Fund Managers under Pressure: Rationale and Determinants of Secondary Buyouts

The fastest growing segment of private equity deals are secondary buyouts sales from one PE fund to another. On a comprehensive sample of 9,575 deals we investigate whether SBOs are valuemaximizing, or reflect opportunistic behavior. To proxy for adverse incentives, we develop buy and sell pressure indexes based on how close PE funds are to the end of their investment period or lifetime, their unused capital, reputation, deal activity, and fundraising frequency. We report that funds under pressure engage more in SBOs. Pressured buyers pay higher multiples, use less leverage and syndicate less suggesting that their motive is to spend equity. Pressured sellers exit at lower multiples and have shorter holding periods. When pressured counterparties meet, deal multiples depend on differential bargaining power. Moreover, funds that invested under pressure underperform. JEL Classification: G35; G32

The objective of this paper is to investigate the impact of PE fund incentives on investment decisions and exit outcomes. Using a combination of fund characteristics, we create a Buy Pressure index and a Sell Pressure index, to identify PE funds more likely to be prone to opportunistic behavior. We then analyze in a large buyout sample the investment and exit choices of funds under pressure, as well as transaction multiples, use of leverage, syndication and performance.
To identify which PE funds are susceptible to conflict of interest, we consider the typical contractual provisions in partnership agreements between GPs and LPs. The GPs are expected to invest during the first five years of the fund's life, called the investment period. The management fees are set to provide incentives to invest early, with GPs being paid a percentage of committed capital during the investment period, and a percentage of net invested capital during the subsequent period, the harvesting period. 3 However, for PE funds with substantial "dry powder" (unspent capital) close to the end of their investment period this provision creates adverse incentives to invest in deals they would have rejected at the start of the fund. This intuition has been formalized in the optimal contracting model of Axelson, Stromberg and Weisbach (2009).
PE sponsors aim to raise a new fund every 3-5 years, and their reputation and track record are critical to be able to do so (Kaplan and Schoar, 2005;Chung, Sensoy, Stern, and Weisbach, 2012).
The pressure of being evaluated during each fundraising cycle is part of the GPs' implicit incentive mechanism. As Chung, Sensoy, Stern, and Weisbach (2012) show, a major part of GPs' lifetime compensation is the expected income from subsequent funds. Prospective LPs not only look at past performance but also at the investment track record of the sponsor's recent funds. If the most recent fund still has a substantial amount of unspent capital near the end of its investment period, the LPs are unlikely to commit capital to a new fund. This puts further pressure on PE funds to invest their dry powder. GPs with substantial unspent capital late in their investment period have distorted incentives, and may be inclined to undertake less attractive investments in order to lock in management fees and boost their investment record. Funds with little reputational capital have more to gain from doing so, and hence have a potentially stronger incentive distortion.
For funds in their harvesting period, the closer the end of their lifetime (typically 10 years) or the more time passed since their last exit, the more exit pressure they face. PE funds with substantial non-exited investments will be tempted to sell quickly and even at relatively low prices, in order to improve their chances to raise a new fund. 4 If a PE fund cannot exit via a trade sale or IPO, it might strike a deal with another PE fund. PE firms with lesser reputation gain more from boosting their record. Funds may also strategically delay exits with modest proceeds in order to collect management 3 Net invested capital is calculated as the cost basis of all investments less the cost basis of realized investments. Metrick and Yasuda (2010) and Robinson and Sensoy (2013) document that the contract terms above are common across all PE funds. 4 On a sample of PE investments in the UK Wang (2012) documents that the seller fund is more likely to exit via a secondary transaction if the GP raises a new fund within two years following the exit. fees for longer periods as documented by Robinson and Sensoy (2013) and those that do so, would presumably find themselves more frequently under pressure when they sell. In either case one would expect to see lower multiples on exits by pressured sellers. When a pressured seller fund trades with a pressured buyer fund, the price is likely to be determined by the relative pressure of the parties.
For our identification strategy it is crucial that there is a dynamic incentive provision story at play over the fund's lifetime. For funds early in their investment period, the pressure from the PE contract is most likely positive and value-enhancing. However, for funds late in the investment period with substantial dry powder, the same contract may potentially create adverse incentives to window dress. As the two-period optimal contracting model of Axelson, Stromberg and Weisbach (2009) demonstrates, funds that invest early only invest in positive NPV projects and will continue to invest only in PNPV projects late in their investment period (in the second period of the model). In contrast, PE funds that have not found positive NPV investments early are willing to lower their investment threshold late in their investment period to keep management fees and improve their fundraising prospects. Axelson, Stromberg, and Weisbach (2009) predict that it is the combination of fund age and dry powder that makes a fund more likely to invest opportunistically, not one or the other alone.
To test these ideas, we extract from S&P's Capital IQ a comprehensive sample of all closed LBO transactions from 1980 to 2010 with targets located in the U.S. and in 12 European countries.
Our sample contains 9,575 LBO transactions involving 8,658 target firms and 957 different PE acquirers. We complement this data with an "event history" of known corporate events (bankruptcies, equity private placements, and mergers) for each LBO company using information from Capital IQ and on Initial Public Offerings (IPOs) from Thomson-Reuters' SDC database. This allows us to identify the type and date of exit of the initial fund for 4,139 exits, of which 1,219 are SBOs.
We construct our Buy Pressure and Sell Pressure indexes from characteristics that jointly identify funds most likely to face adverse incentives and/or conflicts of interests. Following Axelson, Stromberg and Weisbach (2009), we include in our Buy Pressure index proxies for the end of a fund's investment period, and for its unspent capital. We also utilize proxies for the lack of the fund's reputation and the infrequency of the family's fundraising, because opportunistic investment motives are more of a concern for funds without a stellar reputation and for those more uncertain about fundraising success (Gompers and Lerner, 1999;Kaplan and Schoar, 2005;Chung, Sensoy, Stern and Weisbach, 2012). Similarly, our Sell Pressure index captures how close the fund is to the end of its life, its recent exit activity, its lack of reputation and whether its sponsors are infrequent fundraisers.
Our tests are conducted against the alternative hypothesis that SBOs are more likely to be driven by investor value maximization, i.e. buy and sell pressure would not have any bearing on deal outcomes.
Our findings lend support to the view that agency conflicts help explain the behavior of funds under pressure. We document that secondary deals are more likely to involve buyers under pressure.
The regression coefficients indicate that a one standard deviation change in our Buy Pressure index increases the likelihood of SBO from 16.6% to 18%, or about 8.4% of the unconditional probability.
We do not find evidence that fund specialization affects the likelihood of secondary deals. Moreover, buy pressure significantly impacts deal valuations. PE funds under pressure pay more in secondary deals even after controlling for year and industry dummies, firm fixed effects and other fund and deal characteristics. A one standard-deviation shock in the Buy Pressure index increases the purchase multiple (relative to comparable M&A transactions) by about 10.8%. In primary deals, in contrast, buy pressure does not seem to matter: pressured buyers invest at similar multiples as others. These findings are consistent with the prediction of Axelson, Stromberg and Weisbach (2009) that PE funds with substantial dry powder late in their investment period are more likely to invest in less attractive projects.
We also document that exit via SBO is more likely if the seller is under pressure: an increase of one standard deviation in our Sell Pressure index increases the likelihood from 29.5% to 32.1%, an 8.8% increase. Moreover, seller funds under pressure sell at lower transaction multiples in secondary deals with one standard deviation shock in the Sell Pressure index decreasing the sales purchase multiple by about 13.7 percentage points.
The inverse relationship between sell pressure and exit multiples suggests that sell pressure reduces bargaining power and induces funds to accept lower prices. On the other hand, this finding is also consistent with Robinson's and Sensoy's (2013) evidence that PE funds tend to strategically delay exits of investments with modest proceeds and hence, those investments will be sold at lower multiples late in the funds' life. Given our evidence that SBOs are more likely chosen by aging funds, one may wonder whether this strategic delay drives the SBO multiples we find. However, an analysis of the components of our Sell Pressure index shows that besides the fund's age, the seller fund's degree of exit inactivity during the harvesting period and its lack of reputation also significantly negatively impacts the deal multiple.
Sell pressure may also impact the timing of SBO exit decisions. Using a competing hazard model of exit outcomes with a dynamic sell pressure index that varies over the holding period of the portfolio company based on the fund's age, exit and fundraising history, we find that sell pressure is associated with a higher likelihood of SBOs among all other exit routes and a lower expected conditional holding period for SBOs. Our evidence that pressured funds have shorter expected conditional holding periods for portfolio companies exited via SBOs is consistent with the view that funds under exit pressure are anxious to sell their portfolio companies even at less attractive prices.
Our finding on shorter expected conditional holding periods for SBO exits by pressured funds complements the evidence in Sensoy and Robinson (2013) and Wang (2012) on longer unconditional holding periods for portfolio companies exited by PE funds whose fee basis change from committed capital to net invested capital, and for investments exited via SBOs by all PE funds, respectively.
Interestingly, when pressured sellers meet pressured buyers in secondary deals, it is the relative pressure that determines the transaction multiple. In our sample of SBOs we document that less pressured buyers pay lower prices to more pressured sellers and more pressured buyers pay higher prices to less pressured sellers. Our evidence indicates that the difference in the Buy and Sell Pressure indexes affects the two parties' relative bargaining power and the acquisition price. When the two sides face the same pressure, then buy/sell pressure does not matter for pricing of the deal.
The Buy Pressure index also has a negative impact on deal leverage. We find that pressured funds use significantly less leverage (as a fraction of Enterprise Value, Ebitda and Total Assets) and more equity in SBOs. We also document that in SBOs where the lead buyer fund is under pressure to invest, syndicates are smaller with higher deal value per syndicate members. These findings, coupled with the fact that pressured buyers pay higher multiples support the agency story. If pressured buyers pay more for SBOs because (say) such deals are less risky, one should find that value-maximizing GPs would use more leverage in those deals, and not less. 5 Hence, our findings on deal leverage, syndication and transaction multiples seem to suggest that SBOs involving buyers under pressure are driven more by the buyer funds' desire to spend equity than to maximize LP returns.
When GPs use less leverage and pay higher multiples for acquisitions, LPs are likely to get lower returns on their investment. Using IRR data from Preqin over a subsample of completed funds, we document that when GPs made more secondary investments under buy pressure during their investment period, or when GPs more frequently exit via SBO under sell pressure during their harvesting period, their funds yield lower IRR. This evidence suggests that the higher (lower) acquisition (exit) multiples at which PE funds with high Buy Pressure (Sell Pressure) index tend to invest (exit) are associated with less attractive fund performance. While this explanation seems to be the most likely interpretation of our findings, we cannot completely rule out the remote possibility that the SBOs are the better-performing parts of these funds and the underperformance is due to non-SBO investments/exits of the same funds.
Our paper documents agency costs in private equity. Our findings indicate that the agency problems between GPs and LPs are not completely alleviated by the partnership agreement. However, this evidence is by no means inconsistent with optimal contracting. In the presence of effort provision, agency and information problems, optimal contracts between principals and agents are only secondbest. They do not perfectly align the incentives of the principals and agents and result in agency costs and residual losses. 6 The contribution of our paper is to provide new evidence on the nature and 5 In similar spirit, Officer, Ozbas and Sensoy (2010) argue that the pattern they document in PE club deals that transactions with lower multiples obtain more favorable debt financing terms is difficult to reconcile with an information/risk story. 6 See Jensen and Meckling (1976) and Robinson and Sensoy (2013) for more in-depth discussions of agency costs in the context of optimal financial contracting. significance of these agency costs and to identify fund characteristics and transaction types which are associated with economically significant agency costs in private equity investments.
The evidence in our paper raises the question whether the same PE firms repeatedly transact with one another to relieve each other's pressure. 7 Consistent with Wang (2012) we do not find evidence of collusion among PE funds. Wang reports that 6 out of the 247 SBO deals (2.4%) in her sample are "two-way deals", in our sample 83 out of the 1,593 (5.2%) SBOs are "two-way deals". In unreported results we find that (i) the likelihood of a two-way deal is correlated with deal size (not surprisingly because large firms tend to participate in large deals); (ii) a dummy representing two-way deals is uncorrelated with pressure measures and is never significant in any of our regressions. While the lack of evidence of quid pro quo does not imply that such transactions do not occur, based on our findings it is unlikely that this type of cross-subsidization among PE firms is prevalent.
Like in most studies on financial intermediaries, endogeneity is an important concern as the econometric relationships might be driven by unobserved hidden variables. Hence, we designed our analysis to address these concerns in several different ways. Our Buy and Sell Pressure indexes are time-varying and in our data the same GP is typically present in deals with and without pressure. This reduces the impact of unobservable fund-level characteristics. Throughout the analysis we include PE firm fixed effects to pick up any endogeneity that arises from underlying unobservable firm-level characteristics. In addition, we exploit the fact that each SBO has two PE equity funds as counterparties, both of which could potentially be under pressure. We isolate the impact of the Pressure index of the deal counterparty as a reasonably exogenous variation that should be orthogonal to the characteristics of the PE firm and the portfolio company further reducing the concern for endogeneity. Nevertheless, it is not possible to conclusively prove any causal effect of buy/sell pressure in SBOs on transaction multiples and fund performance. While there maybe unobservable factors that both drive the willingness to invest/exit in SBOs and can explain the transaction multiples, the timing of exit, and the patterns of deal leverage, deal syndication and fund performance, it seems difficult to explain our results with simple unobservable aspects of deal risk or quality.
The remainder of the paper is organized as follows. We discuss the literature in Section 1.B and develop our hypotheses in Section 2. Section 3 describes the data and variables. Section 4 and 5 analyze the impact of the Buy and Sell Pressure indexes on the likelihood, valuation, and the holding period of SBOs. Section 6 explores how differences in pressure between buyer and seller funds affect their relative bargaining position. Section 7 investigates the impact of buy pressure on deal leverage, syndication, and fund performance. Section 8 discusses endogeneity issues. Section 9 concludes.

B. Related Literature
A considerable amount of research has been devoted to private equity on a wide variety of questions, including productivity, growth, employment, financial distress and performance of PE as an asset class. 8 Our paper is most closely related to studies on the efficiency of PE investments (e.g. Axelson, Jenkinson, Stromberg and Weisbach, 2012). Since our pressure indexes condition on fund age, our paper also speaks to the effects of limited investment horizon in venture capital and private equity (Kandel, Leshchinskii and Yuklea, 2011;Barrot, 2012). Moreover, our analysis adds new evidence to recent theoretical work on window-dressing by VC/PE funds (e.g. Cornelli and Yosha, 2003;Axelson, Stromberg and Weisbach, 2009). Like Officer, Ozbas and Sensoy (2010), our paper focuses on PE fund's incentives and behavior. Officer, Ozbas and Sensoy (2010) reports that when multiple PE funds jointly conduct LBOs, reduced competition for targets drives prices down relative to sole-sponsored LBOs. Their finding that club deals with lower multiples obtain more favorable debt financing terms is difficult to reconcile with an information/risk story and provides support for strategic motives in club formation.
Similarly, our paper documents strategic motives of pressured buyers/sellers in SBOs and reports that pressured buyer funds pay higher multiples and use less leverage and smaller syndicates in SBOs suggesting that these funds are driven more by the desire to spend equity than to maximize LP returns.
In related papers, Jenkinson, Sousa and Stucke (2013) and Brown, Gredil and Kaplan (2013) document manipulation of NAVs by PE funds during their fundraising cycle. In contrast to the expost manipulation of prior investments, our paper focuses on the GPs' opportunism at the time of making investments. Similar to Brown, Gredil and Kaplan (2013), we also find that top-performing funds are less likely to be driven by adverse incentives than less reputable funds.
Our paper also contributes to a small but growing literature on SBOs. This literature focuses on SBO motives, pricing, operating performance and returns. Wang (2012), Jenkinson and Sousa (2012), and Achleitner et al. (2012) find that SBOs are mainly driven by favorable debt and IPO market conditions. On a hand-collected sample of UK firms, Wang (2012) reports that SBOs are priced higher but there is no pattern of collusion in SBOs among PE funds. While our paper confirms that the debt/IPO market affects deal choices and exit outcomes, it further shows that buy/sell pressure create strategic motives for PE funds to choose SBOs and impacts both prices and deal terms. Bonini (2010), Jenkinson and Sousa (2011), and Wang (2012 present evidence of lower operating performance improvements in SBOs compared to primary deals. Looking at internal rates of return from a proprietary dataset, Achleitner and Figge (2012) and Achleitner et al. (2012) find no 8 See among many others, Kaplan (1989aKaplan ( , 1989b, Smith (1990), Stromberg (2008, Acharya, Gottschalg, Hahn and Kehoe (2013), Gottschalg (2009), Axelson, Jenkinson, Stromberg andWeisbach (2013), Demiroglu and James (2010), Officer, Ozbas, and Sensoy (2010), Boucly, Sraer and Thesmar (2011), Guo, Hotchkiss and Song (2011), and Franzoni, Nowak, and Phalippou (2012 difference in performance between primary and secondary deals. In a contemporaneous paper, Degeorge, Martin and Phalippou (2013) report that funds underperform in SBO investments at the end of their investment period relative to their other investments and their SBOs undertaken earlier. This finding complements our evidence that funds yield lower IRR when GPs are involved in more SBOs under pressure, and is consistent with our deal-level findings on the growth rate of enterprise value.
Unlike their paper, we develop Buy and Sell Pressure indexes from theoretical insights from the PE contracting and compensation literature literature 9 and document incentive distortions in the context of these metrics. We also show that these indexes have a significant effect on the likelihood, pricing and deal terms of SBOs, both on the buy and the sell side.
Finally, this paper connects to empirical work on the impact of PE fund compensation on fund behavior (Gompers and Lerner, 1999;Metrick and Yasuda, 2010;Chung, Sensoy, Stern, and Weisbach, 2012). Robinson and Sensoy (2013) find that, although PE fund fees correlate positively with performance, GP behavior in booms and around certain contractual triggers seems consistent with the existence of agency conflicts. The authors conclude that the agency costs embedded in the GP-LP are bounded away from zero and will manifest themselves in certain circumstances. The findings reported in this paper essentially corroborate this view for funds under pressure.

The PE contract and fund incentives
It is useful to begin with a brief description of how PE funds are organized. The management company, or the GP, sets up the fund and makes all investment (buy) and exit (sell) decisions. Their LPs commit to transfer capital to the fund whenever the GP finds an investment opportunity. The LPs play no active role in the fund's management, and have no specific information upfront of the investments the GP will make (i.e., PE funds are "blind-pool" vehicles). The partnership has a lifetime of 10 years (extendable by one or two more), divided into two distinct periods. During the first 5 years, the investment period, the GP selects investments; during the remaining years, the management or harvesting period, the GP manages and eventually exits from those investments. 10 GPs are compensated by a fixed management fee, typically 1.5% to 2% of committed or net invested capital, and a variable component known as carried interest (or carry), corresponding in most cases to 20% of the fund's profits (often a pre-specified hurdle rate must be reached before the GP can 9 Our Buy and Sell Pressure indexes build on theoretical insights from Axelson, Stromberg andWeisbach (2009), Metrick andYasuda (2010), Kandel, Leshchinskii and Yuklea (2011), Barrot, 2012 andChung, Sensoy, Stern andWeisbach (2012). 10 So called follow-on investments, usually acquisitions made by companies in the fund's portfolio, are allowed during the harvesting period, but typically limited to at most 10% to 15% of the fund and often require LP authorization. receive carry). This convex claim held by the GP aligns the incentives of both parties. Incentives are critically important because LPs are locked in for ten years once the fund closes and have no say in investment and exit decisions. Apart from eventual distributions from exits, the LPs' stake is illiquid: selling it is extremely costly and usually subject to GP approval. The only 'stick' that the LPs possess is the threat not to invest in subsequent funds by the same GPs. This is a strong threat, since the GPs will be out of a job by the end of the current fund's life if future fundraising is unsuccessful.
Based on these contractual features, we develop the following hypotheses regarding the GPs' incentives and investment behavior: (i) Fund incentives and fund activity. The GP-LP relationship can be viewed as a principalagent problem in which an uninformed principal (the LPs) hires a potentially skilled agent (the GP) to trade on his behalf. 11 Investors learn about the GP's ability by observing his past and current deal activity. The agent might therefore engage in suboptimal actions in an attempt to influence the principal's beliefs. In Dow and Gorton (1997), for example, the agent trades too much in order to show activity to his employer. 12 In the PE context, this leads to the prediction that GPs with substantial amounts of unspent capital ("dry powder") will more likely engage in suboptimal acquisitions, to create investment record and use up capital. This theory also implies that funds in the harvesting period which have not shown exit activity for some time will be tempted to engage in suboptimal sales transactions to improve their exit record.
One might argue that the incentives implicit in the partnership agreement should be sufficient to alleviate the agency problem. However, as Axelson, Stromberg and Weisbach's (2009) show, these incentives may not work equally well for all funds. The contract, which provides correct incentives to most GPs, exacerbates distortions for GPs unable to invest early, and such GPs will be willing to spend their capital on negative NPV projects late in their investment period.
(ii) Fund incentives and stage of fund lifecycle. The management fee structure may create adverse incentives to overinvest. Metrick and Yasuda (2010) report that for 84% of buyout funds, the management fee is paid as a percentage of committed capital during the investment period but as a percentage of net invested capital during the harvesting period. They simulate data based on observed PE contracts and find that the fixed compensation represents a large portion (roughly 60%) of the NPV of the GP's income. Therefore, GPs close to the end of the investment period face a tradeoff. If a potential target is somewhat overvalued and the GP invests, the fund's IRR is likely to suffer. On the other hand, by passing up the deal, the GP loses the management fee as their basis shifts from committed capital to net invested capital. Based on Metrick and Yasuda's findings, one would expect that the temptation to overinvest is likely to be particularly strong for funds close to the end of their investment period. 13 For funds in the harvesting period, sell pressure accumulates as the end of the fund's life approaches. Although sometimes investors allow so-called 'zombie' funds to continue, the GP's reputation suffers when one of its funds operates beyond its expected lifetime.
(iii) Fund incentives and fund reputation. The temptation to engage in suboptimal deals discussed above is likely to be more severe for managers that still need to build their reputation. First, the incentive to gamble is highest for funds with little reputational capital (Ljungqvist, Richardson, and Wolfenzon, 2008;Gompers, 1996). Second, high-reputation GPs presumably also have higher skills (e.g. Kaplan and Schoar, 2005), and are able to spot good investment opportunities early in the fund's life. Third, LPs' beliefs about GP ability are less (more) likely to be affected by a single bad deal or a temporary bout of inactivity if the GP's reputation is strong (weak). We therefore predict that funds with less reputation are more likely to engage in suboptimal transactions, both as buyers or sellers.
(iv) Fund incentives and fund raising frequency. Chung, Sensoy, Stern and Weisbach (2012) argue that GPs have strong incentives to maximize the prospects of successful future fundraising, as the flow of fees associated with the future assets under management can be a substantial part of the GPs wealth. The fund raising process for the typical fund lasts between 12 to 24 months, and requires a substantial GP effort in terms of time, monetary outlays, and management attention. Arguably, the expected likelihood of success in raising a future fund is a function of the GP's experience in the fund-raising process. GPs that have completed the process more frequently are more likely to have a large pool of potential investors, a more professional approach to fund raising, and access to intermediaries such as placement agents. Conversely, GPs with infrequent fundraising experience are more likely to be dependent on their track record to woe investors and therefore more tempted to window dress their current performance. This leads to the prediction that infrequent fund raisers are more likely to engage in suboptimal deals, both as buyers and as sellers.

Fund incentives and SBOs
One important feature of SBOs relative to primary deals is that they are easier and quicker to execute. First, a secondary buyer saves on search costs, because the target has already been prescreened by the primary investor. Second, SBOs are faster to complete relative to divisional deals (in 13 In line with Metrick and Yasuda (2010), Robinson and Sensoy (2013) also finds that a substantial fraction of PE funds in their data switch from committed capital to net investment capital as the basis of management fee at the end of the investment period. They document that such funds are more likely to strategically delay exiting their less attractive investments to collect management fees for a longer period. which the decision by the corporate parent might be more convoluted or involve intermediaries), delistings of public firms (in which the buyer has to comply with regulations and possible hold-out by minority shareholders), or sales of stand-alone private firms (often family firms in which the emotional attachment or conflicts within the family can delay the sale). Third, SBOs are probably easier to finance since a substantial amount of information is available from the primary deal (debt documentation, due diligence, financial reporting systems). Furthermore, lenders are likely to be familiar with the target and the same banks may be willing to continue to fund it after the secondary transaction. Exactly the same arguments hold for sellers, making SBOs probably the quickest way to exit a portfolio investment.
In general, SBOs could be done for opportunistic motives or for efficiency reasons.
On one hand, secondary buyouts are attractive for funds that wish to conclude a deal quickly and therefore a prime choice for GPs with misaligned incentives. We hypothesize that GPs are more likely to engage in secondary deals for agency reasons: (i.a) as buyers if they have more dry powder, and (i.b) as sellers if they haven't shown recent exit activity; (ii.a) as buyers if the end of their investment period is close, and (ii.b) as sellers if the end of the harvesting period is close; (iii) if they have not established a strong reputation of a "top quartile" fund for which performance persistence is frequently assumed; (iv) if the GP is an irregular or infrequent fundraiser. In terms of deal pricing we would expect buyers (sellers) doing acquisitions because of distorted incentives to be willing to pay higher (receive lower) transaction multiples. We further predict that, in case two pressured GPs are counterparties in an SBO, their relative opportunism determines the excess multiple of the deal.
On the other hand, SBOs could be efficient transactions for GPs who maximize investor value. First, PE funds might have different skills that are adapted to different target types. For example, organic strategies based on professionalization of business practices are frequent in smaller primary targets, while M&A-driven internationalization strategies are more important for restructuring larger secondary targets (e.g. Acharya et al., 2013). Similarly, PE funds might specialize in certain industries. This leads to the prediction that fund specialization would impact the likelihood of executing an SBO. Second, more reputable funds may have access to more leverage at cheaper rates, so investing in secondary deals would create additional value for these funds. This view would predict that more reputable funds are more likely to invest in SBOs, the opposite to the prediction implied by the agency view. In terms of deal pricing, we would predict that investor-valuemaximizing GPs would pay (receive) no excess multiples in secondary deals relative to primary deals.

Data and empirical testing issues
This section describes in general terms our sample and variables. Appendix A provides the details about the sample construction and Appendix B the full list of variables and their definitions.

Sample construction
We extract from S&P's Capital IQ database all closed LBO transactions with targets located in the U.S. and in 12 European countries (Belgium, Denmark, Finland, France, Germany, Italy, Luxembourg, Netherlands, Spain, Sweden, Switzerland, and U.K.) for the period ranging from January 1st, 1980 to December 31st, 2010. As a first set of filters we exclude targets in financial industries, acquisitions of minority stakes or of remaining interest, deals involving targets with reported negative sales or negative enterprise value, and misclassified non-PE related transactions such as corporate acquisitions, purchases of stakes by hedge funds, and venture capital deals. On this initial sample of 23,032 deals we implement Stromberg's (2008) methodology to obtain an imputed Enterprise Value for transactions without deal value information (roughly 60% of the sample). This involves running a Heckman regression model with the likelihood of a deal having its value disclosed in the first stage, and the determinants of target Enterprise Value in the second stage (see Table A-1 in the Appendix for results and details). This imputed value is used to compute market shares and activity measures of PE fund families required in the analysis.
We then apply a second set of filters excluding: deals without Capital IQ identifiers of buyers and sellers; acquisitions by management teams (management buy-outs) with no evidence of involvement by a PE sponsor; deals in which the target firm is bankrupt or in financial distress; and transactions with a deal value lower than one million dollars. 14 When an acquisition involves multiple stages or transactions, we keep the one in which the buyer acquired most of its stake (typically the first transaction). We also require that we can reasonably trace the purchase to a given fund within a PE fund family under mild assumptions (see below). The final sample contains 9,575 LBO deals involving 8,658 target firms and 957 different PE acquirers.
To obtain the exit of each LBO transaction, we download from Capital IQ data on corporate events related to each target firm (bankruptcies, equity private placements, and mergers) using each firm's unique identifier. We complement this data with information on Initial Public Offerings (IPOs) from ThomsonReuters' Securities Data Corporation (SDC) database. We then construct an "event history" of known corporate actions for each firm after the LBO, allowing us to identify the type and date of exit of the initial LBO investor. We say that an exit takes place if there is evidence of a change 14 All monetary amounts in this paper are in real December 2010 dollars, values in European currency having been converted to U.S. dollars at historical exchange rates. in control (e.g., sale of a majority stake) even if the original buyer funds remain minority shareholders. The final sample contains 4,139 exits, of which 1,219 are secondary LBOs.

Buy Pressure index and related variables
We extract from Capital IQ buyer and seller information that we use to create a unique PE fund family identifier to group fund-level information (again see the Appendix A for details). We identify the "leading buyer" in a multi-buyer transaction as the PE fund family with the highest reputation among the deal's buyers, measured as the dollar market share across all LBO deals made up to that year. 15 For single-buyer transactions, the "leading buyer" (henceforth, the buyer) is the PE fund family of the acquiring fund. We then match each LBO with information on the buyer's existing funds. We check if the acquisition date is within the investment period (e.g., years 1 through 6) of at least one fund in the PE fund family. If it is not the case, we discard the LBO transaction because this is a sign that fund-level information in Capital IQ about the family is incomplete. If the condition is fulfilled, we assume that the deal is executed by the fund family's youngest fund still investing at the time of the deal. The value of the variable Stage equals the number of years between the start of the buyer's fund and the date of the LBO deal.
To capture buy pressure, we create an index from characteristics that are likely to identify funds desperate to invest. The first element of the index is the variable Late Buyer, an indicator variable equal to 1 if the buyer's fund is at the end of its investment period (that is, 4 to 6 years after inception) at the time of the deal, and zero otherwise. The second element is the variable Dry Powder.
For each PE fund family and year, we calculate: (i) the aggregate amount that was raised in the past 3 years and the corresponding median that was raised across fund families in that year; (2) the aggregate dollar value of all investments made during the past three years, and its corresponding median. Using these two quantities, we define Dry Powder, a dummy variable equal to 1 if the buyer's PE fund family is above median in terms of fund raising and below median in terms of deal activity. Third, we define Lack of Reputation, an indicator variable equal to 1 if the buyer is not in the top quartile of funds in terms of deal volume market share. Fourth, we define Infrequent Fundraiser, an indicator variable equal to 1 if the average fundraising frequency of the PE firm until the year of the deal is in the bottom quartile of all PE firms.
The index variable Buy Pressure is the sum of the four dummy variables Dry Powder, Late Buyer Lack of Reputation, and Infrequent Fundraiser. To identify buyers under particularly acute pressure, we define High Buy Pressure, an indicator variable equal to 1 if at least two of the four 15 We also compute market shares of PE fund families using three other backward-looking horizons, 3, 5, and 10 years. In the overwhelming majority of cases the ranking of fund families, and thus the leading buyer, is the same. In very small number of cases in which the different horizons produce different results, we take the buyer with the highest average among the all horizons. components of our Buy Pressure index are equal to 1. The indicator variable Low Buy Pressure takes the value 1 if the reverse holds.
Finally, to control for fund family characteristics, we introduce two variables: Affiliated, a dummy variable equal to 1 if the fund family is affiliated to a financial institution or government agency, and zero otherwise; and Novice, an indicator variable equal to 1 if the buyer is from a PE fund family with 3 funds or less under management at the time of the LBO deal, and zero otherwise.

Sell Pressure index and related variables
As mentioned before, we extract from Capital IQ buyer and seller information that we use to create a unique PE fund family identifier to group fund-level information. The "leading seller" at exit is the PE fund family initially identified as leading buyer, and for consistency we assume that the seller (within the family) is the same fund that made the investment at the time of the initial transaction. 16 To capture sell pressure, we create an index from characteristics likely to identify funds desperate to exit. Our Sell Pressure index is comprised of four binary variables: Late Exit is an indicator variable equal to 1 if three years or more elapsed since the PE fund family last exited an 16 In a few cases LBOs are marked as secondary deals in Capital IQ but no information related to the primary deal exists. In such cases we replicate the process described above for the buyer, that is, we compute market shares among sellers in a deal to select the leading seller, and require that the sale take place during the lifecycle (i.e. years 1 through 10) of at least one fund in the selling family to compute seller-related variables. The selling fund is then defined as the oldest active fund (i.e. less than 11 years old) in the selling fund family.

Other variables
The set of controls in our regression specifications includes several variables. Imputed TEV is the target's enterprise value, that is, the sum of equity market value (valued at the offer price) and the target's pre-deal net debt (financial debt minus cash and marketable securities). Enterprise value, like all monetary amounts in this paper, is measured in real December 2010 U.S. dollars after conversion at historical exchange rates (exchange rates and inflation rates are obtained from the FRED Economic data of the Federal Reserve Bank of St. Louis). Management Participation, U.S. dummy, and Syndicated are dummy variables that indicate, respectively, that management is a shareholder of the acquiring group, the target is a U.S. firm, and there is more than one buyer. To proxy for capital market conditions we include HY Spread, the difference between interest rates on leveraged loans and For tests involving exit, we construct similar variables but as of the time of exit (Exit HY Spread, Exit Cold IPO Market, and Exit Excess Sales (Ebitda) Multiple). One additional variable specific to exit regressions is Add-ons, a dummy variable equal to one if there were significant acquisitions during the time that the buyer held the target firm in its portfolio. valuation measures are only available if the deal value is known and the accounting item entering the multiple is also available). The average Sales Multiple (Ebitda Multiple) for LBOs in our sample is 1.36 (9.37), while their Excess equivalent, net of median transaction multiples, is 0.25 (-1.41).

Summary statistics
Finally, Panels C and D of Table 1 present statistics for exits. Note that 29.5% of the LBOs in our sample are exited through a secondary deal, the second most frequent form of exit after trade sales (48%). About 11% of exited deals involved large add-on acquisitions. About 10% of the sellers last exited a deal three or more years ago, 12.5% sold at the end of the fund's lifetime, 76.7% of exits involved sellers lacking reputation, and 20.8% were infrequent fund raisers. Our Sell Pressure index shows a mean (median) value of 1.20 (1), and 29.9% of the sellers exited under High Sell Pressure.
Turning to valuations, the average Sales Multiple at exit is 1.82 and the average Ebitda Multiple is 10.82, somewhat higher than the corresponding entry valuations. The same holds for the excess sales multiple (excess Ebitda multiple) that reach 0.73 (0.81).

Secondary buyouts: univariate comparisons
To highlight the systematic differences between primary and secondary LBOs, Table 2 displays univariate comparisons of means and medians of our variables between these deals. Panel A shows that relative to non-secondary LBOs secondaries are larger both in mean and in median and management participates more often as equity holder (all differences statistically significant at 1%).
Secondary buyouts are more likely when credit spreads are relatively lower, and when IPO markets are cold (all differences again statistically significant at 1%). Table 2 also presents a preview of the results for our main variables of interest.

Panel A of
We report that secondary LBOs more often involve (1) buyers with dry powder (26.2% of the time versus 22.3% for non-secondaries, statistically significant at 1%), (2) buyers late in their investment cycle (23.9% versus 21.0%, statistically significant at 1%), (3) infrequent fund raisers (27.1% versus 24.8%, statistically significant at 5%). By contrast, there is no perceptible difference in the lack of reputation (69.8% versus 69.5%, which is not statistically significant). The Buy Pressure index is therefore higher for secondaries (1.470 versus 1.375, statistically significant at 1%). The table also shows that affiliated buyers do secondaries relatively more often, and novice funds less often than other deals. This shows the importance of controlling for these fund characteristics in our analysis.
Panel B reports the differences in valuation between the two types of deals. The table shows that secondary deals are more expensive than other LBOs across all valuation measures employed.
For example, the average secondary transaction in our sample was priced at a sales (Ebitda) multiple of 1.589 (10.184), about 22.9% (10.8%) higher than for other deals. The results for the benchmarked excess multiples are similar (all differences statistically significant at the 5% or 1% level).

Secondary buyouts: multivariate analysis
We run a multivariate logit regression to test the hypothesis that PE fund characteristics proxy for investment incentives and predict secondary transactions: The dependent variable, y i , is an indicator that takes value 1 if deal i is a secondary buyout and 0 if it is a primary deal. X represents the matrix of control variables defined in Section 3.4. All regression specifications include industry and year dummies, and we cluster standard errors by deal year. 17 Column 1 of Panel A in Table 3 presents the results from estimating the logistic model for our basic specification. The coefficient of Buy Pressure is positive and highly significant (t-statistic 3.04).
This suggests that pressured buyers do proportionally more secondary deals. The marginal effect of Buy Pressure is 0.014, also significant at 1%, implying that an increase of one standard deviation in Buy Pressure represents an increase in the probability of doing a secondary exit of about 0.014× 0.98 = 1.4%. This is an 8.4% increase relative to the unconditional mean of the likelihood of a secondary buyout (equal to 16.6% from Table 1). Column 2 of Panel A presents results for High Buy Pressure which has a virtually identical level of significance (t-statistic 3.31). Column 3 decomposes the Buy Pressure index into three dummy variables, representing the index being equal to 1, 2, or at least 3.
The coefficients are monotonically increasing as expected, and all three dummies are significant at 1%, 5% or 10%, respectively. Column 4 introduces PE firm fixed effects to account for any unobserved GP heterogeneity that is not controlled for in our specification. The coefficient of Buy Pressure is significant at the 5% level, and slightly larger than the coefficient of column 1. 18 Regarding control variables, the table shows that targets of secondary LBOs are larger than primary deals (with t-statistics between 10 and 20) and have management equity participation more often (t-statistics ranging between 4 and 6). SBOs seem to be somewhat less prevalent in the U.S. than in Europe (t-statistics around -2.5). The results are virtually unchanged when we run the regression on the subsample of LBOs with valuation information (not reported in tables). Table 3 presents results for the individual components of our Buy Pressure index as the main independent variables. All coefficients show a positive loading, with statistically significant t-statistics ranging from 1.7 (Late Buyer), significant at 10%, to 2 (Infrequent Fundraiser), 2.06 (Dry Powder) and 2.3 (Lack of Reputation), the latter three significant at the 5% level.

Likelihood of secondary deals and buyer specialization
So far our findings suggest that buy pressure plays a role in funds' investment decisions but at least two alternative stories can explain why some funds might prefer to engage in secondary deals. To test this hypothesis, we add two proxies for buyer specialization to our model. The first proxy is a set of three indicator variables related to Size Specialization that take the value 1 if the buyer's past deals are focused (more than two thirds) in a particular LBO size category. 19 The second proxy is Industry Specialization, an indicator variable equal to 1 if a significant percentage of the buyer's past deals (33%) are in the same industry as the target. Table 4 presents the estimation of the logit model with the specialization variables in columns 1 through 3. The specification includes the same set of control variables as well as industry and year dummies. The results do not support the specialization hypothesis. All the coefficients of the specialization variables are statistically insignificant with signs opposite to what one would expect.
The first two columns show that funds specialized in an industry are not more likely to make a secondary deal, whether measured by a dummy (Column 1) or the percentage of their deals (Column 2) in their specialist industry, and that funds specialized in large deals are less, rather than more, likely to purchase a secondary target (but the coefficients are not statistically significant). Notably, the coefficient of the Buy Pressure index is positive and statistically significant throughout. Column 3 shows that the coefficients on size specialization are also insignificant.
The second alternative story suggests that the likelihood of secondary deals is determined by industry-or market-specific conditions. Although we control for industry and time effects, some within-industry time varying factors could potentially affect our results. We focus on two possible explanations. The first possibility is that anti-trust concerns might affect the frequency of secondary deals. Secondary LBOs are relatively larger firms, which might face anti-trust hurdles if bought by a trade buyer, while simultaneously not being large enough to be sold through an IPO. Hence in the presence of substantial industry concentration, the only exit route for a seller would be to exit through a secondary LBO. We therefore use as a control Industry Concentration, the geography-and yearadjusted Herfindahl concentration index in the target's industry. The second possibility is that changes in industry capital asset liquidity might explain the likelihood of secondary deals, by changing the pool of available trade buyers over time. We therefore use as control the Asset Liquidity measure proposed by Schlingemann, Stulz and Walkling (2002)

Valuation of secondary buyouts
Our empirical analysis documented that the Buy Pressure index predicts the likelihood of SBOs. Next we focus on deal pricing. To gauge valuation effects, we run the following least-squares regression in the sample of deals with valuation information: where XSMULT, the dependent variable, is one of our two measures of deal valuation (Excess Sales Multiple or Excess Ebitda Multiple) and D SEC is an indicator variable that takes the value 1 if the deal is a secondary transaction and 0 otherwise. We focus our attention on β 2 , the coefficient on the interaction term Buy Pressure × D SEC that measures the impact of our Buy Pressure index for secondary deals. As before, X represents the matrix of control variables, but with one exception: we replace TEV (which is part of our Sales Multiple definition) with an instrument for deal size, represented by Buyer Size, the log value of all deals made by the buyer in the last 5 years. All specifications include industry and year dummies and year-clustered standard errors.
Panel A of Table 5  Purchase prices are positively associated with our proxy for deal size and negatively related to management equity participation, although statistical significance varies.

Determinants of the secondary buyout exit route
We run a multivariate logistic model to test the hypothesis that PE fund characteristics proxy for investment incentives and predict secondary exits: The dependent variable, y 3 is an indicator variable that takes value 1 if deal i is exited via SBO and 0 for other types of exits. We modify the control variable matrix X by adding as controls: Stage, the time in the fund's life when the target was originally bought; Add-On, an indicator of significant build-ups under the seller's control; and dummy variables identifying the original deal's source, that is, whether it was a secondary deal, a divisional buyout, a public-to-private deal, or a sale by a financial institution (the missing category is that of a private-to-private deal). In addition, some of our variables, like Novice, HY Spread, and Cold IPO Market, are now calculated as of the date of exit.
The results are presented in Panel A of Infrequent Fundraiser is insignificant with a positive sign. The control variables exhibit virtually unchanged coefficients and t-statistics.

Holding period of secondary buyouts
Strategic considerations of seller funds are likely to influence several aspects of the exit: the route (sale in SBO versus an alternative), the sale price and the timing. In this subsection we investigate the timing of the exit decision and its interaction with the Sell Pressure index. Exit timing can be synonymously understood as the duration, or holding period, of each portfolio firm.
Let τ represent the time elapsed in years since the PE fund's purchase of portfolio company i, and k ∈ {SBO, OTHER} denote the available exit routes. 20 The hazard rate h k (τ) is the probability that an exit of type k occurs at time τ, conditional on the fact that no exit occurred before τ. Our aim is to 20 The shorthand notation OTHER refers to non-SBO exits such as trade sales, IPOs, etc.
understand how our explanatory variables impact the hazard h k (τ), but the latter is an unobserved latent variable that must be estimated using observed deal durations. We therefore create a panel of deal-year observations with the duration τ of each deal up to the current year. Deals remain in the dataset until an observed exit k* occurs (as in the sample used in the previous subsection), or until the end of the sample period (for unexited deals by the time we stop collecting data). The latter correspond to right-censored observations, but estimating the model including those observations is important because they contain information about implicit choices of exit timing. 21 Our estimation procedure employs a competing risks proportional-hazard duration model (Fine and Gray, 1999), which models the behavior of the hazard rate for our event of interest (an exit via SBO) in the presence of other possible 'competing' events such as trade sales, IPOs, etc. The following model is fitted using maximum likelihood: Several remarks about the model are in order. First, the set of explanatory variables includes both static controls (e.g., deal characteristics at the time of entry) and time-varying variables (e.g., the pressure indexes and variables related to market conditions). In particular, the Sell Pressure index changes over the life of the deal, making the model truly dynamic. As a portfolio company stays longer in the portfolio of a PE fund, the Sell Pressure index evolves according to the fund's history, dropping if the fund exits another portfolio company or manages to raise a new fund, and rising otherwise. Second, the quantity h 0 SBO (τ), which is not estimated, denotes the baseline hazard when all explanatory variables are set to zero. The interpretation of the model is thus made in terms of hazard ratios: a positive coefficient indicates that a one-unit change in a given variable increases the hazard rate relative to its baseline level, therefore making exit more likely (i.e., making the holding period shorter). Conversely, a negative coefficient indicates that a one-unit change in the explanatory variable decreases the hazard ratio and makes exit less likely (i.e., makes the holding period longer). Third, the model incorporates the presence of multiple exit choices, making it particularly adapted to our setting. 22 21 Censoring is a common feature of duration analyses. Estimation is straightforward under the assumption that the censoring event is independent of the exit events (as it is the case here). 22 The competing risks model is most adapted to our setting for two reasons. First, standard duration analyses such as Kaplan-Meier assume that observations exited through competing events are censored observations. But this is incorrect, because if these observations were really censored, the event of interest k could presumably still occur in the future (it is just not observed). In reality the event of interest will never occur, because the exit occurred via a competing event. Second, although the Cox (1972) duration model can accommodate multiple exit types, the interpretation of its coefficients is rendered very difficult because the impact of an explanatory variable on the hazard rate of a given exit type k is a highly nonlinear function of its impact on the hazard rates for all exit types, as well as of their respective baseline hazards. The competing risks model uses a concept known as a subhazard to overcome this issue and provide easily interpretable coefficients. See Cleves et al. (2010) for a discussion. Table 7 presents the estimated coefficients of the model. Column 1 of Table 7 shows that the probability of an SBO exit increases with our Sell Pressure index. A one-unit change in Sell Pressure increases the hazard rate by 8.9% relative to the baseline hazard, a statistically significant change at the 1% level (t-statistic 4.50). 23 Column 2 reports a coefficient with the same sign and strong significance (t-statistic 3.8) for the High Sell Pressure dummy. The coefficient implies that for funds with High Sell Pressure the hazard rate increases by 12.3%. Regarding other control variables, we find that deal size and in recent LBO activity increase the likelihood of exit and lead to shorter expected conditional holding periods, while adverse market conditions (i.e. high credit spreads, low IPO activity) and public-to-private deals are associated with longer expected conditional holding periods (all coefficients significant at the 5% level).
In column 3, we address the history dependence of Sell Pressure. At the time of investment funds might differ in Sell Pressure, and Sell Pressure may rise or fall during the holding period.
Although both the initial and final pressure matter for our story, a more stringent test of our theory addresses the impact of increase in sell pressure on fund behavior. We therefore create two new variables. The first, Sell Pressure at Entry, is the value of the fund's Sell Pressure index at the time of the investment (that is, when τ = 0). This variable is static in our model. The second variable, Increase in Sell Pressure, is a dummy variable that takes the value 1 if the value of Sell Pressure at time τ is higher than the value of Sell Pressure at Entry. This variable is dynamic because it changes over the lifetime of the deal. The results in column 3 show that both the initial value of Sell Pressure and its subsequent increase during the holding period increase the probability of an SBO exit. Both coefficients are statistically significant at the 1% level. Results are, statistically speaking, slightly stronger for Increase in Sell Pressure, indicating that funds become more anxious to exit via SBO when their sell pressure increases.
Finally, one might wonder if the buy pressure faced by the GP at entry also plays a role in divesting the company faster. We therefore replace Sell Pressure at Entry with the value of the Buy Pressure index at the time of the purchase that we call Buy Pressure at Entry. The results in column 4 show that, although higher levels of Buy Pressure at Entry positively affect the hazard rate (coefficient 3%, t-stat. 1.75) and shorten holding periods, the coefficient of Increase in Sell Pressure is still positive and significant at the 1% level. We conclude that the sell pressure of PE funds has a strong impact on the holding period of investments exited via SBOs consistent with our hypotheses. 23 The computation of comparative statics uses the fact that one unit of change in the independent variable changes the hazard ratio by 100×(e β −1) in percentage terms.

Valuation of secondary buyout exits
Our empirical analysis provided evidence that the Sell Pressure index is associated with higher likelihood for secondary deals by PE funds. Next we investigate the relationship between sell pressure of PE funds and deal pricing. We run the following regression model: The dependent variable XSMULT EXIT is a measure of valuation at the time of the exit using either Sales or Ebitda as the basis of the excess multiple. D SECEXIT is an indicator variable that takes value 1 if the LBO is exited through a sale to another PE fund and 0 otherwise. The matrix X includes control variables measured at the time of exit, as in Section 5.1, and we also replace TEV (part of the multiple) with an instrument for deal size, represented by Buyer Size. All specifications include industry dummies, year dummies, and year-clustered standard errors.
The regression results are reported in Table 8 Table 1). Column 2 shows comparable results for the High Sell Pressure index (albeit at 10% significance for the second dependent variable, Exit Excess Ebitda Multiple). In Column 3, the Sell Pressure index is decomposed into three level dummy variables demonstrating a monotonic increase in coefficients throughout. Finally, column 4 shows that our results are robust to the inclusion of PE firm fixed effects.
In Panel B of Table 8 we analyze the four individual components of the Sell Pressure index in columns 1 through 4. The interaction terms of interest are statistically significant for each component in the two panels with the exception of Irregular Fund Raiser × D SECEXIT in Panel B.2 where it is not statistically significant but still has the right sign. We conclude that within the set of secondary exits the pressure to exit depresses valuations in secondary deals. The price impact of Sell Pressure index on the deal price is similar in magnitude but opposite in sign to that of the Buy Pressure index.

Buy pressure and sell pressure: which effect dominates?
Our empirical analysis documented that buyers under pressure are more likely to do SBOs and pay higher prices. We also found that sellers under pressure are more likely to exit at lower multiples via secondary sale. Hence, there is an adverse price effect of both buy and sell pressure in SBOs. An interesting question is what happens when a pressured buyer meets a pressured seller.
Would the party with the higher pressure drive the price higher or lower? If parties are equally pressured, would they offset each other's impact on deal multiples? Are trading partners chosen on the basis of their transaction pressure as trading with equally pressured parties provides opportunity for funds under pressure to hedge against the pricing risk to which the pressured fund is exposed?
In unreported results we find little evidence for assortative matching between buyers and sellers according to their pressure. For the transactions that we can identify both buyer and seller the correlation index of the Buy and Sell Pressure indexes is -1% and statistically indistinguishable from zero. High-pressured sellers match with high-pressured buyers in only 8% of deals. These possible matches represent about 16% of deals made by High Buy Pressure buyers and less than a third of the deals made by High Sell Pressure sellers. Thus there is no evidence for deliberate matching to mitigate the price impact.
Given the opposite signs of the Buy Pressure and Sell Pressure indexes on deal valuation, if buyers and sellers with various pressure indexes are matched, which effect dominates? We estimate a model for the valuation of secondary LBOs within the subsample of exits for which we have information about buyers' status, sellers' status, and deal valuations. We include as explanatory variables our Buy Pressure and Selling Pressure indexes simultaneously, as well as the control variables used in Section 5.2 above.
The results are displayed in Table 9. In column 1, we include the Buy Pressure and Sell Pressure indexes on the sample of SBO deals in which we can identify both parties. The price impact of Buy Pressure is positive and statistically significant at the 1%, whereas the price impact of Sell Pressure is negative, albeit only at 10% of significance. In columns 2 and 3, we interact each pressure index (for buyers in column 2 and for sellers in column 3) with the dummy variables of high and low pressure of the counterparty (respectively, sellers in column 2 and buyers in column 3). We find that when a highly pressured PE fund is matched with a less pressured counterparty, the low-pressure counterparty has the bargaining power and moves the price in its favor. For example, the interaction Buyer Pressure × Low Sell Pressure has a positive and significant coefficient (0.219, t-statistic 4.30) indicating that the impact of Buyer Pressure is particularly strong when the GP faces a low pressure seller. The converse is also true (coefficient of Seller Pressure × Low Buy Pressure negative and significant coefficient at the 1% level). When both parties face high pressure, their bargaining power is equal and neither has an impact on the transaction multiple.
These results are confirmed in column 4 where the high/low pressure dummies are interacted for each of the counterparties to represent all possible pressure outcome pairs (High Buy Pressure × High Sell Pressure, High Buy Pressure × Low Sell Pressure, etc.). The base omitted category is Low Buy Pressure × Low Sell Pressure. Again, we find that the difference in the relative bargaining power of the parties determines the impact of buy/sell pressure on deal prices.
The magnitude of the coefficients and the relatively weaker statistical significance of Sell Pressure in column 1 indicate that seller funds are in somewhat stronger bargaining position relative to buyer funds. This finding also supports the prediction of Axelson, Stromberg and Weisbach (2009) that buyer funds with substantial dry powder late in their investment period are willing to invest regardless of the price and may even invest in negative NPV projects.

Buy pressure and leverage in secondary deals
We document that PE funds under buy pressure are more likely to engage in SBOs and when they do, they pay higher valuation multiples. To further investigate the motives of these funds, we now focus on the relationship between buy pressure and deal leverage in secondary transactions.
Concerning secondary deals, on the supply side lenders face lower information costs for companies that are already LBO targets, and are presumably willing to lend more to firms that were able to carry more debt than in their first LBO. Moreover, secondary deals are often considered less risky, and as such, they can support a higher debt load. On the demand side, as shown in Axelson et al. (2012), Bonini (2012), Wang (2012), and Achleitner et al. (2012), SBO activity is generally high when debt financing costs are low. Moreover, if secondary deals demand higher multiples because they are less risky, these deals would be expected to have even higher debt capacity. Hence, on the basis of the above reasoning, one would predict that secondary deals would be more levered on average. Interestingly, however, existing evidence on deal leverage of SBOs is at best mixed. Among the previous papers investigating SBOs, only Achleitner and Figge (2011) document higher leverage in SBOs whereas Bonini (2012) and Wang (2012) report lower leverage.
If, however, the motive for engaging in secondary deals is the buy pressure of PE funds, then one would expect to find a negative relationship between buy pressure and deal leverage. Pressured buyer funds would be eager to put their capital to use and prefer to execute a deal with spending more of their equity than other PE buyers because their opportunity cost of doing so is lower since their primary motive for completing the deal is to draw down their dry powder and secure management fee.
To decide between these competing hypotheses, we investigate the relation between deal pressure and deal leverage for secondary deals. We use three different leverage metrics: the ratio of senior debt to enterprise value at the onset of an LBO deal, the ratio of debt to Ebitda, and the ratio of debt to total assets. Information on senior debt, defined as the sum of all term debt facilities used in the deal, is obtained from multiple sources including Capital IQ, DealScan, Dealogic, and company filings in the case of public-to-private deals. Table 10 report the results for each of the three different leverage ratios.

Columns 1 to 3 in
Our main variable of interest is the interaction term Secondary × Buy Pressure. It is negative and statistically significant for all three leverage variables, with t-statistics ranging from -1.69 to -2.12.
This evidence provides support for the agency view that secondary deals by pressured buyers are aimed at spending the fund's dry powder rather than maximize their LP returns. 24 Our Buy Pressure index alone is not significant in the regressions suggesting that in primary deals Buy Pressure does not have significant explanatory power for deal leverage. In our sample, secondary LBOs have somewhat higher debt levels than other deals (the coefficient for Secondary is positive and statistically significant in two of the three equations), as expected from the supply side forces at play.

Buy pressure and syndication in secondary deals
After documenting that buy-pressured funds are more likely to engage in SBOs and when they do, pay higher prices and use less leverage, we focus on the relationship between buy pressure and deal syndication. There maybe a positive or inverse connection between the use of leverage and syndication. On the one hand, funds may raise less leverage because they engage larger syndicates to finance a deal, i.e. deal leverage and syndication could be substitutes. On the other hand, funds that aim to spend more of their own equity are likely to employ less debt and use less syndicate financing.
If pressured PE funds invest to lock in management fees and reduce their unspent capital, they should prefer to spend more of their own equity in the SBOs rather than share the deal. Funds under buy pressure should be less likely to syndicate their SBOs, or be inclined to form smaller syndicates with higher deal value per syndicate members. On the other hand, if PE funds are motivated by efficiency considerations, then the equity that they commit to individual deals should determined by deal and firm characteristics, macro-variables, industry, and year effects. Since secondary deals are typically larger than primary deals, one would expect in this case that SBO syndicates are not significantly smaller than syndicates for other deal types.
Our empirical analysis utilizes three measures of syndication: an indicator variable if the deal is syndicated (involves more than one buyer); the size of the syndicate, that is, the number of syndicate members involved in the acquisition; and the deal value per syndicate member, that is, the imputed equity value of the deal divided by the number of buyers. Columns 1 to 3 in Table 11 report the results for each of the three different measures of syndication. Regarding the control variables, 24 Similarly, Officer, Ozbas and Sensoy (2010) find in PE club deals that acquisitions made at lower multiples receive leverage at better terms which, as they explain, is difficult to reconcile with a risk/information story. larger deals and deals with management participation are positively related to all measures of syndication. Affiliated funds also seem to engage in syndicated deals more often.
Our main variable of interest is the interaction term Secondary × Buy Pressure. It is negative and statistically significant for syndicate size at the 5% level (with t-statistics -2.3), positive and statistically significant for value per syndicate member at the 5% level (with t-statistics 2.31), and negative but not statistically significant for likelihood of syndication. The signs of all three coefficients are in line with the agency view that PE funds under pressure are motivated primarily by spending their own equity in SBOs.

Buy pressure and fund performance
Our analysis provides strong evidence that pressured funds are more likely to invest in SBOs, pay higher valuation multiples, use lower leverage and rely less on syndicate finance. This evidence lends support to the prediction of Axelson, Stromberg and Weisbach (2009) that PE funds with substantial dry powder and lesser reputation late in their investment period may be driven to make negative NPV investments. However, we cannot yet exclude that these transactions may still be valuemaximizing for investors. This could be the case for example if pressured funds pay higher multiples in secondary buyouts but are able to sell these companies at even higher multiples when they exit, in other words if pressured SBO deals select high-value companies. Therefore, we consider in our final investigation the effect of SBO activity under buy pressure on fund performance. We explore this issue by analyzing overall fund returns. We collect fund-level performance data from Preqin. After making sure we can obtain the variables required for our analysis, we are left with a sample of 281 funds. We use two metrics for PE fund performance, both commonly used in the industry: the fund's internal rate of return (IRR); and the Multiple (the money-on-money multiple of the fund's equity distributed upon deal exits relative to the equity committed when the fund made investments). To avoid the influence of extreme winners or losers, we compute the percentile rank of each fund in terms of IRR and Multiple, and also use those as dependent variables. Our specification and set of controls originate in the literature on PE performance (e.g., Kaplan and Schoar, 2005). We control for, among other things: the fund's size and sequence; whether the fund family is highly reputable, that is, it appears in the Top 50 list of the Private Equity International magazine (e.g. Demiroglu and James, 2010); and whether the firm is specialized in deals of a certain size or industry as in section 4.3. The specification also uses vintage dummies to control for differences in performance across fund vintages. The table shows that in all cases this variable has a negative impact on fund performance, and that this impact in statistically significant in 3 out of 4 cases: at the 5% level (IRR and Multiple) and at the 10% level (Multiple rank). In terms of comparative statics, the coefficients indicate that a one standard-deviation increase in Average Buy Pressure (0.8 in this sample) leads to a decrease in IRR of about 2%. This indicates that investment decisions possibly influenced by pressure seem to be associated with weaker fund performance.
However, our story is not about pressure per se but how SBOs are the most likely type of investment decision with negative effects of pressure. To address this issue, we introduce two further variables in the analysis: the value-weighted percentage of secondary deals made by the buyer, denoted Percentage SBOs; and an interaction term of our two variables of interest, Percentage SBOs×Average Buy Pressure. Panel B of Table 12 shows the results. First, we find that Percentage SBOs has a positive and statistically significant loading in 3 out of our 4 specifications. This indicates that SBOs seem to be associated with positive rather than negative fund performance.
Secondly, and more importantly for our purposes, we find that the interaction term Percentage SBOs×Average Buy Pressure is negative and significant in all specifications, while the coefficient of Average Buy Pressure is negative but no longer significant at conventional levels. This result indicates that the negative effect of pressure is mostly associated with SBO investments. To gauge the magnitude of the effect, if we take the sample's average of secondary deals made by these buyers (14.9%) and compute the marginal effect of Average Buy Pressure using the interaction specification, we find that the effect of Average Buy Pressure on performance roughly doubles. 25 In other words, SBOs explain around 50% of the negative impact of Buy Pressure on performance, and this fraction is the one that seems most statistically significant.
While we can exclude the possibility that some unobservable factors could influence these results, however these findings, along with those reported in the rest of the paper, provide a coherent picture that points in the same direction. Funds taking decisions under pressure and choosing the path of least resistance by doing SBO deals seem to face adverse price impact and show weaker fund performance.

Endogeneity issues
In most studies on the behavior of PE funds or financial intermediaries, endogeneity is an important concern as the econometric relationships might be driven by unobserved hidden variables.
For example, taking the result that secondary buyouts typically have higher valuations (e.g. Wang,25 The marginal effect of Average Buy Pressure for (say) IRR for the average buyer in terms of percentage of SBOs in its portfolio would be -1.473 × (-8.744×0.149) = -2.775, or roughly the double of the base coefficient Average Buy Pressure. 2012), a potential concern is that some funds could have a simultaneous preference for secondary buyouts and for relatively expensive deals, and hence be spuriously identified as willing to invest in SBOs and pay relatively high prices, or that secondary buyouts select companies that are less risky, and hence command higher prices. Then the link between value and SBO would not be causal but would reflect unobserved GP or company characteristics.
We cannot rule out this scenario, but we point out that our research design addresses many of these concerns. First, we do not explain SBO activity per se, but only its relationship to our measures of incentive distortions. Our Buy Pressure and Sell Pressure indexes are time-varying and in our data the same GP is typically present in deals with and without pressure. This reduces the possible impact of unobservable fund-level characteristics on our relationship of interest. Similarly, even though certain companies might be preferred in SBOs, there is little reason to believe that they are systematically different in SBOs under pressure compared with SBOs without pressure.
Second, we use in our specifications a large number of control variables capturing a wide array of portfolio company characteristics, fund characteristics, and market conditions. We find that these covariates, while often significant, do not alter the relationship between our variables of interest.
Third, and more importantly, throughout the analysis we include PE firm fixed effects, taking advantage of the panel nature of our data. These fixed effects should be able to pick up any endogeneity that arises from underlying unobservable firm-level characteristics jointly generating the relationships explored in this paper. Our findings are robust to the inclusion of these fixed effects. We find that the significance level of our covariates of interest rarely weakens. 26 In addition, we exploit the fact that each SBO has two PE equity funds as counterparties, both of which could potentially be under pressure. Focusing on our valuation relationships (this analysis cannot be done for the logit regressions), we look at the pressure index "on the other side" of the deal by controlling for the pressure index "on this side" of deal. As discussed in Section 6, both the buyer and the seller fund might be under pressure and the difference in the buy and sell pressure influences the deal valuation. Most importantly, we find no evidence in favor of assortative matching between sellers and buyers according to their pressure status. In other words, in an SBO the probability that a buyer is under pressure seems to be orthogonal to the probability that the seller is under pressure. Therefore, we isolate the impact of the pressure index of the deal counterparty as a reasonably exogenous variation that should be orthogonal to the characteristics of the PE firm and the portfolio company. Thus, the impact of the randomly drawn pressure index of the counterparty should be much less concern for endogeneity.

Conclusion
This paper investigates the private equity contract's impact on the investment decisions and exit outcomes of funds. We focus on secondary buyouts, an increasingly important and growing segment of the LBO market. We investigate to what extent these deals are motivated by efficiency considerations or opportunistic behavior induced by the PE contract. We stipulate that contract incentives affect PE funds' propensity to engage in secondary deals, the valuation multiples they pay/accept, the deal leverage and syndicate size. We develop Buy Pressure and Sell Pressure indexes that reflect the end of a fund's investment period/life, its unspent capital/deal inactivity, its reputation, and how often its fund family raises funds from investors.
We construct a comprehensive sample from one of the largest available databases of PE transactions; it involves a diverse set of GPs and LPs and hence reduces the risk of selection bias inherent in proprietary datasets obtained from a few highly reputable LPs. Using a sample of 9,575 U.S. and European LBO deals, we find strong evidence that pressured buyers and sellers are more likely to engage in SBOs. Participation of a pressured seller (buyer) moves the transaction multiple significantly lower (higher) even after controlling for market conditions and firm characteristics.
When pressured buyers meet pressured sellers in a secondary transaction, it is the relative pressure between the two that affects their relative bargaining power and the deal price. We also find that pressured buyers use less leverage and rely on smaller deal financing syndicates. Finally, on a subsample of completed deals we report that funds with higher average buy pressure have lower IRR and return multiple over the life of the fund. Our findings provide strong support for the prediction of Axelson, Stromberg and Weisbach (2009) that PE funds with substantial unspent capital late in their investment period are more likely to make negative NPV investments.
By providing an in-depth analysis of the impact of the incentives from PE contractual provisions on deal type, deal prices and deal terms, our paper contributes to the literature on PE investments, as well as to the literature on agency conflicts between fund managers and investors. We show how positive incentives early in a fund's life turn into adverse incentives later, corroborating the notion that agency costs are inevitable even in a sophisticated contractual environment such as the one of private equity (Robinson and Sensoy, 2013). Our findings should also have implications for investment and exit decisions in other segments of the ever-growing fund management industry. The more illiquid the asset base of a fund, the more likely should be the occurrence of the agency-driven effects that we document for the case of secondary buyouts.

Table 1 Summary Statistics
This table presents summary statistics for the sample used in this study. Panel A presents the summary statistics data for our deal-entry level panel. The definitions of the main variables are as follows (for other variables please see Appendix B). Secondary is an indicator variable with value 1 if the seller in a deal is a PE fund. Imputed TEV of the target firm is the sum of the target's equity market value valued at the LBO offer price and the target's net debt. For transactions without deal value information, we compute an estimate of deal value using the methodology of Stromberg (2008). Buy Pressure is the sum of variables Dry Powder, Late Buyer, Lack of Reputation, and Infrequent Fund Raiser. Dry Powder is an indicator variable equal to 1 if the buyer is above median in terms of fund raising and below median in terms of deal activity (see Appendix B for details). Late Buyer is an indicator variable with value 1 if the buyer's most recent fund at the time of the deal is at the end of its investment period (4 to 6 years after inception). Lack of Reputation is an indicator variable equal to 1 if the buyer is below the 75th percentile in terms of rank of value-weighted market share, computed using all deals made from 1980 up to deal year. Infrequent Fund Raiser is an indicator variable equal to 1 if the buyer's average time between fund raisings is more than 3 years. Panel B presents the variables used to measure valuation of the deal at the time of entry. Sales (Ebitda) Multiple is the ratio between TEV and latest available yearly sales (Ebitda) for the target firm at the time of the LBO. These variables only take non-missing values when deal value is known. Excess Sales (Ebitda) Multiple is the difference between the target's Sales (Ebitda) Multiple and a valuation benchmark constructed as a follows. For every year, geography (U.S. versus Europe), industry (Fama-French 12-industry classification) and public status (public or private), we compute the median sales (Ebitda) multiple for all merger transactions with value larger than 1 million dollars involving a majority stake over the previous two years relative to the date of the LBO. Panel C describes the characteristics at the time of exit. Secondary Exit is an indicator variable equal to 1, if the exit route of the LBO is a sale to another PE fund or group of PE funds. Add-ons is equal to 1, if there were significant acquisitions during the time that the buyer held the target firm in its portfolio. Stage is the number of years elapsed since fund raising for the youngest fund in the PE fund family whose investment period overlaps with the LBO deal date (that is. whose deal date falls within years 1 through 6 of the lifetime of the fund). Sell Pressure is the sum of variables Exit Pressure, Late Seller, Lack of Reputation, and Infrequent Fund Raiser. Last Exit is equal to 1 if three or more years have elapsed since the PE fund family last exited an LBO deal. Late Seller is an indicator variable with value 1, if the sale takes place in years 9 or 10 of the PE fund family's oldest active fund (i.e. less than 11 years old). Panel D describes the valuation variables at the time of exit (see Panel B above for a description of the variables

Table 4 Likelihood of a Secondary Deal: Fund Specialization
This table presents a logistic regression in which the dependent variable is an indicator variable with value 1 if the seller in a deal is a PE firm. The explanatory variables of interest are the industry specialization of the buyer (Industry Specialization Dummy, Industry Specialization %), the size segment specialization of the buyer (specialization in small, mid, or large LBOs), and industry related variables (Concentration, Asset Liquidity) and the Buy Pressure index. Note that in column 3 the base omitted category is the group of PE funds that are not "size-focused" in a particular size category. All variables are defined in the caption of Table 1 and in Appendix B. Standard errors are clustered by deal years. t-statistics are reported in parentheses and ***, **, * denote significance at 1%, 5% and 10%.

Table 7 Holding Period of Secondary Exits
This table presents results of a competing risks regression model, in which the dependent variable is the holding period of an LBO, measured in years. A coefficient larger than one indicates a shorter expected conditional duration. The sample includes both deals with known exit and those not yet exited at the end of the sample period. The main explanatory variables are: Sell Pressure, the sum of Exit Pressure, Late Seller, Lack of Reputation, and Infrequent Fund Raiser at any given year during the holding period of a deal; Sell Pressure at Entry, the value of the Sell Pressure index of the buyer at the time of the investment; Increase in Sell Pressure, an indicator equal to 1 if the Sell Pressure index increases during the holding period (relative to "Sell Pressure at Entry"); Buy Pressure at Entry, the value of the Buy Pressure index for the buyer at the date of purchase of the LBO. All other variables are defined in the caption of Table 1 and in Appendix B. Standard errors are clustered by buyer. T-statistics are reported in parentheses. The symbols ***, **, * denote significance at 1%, 5% and 10%.

Table 11 Pressure and Syndication in Secondary Deals
This table presents regression results of the relation between deal syndication and Buy Pressure, using three measures of the former: (i) an indicator variable if the deal is syndicated (involves more than one buyer); (ii) the number of buyers involved in the deal; (iii) the deal imputed equity value divided by the number of buyers. All variables are defined in the caption of Table 1 and in Appendix B. Standard errors clustered by deal year. tstatistics are reported in parentheses and the symbols ***, **, * denote significance at 1%, 5% and 10%.  (ii) the rank in terms of IRR of a PE fund, computed across funds with the same vintage (year of fund closing) and geography (U.S. vs. non-U.S.) (a rank of 1 indicates the fund with the lowest rank; (iii) the moneyon-money fund multiple; (iv) the rank of Multiple of a PE fund, computed across funds with the same vintage (year of fund closing) and geography (U.S. vs. non-U.S.) (1 indicates the fund with the lowest rank). Standard errors clustered by buyer fund family. All variables are defined in Appendix B. t-statistics are reported in parentheses. Symbols ***, **, * denote significance at 1%, 5% and 10%.  IPO values are measured in real 2010 U.S. dollars, after conversion at historical exchange rates.

Stage
Number of years elapsed, relative to the deal date, since fund raising for the youngest fund in the PE fund family whose investment period overlaps with the LBO deal date (that is. whose deal date falls within years 1 through 6 of the lifetime of the fund).

Divisional
Indicator variable equal to 1 if the seller is a corporate entity, and zero otherwise.

Financial seller
Indicator variable equal to 1 if the seller is a financial institution, and zero otherwise.
Public to private Indicator variable equal to 1 if the target firm is a publicly listed company, and zero otherwise.
Private to private Indicator equal to 1 if the seller is a non-PE investor group or individual, zero otherwise.

Sales Multiple
Ratio between TEV and latest available yearly sales for the target firm at the time of the LBO. involving a majority stake over the previous two years in the same geography (U.S. versus Europe), industry (Fama-French 12-industry classification) and public status (public or private).

Industry Specialization
Dummy variable that takes the value 1 if the PE fund family has done more than one-third of its past deals in the same industry group as the target's industry.
Industry Specializ. (%) Percentage of the PE fund family's past deals in the same industry group as the target's industry.
Size Specialization Set of three indicator variables equal to 1 if the PE fund family has done more than two-thirds of its deals in the small (medium) [large] size category, defined as deals with an imputed enterprise value < 50 million (between 50 and 250 million) [> 250 million] real 2010 U.S. dollars.

Industry Concentration
Herfindahl index, by geography (U.S. vs. Europe) and year, of public firms with the 48 Fama-French industry code as the target firm.
Asset Liquidity Target industry's ratio of the value of corporate transactions (excluding LBOs) to the value of the total assets of public firms in that industry (Schlingemann, Stulz and Walkling, 2002).
Debt to Enterprise Value Ratio of Senior Debt to Total Enterprise Value. This ratio is only computed for deals in which total enterprise value is observed. Senior Debt, defined as the sum of all term debt facilities used in the deal, is obtained from multiple sources including Capital IQ, DealScan, Dealogic, and company filings in the case of public-to-private deals.
Debt to Ebitda Ratio of Senior Debt to Ebitda for the target firm at the time of the LBO. Senior Debt is defined as above (see "Debt to Enterprise Value").
Debt to Assets Ratio of Senior Debt to Total Assets for the target firm at the time of the LBO. Senior Debt is defined as above (see "Debt to Enterprise Value").

Syndicated
Indicator variable equal to 1 if there is more than one buyer, and zero otherwise. Fund belonging to the same fund family are counted as a single buyer.

Variable: Definition:
Buy Pressure Sum of variables Dry Powder, Late Buyer, Lack of Reputation, and Infrequent Fund Raiser.

High Buy Pressure
Indicator variable equal to 1 if Buy Pressure is equal or greater than 2.

Dry Powder
Indicator variable equal to 1 if the buyer is above median in terms of fund raising and below median in terms of deal activity. These criteria are computed as follows. First, for each PE fund family and year, we calculate the aggregate funds raised in the past 3 years, and the corresponding median across fund families in that year. Second, for each PE fund family and year, we compute the aggregate dollar value of all investment made during the past three years, and its respective median across fund families in that year. All monetary values are measured in real December 2010 U.S. dollars, after conversion at historical exchange rates.

Late Buyer
Indicator variable equal to 1 if the buyer's most recent fund at the time of the deal is at the end of its investment period (4 to 6 years after inception), and zero otherwise.

Lack of Reputation
Indicator variable equal to 1 if the buyer is below the 75th percentile in terms of rank of valueweighted market share, computed using all deals made from 1980 up to deal year.

Infrequent Fund Raiser
Indicator variable equal to 1 if the buyer's average time between fund raisings is more than 3 years. Time between raisings is computed dividing the total number of funds raised by a firm up to the observation year by the age of the firm (time since the firm's first deal in the database).

Affiliated
Indicator variable equal to 1 if the buyer is affiliated to a financial institution or government agency, and zero otherwise.

Novice
Indicator variable equal to 1 if the buyer is a PE fund family with 3 funds or less under management at the time of the LBO deal, and zero otherwise.

Buyer Size
Log of the value of deals done by the buyer in the last five years.

Variable: Definition:
Secondary Exit Exit route of the LBO is a sale to another PE fund or group of PE funds.

Down Exit
Exit route of the LBO is a sale to management, a distressed merger transaction, or bankruptcy.