Does shareholder coordination matte

,n

ool, U

Article history: We propose a new role for private investments in public equity (PIPEs) as a mechanism

result obtains even after controlling for alternative explanations such as information

asymmetry and access to public markets. Improved equity coordination following a

private placement leads to favorable debt renegotiations within one year of issuance.

Mitigating coordination frictions among shareholders ultimately decreases the odds of

rm default in half.

& 2012 Elsevier B.V. All rights reserved.

f private equityent return. For

age daily returne price reactionment returns ofplies that PIPEs

o existing share-holders. This is even more surprising considering that the

period) and results in signicant dilution of the holdings

Contents lists available at SciVerse ScienceDirect

journal homepage: www.e

Journal of Financial Economics

We thank Bill Schwert (the editor), Ilya Strebulaev (the referee),

and seminar participants at the Lone Star Conference 2011 and the

University of North Carolina at Chapel Hill for helpful comments.

Journal of Financial Economics 108 (2013) 2132300304-405X/$ - see front matter & 2012 Elsevier B.V. All rights reserved.

http://dx.doi.org/10.1016/j.jneco.2012.10.001n Corresponding author. Tel.: 1 919 962 4926;fax: 1 919 962 2068.

E-mail addresses: ichakraborty@cox.smu.edu (I. Chakraborty),

gantchev@unc.edu (N. Gantchev).

1 The total volume of private equity issuance during the sample

period was $164 billion versus $715 billion of public equity offerings

(see Table 1).average private equity placement is offered at a largediscount to current market prices (13% in our sample

Christos Cabolis, Paolo Fulghieri, Diego Garcia, Pab Jotikasthira, Swami-

nathan Kalpathy, William Maxwell, Paige Ouimet, Christopher Parsons,

Rex Thompson, Anil Shivdasani, Johan Sulaeman, Kumar Venkataraman,the unregistered sale of publicly traded securities such ascommon or preferred stock and convertibles to a smallgroup of sophisticated private investors. Despite theirmore complex contract structure, frequently includingreset provisions and warrants, PIPEs have become anincreasingly important means of raising equity for troubled

in 2007.1

One of the most puzzling features oplacements is their positive announcemexample, the (3, 1) cumulative averduring 19952007 is 2.12%. This positivcontrasts with the negative announcesecondary equity offerings (SEOs) and imare viewed by the market as benecial t

$Private investments in public equity (PIPEs) involveresult, the share of private placements in secondaryequity issuance has increased from 4% in 1995 to 27%Received in revised form

6 June 2012

Accepted 11 June 2012Available online 13 October 2012

JEL classication:

G32

G33

G34

Keywords:

Private placements

Equity issuance

Shareholder coordination

Debt renegotiation

Firm distress

1. Introduction rms with limited access to the public equity market. As aReceived 16 April 2012 to reduce coordination frictions among existing equity holders. We establish a causal

link between the coordination ability of incumbent shareholders and PIPE issuance. Thisprivate placements$

Indraneel Chakraborty a, Nickolay Gantchev b

a Southern Methodist University, Cox School of Business, United Statesb The University of North Carolina at Chapel Hill, Kenan-Flagler Business Sch

a r t i c l e i n f o a b s t r a c tr? Evidence from

nited States

lsevier.com/locate/jfec

of incumbent equity holders (30% on average in 19952007).2

The existing literature has provided several competinginterpretations of the positive announcement effect ofPIPEs. Wruck (1989) establishes a relation between themarkets positive reaction to private placements and theincrease in ownership concentration following PIPE issu-ance. She interprets the positive price effect of PIPEs asevidence that changes in ownership concentration betteralign the interests of managers and shareholders. Hertzeland Smith (1993) consider the role of private placementsin resolving asymmetric information problems about rmvalue. They view a private issue as a seal of approval by

Reduced coordination frictions among shareholders fol-lowing PIPE issuance substantially decrease the odds ofdefault of PIPE rms compared with matched controls.PIPE issuers are also more likely to experience favorabledebt renegotiations resulting in lower interest spreads

I. Chakraborty, N. Gantchev / Journal of Financial Economics 108 (2013) 213230214sophisticated institutional investors on the current valua-tion of a rm.3

Typical PIPE issuers are troubled rms with moredispersed shareholders and more concentrated debt-holders than the average rm. Building on the Wruck(1989) contribution, this paper argues that PIPE issuanceallows dispersed equity holders to concentrate theircontrol rights by bringing in a new blockholder with alarge incentive to improve rm value. However, unlike theWruck (1989) emphasis on improved monitoring redu-cing agency conicts within the rm, we focus on analternative channel whereby private placements serve asa mechanism to mitigate coordination frictions amongexisting equity holders in their choice of rm policy.

A distressed rm is likely to experience a shift ofcontrol rights from equity to debt, in which case anychange in existing rm policy could require negotiationsbetween equity holders and debtholders. We claim thatPIPE issuance improves the coordination ability of equityholders and facilitates negotiations of rm policy withdebtholders. We focus on debt renegotiation as a specicexample of a major policy, which benets from improvedability of a rms stakeholders to come to an agreement.4

Debt renegotiations are especially important for privateplacement rms because of their high level of distress andreduced ability to access public markets.

Two main contributions of this paper deserve atten-tion. First, we use instrumental variables (IV) analysis toestablish a causal link between the coordination ability ofincumbent equity owners and PIPE issuance. This resultobtains even after propensity score matching on alter-native explanations of private equity issuance. Second, weshow the effect of the coordination channel on a rmspost-issuance debt renegotiation and default likelihood.

2 Hertzel, Lemmon, Linck, and Rees (2002) discuss the positive price

effect of private placements and their negative long-run performance.

Huson, Malatesta, and Parrino (2010) investigate the recent decline in

the PIPE discount.3 Both Hertzel and Smith (1993) and Wu (2004) provide cross-

sectional evidence at odds with the Wruck (1989) monitoring hypoth-

esis. Barclay, Holderness, and Sheehan (2007) interpret the PIPE discount

as compensation to investors for their implicit support of management

entrenchment.4 The coordination hypothesis we propose builds on previous

theoretical work, which considers the role of debt contracts in transfer-

ring state-contingent control rights to creditors (e.g., Aghion and Bolton,

1992; Dewatripont and Tirole, 1994). Recent empirical work has

explored the importance of control right dynamics for rm policy (see

Chava and Roberts, 2008; Nini, Smith, and Su, 2009).and larger loan principals within one year of issuance.Our empirical approach aims to differentiate the coor-

dination channel proposed in this paper from the infor-mation asymmetry and monitoring hypotheses in theexisting literature. Ideally, we would be able to conducta randomized experiment in which rms with differentcoordination ability of incumbent equity holders arerandomly chosen to issue equity in the secondary publicmarket (SEO) or to private investors (PIPE). In the absenceof such randomization, we need to effectively control forthe potential selection bias resulting from the effect ofrm characteristics (such as information asymmetry,access to public markets, and distress) on the choice ofequity nancing.

We use propensity score matching techniques toreduce the confounding effects of rm attributes onthe mode of equity issuance. We look for conditioningvariables among the rm characteristics suggested byalternative explanations of private equity issuance. Spe-cically, we compare each PIPE issuer to its SEO counter-parts in terms of pretreatment differences in informationasymmetry, access to public markets, and predicteddefault probability. Our propensity score analysis correctsfor selection bias in terms of observable characteristicsthat could affect the decision to issue private equity. Wealso use instrumental variables analysis to address poten-tial self-selection concerns in terms of unobservable rmheterogeneity.

Our measure for shareholder concentration directlyreects the level of coordination necessary to reach adecision based on shareholder voting. We use a rmstotal Shapley value to proxy for existing coordinationfrictions among incumbent equity holders. The Shapleyvalue captures the relative importance of each votingshareholder in terms of her expected ability to have apivotal vote in changing rm policy.5 A low Shapley valueof current shareholders suggests larger coordination ben-ets from adding a PIPE investor. Our univariate resultsshow that PIPE issuers have 51% lower Shapley values ofincumbent equity than their non-PIPE counterparts.

To account for the pre-issuance balance of powerbetween equity holders and debtholders, we also measurea rms concentration of public debt claimants by theHerndahl Index of its bond issues. This proxy capturesthe distribution of par values of outstanding bonds. A higherbond Herndahl Index indicates more concentrated bond-holders, which increases the benet of improving thecoordination ability of a rms equity holders. We observethat PIPE rms have 33% more concentrated bondholdersthan SEO rms.

5 Using Shapley value instead of alternative measures such as total

institutional ownership also differentiates our coordination mechanism

from the Wruck (1989) monitoring hypothesis.

equity coordination after PIPE issuance raises a rmslikelihood of a favorable debt modication.

The rest of the paper proceeds as follows. Section 2describes the data and introduces our coordination proxiesas well as our main controls. Section 3 reviews the empiricalevidence supporting the coordination hypothesis. Section 4

I. Chakraborty, N. Gantchev / Journal of Financial Economics 108 (2013) 213230 215Our rst set of results shows that both the coordina-tion ability of a rms incumbent equity holders (mea-sured by their Shapley value) and the concentration ofits public debtholders (proxied by the bond HerndahlIndex) are highly statistically signicant in predictingPIPE issuance. Using a comprehensive US sample ofprivate equity placements and secondary equity offeringsbetween 1995 and 2007, we show that a rm experiencesa 38% increase in the odds of a private placement with aone standard deviation decrease in the coordinationability of its shareholders. In addition, we observe a 20%increase in the odds of a private placement with a onestandard deviation increase in the concentration of arms debtholders.

Using the initial-year industry Shapley value as aninstrument for the rm-specic Shapley value of anissuer, we establish a causal relation between a rmsequity holder coordination and PIPE issuance. This causallink remains signicant even after propensity scorematching on other determinants of the choice of externalnancing such as information asymmetry and access topublic markets. Our results demonstrate that the coordi-nation mechanism plays an important role in explainingthe choice to issue private equity.

How do current shareholders and new PIPE investorsshare the surplus realized by coordination improvement?We nd that a one standard deviation increase in theShapley value of incumbent equity holders decreasesthe discount offered to PIPE investors by 14%. This resultimplies a statistically and economically signicant rela-tion between the benets of reducing equity coordinationcosts and the discount that new PIPE investors receive.We also show that high bond concentration increasesthe gains from improved equity coordination. Firms withabove-median bond concentration have 6% higher PIPEdiscounts compared with rms with below-median bondconcentration. These additional tests provide compellingevidence in support of the coordination hypothesis wepropose in this paper.

Our second set of results demonstrates that privateequity issuance is highly signicant in predicting areduced likelihood of default even after propensity scorematching on the typical determinants of default as well asinformation asymmetry and amount of capital infusion.We instrument PIPE issuance by our two coordinationproxies (Shapley value and bond Herndahl Index) andnd that a one standard deviation increase in predictedPIPE decreases the odds of default in half. In fact, PIPEissuance has higher economic signicance than any of thecommon bankruptcy predictors including Z-score.

To provide direct evidence that PIPE rms improvetheir nancial health post-issuance, we examine whetherprivate placements facilitate debt renegotiation in prac-tice. We use difference-in-differences analysis to estimatethe probability of favorable debt renegotiation within oneyear of PIPE issuance. Compared with rms matched onsize, equity coordination, and distress, PIPE rms partici-pate in fewer loan amendments following issuance butachieve a 40% higher incidence of favorable outcomessuch as lower interest spreads and larger principals. Ourempirical results provide strong evidence that improvedfrom the Center for Research in Security Prices (CRSP)/Compustat Merged Database by rst matching tickersymbols from PlacementTracker to PERMNOs (permanent

6 Reg-S PIPEs are placed with foreign institutional investors. 144-A

issuances are subject to different regulatory requirements and generally

are not considered PIPEs.7 Chaplinsky and Haushalter (2010) show that issuers of warrant

contracts achieve similar risk-adjusted returns as issuers of reset

contracts. However, issuers of warrant-only contracts are more dis-

tressed whereas issuers of resets have more volatile returns.relates private placements to reduced default and favorabledebt renegotiation. Section 5 discusses robustness tests.Section 6 concludes.

2. Data and variable denitions

2.1. Data sources

This study uses data on US private equity placementsbetween 1995 and 2007 from the PlacementTrackerdatabase by Sagient Research. After excluding foreign,144-A, and Regulation-S (Reg-S) issuers, the Placement-Tracker data set includes 6,442 unique rms involved in10,765 transactions.6 PIPE issuance has become a vitalsource of equity nancing for most rms. The totalvolume of private equity issues during the sample periodis $163.86 billion, compared with $714.54 billion of publicequity offerings. As seen in Table 1, the share of privateplacements in secondary equity issuance has increasedfrom 4% in 1995 to 27% in 2007.

PlacementTracker contains detailed information aboutthe terms of each PIPE contract. We collect data on the typeof private placement, legal structure, gross proceeds, dilu-tion, discount to market price, warrant coverage, and otherspecics. Table 1 reports the percentage dilution of existingequity dened as one minus the ratio of old equity to thesum of old and new equity. The average dilution of existingequity is 30% and does not vary signicantly between 1995and 2007. The mean discount to market price is 13% butshows a decreasing trend during the sample period. Huson,Malatesta, and Parrino (2010) relate the decrease in thePIPE discount (especially in the latter part of the period)to changes in the characteristics of PIPE issuers and thecontracting environment.

As in Brophy, Ouimet, and Sialm (2006), we classifycommon stock and xed convertible issues as traditionalPIPEs. Structured PIPEs are common stock or convertibleissues with reset provisions, structured equity place-ments, or oating convertibles. As seen in Table 1, 22%of the PIPEs in the sample period are categorized asstructured. We also observe a trend away from structuredissues to placements with higher warrant coverage, espe-cially after 2002.7

We obtain quarterly accounting and stock price data

Table 1Private placement transactions, 19952007.

The table reports statistics on the distribution of private placements in Sagient

non-US, 144-A, and Regulation S issuers. Column 1 presents gross proceeds from se

private investment in public equity (PIPE) issues. Data on public equity offerings ar

olumn

overag

tibles.

PIPE

ssues

(4)

114

306

456

440

691

,254

,036

756

880

,285

,325

,346

876

0,765

828

I. Chakraborty, N. Gantchev / Journal of Financial Economics 108 (2013) 213230216security identication numbers) using the CRSP historicalle of rm names. Then, we match issuers in Placement-Tracker to CRSP/Compustat data by PERMNOs. Matchingby PERMNOs instead of issuer tickers signicantlyimproves the match to approximately 95% of PIPE issuers.Many private placements consist of multiple trancheswithin several weeks of each other. To make themcomparable to SEOs, we combine multiple PIPE transac-tions for a rm within a quarter, which results in 5,610

Column 3 shows the percent of equity proceeds coming from PIPE issues. C

structured PIPEs). Column 7 reports the percent of PIPE issues with warrant c

stock or convertible issues with resets, structured equity or oating conver

Gross SEO Gross PIPE % Equity

Year proceeds, $B proceeds, $B proceeds i

(1) (2) (3)

1995 31.65 1.33 4.03

1996 43.09 4.08 8.65

1997 47.00 4.75 9.18

1998 45.28 3.00 6.21

1999 63.72 10.30 13.92

2000 71.37 24.40 25.48 1

2001 51.29 14.60 22.16 1

2002 48.52 12.10 19.96

2003 49.34 11.60 19.04

2004 68.19 13.70 16.73 1

2005 59.26 16.90 22.19 1

2006 68.78 22.40 24.57 1

2007 67.05 24.70 26.92

Total 714.54 163.86 1

Mean 54.96 12.60 18.65rm-quarter PIPE observations.We calculate our measure for pre-issuance equity holder

coordination (Shapley value) using institutional ownershipdata from Thomson Reuters Institutional Holdings (13F)database. (Section 2.2 discusses Shapley value and itsrelevance in the context of this paper.) We proxy for thebargaining power of debtholders by calculating a HerndahlIndex of outstanding public bond issues using data fromMergent Fixed Income Securities Database (FISD).

To study the effect of PIPE issuance on debt renegotia-tion, we collect data on bank loan facilities and amend-ments from Thomson Reuters LPCs Dealscan database.The data consist of private loans made by bank andnonbank lenders to US corporations.8 The basic unit ofobservation in Dealscan is a loan facility. We obtain theoriginal terms of all bank loans by PIPE rms in the period19952007 and track changes in their maturity, interestspread, and loan amount for one year after PIPE issuance.

Our main control group in most tests consists of allrms with SEOs during the sample period. We obtain dataon public equity issues from Thomson Reuters Securities

8 We merge Dealscan to Compustat data using the Roberts

DealscanCompustat Linking Table from Wharton Research Data Ser-

vices (WRDS). See Chava and Roberts (2008) for further information.Data Company (SDC) Platinum. The data set contains4,841 rm-quarter observations for 2,888 unique rms.In robustness tests, we also compare PIPE issuers to theaverage CRSP/Compustat rm.

We control for a rms information asymmetry, dis-tress level, and access to public markets as alternativeexplanations of PIPE issuance. Following Wu (2004), weproxy for information asymmetry using analyst coverage,trading volume, and the ratio of research and develop-

s PlacementTracker database. The sample covers 19952007 and excludes

condary equity offerings (SEOs), and Column 2 reports gross proceeds from

e obtained from Thomson Reuters Securities Data Company (SDC) Platinum.

s 5 and 6 include only common stock private equity placements (i.e., non

e. Column 8 lists the percent of structured PIPEs in the sample, i.e., common

$B denotes billions of dollars.

% Current % Equity % with % Structured

discount dilution Warrants PIPEs

(5) (6) (7) (8)

25.19 20.36 4.19 30.70

25.94 28.03 8.71 52.94

16.18 28.42 7.01 57.24

12.47 25.41 7.96 55.23

14.39 22.58 13.59 28.08

10.42 25.14 21.84 31.26

7.33 25.95 16.08 21.91

5.91 29.06 14.73 15.34

14.16 25.86 20.60 6.70

11.70 22.30 29.48 12.45

7.99 25.14 31.77 15.16

8.02 28.89 38.11 15.97

7.61 34.20 30.72 11.19

12.87 30.07 23.06 21.95ment (R&D) expense to total rm assets. We measureanalyst coverage with data from Thomson Reuters I/B/E/Sand trading volume using data from CRSP. We calculate arms access to public markets and predicted defaultprobability using data from CRSP and Compustat. Seethe Appendix (Table A1) for all variable denitions anddata sources.

2.2. Shapley value and equity coordination

We consider private equity issuance as a mechanismto improve the coordination ability of equity holders andfacilitate their negotiation of rm policy with bond-holders. This coordination hypothesis builds on previoustheoretical work, which examines the role of debt con-tracts in transferring state-contingent control rights tocreditors (e.g., Aghion and Bolton, 1992; Dewatripontand Tirole, 1994). A distressed rm experiences a shiftof control rights from equity to debt, in which case anychange in existing rm policy typically requires negotia-tions between equity holders and debtholders. This paperuses the context of private placements to study the stepsthat incumbent equity holders take to improve theircoordination ability and strengthen their control rights.

Recent empirical work has explored the importance ofcontrol right dynamics on rm policy. Chava and Roberts

(2008) show that loan covenant violations transfer controlrights to debtholders who subsequently cause a reductionin rm investment. Nini, Smith, and Su (2009) demon-strate that conicts of interest between creditors andborrowers have a signicant impact on a rms invest-ment policy. We hypothesize that PIPE issuance allowsincumbent equity holders to concentrate their controlrights by bringing in a new blockholder with a large

by the probability fi that

wPxiyircrwPxiwiyi, 1where c is the pivotal vote (in our case, 50.01% of totalshares). The total Shapley value of all shareholders of therm (M large shareholders and an innite number ofsmall ones) is by denition one:

I. Chakraborty, N. Gantchev / Journal of Financial Economics 108 (2013) 213230 217incentive and ability to improve rm value. In particular,we suggest that favorable renegotiations of debt policyare easier to achieve in rms whose equity holders havelower coordination frictions.

We measure the coordination ability of existingshareholders by the Shapley value of current institutionalblockholders.9 Other measures such as total institutionalownership or number of blockholders do not directly reectthe level of coordination necessary to reach a decision andsuffer from alternative interpretations as proxies for infor-mation asymmetry or institutional monitoring.

Consider the following example. Compare rm A,whose three equity holders have 49%, 49%, and 2% own-ership stakes, with rm B, whose shareholders hold 51%,47%, and 2% equity stakes. In rm A, two equity holdershave to vote together to reach a majority, and there arethree such combinations. Hence, the Shapley value of the2% stakeholder in rm A is 1/3 (same as the Shapleyvalues of the other two equity holders). The Shapley valueof the 2% shareholder in rm B is zero because the 51%stakeholder has full control and a Shapley value of one.Alternative proxies for shareholder concentration such asthe Herndahl Index do not directly measure the controlrights of shareholders.10 For instance, the HerndahlIndex of rm A is almost identical to that of rm B:0.4806 versus 0.4814.

As in Milnor and Shapley (1978), we use the generalizedpivotal player approach for innite-person games to com-pute the Shapley values of a rms institutional blockholderswho own at least 3% of its outstanding shares. In thisapproach, an equity holders Shapley value is the probabilitythat in a randomly permuted ordering of all shareholders,the equity holder and her predecessors together havea majority vote but her predecessors alone do not. Thisdenition captures the expected importance of each playerin deciding rm policy through a majority vote.

Specically, let x1, . . . ,xm be the major shareholders ofa rm who each own a fraction wi 2 0,1 shares. M is thetotal number of major shareholders. Pxi denotes thenite set of major shareholders x1, . . . ,xi1 who are thepredecessors of player i, where i 2 M. Let the small playerspreceding major player i make up a mass of yi 2 0,a,where a represents the total weight of all small share-holders. In this case, the Shapley value of player i is given

9 In the recent literature, the Shapley value approach has been used

by Zingales (1994) and Nenova (2003) to determine the value of voting

rights.10 A bond Herndahl Index is an appropriate measure of public debt

concentration because a simple bondholder majority is generally not

sufcient to reach a decision due to the more complex features of bond

contracts.fMF 1, 2whereF is the total Shapley value of all small shareholders.11

Using a simulation methodology, we compute theShapley values of all major shareholders each quarterand then add the individual Shapley values to obtain thetotal Shapley value of a rm, fM, which is our proxy forthe coordination ability of its current shareholders.

By denition, the Shapley value captures the impor-tance of each voting shareholder in terms of her expectedability to have a pivotal vote in changing rm policy. Thetotal Shapley value of a rms current shareholdersprovides information about the maximum relative votingshare that a new PIPE investor can obtain. In other words,the smaller the total Shapley value of incumbent equityholders, the larger the coordination benets that anew blockholder can bring. Hence, we expect a negativecorrelation between a rms Shapley value and the like-lihood of a private placement.

Gertner and Scharfstein (1991) show that dispersedpublic debt could lead to holdout problems and inefcientliquidations. Consequently, having more concentratedpublic debt increases the potential benets of improvingequity coordination. We measure the concentration of arms public debt claimants by the Herndahl Index ofbond issues (as in Davydenko and Strebulaev, 2007). DebtHerndahl Index captures the distribution of par values ofa rms outstanding public bonds and is dened as thesum of the squared face values of all bonds divided bythe square of the sum of the face values.12 We expect apositive correlation between our measure of debt con-centration and the incidence of private placements.

2.3. Information asymmetry and access to public markets

Our empirical approach aims to differentiate the coordi-nation hypothesis proposed in this paper from other expla-nations of the choice of external nancing. Hertzel and Smith(1993), Chemmanur and Fulghieri (1999), and Wu (2004)argue that rms with high information asymmetry useprivate placements instead of public equity as a mechanismto reduce this asymmetry. In addition, Bolton and Freixas(2000) and Lemmon and Zender (2010) relate access topublic markets to a rms ability to issue public-rated debt.

We use propensity score matching techniques toreduce the confounding effects of rm attributes on thechoice of external nancing. A propensity score index

11 The denition of Shapley value does not require the presence of

small players with mass y. A rm could be entirely composed of major

shareholders who behave in a strategic manner.12 In terms of concentration, bank debt can be viewed as a limiting

case of bond debt. Most loan facilities are originated by loan syndicates

whose lead arranger eliminates most coordination frictions among debt

holders.

allows us to compare each PIPE issuer with its SEOcounterparts in terms of pre-issuance differences ininformation asymmetry, access to public markets, andpredicted default probability.

have 33% more concentrated debtholders (0.59 versus

I. Chakraborty, N. Gantchev / Journal of Financial Economics 108 (2013) 213230218Specically, we control for the information hypothesisof private placements with three frequently used mea-sures of information asymmetry: analyst coverage, trad-ing volume, and R&D ratio (as in Wu, 2004). Analystcoverage is the number of equity analysts following a rmon an annual basis. We expect a negative correlationbetween analyst coverage and the likelihood of a privateplacement. Trading volume is the ratio of trading volumefrom CRSP divided by the average number of outstandingshares over the previous two years. Firms with lowervolume are more likely to issue private equity. R&D ratiois research and development expenditures to total assetsfrom Compustat. We expect that the R&D ratio is posi-tively correlated with the likelihood of a private place-ment. All information asymmetry measures are lagged toaddress potential simultaneity concerns.

We also take into account a rms access to publicmarkets by estimating its predicted probability of havinga long-term bond rating. Lemmon and Zender (2010) usethe cross-sectional heterogeneity in debt capacity to showthat rms with lower debt capacity (typically small, highgrowth rms with lower return on assets) choose to issueequity to alleviate their nancing decits. Following theirapproach, we use a rms predicted likelihood of having abond rating (not the actual presence of a bond rating) as aproxy for its access to public markets.

We estimate a rms predicted probability of having abond rating using a multinomial logit model, in which thedependent variable equals one if a rm has a Standard &Poors (S&P) rating in a given year and zero otherwise. Ourexplanatory variables are rm size (dened as the log oflagged rm assets), protability (operating income beforedepreciation divided by lagged assets), asset tangibility (ratioof net property, plant, and equipment to lagged assets),market-to-book (total assets less book equity plus marketequity over lagged assets), leverage (long-term debt and debtdue in one year divided by lagged assets), standard deviationof daily stock returns (lagged), and rm age (log of yearssince rst Compustat record). We group rms in each yearinto terciles based on their predicted likelihoods of having abond rating and use these terciles in our default analysis.13

Most PIPE issuers are distressed rms. In most tests,we control for a rms level of distress by estimating itspredicted probability of default. As seen in Table 8, wepredict default using a standard bankruptcy regression, inwhich the dependent variable equals one if a rm experi-ences default or bankruptcy, and the independent vari-ables are the log of total rm assets, earnings beforeinterest, taxes, depreciation, and amortization (EBITDA)ratio (EBITDA/Assets), book leverage, Altmans Z-score,and debt capacity (calculated following the approach ofLemmon and Zender, 2010).

13 Our data come from CRSP and Compustat and cover the period

19862010. The model t is consistent with the results in Lemmon and

Zender (2010), with McFaddens R2 of 54.9% and McKelvey and Zavoinas

R2 of 78.1%.0.44 for SEO rms). A high concentration of bond clai-mants implies greater benets from improved equityholder coordination. The comparison with the averageCRSP/Compustat rm (presented in Panel B of Table 3)conrms these ndings. Private placement rms have 31%lower coordination of incumbent equity (0.11 versus 0.16for non-PIPE rms) and 20% higher debt concentration(0.59 versus 0.49 for non-PIPE rms).In sum, our propensity score analysis allows us todifferentiate the coordination hypothesis from alternativeexplanations of the choice of nancing by correcting forselection bias in terms of observable rm characteristics.We also use instrumental variables analysis to addresspotential self-selection concerns resulting from unobser-vable rm heterogeneity.

3. Empirical results

3.1. The PIPE sample

The coordination hypothesis we propose regards privateplacements as a mechanism to reduce coordination frictionsamong existing equity holders in their choice of rm policy.This channel is more valuable for rms with dispersedshareholders and concentrated debtholders because suchrms gain the most by reducing coordination costs. In thissubsection, we present some preliminary univariate evi-dence that differentiates the coordination mechanism fromthe information asymmetry and limited market accessexplanations of PIPE issuance.

Table 2 reports the pairwise correlation matrix betweencoordination proxies, information asymmetry measures,and PIPE issuance. All variables are lagged by one quarterrelative to the occurrence of a private placement. Ourmeasure of equity coordination frictions (Shapley value)has the highest negative correlation with R&D ratio (0.07)and the highest positive correlation with analyst coverage(0.11). Our proxy for bond concentration (Debt HerndahlIndex) has the highest negative correlation with analystcoverage (0.22) and the highest positive correlation withR&D ratio (0.09). The relatively low correlations betweencoordination proxies and information asymmetry measuressuggest that the coordination mechanism could provideadditional insights into the motivation to issue privateequity.

Table 3 describes PIPE rms in more detail. It reportst-tests for differences in means between PIPE issuers andtwo sets of control rms. Panel A uses our main controlgroup consisting of rms with SEOs, and Panel B uses allrms in the CRSP/Compustat universe. Columns 2 and 4report the 25th and 75th percentiles of the variablesof interest. Columns 5 and 6 show differences in meansbetween PIPE issuers and control rms, with their respec-tive standard errors.

The rst interesting observation from Panel A is thatPIPE issuers have 51% lower equity Shapley values (0.11versus 0.22 for SEO rms). A low Shapley value of currentinstitutional blockholders suggests larger coordinationbenets from adding a PIPE investor. Also, PIPE issuers

Table 2Correlation matrix of coordination and information asymmetry proxies.

Pairwise correlation between coordination measures and common informati

coordination is measured by the total Shapley value of its incumbent equity ow

of outstanding public bond issues. Information asymmetry variables include

ccurre

I. Chakraborty, N. Gantchev / Journal of Financial Economics 108 (2013) 213230 219trading volume. All measures are lagged by one quarter relative to the o

Variable Private Equity

placement coordination

Private placement 1.0000

Equity coordination 0.0427 1.0000Debt Herndahl 0.0180 0.0073R&D ratio 0.0882 0.0723Analyst coverage 0.0486 0.1123The majority of PIPE issuers are distressed rms. Asseen in Panel A, the 75th percentile of PIPE Z-scores islower than the 25th percentile of SEO Z-scores. In fact, 82%of all PIPE rms in the sample period have Z-scores lowerthan the mean SEO Z-score. The average Z-score of PIPEissuers is 2.71 versus 0.27 for their SEO counterparts.PIPE rms are likely to experience a shift in control rightsfrom equity to debt due to their high distress levels.14

Trading volume 0.0547 0.0950

Table 3Comparison of PIPE issuers with non-PIPE rms.

The table reports results of two-sample t-tests with unequal variances. The sam

secondary equity offerings (SEOs). Panel B compares PIPE rms with the average C

value of incumbent equity holders. Debt concentration is proxied by the Herndah

and 6 report the difference in means between PIPE and non-PIPE rms. Stars den

Panel A: Comparison with SEO rms

PIPE rms

Mean 2575% Mean

Variable (1) (2) (3)

Equity coordination 0.109 0.0000.150 0.223

Debt Herndahl Index 0.592 0.3381.000 0.444

R&D ratio 0.073 0.0130.093 0.023

Analyst coverage 1.309 0.6931.946 2.023

Trading volume 18.710 6.09821.574 18.072

Market-to-book 2.882 0.7493.415 2.508

Altmans Z-score 2.711 4.543 to 0.340 0.271Debt capacity 0.066 0.0010.023 0.370

Panel B: Comparison with all CRSP/Compustat rms

PIPE rms

Variable Mean 2575% Mean

Equity coordination 0.109 0.0000.150 0.157

Debt Herndahl Index 0.592 0.3381.000 0.494

R&D ratio 0.073 0.0130.093 0.029

Analyst coverage 1.309 0.6931.946 1.695

Trading volume 18.710 6.09821.574 11.812

Market-to-book 2.882 0.7493.415 2.733

Altmans Z-score 2.711 4.543 to 0.340 0.088Debt capacity 0.066 0.0010.023 0.280

14 The average interest coverage (EBIT/Interest Expense) of PIPE

issuers is 2.18 versus 11.87 for their SEO counterparts. About 80% of all

PIPE rms in the sample period have interest coverage lower than 1.on asymmetry proxies. The sample period is 19952007. A rms equity

ners; a rms debt concentration is proxied by the bond Herndahl Index

research and development (R&D) expense ratio, analyst coverage, and

nce of a private placement.

Debt R&D Analyst Trading

Herndahl ratio coverage volume

1.0000

0.0930 1.0000

0.2180 0.1014 1.0000Panel A of Table 3 also presents a comparison of PIPE andSEO rms in terms of information asymmetry. Privateplacement rms have signicantly higher R&D ratios (0.07versus 0.02) and lower analyst coverage (1.31 versus 2.02),suggesting higher information asymmetry. The difference intrading volumes between PIPE and SEO rms is marginallystatistically signicant but not economically signicant.

We use a rms predicted likelihood of having a bondrating to control for its debt capacity (i.e., access to publicmarkets). As seen in Table 3, private placement rms have81% lower debt capacity (0.07 versus 0.37) than SEO rms.In fact, the 75th percentile of debt capacity of PIPE issuersis lower than the 25th percentile of debt capacity of SEO

0.0036 0.0449 0.2508 1.0000

ple period is 19952007. Panel A compares PIPE issuers with rms with

RSP/Compustat rm. Equity coordination is measured by the total Shapley

l Index of bond issues. All variables are dened in the Appendix. Columns 5

ote statistical signicance (nnnpo0:01, nnpo0:05, and npo0:1).

SEO rms Difference

in means

Standard error

of difference

2575%

(4) (5) (6)

0.1050.273 0.114nnn 0.0020.2010.506 0.147nnn 0.015

0.0000.029 0.050nnn 0.002

1.3862.708 0.714nnn 0.0197.21222.964 0.638n 0.382

0.9472.440 0.374nnn 0.078

0.0551.099 2.982nnn 0.0390.0400.704 0.304nnn 0.004

All rms Difference

in means

Standard error

of difference

2575%

0.0400.200 0.047nnn 0.0020.2091.000 0.098nnn 0.015

0.0000.034 0.044nnn 0.002

1.0992.485 0.386nnn 0.0193.35814.334 6.898nnn 0.375

0.6562.097 0.149 0.100

0.1261.200 2.798nnn 0.0390.0090.532 0.215nnn 0.003

ood.

differ

lic equ

hed co

search

g rm

ed by

rs den

I. Chakraborty, N. Gantchev / Journal of Financial Economics 108 (2013) 213230220rms. We control for debt capacity in our multivariateanalysis by including it as an additional covariate inpredicting an issuers probability of default.

In Table 4, we report results of propensity scorematching on pretreatment differences in informationasymmetry and predicted probability of default. We proxyfor information asymmetry by R&D expense ratio, analystcoverage, and trading volume. We predict default by astandard bankruptcy regression using rm-level charac-teristics, including debt capacity.

Panel A of Table 4 compares PIPE issuers with SEO rms,and Panel B uses all CRSP/Compustat rms as a controlgroup. Both Shapley value (measuring equity coordinationfrictions) and bond Herndahl Index (proxying for debtconcentration) remain signicantly different across PIPE and

Table 4Propensity score matching on information asymmetry and default likelih

The table presents results of propensity score matching on pre-issuance

sample period is 19952007. Panel A compares private investment in pub

Panel B compares PIPE rms with the average CRSP/Compustat rm. Matc

index to control for differences in information asymmetry (proxied by re

and predicted default (estimated by a standard bankruptcy regression usin

Shapley value of institutional equity holders. Debt concentration is proxi

denitions and estimation procedures are discussed in the Appendix. Sta

respectively).

Panel A: Firms with secondary equity offerings

Sample PIPE

Equity coordination Unmatched 0.160

Matched 0.160

Debt Herndahl Index Unmatched 0.560

Matched 0.560

Panel B: All CRSP/Compustat rms

Equity coordination Unmatched 0.160

Matched 0.160

Debt Herndahl Index Unmatched 0.560

Matched 0.560non-PIPE rms after propensity score matching on informa-tion asymmetry and predicted default probability. If thecoordination channel had no independent explanatorypower, we should have observed no signicant differencesin the coordination measures of PIPE and non-PIPE rms.The evidence in Table 4 suggests that the coordinationmechanism is distinct from the information asymmetryand limited market access explanations of private equityissuance.

The univariate analysis so far supports the coordina-tion hypothesis. Relative to their SEO counterparts, pri-vate placement rms have 51% lower coordination amongincumbent equity holders as well as 33% more concen-trated debtholders. Notably, both Shapley value and bondHerndahl Index remain signicantly different betweenPIPE and non-PIPE rms matched on information asym-metry and predicted default probability.

3.2. The coordination hypothesis

The literature has examined the choice of privateequity issuance in terms of information asymmetry andaccess to public markets. We propose a new role forprivate equity placements as a mechanism to reducecoordination frictions among incumbent equity holdersin their negotiations with debtholders. In this subsection,we establish that the coordination channel plays animportant role in explaining the choice to issue privateequity, even after taking into account the informationasymmetry and market access hypotheses.

Fig. 1 plots the incidence of private placements as afunction of the (lagged) Shapley value of equity holders.The plot establishes a clear negative relation betweenShapley value and the number of private equity issues. Infact, most PIPEs are issued by rms with low Shapleyvalues reecting the low ability of their incumbent equityholders to coordinate. These rms gain the most byattracting a new PIPE investor with a concentrated stake.

ences in information asymmetry and predicted probability of default. The

ity (PIPE) issuers only with rms with secondary equity offerings (SEOs).

mparisons of PIPE issuers to non-PIPE rms use a single propensity score

and development expense ratio, analyst coverage, and trading volume)

characteristics). Incumbent equity coordination is measured by the total

the Herndahl Index of par values of outstanding bond issues. Variable

ote standard statistical signicance (nnnpo0:01, nnpo0:05, and npo0:1,

No PIPE Difference Standard error

0.226 0.066nnn 0.0070.232 0.072nnn 0.0100.478 0.082nnn 0.032

0.464 0.096nnn 0.035

0.223 0.063nnn 0.0050.211 0.051nnn 0.0050.459 0.101nnn 0.026

0.502 0.058n 0.026In Fig. 2, we plot the average change in a rms Shapleyvalue after PIPE issuance as a percentage of the startingShapley value one quarter before issuance. Here, thex-axis is the mean Shapley value one quarter before PIPEissuance, and the y-axis is the mean percentage change inShapley value after issuance. We observe that the changein Shapley value is positive. As expected, rms with thelowest Shapley values before issuance see the largestpercentage increase in their Shapley values after issuance.

We also study the evolution of Shapley value followinga private placement. The mean percentage increase inShapley value from the quarter before to the quarter afterissuance is 48.92%, with little variation in the subsequentquarters. The increase in Shapley value is very persistentin the year after issuance, implying that PIPE investors arenot just short-term liquidity providers.

Table 5 continues our tests of the coordination hypoth-esis in a multivariate setting. We estimate the probabilityof PIPE issuance as a function of our measures for equitycoordination frictions and debt concentration. Column 1uses information asymmetry proxies (R&D ratio, analystcoverage, and trading volume) as well as predicted defaultprobability estimated by a standard bankruptcy

I. Chakraborty, N. Gantchev / Journal of Financial Economics 108 (2013) 213230 221Fig. 1. Private investment in public equity (PIPE) issues and incumbentregression on rm-level characteristics (as in Column 1 ofTable 8). Both information asymmetry and distress areimportant in determining the choice to issue privateequity conrming ndings in the previous literature.

Columns 2 and 3 report results of multivariate logisticregressions predicting PIPE issuance with our coordina-tion proxies: Shapley value (equity coordination frictions)and bond Herndahl Index (debt concentration). Column 2restricts the comparison sample to SEO rms, and Column 3uses all CRSP/Compustat rms. All explanatory variables arelagged by one quarter relative to the occurrence of a privateplacement. We include year xed effects and cluster stan-dard errors by rm.

The results in Table 5 demonstrate that rms withlower Shapley values of equity owners and higher debtconcentration are more likely to use private equity

equity coordination. This gure plots the incidence of PIPE issues with

respect to the coordination ability of incumbent equity holders, mea-

sured by the Shapley value of current institutional owners. The Shapley

value captures the probability that a blockholders vote will be pivotal

in reaching a majority decision. The plot excludes PIPE issuers with

Shapley values lower than 1%.

Fig. 2. Change in Shapley value following PIPE issuance. This gureshows the average change in a rms Shapley value following a PIPE

issue. The Shapley value measures the coordination ability of a rms

incumbent equity holders in reaching a majority decision. The reference

quarter is the quarter before the private placement. The x-axis is the

mean Shapley value before PIPE issuance; the y-axis is the mean

percentage change in Shapley value after PIPE issuance.placements. The regressions in Columns 2 and 3 producesimilar results: Equity Coordination is signicant at 1%,and Debt Herndahl Index is signicant at 5% in the SEOsample and 1% in the full sample. A low Shapley value ofcurrent shareholders suggests larger coordination benetsfrom adding a new PIPE investor. A higher bond concen-tration increases the need to improve the coordinationability of incumbent equity by PIPE issuance. Conse-quently, a private placement shifts the balance of powerbetween shareholders and debtholders in PIPE rms.

Both coordination proxies also have very high eco-nomic signicance. Based on the restricted comparison ofPIPE issuers with SEO rms (Column 2), a one standarddeviation increase in equity Shapley value decreases theodds of a private placement by 30% and a one standarddeviation increase in debt concentration increases theseodds by 33%. Of the other explanatory variables, onlypredicted default probability has high economic signi-cance. Some of the information asymmetry variables losestatistical signicance after the inclusion of the coordina-tion proxies.

Fig. 3 plots the estimated probability of PIPE nancingconditional on the issuers lagged Shapley value (as inColumn 2 of Table 5). We focus on rms with Shapleyvalues less than 25%, which represent about 90% of thetotal PIPE sample. We observe again a denite negativerelation between the probability of PIPE issuance andthe coordination ability of incumbent equity holders.Intuitively, the higher the need to reduce coordinationcosts, the bigger the benets from bringing in a PIPEblockholder.

To differentiate the coordination channel from alter-native hypotheses of private equity issuance, we performpropensity score matched logistic regressions, in whichwe rst match PIPE rms to non-PIPE rms based onpretreatment differences in information asymmetryand predicted default probability (controlling also foraccess to public debt markets). Our propensity scoreanalysis corrects for selection bias in terms of observablecharacteristics.

Columns 4 and 5 of Table 5 present our propensityscore matched regression results. Both Shapley value andbond Herndahl Index have the expected signs and arehighly statistically and economically signicant. Based onthe restricted comparison of PIPE issuers with SEO rms(Column 4), a one standard deviation increase in Shapleyvalue decreases the odds of a private placement by 38%and a one standard deviation increase in bond concentra-tion increases these odds by 20%. None of the informationasymmetry or default variables is signicant, implyingthat the matching procedure is successful in minimizingthe differences between our treatment and control groupsin terms of these characteristics.

Our previous analysis has not addressed the role ofmanagement in the decision to issue private equity.Dispersed equity holders could lack the ability to cometo an agreement, in which case a private placement isimpossible without the help of management. In Section 5,we examine the role of managerial ownership and incen-tives in PIPE issuance using the delta and vega of ChiefExecutive Ofcer (CEO) stock and option holdings. We

ction

2007.

ms. In

dahl I

analys

in th

e log

. Stan

sions

I. Chakraborty, N. Gantchev / Journal of Financial Economics 108 (2013) 213230222Table 5Probability of PIPE issuance as a function of coordination proxies.

The table presents estimation of the probability of PIPE issuance as a fun

information asymmetry and predicted default. The sample period is 1995

equity offerings (SEOs), and Columns 3 and 5 use all CRSP/Compustat r

institutional equity holders. Debt concentration is proxied by the Hern

information asymmetry proxies [research and development (R&D) ratio,

bankruptcy regression using rm characteristics). All variables are dened

a rm issues private equity). Columns 4 and 5 present propensity scor

information asymmetry, predicted default, and access to public markets

denote signicance levels (nnnpo0:01, nnpo0:05, and npo0:1).

Logistic regresnd that managerial incentives have a positive effect onthe probability of PIPE issuance in distressed rms andconrm our results in this setting (see Table 11 fordetails).

To establish a clear causal relation between a rmsequity holder coordination and PIPE issuance, we alsoconduct an instrumental variables analysis. In the spirit ofLaeven and Levine (2009), we use the initial year (i.e.,1995) mean industry Shapley value as an instrument forthe rm-specic Shapley value of an issuer. Using theinitial year industry Shapley value also mitigates anysimultaneity concerns. We verify that the mean industry

Independent variables All rms SEO rms

(1) (2)

Equity coordination 1.744nnn(0.636)

Debt Herndahl Index 0.856nn

(0.636)

R&D ratio 5.744nnn 0.398

(0.850) (2.777)

Analyst coverage 0.516nnn 0.521nnn(0.040) (0.133)

Trading volume 0.006nnn 0.010nn

(0.001) (0.005)

Predicted default 11.666nnn 6.078nnn

(0.578) (0.846)

Amount equity raised 0.239

(0.188)

McFaddens R2 14.56% 15.26%

Number of observations 41,143 6,834

Fig. 3. Probability of PIPE issuance as a function of Shapley value. Thisgure plots the estimated probability of PIPE issuance as a function of a

rms lagged Shapley value. The Shapley value measures the coordina-

tion ability of a rms incumbent equity holders in reaching a majority

decision. 95% CI refers to the 5th and 95th condence intervals. Excluded

are rms with Shapley values exceeding 25% (top 10% of the PIPE

sample).Shapley value is a valid instrument. Intuitively, it iscorrelated with a rms Shapley value but is unlikely todirectly affect the propensity of PIPE nancing except

of equity coordination and debt concentration, controlling for an issuers

Columns 2 and 4 restrict the comparison sample to rms with secondary

cumbent equity coordination is measured by the total Shapley value of

ndex of par values of outstanding bond issues. Column 1 includes only

t coverage, and volume] and predicted default (estimated by a standard

e Appendix. Columns 2 and 3 report results of logistic regressions (Y1 ifistic estimation in which PIPE issuers are rst matched to controls on

dard errors are clustered by rm. Year xed effects are included. Stars

Propensity score matching

All rms SEO rms All rms

(3) (4) (5)

1.227nn 2.111nnn 1.569nn(0.614) (0.704) (0.661)

0.953nnn 0.667n 0.634n

(0.348) (0.380) (0.373)

2.380 1.357 1.039(2.926) (1.967) (1.753)

0.493nnn 0.015 0.101(0.124) (0.155) (0.142)

0.014nnn 0.008 0.004

(0.004) (0.05) (0.004)

6.594nnn 1.053 1.426

(0.769) (0.978) (0.823)

0.087(0.180)

17.37% 4.23% 3.83%

15,877 699 1,129through its indirect effect on an individual rms Shapleyvalue. Thus, both the inclusion and exclusion restrictionsare satised by this instrument.

In Table 6, we replicate Table 5 using the meanindustry Shapley value at the start of the sample periodas an instrument. The high statistical and economicsignicance of Shapley value conrms the causal effectof equity coordination on the probability of PIPE issuance.

Our multivariate results provide strong evidencethat the coordination mechanism is an important deter-minant of a rms choice to issue equity in a privateplacement. The propensity score matched estimationsuggests that the coordination channel is distinct fromthe alternative hypotheses discussed in the literature. Inaddition, our instrumental variables analysis conrmsthat unobservable rm attributes do not seem to bedriving our results.

3.3. The PIPE discount: division of coordination gains

How do current owners and new PIPE investors sharethe surplus realized by coordination improvement? Inthis subsection, we analyze whether coordination proxiescan help explain the observed variation in the discounts,at which private equity is issued.

In addition to providing further evidence in support ofthe coordination hypothesis, our study of the PIPE discounthelps address potential self-selection concerns in terms ofunobservable rm characteristics not controlled for in our

dinati

proba

t the

5) me

y the

cludes

a stan

1 if a

on asy

rd sig

ions

s

n

I. Chakraborty, N. Gantchev / Journal of Financial Economics 108 (2013) 213230 223propensity score estimation. The discount is determined bya bargaining game between old and new owners, whichtakes into account unobservable rm attributes. The higher

Table 6Instrumental variables (IV) approach: PIPE issuance as a function of coor

The table presents an instrumental variables (IV) estimation of the

concentration. The sample period is 19952007. Columns 2 and 4 restric

Columns 3 and 5 use all CRSP/Compustat rms. The initial year (i.e., 199

Shapley value of an issuer. Incumbent equity coordination is measured b

proxied by the Herndahl Index of outstanding bond issues. Column 1 in

ratio, analyst coverage, and volume] and predicted default (estimated by

dened in the Appendix. Columns 2 and 3 report probit regressions (Yestimation in which PIPE issuers are rst matched to controls on informati

are clustered by rm. Year xed effects are included. Stars denote standa

Probit regress

Independent variables All rms SEO rm

(1) (2)

Equity coordination 5.052nn(0.459)

Debt Herndahl Index 0.131(0.238)

R&D ratio 3.149nnn 1.636(0.406) (1.694)

Analyst coverage 0.240nnn 0.001(0.018) (0.082)

Trading volume 0.003nnn 0.002

(0.001) (0.003)

Predicted default 5.230nnn 1.874

(0.265) (1.528)

Amount equity raised 0.041

(0.277)

Pseudo R2 20.25%

Number of observations 41,143 6,969the benets incumbent equity holders expect from PIPEissuance, the larger the discount they are willing to offer tonew PIPE investors. If improved equity holder coordinationis one of the potential benets of a private placement, thenthe coordination proxies should explain (some of) thevariation in the PIPE discount.

Panel A of Table 7 presents a univariate comparison ofPIPE discounts between rms with high and low levels ofthe coordination proxies. We divide the sample of PIPEissuers into those with below- and above-median Shapleyvalue and bond Herndahl Index, respectively. We thenreport t-tests for differences in the mean discounts forlow and high values of each measure. This analysisfocuses on the most distressed rms (i.e., rms in thelowest tercile of Altmans Z-score).

Panel A of Table 7 shows that the PIPE discount variessignicantly with our measure for equity coordination.PIPE issuers with low coordination ability of currentblockholders (low Shapley values) issue equity at higherdiscounts. The difference in the mean discount betweenthe samples with low and high equity coordination isabout 5%. Given that the mean discount of the lowShapley value group is approximately 18.4%, this impliesa 27% reduction in the discount between the two groups.In addition, we see a 52% increase in the discount whengoing from the sample with low debt concentration to thegroup with high debt concentration. This suggests thathigh bond concentration increases the benets fromimproved equity coordination.Panel B of Table 7 reports the results of four multivariateregressions estimating the average PIPE discount as afunction of equity coordination and debt concentration.

on proxies.

bility of PIPE issuance as a function of equity coordination and debt

comparison sample to rms with secondary equity offerings (SEOs), and

an industry Shapley value is used as an instrument for the rm-specic

total Shapley value of institutional equity holders. Debt concentration is

only information asymmetry proxies [research and development (R&D)

dard bankruptcy regression using rm characteristics). All variables are

rm issues private equity). Columns 4 and 5 present propensity score

mmetry, predicted default, and access to public markets. Standard errors

nicance levels (nnnpo0:01, nnpo0:05, and npo0:1).

Propensity score matching

All rms SEO rms All rms

(3) (4) (5)

5.035nn 4.774nnn 4.740nn(1.357) (0.771) (1.908)

0.003 0.645 0.562(0.169) (1.037) (0.773)

0.348 2.672nn 1.103(1.799) (1.304) (2.429)

0.035 0.017 0.067(0.038) (0.086) (0.109)

0.006nnn 0.001 0.0003(0.002) (0.004) (0.004)

3.053nnn 0.544 0.430

(0.924) (0.876) (0.762)

0.246nn(0.112)

16,567 703 1,137We cluster standard errors by rm and include year xedeffects. The base regression in Column 1 demonstrates thatShapley value is signicantly negatively correlated with theobserved discount. A one standard deviation increase inShapley value decreases the offering discount by 14%. Interms of the average dollar discount of $54.70 million, thisrepresents $7.66 million lower discount.

Column 2 includes an indicator for high debt concen-tration as an explanatory variable, which is statisticallyand economically signicant. Firms with above-mediandebt concentration have 6% higher discounts comparedwith rms with below-median debt concentration.

Columns 3 and 4 add controls that could impact thedetermination of the PIPE discount. We include an indicatorfor rm distress, which identies distressed issuers as thosewith below-median Altmans Z-score. We also control forthe amount of capital proceeds as proportion of current rmvalue and the percentage of an offering covered by warrants.Shapley value retains its strong statistical and economicsignicance. The magnitude of the effect is also robust tothese additional controls. A one standard deviation increasein Shapley value decreases the PIPE discount by 13% (basedon the regression in Column 4).

Our examination of the PIPE discount offers furthersupport for the coordination hypothesis. We nd thatShapley value as a measure of equity coordination helpsexplain the observed variation in the offering discount. Inaddition, we provide evidence that high debtholder con-centration increases the PIPE discount.

Table 7Estimation of the PIPE discount to market price.

The table presents analysis of the PIPE discount to pre-issue market price.

discount between rms with above- and below-median levels of Shapley va

(measuring debt concentration). Panel B presents ordinary least squares (OLS) r

PIPE discount. High Debt Herndahl Index is an indicator for above-median bond

signi

Stan

3

I. Chakraborty, N. Gantchev / Journal of Financial Economics 108 (2013) 2132302244. Firm default

4.1. Reduced default probability

The coordination hypothesis claims that private place-

errors are clustered by rm. Year xed effects are included. Stars denote

Panel A: Variation of PIPE discount with coordination proxies

Mean

Low equity coordination 18.356

High equity coordination 13.391

Low Debt Herndahl Index 10.792

High Debt Herndahl Index 16.379

Panel B: OLS regressions of PIPE discount

Equity coordination 11.309nnn(1.634)

High Debt Herndahl Index

Distressed rm

Proceeds to market value

Warrant coverage (percent)

McFaddens R2 10.87%

Number of observations 1,971ments improve the coordination ability of equity holdersand facilitate negotiations with debtholders. If this is thecase, we expect a reduced post-issuance default likelihoodof PIPE issuers compared with matched non-PIPE rms,even after controlling for their information asymmetryand default probability. To test this hypothesis, we usedata on both bankruptcies and bond defaults in the period19952007 from Mergent FISD. Data availability restrictsour analysis to rms with FISD bond data.

Table 8 reports ve bankruptcy prediction models. Theyestimate multivariate logistic regressions of default, inwhich the dependent variable equals one if a rm experi-ences default or bankruptcy and zero otherwise. We includestandard covariates used in bankruptcy prediction: the logof total rm assets, EBITDA ratio (EBITDA/Assets), bookleverage, Altmans Z-score, and debt capacity (predictedaccess to public bond markets). All independent variablesare lagged by one quarter. We use the tercile ranks of arms Z-score and debt capacity (instead of their raw values)because both Z-score and debt capacity already use leveragein their estimation. We cluster standard errors by rm andinclude year xed effects.15

15 In unreported results, we choose alternative proxies for size,

protability, and leverage. We also include industry xed effects. The

results remain substantially unchanged.Column 1 conrms the standard result that a rmslikelihood of default is decreasing in its size, protability,and Z-score but increasing in book leverage and debtcapacity. All covariates are highly statistically signicant.The R2 of the model is 8.6%.

The sample period is 19952007. Panel A compares the average PIPE

lue (measuring equity coordination costs) and bond Herndahl Index

egressions in which the dependent variable is the (absolute) value of the

Herndahl Index. Distressed rm denotes below-median Z-score. Standard

cance levels (nnnpo0:01, nnpo0:05, npo0:1).

dard error Difference Standard error

0.399 4.965nnn 0.5070.313

1.155 5.586nnn 1.186

0.270

10.950nnn 11.151nnn 9.283nnn(1.623) (1.607) (1.597)

.489nnn 3.006nnn 2.432nn

(1.290) (1.204) (1.198)

3.196nnn 2.618nnn

(0.718) (0.710)

2.700n 2.091

(1.506) (1.554)

5.908nnn

(0.998)

10.86% 13.27% 15.96%

1,971 1,971 1,971Fig. 4 plots the change in predicted default probabilityfollowing PIPE issuance versus the change in the issuersequity holder coordination (measured by Shapley value).We nd that the higher the change in Shapley value, thegreater is the reduction in the default probability of theissuer. Given that the unconditional probability of defaultis about 16%, PIPE nancing reduces this probability by25%, which is highly economically signicant.

Columns 2 and 3 of Table 8 estimate instrumentallogistic regressions, in which the rst stage predicts theprobability of private equity issuance using equity coor-dination (measured by Shapley value) and debt concen-tration (proxied by bond Herndahl Index). We alsoestimate in a rst-stage regression a rms informationasymmetry by its analyst coverage, trading volume,and R&D ratio. Both rst-stage regressions use ordinaryleast squares (OLS), which ensures that our estimatesare consistent and unbiased. We include year xedeffects and cluster observations by rm in the rst-stageregressions.

Column 2 of Table 8 includes PIPE and SEO rms, andColumn 3 looks at all CRSP/Compustat rms. As additionalcontrols we use the standard bankruptcy covariates fromColumn 1 and the amount of equity raised (in a PIPE or anSEO). We control for the amount of capital infusion toensure that our conclusions are not mechanically drivenby a rms capital structure rebalancing. We nd that thepredicted PIPE covariate is signicant at 1%. Its economic

as a f

. Colu

osts) a

ession

I. Chakraborty, N. Gantchev / Journal of Financial Economics 108 (2013) 213230 225Table 8Estimation of post-issuance default.

The table reports predictive regressions estimating default probability

rm experiences bankruptcy or default). The sample period is 19952007

PIPE is instrumented by Shapley value (measuring equity coordination c

asymmetry is estimated in a rst stage ordinary least squares (OLS) regrsignicance is also very high. A one standard deviationincrease in predicted PIPE leads to a 96% decrease inthe odds of default (in Column 2). The economic signi-cance of predicted PIPE is higher than any of the otherexplanatory variables including Z-score. A one standard

volume. See the Appendix for variable denitions. Year xed effects and clust

Columns 2 and 4 restrict the comparison sample to rms with secondary equi

Columns 4 and 5 present propensity score logistic estimation in which PIPE

issuance differences in information asymmetry, predicted default, and accessnnpo0:05, and npo0:1).

(Instrumental) logistic regr

Independent variables All rms SEO rms

(1) (2)

Private placement 123.889nnn(46.234)

Log(Assets) 0.361nnn 0.565n(0.066) (0.301)

EBITDA/Assets 4.299nnn 7.888nn(1.420) (3.773)

Book leverage 0.603nnn 0.553(0.245) (1.021)

Z-score tercile 0.499nnn 0.395(0.105) (0.369)

Debt capacity tercile 0.937nnn 0.804n

(0.149) (0.482)

Amount equity raised 0.358

(0.525)

Information asymmetry 19.833

(14.114)

McFaddens R2 8.62% 9.72%

Number of observations 39,039 6,835

Fig. 4. Change in default probability versus change in Shapley valueafter PIPE issuance. This gure plots the change in predicted default

probability after PIPE issuance versus the change in the issuers Shapley

value. The Shapley value measures the coordination ability of a rms

incumbent equity holders in reaching a majority decision. Default is

estimated by a logistic regression, in which the dependent variable

equals one if a rm suffers default or bankruptcy, and the independent

variables are log of rm assets, EBITDA/assets, book leverage, Altmans

Z-score, and debt capacity. 95% CI refers to the 5th and 95th condence

intervals.unction of private investment in public equity (PIPE) issuance (Y1 if amn 1 presents a standard default regression on rm-level characteristics.

nd bond Herndahl Index (measuring debt concentration). Information

based on research and development (R&D) ratio, analyst coverage, and

ered standard errors by rm are included in the rst-stage regressions.

ty offerings (SEOs), and Columns 3 and 5 use all CRSP/Compustat rms.

issuers are rst matched to either SEO rms or all rms based on pre-

to public markets. Stars denote standard signicance levels (nnnpo0:01,

essions Propensity score matching

All rms SEO rms All rms

(3) (4) (5)

105.563nnn 78.204nnn 52.316nnn(32.267) (31.443) (21.605)

0.358nnn 0.215 0.058(0.138) (0.225) (0.119)

8.418nnn 2.128 3.808(2.741) (3.649) (2.875)

0.111 0.247 0.067(0.537) (0.898) (0.555)

0.450nn 0.033 0.033(0.205) (0.356) (0.206)

0.783nnn 0.008 0.048

(0.303) (0.406) (0.292)

0.081

(0.572)

19.474n 12.595 10.149(10.485) (12.511) (8.127)

8.97% 4.41% 1.81%

16,144 794 1,822deviation increase in Z-score reduces the odds of defaultby only 25%.

Columns 4 and 5 present propensity score matchedlogistic estimation in which PIPE rms are rst matchedto either SEO rms or all rms based on pre-issuancedifferences in information asymmetry, predicted defaultprobability, and access to public markets. Predicted PIPEissuance remains statistically and economically signi-cant. When compared with our main control group ofSEO rms, the odds of default decrease by 48% for aone standard deviation increase in the likelihood of PIPEissuance. Also, none of the information asymmetry ordefault variables is signicant, implying that the matchingprocedure is successful in minimizing the differencesbetween our treatment and control groups with respectto the aforementioned characteristics.

The results of the default prediction regressionsdemonstrate that private equity issuance plays an impor-tant role in reducing the probability of default of PIPErms, even after controlling for their information asym-metry and default probability. In the next subsection, weexamine whether private placements facilitate debt rene-gotiations in practice and provide evidence that improvedequity coordination post-issuance raises a rms likelihoodof favorable debt modications.

4.2. Bank debt and loan renegotiation

As discussed in Franks and Sussman (2005), concen-trated bank debt creates a trade-off. On the one hand, it

Table 9Probability of PIPE issuance (controlling for bank debt).

The table presents estimation of the probability of PIPE issuance as a function

information asymmetry, predicted default, and proportion of bank debt. The sample

with secondary equity offerings (SEOs), and Columns 46 use all CRSP/Compustat

ahl In

ebt ba

e top t

tars de

1

5

I. Chakraborty, N. Gantchev / Journal of Financial Economics 108 (2013) 213230226could reduce the lenders incentives to restructure a

of institutional equity holders. Debt concentration is proxied by the Hernd

calculated as the proportion of bank debt to the sum of bank and bond d

Thomson Reuters LPCs Dealscan. High ratio of bank debt indicates rms in th

in the Appendix. Standard errors are clustered by rm. Year xed effects. S

Independent variables SEO rms

(1) (2)

Equity coordination 1.716nnn 1.660nnn(0.634) (0.625)

Debt Herndahl Index 0.843nn 1.042nnn

(0.353) (0.364)

R&D ratio 0.425 0.653(2.780) (2.582)

Analyst coverage 0.522nnn 0.456nnn(0.133) (0.134)

Trading volume 0.010nn 0.007

(0.005) (0.005)

Predicted default 6.074nnn 5.967nnn

(0.845) (0.832)

Amount equity raised 0.237 0.321n

(0.188) (0.184)

High ratio of bank debt 2.380nnn(0.702)

High ratio of bank debt

Equity coordinationMcFaddens R2 15.28% 17.22%

Number of observations 6,793 6,771distressed rm (so-called lazy banking hypothesis). Onthe other hand, more concentrated bank debt makes thelenders vulnerable to strategic renegotiation by the rm(known as the soft banking hypothesis). We nd resultsconsistent with both hypotheses.

To study the effect of private debt on PIPE issuance, wefocus on the ratio of private debt to the sum of private andpublic debt.16 We create an indicator variable High ratio ofbank debt equal to one if a rm is in the top tercile interms of its bank debt ratio. Our sample consists of theintersection of rms in Mergent FISD and LPCs Dealscan.Table 9 presents our results.

High bank debt is negatively correlated withPIPE issuance. This is consistent with the lazy bankinghypothesis, i.e., the seniority of bank debt makes a lenderless likely to work with a distressed rm, whichreduces the coordination benets of PIPE issuance. How-ever, the interaction effect between high bank debtand incumbent equity coordination is also negativeand highly statistically signicant. We interpret this resultas evidence that a high ratio of bank debt (i.e., highlikelihood of strategic debt renegotiation with lendersaccording to the soft banking hypothesis) increasesthe benets of improving equity coordination. Theresults below conrm that PIPE issuance leads to

16 Most private debt is syndicated. A lead arranger bank has

substantial decision power even though it could be providing only a

fraction of the total nancing. Consequently, private debt can be viewed

as very concentrated.favorable debt renegotiations within one year of a private

of equity coordination and debt concentration, controlling for an issuers

period is 19952007. Columns 13 restrict the comparison sample to rms

rms. Incumbent equity coordination is measured by the total Shapley value

dex of par values of outstanding bond issues. A rms ratio of bank debt is

sed on the intersection of Mergent Fixed Income Securities Database and

ercile of all rms in terms of their bank debt ratios. All variables are dened

note signicance levels (nnnpo0:01, nnpo0:05, and npo0:1).

All rms

(3) (4) (5) (6)

1.642nnn 1.212nn 1.131n 1.120n(0.625) (0.613) (0.609) (0.614)

.046nnn 0.940nnn 1.195nnn 1.297nnn

(0.364) (0.349) (0.367) (0.350)

0.634 2.418 1.492 1.312(2.582) (2.933) (2.626) (2.544)

0.456nnn 0.496nnn 0.443nnn 0.350nnn(0.134) (0.124) (0.124) (0.117)

0.007 0.014nnn 0.011nnn 0.004nnn

(0.005) (0.004) (0.004) (0.001)

.960nnn 6.581nnn 6.423nnn 6.595nnn

(0.832) (0.769) (0.753) (0.743)

0.325n

(0.184)

2.470nnn 2.432nnn 2.849nnn(0.703) (0.700) (0.870)

3.634n 6.920nnn(2.001) (2.675)

17.24% 17.40% 19.16% 19.01%

6,771 15,810 15,767 15,783placement.To provide direct evidence that PIPE rms improve their

nancial health post-issuance, we examine whether privateplacements facilitate debt renegotiation in practice. Weexpect that the improved coordination of equity holdersfollowing a private placement manifests itself in betterrenegotiation outcomes. Specically, we compare the debtmodications of PIPE and control rms matched on size(total assets), equity coordination (Shapley value), and dis-tress level (Altmans Z-score) and look for favorable outcomesin terms of maturity, interest spread, and debt amount.

Unlike bond agreements that have a very low fre-quency of renegotiations, about three-quarters of bankloans are modied before their maturity (see Roberts andSu, 2009). We use bank loan data from Thomson ReutersLPCs Dealscan for the period 19952008. We include onlyfacilities whose terms are modied within a year of PIPEissuance. For rms with multiple PIPE issues, we consideronly their rst private placement.

As discussed in Roberts and Su (2009), the mostcommon renegotiation outcomes are maturity extensions(57% of their sample), amount increases (56%), andchanges in spreads (55%). Following their approach, weclassify favorable loan modications as those that resultin a decrease in interest spread without a decrease inloan amount or an increase in loan amount without anincrease in spread. Unfavorable loan modications arethose that result in an interest spread increase without anincrease in loan amount or a decrease in amount withouta decrease in spread.

uers a

2007

r of th

IPE

and a

lassi

rando

I. Chakraborty, N. Gantchev / Journal of Financial Economics 108 (2013) 213230 227Table 10Loan renegotiations of PIPE issuers and matched control rms.

The table reports a comparison of loan renegotiations between PIPE iss

value), and distress level (Altmans Z-score). The sample period is 1995

Dealscan and include all facilities whose terms are modied within a yea

with multiple PIPE issues. Panel A presents univariate comparisons of P

Column 3 reports the difference in loan characteristics of PIPE rms before

after issuance and matched non-PIPE rms. Favorable amendments are c

difference-in-differences analysis, in which control rms are assigned annpo0:05, and npo0:1).

Panel A: Univariate comparisons of loan renegotiations

PIPE rms

After issuance Before issuance

Variable (1) (2)

Number of amendments 3.279 4.218

Favorable amendments 0.457 0.358

Decreased spread 0.737 0.623

Increased amount 0.308 0.298

Increased maturity 0.254 0.273

Panel B: Difference-in-differences PIPE versus matched controls

Before issuance

PIPE rms Control rms

Variable (1) (2)

Number of amendments 4.218 4.678

Favorable amendments 0.358 0.356

Decreased spread 0.623 0.715

Increased amount 0.298 0.248

Increased maturity 0.273 0.322Panel A of Table 10 presents results of t-tests fordifferences in means between PIPE and matched controlrms. Column 3 reports differences in the loan terms ofPIPE rms before and after the private placement. PIPErms have fewer loan amendments post-issuance (sig-nicant at 1%). However, they are more likely to achievefavorable changes in loan terms (signicant at 5%) andreduce their interest spreads (signicant at 1%). PIPEissuers improve their likelihood of achieving a favorableoutcome by 28% within one year of issuance.

Column 5 presents a similar comparison of PIPE rmsafter issuance and matched non-PIPE rms. PIPE issuershave a one-third higher likelihood of favorable debtrenegotiations and experience statistically signicantdecreases in spreads and increases in loan amounts (bothsignicant at 5%). The univariate results in Panel A suggestthat PIPE issuance improves a rms likelihood of apositive loan modication, resulting in a lower interestspread and a larger loan principal.

Panel B of Table 10 presents a difference-in-differencesanalysis, in which control rms are assigned a random(placebo) issuance date in the sample period. Columns 1and 2 compare the average renegotiation outcomes ofPIPE and control rms in the year before issuance.Columns 3 and 4 present loan changes in the year afterissuance. Column 5 reports the difference-in-differencescomparison between PIPE and matched rms before andafter issuance. Conrming the univariate results, we ndthat PIPE issuers are much more successful in achievingfavorable revisions of interest spreads (signicant at 1%).nd control rms matched on size (assets), equity coordination (Shapley

. Data on loan amendments are obtained from Thomson Reuters LPCs

e issuance date. Only the rst private placement is considered for rms

rms (before and after issuance) and the average matched control rm.

fter the private placement. Column 5 presents a comparison of PIPE rms

ed following the approach in Roberts and Su (2009). Panel B presents a

m (placebo) issuance date. Stars denote signicance levels (nnnpo0:01,

Difference

after-before

Matched

control rms

Difference

PIPE after-Controls

(3) (4) (1)-(4)

0.939nnn (0.232) 4.637 1.358nnn (0.208)0.099nn (0.050) 0.335 0.122nnn (0.044)

0.114nnn (0.038) 0.666 0.071nn (0.033)

0.010 (0.037) 0.239 0.069nn (0.032)

0.019 (0.036) 0.289 0.035 (0.032)

After issuance Difference-in-differences

PIPE rms Control rms

(3) (4) (5)

3.279 4.597 0.858nnn (0.414)0.457 0.312 0.143nn (0.072)

0.737 0.617 0.211nnn (0.057)

0.308 0.230 0.028 (0.054)

0.254 0.255 0.048 (0.055)In terms of the 36% unconditional probability of a favor-able amendment, PIPE rms have a 40% higher probabilityof a positive outcome compared with their matched peers.

Both the univariate results and the difference-in-differences analysis suggest that PIPE issuance leads tofavorable debt renegotiations within one year of a privateplacement. These loan modications are generally asso-ciated with favorable changes in interest spreads and loanprincipals.

5. Robustness analysis

Our analysis differentiates the coordination mechanismfrom other explanations of the choice of external nancingsuch as information asymmetry and access to public mar-kets. We use propensity score matching techniques tocorrect for selection bias in terms of observable rm attri-butes. These conditioning variables come from the alterna-tive hypotheses of private equity issuance in the existingliterature. Specically, we compare each PIPE issuer with itsSEO counterparts in terms of pretreatment differences ininformation asymmetry, access to public markets, and pre-dicted default probability.

The propensity score estimation does not take intoaccount omitted variables that could inuence an investorschoice to buy the private equity of a specic rm. Conse-quently, we conduct an IV analysis, in which we use theinitial year mean industry Shapley value as an instrumentfor the rm-specic Shapley value of an issuer. The highstatistical and economic signicance of the (instrumented)

shareholder maximization].

ction of equity coordination and debt concentration, controlling for an issuers

e sample period is 19952007. Columns 13 restrict the comparison sample to

RSP/Compustat rms. Incumbent equity coordination is measured by the total

d by

on. CE

vega

of all

cluded

4

8

I. Chakraborty, N. Gantchev / Journal of Financial Economics 108 (2013) 213230228Table 11Probability of PIPE issuance [Chief Executive Ofcer (CEO) incentives for

The table presents estimation of the probability of PIPE issuance as a fun

information asymmetry, predicted default, and management incentives. Th

rms with secondary equity offerings (SEOs), and Columns 46 use all C

Shapley value of institutional equity holders. Debt concentration is proxie

and vega proxy for managements incentives for shareholder maximizati

rm (including equity and options) for a given change in stock price. CEO

High CEO delta and High CEO vega indicate that a CEO is in the top tercile

Appendix. Standard errors are clustered by rm. Year xed effects are in

Independent variables SEO rms

(1) (2)

Equity coordination 1.760nnn 1.917nnn(0.622) (0.635)

Debt Herndahl Index 0.729nn 0.687n

(0.341) (0.351)

R&D ratio 0.277 1.113

(2.604) (2.591)

Analyst coverage 0.412nnn 0.373nnn(0.127) (0.129)

Trading volume 0.008n 0.008n

(0.005) (0.005)

Predicted default 5.208nnn 4.203nnn

(0.832) (0.953)

Amount equity raised 0.151 0.062

(0.194) (0.186)

High CEO delta 1.016nnn 1.382nnn(0.272) (0.306)

Predicted defaultHigh 8.526nnnCEO delta (2.583)

High CEO vegaShapley value suggests that unobservable rm attributes donot seem to be a signicant concern in our analysis.

We also address potential issues of unobserved het-erogeneity by studying the PIPE discount. This discount isa result of a bargaining game between incumbent equityholders and new investors and takes into account rmattributes unobservable to the econometrician but obser-vable by the agents. If improved equity holder coordina-tion is not one of the potential benets of a privateplacement, then we should not be able to explain thePIPE discount by the variation in our coordination proxies.However, we nd a high correlation between the PIPEdiscount and Shapley value.

Two additional robustness tests support our conclu-sions. First, we examine cases in which a private equityissue is a part of a debt renegotiation package. In thatcase, a PIPE might be in anticipation of or a preconditionto getting the support of debtholders in avoiding default,which implies reverse causality. Using information onbond issues from Mergent FISD and loan facilities andamendments from LPCs Dealscan, we nd that a PIPE isa part of a renegotiation package in less than 2% ofall transactions. In unreported results, we reestimatethe probability of PIPE issuance as a function of equitycoordination and debt concentration, excluding PIPEswith a debt contract in the 60-day window around theprivate placement. The exclusion of these packaged PIPEdeals has virtually no effect on the estimated coefcients.We conclude that PIPE issues are not typically part of a

McFaddens R2 17.0% 18.0%

Number of observations 6,834 6,834the Herndahl Index of par values of outstanding bond issues. CEO delta

O delta is dened as the change in the value of a managers stake in the

captures the sensitivity of managerial wealth to stock return volatility.

rms in terms of managerial incentives. All variables are dened in the

. Stars denote signicance levels (nnnpo0:01, nnpo0:05, and npo0:1).

All rms

(3) (4) (5) (6)

1.922nn 1.288nnn 1.503nn 1.507nnn(0.634) (0.604) (0.624) (0.623)

0.684n 0.781nn 0.660n 0.654n

(0.352) (0.345) (0.353) (0.353)

1.107 2.517 3.498 3.493

(2.587) (2.766) (2.637) (2.632)

0.371nnn 0.342nnn 0.308nnn 0.307nnn(0.130) (0.123) (0.126) (0.126)

0.008n 0.012nnn 0.010nnn 0.010nnn

(0.005) (0.004) (0.004) (0.004)

.201nnn 5.339nnn 4.119nnn 4.114nnn

(0.953) (0.803) (0.898) (0.898)

0.059

(0.187)

1.688nnn 1.271nnn 1.680nnn 2.066nn(0.648) (0.285) (0.301) (0.621)

.536nnn 10.783nnn 10.836nnn

(2.607) (2.306) (2.315)

0.315 0.396

(0.660) (0.615)packaged renegotiation with debtholders, which miti-gates concerns of reverse causality in our estimation.

Second, we study the role of management in thedecision to issue private equity. Dispersed equity holderscould lack the ability to come to an agreement, in whichcase a private placement is impossible without the help ofmanagement. To measure managements incentives forshareholder maximization, we use the CEO incentivemeasures from Kalpathy (2009), who applies the methodin Core and Guay (2002). The incentive-alignment mea-sures are CEO delta and CEO vega. CEO delta measures thechange in the value of a managers stake (including equityand options) for a given change in stock price. CEO vegacaptures the sensitivity of managerial wealth to stockreturn volatility.17 We observe that PIPE rms have higherCEO delta and vega (signicant at 1%) than comparablerms suggesting that PIPE managers have higher incen-tives to maximize shareholder value.

In Table 11, we add controls for managerial incentivesin the regressions predicting PIPE issuance. We dene twodummy variables (High CEO delta and High CEO vega) toindicate whether a CEO is in the top tercile of therespective measure for managerial incentives. PIPE rmsare compared with SEO rms in the rst three columns

18.0% 19.7% 21.2% 21.2%

6,834 15,877 15,877 15,877

17 Due to changes in disclosure rules in 2006, we can use a rms

detailed