez-de-Silanes et al., 2015).Finally, while our results are arguably consistent with performance persistence in PE, this is not the central focus of our study (Braun et al., 2017; Buchner et al.,  2016).We recognize these limitations and leave more detailed analyses of these issues for future research.
such Harvard connections are indeed present at this level.From this observation, we infer that Crimson ties are prevalent not just among PE co-investors, but also between PE co-investors and MBO management.The Crimson alumni network appears to function at various levels to yield exceptional performance, although the evidence supporting this pathway is merely suggestive due to the limited number of firms involved. 41Table 3 in the Appendix presents a summary of statistics related to deal-specific, team-specific variables, and proxies for agency costs among co-investing groups with HBS alumni.When we compare this data with that in Figure 1, we observe a temporal trend between HBS-HBS pairings and the U.S. business cycle.This suggests that deal reciprocity could be a potential driving factor in HBS-HBS pairings.While our data doesn't allow us to conclusively capture deal reciprocity, we can't entirely dismiss this hypothesis.However, the incorporation of year fixed effects in our regression analyses should help alleviate this concern to some extent.This table presents means and medians on deal-and management team-specific characteristics for syndicated and non-syndicated investments in LBO.Deal value, capital invested, and fund size are in U.S. million dollars.Geographic distance is defined as the distance between the capital city of the portfolio company and the investment firm.Firm Experience is the difference between the founding year of the investment firm and the acquiring year of the portfolio company.Except for team size, investment team characteristics are proxied using density variables (i.e., number of investment professionals who have specific characteristics scaled by size of the investment team within a PE firm).Skill diversity is the standard deviation on three variables of skills, i.e., Law, Business, and Engineering.This table reports pair-wise correlation coefficients between the selected variables for LBO syndication likelihood.Syndication is a dummy variable which is assigned the value 1 for syndicated investments and 0 otherwise.Deal value, capital invested, and fund size are in US million dollars.Geographic distance is measured by the distance between the capital city of the portfolio company and that of the investment firm.Firm Experience is the difference between the founding year of the investment firm and the acquiring year of the portfolio company.All these variables are transformed by logarithm.Management buyout is a dummy variable which is assigned the value 1 if the investment team is involved with the portfolio company management and 0 otherwise.The investment team characteristics are proxied by using density variables, i.e., defined as the number of the professionals who have specific characteristics, scaled by the number of the whole investment team members within a firm.The skill diversity variable is the standard deviation of the three variables of skills, i.e., Law, Business, and Engineering.* P ≦0.05.

Fig. 1 .
Fig. 1.Illustration of the Relationship between Investment Type and Performance.
(v) Partner selection and deal flow: To address concerns about the selection of partners and deal flow, we choose repeated pairings in LBO co-investments (47 versus 103 investments) and check if the impact of Harvard alumni connections remains robust with unique pairings.The coefficients on GIRR and multiple in this setup, which uses the model in Column 12 of

Table 1
(continued ) criteria: (1) acquiring year from 1991; (2) buyout related; (3) already exited.PanelsA and B show size distribution in terms of deal value and capital input (in $U.S. million), respectively.Panel C shows sample time trend in calendar time.Panels D and E report geographic distribution and industrial orientation of portfolio companies on sample investments according to SIC code classifications in Chidambaran und Prabhala (2003).12Traditionaldatabaseslike Thomson Ventures Economics do not offer investment-level PE data for managers, and inconsistencies in fund sequences make comparisons challenging.13Thedataset we've put together is based on one compiled by Lopez-de-Silanes et al. (2015), and Degeorge et al. (2016).Our sample of private equity (PE) investments: (a) are comparable to PE deals found in Capital IQ and Thomson Reuters; (b) provide superior coverage outside the U.S. and prior to 2000; and (c) exhibit no significant survivorship or selection bias compared to commercially available PE databases.

Table 2
Summary statistics: management teams

Table 3
Determinants of LBO syndication Likelihood

Table 4
MBA selection of LBO syndication partners: networks and agency Costs

Table 5
Summary statistics: LBO performance

Table 6
Syndication and performance

Table 7
Summary statistics: Harvard MBA and LBO performance

Table 8
MBA Team attributes, syndication, and performance Table 9, are statistically significant at the 0.01 level.Thus, deal flow does not explain

Table 10
Sensitivity test of Harvard MBA selection of syndicated partners and performance This table shows how the impact of Harvard MBA match on investment performance of the subsample of investments co-invested by only two investors might be affected by potentially other known agency costs.The dependent variable is investment performance, proxied by multiple (Panel A) and gross internal rate of return (Panel B), both winsorized at the 5% level.The main explanatory variable, MBA(f), is a density variable, defined as the number of the professionals who have (specific) MBA degrees, scaled by the number of the whole investment team members within the firm.High leverage amount is a dummy variable equal to the value of one if debt value is more than the 75 percentile of this subsample (278 million dollars), and zero otherwise.High leverage ratio is a dummy variable equal to the value of one if debt over deal value ratio is at least 80%, and zero otherwise.High PE reputation (d) is a dummy variable equal to the value of one if the PE firm has conducted at least 15 deals in the whole sample, and zero otherwise.High PE reputation (count) is the log of the number of deals conducted by the PE firm in the whole sample.Large fund size is a dummy variable equal to the value of one if fund size is more than the 75 percentile of this subsample (1000 million dollars), and zero otherwise.Early exit is a dummy variable equal to the value of one if the difference between the acquiring year and the exiting year is no more than two.The control variables include size, proxied by the capital invested in US million dollars.Distance is measured by the distance between the capital city of the portfolio company and that of the investment firm.Experience is the difference between the founding year of the investment firm and the acquiring year of the portfolio company.All these control variables are transformed by logarithm.Each specification provides coefficient estimates for different subgroups of MBA graduates.Standard deviations are clustered at the industry level.T-values are reported in parentheses, and the symbols *, **, and *** represent statistical significance at the 0.1, 0.05, and 0.01 levels, respectively.thatcompletedatleastone deal annually from 1991 to 2004, and zero otherwise, while in Column 4 it represents the raw count of deals.The coefficients on both PE fund reputation proxies are not statistically significant, whereas the coefficient MBA(f)*MBA(i) is significant at the 0.05 level or lower.In Column 5 of Table10, we add a proxy for PE overfunding, in accordance with Metrick and Yasuda (2010) and

Table 2
Correlation Matrix of Selected Variables for LBO Syndication Likelihood