SYNDICATE STRUCTURE, OVERALLOCATION, AND SECONDARY MARKET OUTCOMES IN CORPORATE BOND OFFERINGS

Hendrik Bessembinder W.P. Carey School of Business, Arizona State University [email protected]

Stacey Jacobsen Cox School of Business, Southern Methodist University [email protected]

William Maxwell Cox School of Business, Southern Methodist University [email protected]

Kumar Venkataraman Cox School of Business, Southern Methodist University [email protected]

Current Draft: May 2021

------* We thank Bernard Dumas, Paul Schultz, Jonathan Sokobin and seminar participants at INSEAD, Southern Methodist University, and the 2020 Shanghai Financial Forefront Symposium for helpful comments. We also thank Andrew Karp, Sonali Thiessen, Rachel Wilson, and Larry Wolfson for helping us understand institutional aspects of the bond issuance process, as well as the Financial Industry Regulatory Authority (FINRA) for provision of the data and in particular, Alie Diagne, Elliot Levine, Ola Persson, and Jonathan Sokobin for their support of the study. FINRA screened the paper to ensure that confidential dealer identities were not revealed. None of the authors received financial support specific to this project. Bessembinder, Maxwell, and Jacobsen have no conflicts of interest to report. Venkataraman is a visiting economist at the FINRA Office of Chief Economist and acknowledges financial support for other projects.

SYNDICATE STRUCTURE, OVERALLOCATION, AND SECONDARY MARKET LIQUIDITY IN CORPORATE BOND OFFERINGS

Abstract

We study corporate bond offerings, including underwriting syndicate structure, primary placement transactions, and secondary market outcomes. Although bond offerings rarely include “Greenshoe” options, the syndicate “overallocates” many issues, thereby attaining net short positions. Overallocations are economically substantive, facilitate the syndicate’s price stabilization efforts, and are largely offset within a few days after issuance. Secondary market liquidity is better and retail participation is higher for overallocated issues, and these issues appreciate less in the aftermarket, i.e., are less underpriced, despite syndicate purchases that temporarily support the bond price. The results indicate that the syndicate overallocates issues when uncertainty regarding secondary market order flow is high and that overallocation facilitates redistribution of the offer toward retail traders. 1. Introduction Primary issuance markets, where companies raise capital from investors, are crucial to allocational efficiency in market-based economies. While dozens of research papers have studied initial and secondary offerings of common equity, issuances of corporate bonds have received less attention, despite the fact that corporations raise significantly more capital through bond than stock issuances.1 In particular, we know little about underwriters’ activities prior to the commencement of secondary market trading, or relations between underwriter decisions and secondary market outcomes. In this paper, we study the role of the underwriting syndicate throughout the issuance process, from syndicate formation, to primary allocation, and finally to secondary market trading, with a particular focus on underwriter activities intended to manage aftermarket demand uncertainty.

Underwriters typically seek to ensure that the initial secondary market price move is positive, i.e., that primary investors earn a return on their allocation, but modest in magnitude, i.e., that the cost to the issuer is not excessive. We consider the role of uncertainty regarding secondary market order imbalances following the commencement of trading. During the bookbuilding process the underwriter gains information regarding both the total primary demand for the offering (i.e., the degree of oversubscription) and the composition of the primary investor base. As Fishe (2002) argues, some primary investors are more likely to “flip” their allocations by quickly selling in the secondary markets, while others (e.g., insurance companies) are more likely to be long-term investors. Based on this information, the underwriter can make subjective forecasts regarding both the likely strength and uncertainty in secondary market order flow.2

1 Bessembinder, Spatt, and Venkataraman (2020) report that the dollar amount of debt issuances by U.S. Corporations in 2017 was nearly eight times as large as equity issuances. U.S. corporate bond issuances reached a record level of $1.92 trillion in 2020, surpassing the previous annual record of $1.9 trillion set in 2017. See https://www.ft.com/content/a59c2a9d-5e0b-4cbc-b69e-a138de76a776. 2 The composition of the primary order book in terms of long-term investors versus potential “flippers” has recently been the focus of attention in the Financial press. With regard to sovereign issues, the CEO of the French public debt and treasury management recently stated: “The relationship between what could be considered as a good transaction and the sheer size of the order book is broken. One may be confronted with a skyrocketing order book, but which is in fact of poor quality. In that case, tightening the price might be challenging or risky if you are

We report evidence supporting the reasoning that underwriters accommodate uncertainty regarding secondary market order flow by preemptively “overallocating” issues with greater uncertainty, such that the syndicate attains a net short position. While overallocation is also common in equity market offerings, the mechanics differ. Most equity offerings include a “Greenshoe” option which, if exercised, allows the syndicate to cover their short position with additional newly issued shares purchased at the offer price, with the combined effect of the overallocation and option exercise being that underwriters can elect to increase the size of the offering. The Greenshoe option affords the syndicate flexibility to respond to market conditions. If order flow is strong, the syndicate need not make secondary market purchases and can exercise the Greenshoe option to cover the overallocation. If order flow is weak the syndicate can stabalize prices through direct secondary market purchases and need not exercise the

Greenshoe option. In contrast, corporate bond offerings rarely include Greenshoe options, implying that overallocation results in a naked short position and effectively commits the syndicate to secondary market purchases, whether prices are weak or not. Since secondary market prices are on average greater than the offer price, i.e., bonds are underpriced, these short-covering transactions are potentially costly to the syndicate. We document that overallocation is nevertheless common in bond offerings, particularly for high yield issues, and assess the effects of overallocation on syndicate profits and secondary market outcomes.

Our sample includes 5,573 bond issuances during the period March 2010 (when FINRA began to collect information on primary market allocations) through March 2018. For each issue, we merge data on bond characteristics from the Mergent Fixed Income Securities Database (FISD), syndicate structure and underwriting fees from the Securities Data Company (SDC), and primary and secondary market transactions from the Trade Reporting and Compliance Engine (TRACE). In addition to the data

targeting a transaction of [a] certain size.” See https://www.wsj.com/articles/hedge-funds-face-backlash-from- europe-in-bond-market-11620639114.

4 contained in the academic version of FINRA’s TRACE dataset, including masked dealer identifiers, uncapped secondary market transaction sizes, and the primary placement transactions associated with each bond issue, we obtain from FINRA additional information to link the syndicate members identified by the SDC database to individual primary and secondary market transactions completed by thirty-four prominent dealer firms.

We find evidence that the underwriting syndicate is structured in anticipation of deal complexity and uncertainty. Large deals and those with multiple tranches are more widely distributed across bookrunners. However, issue characteristics and market conditions associated with greater uncertainty result in a more concentrated syndicate. In particular, issues associated with first-time issuers and issuers that have not issued in the last two years are associated with fewer bookrunners. Bonds issued when VIX levels and the dispersion of first-day returns for related issues in the prior week are high and when recent bond index returns are low are also associated with fewer bookrunners.

We provide evidence that the syndicate preemptively overallocates issues when uncertainty regarding aftermarket order flow is high, implying a higher probability that temporary price support will be required. We measure the extent of overallocation by comparing the sum of the primary placement quantities to the issue amount. We focus in particular on issues overallocated by at least two percent of the issue size, documenting that over three-quarters of high yield issues, for which the long-term investor base is narrower, have overallocation that exceeds this threshold, compared to less than one-third of the less risky investment grade issues. Large issues are more likely to be overallocated than small issues.

Importantly, the decision to overallocate is related to market conditions that may exacerbate aftermarket demand uncertainty. Issues are more likely to be overallocated when the dispersion in issuance day prices for related offerings, VIX levels, and changes in VIX are high, and when recent bond index returns are low.

Among issues overallocated by the threshold of two percent, overallocation amounts are economically substantive -- the median overallocation is $20.0 million for investment grade and $26.3 for high yield issues. On average, the overallocation represents more than one-third of the issues’ secondary

5 market trading volume over the first two trading days, indicating the short covering transactions are likely to affect secondary market prices, i.e., to be effective for price stabilization purposes. We document that syndicate members are aggressive net purchasers in the aftermarket for overallocated issues, as the mean net purchase by the syndicate exceeds $20 million in the first two days of trading. The short position is quickly offset; more than 70% of the short positions are fully offset within two days, indicating price stabilization activities begin immediately following the commencement of trading and are short-lived.

These results indicate that corporate bond underwriters anticipate price stabilization needs at the time of the primary allocation decision, then directly participate in the secondary market with purchases to cover short positions.

We also contribute to the literature that studies corporate bond underpricing. Following the methodology in Ruud (1993), we find that the first-day return distribution is largely censored at zero for issues with large overallocation, indicating underwriters are typically successful in deterring temporary order imbalances from driving secondary market prices below the offer price. Importantly, we document that overallocated issues tend to appreciate less, i.e., are less underpriced than other offers, despite the fact that the issuing syndicate makes aggressive secondary market purchases. This outcome supports the reasoning that the syndicate overallocates issues when they anticipate weaker or more uncertain secondary market buy vs. sell order flow. We document that the syndicate incurs losses on the secondary market trades undertaken to cover the short positions generated from the overallocation. However, these trading losses are modest relative to the magnitude of underwriting commissions.

Finally, we assess the impact of the overallocation on secondary market outcomes during the first month after issuance. The period immediately after issuance is particularly important for bonds, because trading activity tends to decline markedly as bonds are absorbed into long-term portfolios. We show that bid-ask spreads are substantially narrower and retail participation is markedly higher for issues with large overallocation. Consistent with the syndicate’s ex ante expectation of a need to provide price support, institutions and other dealers are net sellers of overallocated bonds and effectively supply the bonds to syndicate members covering short positions.

6 Prior research focused on equity offerings, (e.g., Booth and Chua, 1996), indicates that the underwriting syndicate can adjust offer prices to attract a broader investor base and improve secondary- market liquidity. Our results suggest that the overallocation mechanism used in the absence of a

Greenshoe option can offer similar benefits. In addition to the price stabilizing benefits of the short covering purchases, which we show are associated with improved secondary market liquidity, the overallocation in combination with secondary trading effectively repositions a portion of the issue from primary investors who “flip” their allocations towards other, including retail, investors, thus achieving greater ownership dispersion while most often maintaining prices at or above the offer level.

2. Related Literature, Institutional Background and Testable Predictions

In this section, we provide an overview of the related literature and the corporate bond issuance process. While some issues overlap with the extensive literature on primary and secondary equity offerings (surveyed by Eckbo, Masulis and Norli, 2007), our focus is mainly on the smaller literature considering bond offerings.

2.1. The Related Literature

Our paper is related to the literature that studies bond issuance costs. Brugler, Comerton-Forde, and Martin (2019) consider the impact of secondary market transparency, while Gande, Puri, Saunders, and Walter (1997) link issuance costs to underwriter certification. Our paper is also related to studies that consider underwriter behavior and bond allocation decisions. Using NAIC data that identifies those primary investors that are also insurance companies, Nagler and Ottonello (2018), Nikolova, Wang, and

Wu (2019), and Flanagan, Kedia, and Zhou (2019) document that the syndicate is strategic in its primary market allocations to affiliated or favored investors. Auh, Kim, and Landoni (2019) document that underwriters are net purchasers of issuers’ previously issued bonds prior to the current issuance event, which they interpret as a form of strategic price support. Goldstein, Hotchkiss and Nikolova (2021) study post-offering transaction prices for institutional and retail customers, and document that non-underwriter dealers capture rents from selling recently-issued bonds at higher and more disperse prices. The literature

7 has documented that corporate bonds, like equity IPOs, are underpriced on average.3 We extend this literature by relating underpricing to the decision to overallocate the offering.

Our study complements the literature that studies underwriter activities and price stabilization in equity issuances, including Ellis, Michaely, and O’Hara (2000), who study 306 equity IPOs between 1996 and 1997, reporting that underwriters are important providers of aftermarket liquidity and engage in stabilization activities for less successful offerings. Aggarwal (2000) studies 137 equity IPOs over a three-month period in 1997, reporting that underwriters overallocate issues and cover the resulting short positions with a combination of Greenshoe option exercises and secondary market purchases. In contrast, the absence of the Greenshoe option in the bond market implies that overallotment is effectively a commitment by the syndicate to make short-covering purchases in the aftermarket.

2.2. Book Building and Overallocation Corporate bonds are issued more frequently than equity, and some large firms have in excess of one hundred distinct bonds outstanding. Similar to equity issues, corporate bond issues are brought to the market by a syndicate of “bookrunners” and co-managers, who have a less active role. While an investor can place orders with any syndicate member, one of the bookrunners, referred to as the “bill and deliver” agent, makes the actual allocation of bonds to initial investors.

Prior to the issuance date, the issuer and the syndicate agree on broad terms, including initial ranges for issue amounts and price, and negotiate the fees or commissions paid by the issuer to the syndicate. While an equity offering may take weeks or months to complete, the issuances process for corporate bonds is often much faster. Wang (2020) documents that over 95% of the bonds in her sample of 2,323 bond offerings from 2016 to 2018 are “drive by” offerings, where the bond is priced, allocated to

3 See Ederington (1974), Lindvall (1977), Weinstein (1978), Sorenson (1982), Fung and Rudd (1986), Wasserfallen and Wydler (1988), Datta, Isakndar-Datta and Patel (1997), Helwege and Kleiman (1998), Cai, Helwege and Warga (2007), Hale and Santos (2009), Helwege and Wang (2019), Brugler, Comerton-Forde, and Martin (2019), and Goldstein, Hotchkiss, and Nikolova (2021). Ellul and Pagano (2006) propose that underpricing reflects compensation for secondary market illiquidity.

8 investors, and secondary market trading commences all within a single day.4 The relative speed of the corporate bond issue process reflects in part that many bond issuers are already known to the market because of prior issuances, and for firms with publicly traded equity, are facilitated through shelf registration, per SEC Rule 415. While many corporate bonds are issued to the public, a significant portion are privately issued (144A) bonds, from which retail investors are largely excluded.5

On the issue date, the syndicate circulates information to potential investors regarding the issue amount and price range and solicits indications of interest. Based on the investor response, the syndicate, in consultation with the issuer, circulates a tighter price range and solicits firm investor commitments.

Wang (2020) reports that the bookbuilding process is important to price formation, as updates in the offering yield between initial pricing and final pricing are substantive, averaging about 15 basis points.

The syndicate’s goal is typically that the bond should trade slightly above the issue price in the secondary market, so that primary investors earn a positive return on their allocation, but not so far above par that the cost to the issuer is excessive.6 As in equity issuances, underwriters are implicitly understood

(though they are not contractually obligated) to assume a commitment to stabilize prices should secondary market demand fail to support trading at or above the offer price, and the syndicate can choose to

“overallocate,” i.e. to allocate a quantity of bonds at the issue price that exceeds the issue size. However, unlike equity issuances, bond offerings do not typically include a “Greenshoe” option that allows the syndicate to purchase additional bonds from the issuer at the issue price. In the absence of a Greenshoe

4 The Credit Roundtable (2015) offers additional perspective on the speed of corporate bond offerings, noting that the issue can close as soon as 15 minutes after the announcement, with an average elapsed time of one to two hours. Their report also notes that books can often close with short notice, inserting a sense of urgency into the process. 5 Private bonds need to comply with fewer regulatory requirements, but the sale is restricted to Qualified Institutional Buyers (QIBs). Many investment grade144A bonds are issued by large foreign financial and industrial firms that do not wish to meet SEC listing requirements. Some 144A bonds are issued with registration rights and become public at a later date. Han, Huang, Kalimipalli and Wang (2019) report that the public registration of 144A bonds improves market liquidity. Other firms avoid the regulatory costs and disclosure requirements and issue “144A for life” bonds. 6 Unlike corporate bonds, where initial allocations are reported on TRACE at a single price on the offering date, Green, Hollified and Schorhoff (2007) and Schultz (2012) show that the lead underwriter in municipal issues sells the bonds to syndicate members at a slight mark-up to the offering price (i.e., takedown price) and the syndicate sells the bonds to the public at a higher, re-offering price, over the next week. Schultz (2012) shows that initiation of transaction reporting by MSRB results in a sharp reduction in the dispersion of secondary market transaction prices.

9 option, the syndicate must purchase bonds from the secondary market to cover any residual short position, and will incur a trading loss if these secondary market purchase prices are higher than the issue price.

Based on discussions with issuers and underwriters, there are at least two reasons that Greenshoe options are rarely employed for corporate bond issues.7 First, there is typically less uncertainty about fundamental value as compared to equity issues. While price stabilization may be necessary for bond issues, it is more likely to be in response to temporary trading shocks or “flipping” activity, and less likely to reflect a lower long-term market valuation. Second, and perhaps more important, issuers generally have specific funding objectives, in part because rating agencies seek clarity on the magnitude of the principal and interest obligations that will result from the issue. Bond issuers may elect in some cases to

“upsize” the issue during the bookbuilding process, but this decision is at the discretion of the issuer, not

(as with Greenshoe options) at the discretion of the underwriters.8

Wang (2020) documents that the syndicate often responds to demand that is stronger than anticipated by reducing the offer yield, but rarely (only 5.5% of the observations in her sample) responds to weak demand by increasing offer yields. Further, conversations with industry participants indicate that offer sizes are rarely reduced in response to weak demand, although in some cases offers are canceled entirely.9 Fishe (2002) argues that the bookbuilding process not only provides the syndicate with information about overall demand, but also regarding demand from potential flippers versus longer-term investors. His model also implies that flipping will tend to occur more frequently in issues with weaker overall demand. The likelihood that secondary market price support will be required increases with either

7 The rarity of Greenshoe options in corporate bond issues has been noted by the financial press, e.g. see https://www.reuters.com/article/ubs-bonds/ubs-turns-to-greenshoe-for-coco-deal-pricing- idUKL5N0VR45N20150219. 8 The SDC data we employ includes fields for both the presence and the amount of a Greenshoe options (referred to there as “overallotment options”). However, this field was populated for only 31 U.S. issuers during our sample period and for only one of the 14,409 corporate issues in our initial sample (as described in Section III). This issue with a Greenshoe option was a financial firm, and thus excluded from our final sample of 5,573 issues. 9 We provide limited direct evidence on this issue. Following Wang (2020), we pulled Bloomberg announcement, guidance, and final launch pricing and size data, for a random sample of 100 tranches (25 IG and not overallocated, 25 IG and overallocated, 25 HY and not overallocated, and 25 HY and overallocated). Among those with Bloomberg data, only a single tranche (which was not in the overallocated sample) was downsized.

10 weaker forecasts of, or greater uncertainty regarding, net secondary market order flow, inclusive of the potential quantity of “flipping” activity.

This simple framework points to testable predictions. First, since the absence of a Greenshoe option implies that the syndicate will need to cover their short position with secondary market purchases, the syndicate will overallocate selectively, when it forecasts weak or uncertain secondary market net order flow. That is, while researchers cannot directly observe the quality of an offering’s order book, overallocation comprises an empirically observable indicator that the syndicate perceives weakness or uncertainty regarding secondary market order flow. Second, while issues with large overallocation may still be underpriced on average, they will be less underpriced than issues without overallocation, which reflects that uncertainty regarding net order flow and flipping activity tends to be greatest when fundamental demand is also weaker (Fishe, 2002). Third, since the syndicate enters into a short position on overallocated issues even while the offer is underpriced, overallocation will be associated with secondary market trading losses to the syndicate. However, an alternative hypothesis advanced by

Corrigan (2020) is that the syndicate overallocates issues with weak demand to obtain a short position that will profit from price declines.

Investment grade (“IG”) issuers are typically larger firms with more prior issues, and with a broad and dispersed investor base, relative to high yield (“HY) issuers. The larger pool of potential buyers for

IG issues in combination with the fact that many long-term investors (e.g. insurance companies and pension funds) are required to hold only IG issues, reduces the risk that flipping activity will push the secondary market price below the offer price. For HY bonds, in contrast, less frequent issuance activity, higher uncertainty regarding valuation, and a narrower base of potential investors leads to heightened risk that weak secondary market demand or significant flipping activity will cause downward price pressure.

If, as we hypothesize, overallocation reflects the syndicate’s perception of secondary market trading uncertainty, this reasoning predicts that overallocation will be more common for HY than IG issues.

The primary issuance process is sequential in nature. We begin by analyzing syndicate structure

(e.g., the number of syndicate firms), move to primary market outcomes such as overallocation, and end

11 with secondary market outcomes such as underwriter trading activities, trade execution costs, institutional trading, and retail participation. In a broad sense, many of the variables studied herein are endogenous.

However, the decision variables at one stage of the process (e.g., the decision to overallocate) can be considered predetermined when used as explanatory variables for subsequent outcomes (e.g. the liquidity of the secondary market).

3. Data & Sample Characteristics

3.1. Data Sources and Methods We rely on Securities Data Company (SDC) for information regarding the syndicate that oversees each bond issue, FINRA’s enhanced TRACE database for information on bond transactions, and FISD and SDC for basic issue characteristics. SDC reports the names of the bookrunners and managers in the syndicate. For most public (non 144A) issuances, SDC also contains information on the underwriting commitment to each bookrunner and manager, as well as the percentage fee or “gross spread” collected as a commission by the syndicate.

The public version of the TRACE data includes information on bond CUSIP, the date and time of execution, the transaction price and volume (in dollars of par, up to a maximum “cap” size), and whether the trade represents a customer sale or purchase, or a trade between two dealers. The enhanced TRACE data made available to academics by FINRA includes information on both publicly disseminated and non- disseminated historical transactions, including those in non-registered 144A bonds, unmasked trade sizes, masked identification numbers for individual dealers on each transaction, and the primary placement transactions associated with each bond issue (flagged as “P1” trades). In addition, FINRA provided us with information that is sufficient to link individual TRACE primary and secondary transactions for 34 prominent dealers examined by Bessembinder, Jacobsen, Maxwell, and Venkataraman (2018) to specific syndicate members identified by SDC.10 In particular, for each issue, we have information on whether

10 The thirty-four dealers are generally the most active in the corporate bond market. More specifically, they are the dealers at the intersection of the “Constant Dealer” and “Top 70%” samples described by Bessembinder, Jacobsen,

12 each of these thirty-four dealers was a syndicate member or bookrunner, and in the case of HY issues, whether the dealer was the “lead left”. The data begins March, 2010 (when FINRA began to collect information on primary market allocations) and runs through March, 2018.11

Panel A of Table I reports summary data regarding the sample. We rely on FISD to identify bond issuances by corporations. More specifically, we focus on non-puttable or convertible U.S. Corporate

Debentures and U.S. Corporate Bank Notes (bond type=CDEB or USBN) with complete issuance information (offering date, amount and maturity), resulting in an initial sample of 14,409 issuances.12 We require that the FISD data can be matched with both SDC and TRACE data, which reduces the sample to

9,958 issues. We retain issues with at least one trade in both the primary and secondary market in the year following the offer. We exclude bonds with variable interest rates and Yankee bonds (those with a non-US issuer) or that are identified as financial firms by FISD (industry group=2). We also exclude issues with any bookrunner that is not listed as a member on FINRA’s website (and thus not subject to

TRACE reporting requirements) or those issues where none of the aforementioned thirty-four large dealers participated as bookrunners, leaving 5,971 issues.13

To measure the extent to which a given issue is overallocated, we sum dealer sell quantities across all primary placement (“P1”) transactions in the enhanced TRACE data, and compare to the issue size. We exclude P1 trades between syndicate members in this computation. Our final sample filter excludes issues with overallocation greater than 115% or less than 95% of the offering amount, as these are likely indicative of substantive data errors, such as misreported trade quantities or S1 trades reported

Maxwell, and Venkataraman (2018). At least one of these thirty-four dealers is present for more than 99% of sample issues. 11 Our sample period overlaps with the implementation of the Volcker provision of the Dodd-Frank Act which restricted U.S. banks from certain speculative trading activities. However, as noted by Allahrakha, Cetina, Munyan and Watugala (2019), the Volcker rule contains an exemption for members of the underwriting group that provides greater safe-harbor on trading new issues during the “distribution” period, which they assert to not exceed 90 days. 12 Specifically, we exclude the following types of debt: retail notes, foreign government, agency, municipal, pass through trusts, pay in kind, strips, zeros, Eurobonds/Euronotes, asset and mortgage backed, insured and guaranteed by letters of credit, medium term notes/zeros, convertible, and foreign currency. 13 The list of FINRA member firms is available at: https://www.finra.org/about/firms-we-regulate.

13 as P1 trades (or vice versa). This restriction reduces our final sample to 5,573 issues.14 We are primarily interested in issues with economically substantive overallocation, so we divide the sample into two groups: those with measured overallocation greater than the full sample median of two percent of the issue size (we call these “large overallocation” issues), and other issues.

3.3. Issue Characteristics Panel B of Table I contains descriptive information regarding issue characteristics, separately for the IG and HY samples. The median issue size for the full sample is $500 million. The majority (3,611, or 65%) of the sample issues are IG. About 69% of IG issues are multiple-tranche deals and 57% of the deals, summed across tranches, are greater than $1 billion.15 By comparison, only 14% of HY issues have multiple tranches, and only 19% of HY issues are larger than $1 billion. IG issues have a longer average maturity (10 years) than HY issues (7 years). Only a minority, 23% for HY and just 4% for IG, of bond issuances in our sample are IPOs, i.e., the first offering by the issuing firm in FISD. Over eighty percent of HY issues are non-public 144A bonds, compared to just 11% of IG issues. Since IG and HY issues differ substantively in terms of issuance characteristics and underwriter practices (discussed in the next section), we report separate results for the IG and HY samples.

4. Syndicate Structure and Primary Market Outcomes

4.1. Syndicate Structure IG issuances typically include co-equal bookrunners, each of which is allocated the same amount of the issuance (including overallocation), and since compensation is typically proportional to the issue amount, each receives the same compensation.16 Each bookrunner also participates in any secondary

14 Conversations with industry participants indicate that issues are not underallocated. In a few cases, the total volume of reported P1 transactions is less than the offer amount. We therefore attributed these cases to data errors (such as a missing P1 transaction) and set the computed overallocation to zero. 15 For example, in March of 2016, Exxon issued $2.5 billion in five-year bonds, $1.25 billion in seven-year bonds, $2.5 billion in ten-year bonds, and $2.5 billion in thirty-year bonds, all on the same date. 16 The descriptions in this section are based in part on our discussions with investment bankers involved in underwriting syndicates, institutional investors who purchase bonds in the primary market, and CFOs of firms who issue corporate bonds.

14 market stabilization trading. One bank (typically rotated across deals) is selected to coordinate the process and serve as the “bill and deliver” agent.

In the case of HY issues, a single bookrunner typically controls and manages the issuance process. As in stock IPOs, this bank is referred to in HY issues as the “lead left” underwriter, and typically receives an extra five to ten percent of fees above that earned based on their share of the underwriting commitment. Potential investors typically contact only the lead bookrunner to request allocations. In HY issues the lead bookrunner serves as the “bill and deliver” agent, chooses whether and to what extent to overallocate, and engages in stabilizing trades in the secondary market.

We report descriptive statistics regarding the underwriting syndicate in Table II. The syndicate includes a median of eleven managers for IG issuances and seven managers for HY issuances. For both groups, the median number of bookrunners is four. The SDC database includes the underwriting commitment to each bookrunner for most public deals. The median percentage of the total issue size underwritten by a bookrunner is 18% for IG issues and 17% for HY issues. HY bookrunners are more often (50%) “Top 10” dealers (based on total issue volume within the sample) as compared to IG issues

(29%).17

4.2. Primary Market Allocations Table III provides descriptive statistics on primary market trades. As noted, a single dealer known as the “bill-and-deliver” agent makes the actual allocation of bonds to initial investors, and each primary trade represents the entire allocation received by a distinct investor.18 A median of 91 investors

(for both HY and IG issues) receive primary allocations. The size of the typical allocation supports that

17 The bookrunners that appear most frequently are Bank of America, J.P. Morgan, Wells Fargo and Citibank, all of whom have a large U.S. commercial banking presence. In contrast, equity IPO league tables for 2018 list the traditional investment banks Morgan Stanley and Goldman Sachs first and second, while Wells Fargo ranks only fifteenth. See https://data.bloomberglp.com/professional/sites/10/Bloomberg-Global-Equity-Capital-Markets- League-Tables-FY-2018.pdf. 18 While an investor may place orders with multiple bookrunners, it will receive its entire allocation in a single P1 trade from the “bill-and-deliver” agent. The median percent of P1 transactions completed by a single dealer is 98% for IG issues and 100% for HY issues. Under FINRA Rule 2090, a Bookrunner must meet a “Know Your Customer” rule to transact with that customer. The “bill and deliver” agent may on occasion ask other bookrunners to handle allocations for (typically smaller) investors that do not meet this requirement.

15 these are primarily institutions; median sizes of primary trades are $5.8 million (1.2% of issue size) for

HY issues and $6.7 million (1.1% of issue size) for IG issues. Some primary trades are very large, with the largest individual placements averaging $49.9 million (8.4% of the offer size) for IG offerings and

$45.0 million (10.0% of the offer size) for HY offerings.

In Panel B of Table III, we report on overallocation statistics. The median (mean) overallocation for IG bonds is $2.5 million ($10.1 million) or 0.5% (1.4%) of the issue size, reflecting that many IG issues are not meaningfully overallocated. In contrast, overallocation is more important for HY issues, where the median (mean) overallocation is $19.7 million ($27.3 million) or 4.5% (4.9%) of the issue size.

Figure 1 displays the frequency distribution of overallocation by credit quality and issue size.

We compile a number of empirical outcomes for issues that are meaningfully overallocated, which we define as overallocation that exceeds two percent (i.e., the sample median) of the issue size.

Panel C of Table III reports on the 30% of sample IG issues and 77% of sample HY issues that meet this criterion. We hereafter refer to this subsample as “issues with large overallocation.” For this subsample, the dollar overallocation is economically large. The median overallocation is $20.0 million for IG issues and $26.3 million for HY issues. We compare the overallocation amount to the average secondary market dollar trading volume on the offer date and day after the offering date. The median overallocation relative to volume on the offer date and day after is 33% and 27% for IG issues and 35% and 37% for HY issues.

We conclude that the dollar amounts involved for issues with large overallocation are sufficiently large such that short covering transactions are likely to meaningfully affect market outcomes.

We next report on tests intended to assess the extent to which issues with large overallocation differ systematically from other issues. Table IV reports median (and, in some cases mean) outcomes for each subsample, as well as p-values from Wilcoxon tests of the hypothesis that outcomes for each sample are drawn from the same distribution. The results reveal that issues with larger overallocation have a median issue size that is $50 million greater than other issues, for both HY and IG offers. The mean gross spread (i.e. the commission rate) is slightly larger for overallocated vs. other issues, though the difference is not significant for HY issues. Gross spreads are substantially greater for HY than IG issues, likely

16 reflecting the higher risk and greater uncertainty regarding HY demand.19 Issues with large overallocation are placed with more customers (93 vs. 90 primary trades for IG issues and 92 vs. 90 primary trades for IG issues). However, despite the broader distribution, the median primary trade size is greater for issues with large overallocation in both markets, and the placement of overallocated issues is more concentrated, as revealed by a higher median Herfindahl index. This final result is suggestive that certain institutional customers may be subject to a selection bias whereby they receive larger allocations in those issues where the syndicate perceives weak or uncertain demand. Overall, offering statistics vary substantively across large overallocation and other issues.

4.3. Determinants of Syndicate Structure and Primary Market Allocations

We next report on a systematic assessment of the variables that are relevant in explaining cross- deal variation in the number of bookrunners involved in each deal, the extent of overallocation relative to the issue size, and the pricing (yield spread over treasury) of the issue. Results are based on issue-level observations, and are estimated separately for IG and HY issues. We assess the effect of market conditions by including the dispersion of secondary market prices on the offering date and day after the offering date for corporate bond offerings in the previous quarter, both the average level of the VIX index over the five days preceding the offer, and the change in the VIX index on the offer day relative to the average VIX index level the prior five days, the trailing five-day average corporate bond market index return, the number of corporate bond issues over the prior five days, and the average three-month LIBOR interest rate over the prior five days. The corporate bond offering data is computed separately for IG and

HY issues. We also employ control variables for various issue characteristics, including the (natural log of) the issue size, the time to maturity, the credit rating (with a higher number for weaker ratings), and an indicator variable that identifies issues by industrial firms, as defined by FISD. We also employ indicator variables for firms with publicly listed stock, for non-public 144A bond offerings, and for issues that

19 However, these results are based on only a subset of issues. Specifically, gross spread data is available for 79% of non-overallocated and 89% of overallocated IG issues. For high yield issues, gross spread data is available for 11% of non-overallocated and 22% of overallocated issues.

17 contain multiple tranches. Finally, we also include indictor variables that identify the first issue reported in FISD for the firm, i.e., that indicates the issue to be an IPO, and that identify issuers that have not completed an offering within the past two years.

Outcomes obtained by estimating the multivariate regressions are reported on Table 5. For columns 1 and 2, the dependent variable is the number of bookrunners, for IG and HY issues, respectively. Columns 3 and 4 contain outcomes for the amount of overallocation, while columns 5 and 6 pertain to the offering interest rate spread over Treasury yields. Results in Columns 1, 2, 5, and 6, where the dependent variables are the number of bookrunners and the natural log of the offering spread over the benchmark treasury are based on an OLS regression framework. Results in Columns 2 and 3, where the dependent variable is the dollar overallocation scaled by the offering amount, are based on a Tobit regression framework. We report dependent variable averages above the regression results and standard errors estimated using the Huber-White sandwich estimator below the coefficients.

Focusing first on the number of bookrunners in the syndicate, it can be observed that issues associated with first-time issuers and issuers that have not placed bonds in the last two years are associated with fewer bookrunners. Further, bonds issued during times of greater uncertainty – when

VIX levels and the dispersion of first-day returns for issues in the prior week are high, and when bond index returns are low, are also associated with fewer bookrunners. These results indicate that issuances that are associated with greater uncertainty due to a lack of prior issuances, or those that occur during periods of higher market uncertainty, are associated with smaller syndicates. This may reflect that smaller bookrunners are less willing to participate in those issuances.

Turning to the multivariate results focused on overallocation, larger and public bond (non-144A) issues are more overallocated, for both IG and HY deals. We find the overallocation decision is related to market conditions that may exacerbate uncertainty regarding aftermarket demand. Economic uncertainty, as measured by the standard deviation of recent offer-date prices, the recent level of VIX, and the offer- date change in VIX are all positively related to overallocation for both IG and HY issues. Issues with larger syndicates tend to be more overallocated.

18 The results reported in columns (5) and (6) show that offering spreads are reliably greater for

144A issues (potentially due to fewer disclosure requirements and reduced secondary market liquidity) and for firms that have not issued within the prior two years, that do not have public stock, and (not surprisingly) that have worse credit ratings. Among market condition variables, for both IG and HY issues, offerings spreads are sensitive to the recent level of both interest rates and VIX.

It is informative to compare and contrast the variables that explain overallocation with those that explain yield spreads. IPOs are less overallocated for both IG and HY issues, as are secondary offerings where the firm has not issued a bond in the prior two years, while these same issues have higher offering yields. The offer-date change in VIX has only weak explanatory power for the offering spread, while the volatility of recent offer date prices is significant in explaining HY spreads, but not IG spreads, even while these variables have substantive explanatory power for overallocation.

We posit that 144A issues, issues that are IPOs or are the first within two years, and issues by firms without public stock are likely associated with more uncertainty regarding fundamental value, and note that these issues are associated with higher yields, but are not positively related to overallocation.

These findings support that changes in offering spreads and overallocation are not perfect substitutes; the underwriting syndicate responds to uncertainty regarding fundamental value by increasing offer yields, while responding to uncertainty regarding secondary market order flow by overallocating the issue.

5. Secondary Market Outcomes

We next turn to assessments of the relations between the syndicate’s decision to overallocate the offering and secondary market outcomes, including net syndicate purchases and changes in bond prices subsequent to the offer date, while controlling for bond characteristics and market conditions.

5.1. Overallocation and Syndicate Trading in the Secondary Market We posit that underwriters overallocate certain bond issues in anticipation of making secondary market purchases to support prices. The purchases undertaken to support the price offset the short position created by the overallocation. To assess this hypothesis, we examine relations between the

19 underwriting syndicate’s overallocation decisions and their net trading in the secondary market. To do so, we rely (as discussed in Section 3.1) on information provided by FINRA that allows us to link syndicate members to specific primary and secondary TRACE transactions completed by 34 prominent dealers. For

IG bonds we study secondary market trades for the entire underwriting syndicate, while for HY bonds we study the trades of the “lead left” bookrunner, reflecting the industry practice whereby this bookrunner is alone responsible for stabilizing HY issues. In each case, we calculate the bookrunner signed net position change as their quantity of secondary market purchases minus secondary market sales, by trading day and relative to the offer date for each issue. We also tabulate for each issue the percent of the initial overallocation (i.e. syndicate short position) that has been offset by net secondary market purchases on a daily basis, and create an indicator variable for each issue that is set to one at the date when the cumulative net secondary market position change exceeds the initial short position, i.e., when the overallocation has been fully offset.

Figure 2 reports information regarding the speed at which the overallocation is covered and the net syndicate position change from the offering date (day 1) to twenty-one days following the offering date (day 21). The shaded portion represents the 95% confidence interval. Panel A shows that the net syndicate secondary market position change is positive on the issue date as well as the following date, for both large overallocated and other issues. The fact that the syndicate is, on average, a net purchaser in the aftermarket even for issues without large overallocation reflects that net secondary market order flow demand cannot be forecast perfectly, and that some issues will trade below the offer price even if weak or uncertain demand was not anticipated. However, the quantity of syndicate net purchases is notably greater for large overallocated as compared to other issues, throughout the month after issue.

For large overallocated IG deals, the net syndicate position change averages just under $25 million over the first two trading days, and is relatively flat thereafter. For large overallocated HY issues the lead left position change is also highly positive over the first two trading days, averaging about $20 million, but continues to drift upward to approximately $35 million by ten days after the issue.

20 Panels C of Figure 2 is based on the large overallocated sample and shows the mean percent of the overallocation offset each day for the investment grade and high yield sample. For large overallocated IG issues, about 80% of the initial short position is covered, on average, by the second day of trading. For large overallocated HY issues, on average about 65% of the initial short position is covered by the second day of trading and 80% is covered by the fifth day of trading. Panel D of Figure 2 is based on the large overallocation sample and shows the percentage of issues with overallocations that are fully offset each day for the investment grade and high yield sample. Panel D of Figure 2 shows that the original short position is fully covered by the second day of trading for about 60% of large overallocated IG deals and 40% of HY deals.

On balance, the data displayed on Figure 2 supports the conclusion that the underwriting syndicate (or lead-left underwriter) supports overallocated deals with aggressive secondary market purchases. For IG issues, the initial short position is largely covered by secondary market trades on the offer date and the first subsequent date. The lead left underwriter is an aggressive purchaser in the secondary market during the first two days for overallocated HY issues as well, but the net purchasing persists for a longer period relative to overallocated IG issues.

In Figure 3 we display the frequency distribution of returns (percentage price changes relative to the offer price) for the sample of IG and HY issues with large overallocation, based on the weighted average trade price on the first, second, and fifth days of secondary market trading. The solid blue bar denotes returns that exceed zero, the patterned red bar denotes returns that are less than zero.

The most notable features of the one-day return distributions displayed in Panels A and Panel B are the very small number of issues with negative first-day returns and the large numbers of observations with returns that only slightly exceed zero. This pattern is similar to that displayed by Ruud (1993) for equity IPOs, and supports the reasoning that underwriters effectively support bond issues to mostly avoid instances where the bond trades in the secondary market at prices below the offer price. However, negative returns (relative to the offer price) are somewhat more frequent on the second day (Panels C and

21 Panel D), and notably more frequent on the fifth day (Panels E and Panel F), once the syndicate’s price- supporting purchases tend to be complete.

We next turn to a multivariate assessment of the determinants of dealer net purchasing around the commencement of secondary market trading and report the results in Table VI. Since the results in Figure

2 indicate that the majority of syndicate short covering activities occur quickly, we focus on the first two days following the issue date. Dependent variables are net signed position change for IG and HY issues in columns 1 and 2, respectively, and the percent of total volume comprised of syndicate buys from customers or other dealers for IG and HY issues in columns 3 and 4, respectively. Explanatory variables are the same deal characteristic and market condition measures as employed for the results reported in

Table 5, with three additions. These include (i) an indicator variable that identify issues with large overallocation, (ii) an indicator variable that identifies trades by underwriter dealers, and (iii) the product of the two indicators. For IG issues, the second indicator variable identifies dealers who are listed as syndicate members by SDC and are included among the 34 large dealers examined by Bessembinder et al.

(2018), as well as the dealer who engages in P1 trades (if that dealer is not already identified). For HY issues the second indicator identifies only the “lead left” dealer who engages in P1 trades. These OLS regressions include two observations per issue: the trading activity of the syndicate (signed volume summed across members) or lead left dealer, and the trading activity of all other non-syndicate dealers.

We report dependent variable means above the regression results, and we report standard errors that are clustered at the issue level. Net signed position changes in Columns 1 and 2 are based on dollars in thousands.

Coefficient estimates reported on Table VI for the indicator variable denoting that the dealer is a member of the underwriting syndicate (or is the lead left underwriter for HY offers) are uniformly positive and significant, verifying that syndicate members are on average net purchasers in the secondary market after controlling for market conditions and issue characteristics, even for issues without large overallocation. Coefficient estimates on the overallocation indicator itself are negative and significant when explaining signed position change (columns 1 and 2) and insignificant when explaining buying

22 volume as a percentage of total volume (columns 3 and 4). These results imply that non-underwriter dealers, who do not attain a short position due to an overallocation, do not on average purchase overallocated issues in the secondary market.

The key finding that can be observed on Table VI is that the coefficient on the product of the large overallocation indicator and the syndicate/lead left membership indicator is positive and statistically significant in every column. These estimates imply that syndicate members (or the lead left in the case of high yield issues) make significantly larger net secondary market purchases in issues with large overallocation over the first two trading days, after controlling for bond attributes, syndicate attributes and market conditions. These coefficient estimates are economically large, equal to $11.9 million (IG, column 1), $18.9 million (HY, column 2), 7.6% of volume (IG, column 3) and 10.9% of volume (HY column 4). These empirical estimates strongly support the reasoning that the syndicate overallocates deals with the intention that the resulting short position will be quickly covered by aggressive secondary market purchases that will also tend to support bond prices.

5.2 Overallocation and Underpricing

Prior studies (e.g., Datta, Iskandar-Datta, Patel, 1997; Cai, Helwege and Warga, 2007; Helwege and Wang, 2016; Brugler, Comerton-Ford, and Martin, 2016; Goldstein, Hotchkiss, Nikolova, 2021) show that corporate bonds trade in the secondary market at prices higher than the offer price on average, though the magnitude of such “underpricing” is less than in the case of equity IPOs. The results in our sample also indicate underpricing of bond offerings. The key innovation of our study with regard to corporate bond underpricing stems from our analysis of underwriter incentives to overallocate some issues. As noted, our analysis predicts that issues with large overallocation are associated with weaker or more uncertain expected investor demand, and as a consequence, these issues should experience smaller aftermarket price increases, i.e. less underpricing, on average.

23 Figure 4 displays the relative price paths for the first 21 days following the offering (day 1) for n

= 1 to 21 days. Relative prices are the weighted average raw price on day n scaled by the offering price.20

We report cross-sectional mean relative prices by day for issues with large overallocation and for other issues. Panel A pertains to IG issues, while Panel B pertains to HY issues. The 95% confidence interval around the mean price is displayed by the shaded area around each curve. The blue solid line represents the sample average relative prices for the large overallocation sample and the red dashed line represents the relative prices for the sample without large overallocation.

The data on Figure 4 shows that large overallocation issues are less underpriced as compared to other issues, for both IG and HY deals. However, the differential in underpricing is smaller for IG issues, and the difference in mean prices is no longer statistically significant (as confidence intervals grow with time) after about ten days of trading. For HY issues, the difference in underpricing across large overallocation and other issues is larger, equal to about fifty basis points, and is statistically significant (as the 95% confidence intervals remain quite distinct) throughout the first twenty-one trading days.

On Tables VII and VIII, we report the results of formal statistical tests pertaining to underpricing for large overallocation and other issues. To measure underpricing at various horizons, we compare the weighted average secondary market price expressed as the sum of the flat price and accrued interest n days after the offer, for n = 1 to 21 days, to the offer price. To enhance statistical power for these tests, we adjust for general bond market conditions by subtracting the cumulative bond index return over the same n days. For IG issues we subtract the Bank of America Merrill Lynch U.S. Corporate Total Return

Index, and for HY issues we subtract the Bank of America Merrill Lynch U.S. High Yield Total Return

Index. Underpricing measures have been winsorized at the 1% and 99% levels.

Table VII reports on parameters of the underpricing distribution for large overallocation and other issues, including the cross-sectional mean, standard deviation, 10th and 90th percentiles, as well as the skewness and kurtosis coefficients. We also report Wilcoxon p-values for the hypothesis that the

20 If a bond does not trade on a particular day, we retain the weighted average price from the previous trading date. The Figure obtained when we focus only on those bonds with trades is very similar.

24 underpricing outcomes large overallocation and other offers are drawn from the same distribution. The

Wilcoxon tests reject the hypothesis that the underpricing data is drawn from a common distribution in all cases, with the lone exception of the underpricing of IG issues at the 21-day horizon.

Consistent with prior studies, corporate bond offerings in our sample are underpriced on average, and HY issues are more underpriced than IG issues. The underpricing of IG bonds ranges from an average of 0.35% at the first day of trading to 0.86% at the 21st day of trading, while the underpricing of

HY bonds ranges from an average of 1.01% at the first day of trading to 1.77% at the 21st day of trading.21

The results for both IG and HY issues verify that offerings with large overallocation are associated with smaller underpricing, i.e. less secondary market appreciation, than offerings with small overallocation. Focusing on underpricing measured two days after the issue, IG offerings with large overallocation are underpriced by an average of 0.37%, versus 0.48% for issues with small overallocation.

The differential in underpricing for HY issues is larger. Again, focusing on underpricing measured two days after the issue, HY offerings with large overallocation are underpriced by an average of 0.99%, compared to 1.49% for offerings with small overallocation. By the fifth day of trading, 26% of IG and

17% of HY issues with large overallocation are trading below the offer price, whereas 23% and 10% of other issues trade below the offer price.

In Table VIII we report results that examine the determinants of underpricing in a multivariate setting. Specifically, this table reports results of OLS regressions with observed underpricing as the dependent variable, relying on the same control variables as in Tables V and VI, while including indicator variables to identify high yield issues and issues with large overallocation. Columns 1 to 4 report results when underpricing is measured based on secondary market prices on the first, second, fifth, and twenty first day of trading, respectively. The sample mean of each dependent variable is reported at the top of the table as a basis for comparison.

21 The average underpricing we report for high yield issues is somewhat larger as compared to Nikolova, Wang, and Wu (2019). However, when we follow them in excluding 144A offers and focus on a matched sample period we obtain mean underpricing of 0.85%, quite similar to the 0.74% that they report.

25 The results in Table VIII show that issues with large overallocation experience 0.22% less underpricing on the first trading day as compared to non-overallocated issues, after allowing for control variables.22 This estimate is substantive relative to the sample mean underpricing measured during the first trading day, which is 0.58%. Similar results are observed when the dependent variable is underpricing measured over horizons of 2, 5 and 21 days after the issue. In particular, the coefficient on the large overallocation indicator decreases to -0.24%, -0.25%, and -0.30%, respectively, as the horizon is increased, and each estimate is statistically significant.

These results imply that offerings with large overallocation in the primary market trade at a smaller premium to offer price in the secondary market, despite large syndicate purchases subsequent to the offer. This outcome supports the hypothesis that the underwriting syndicate chooses to overallocate those issues where they detect weak or uncertain secondary market demand. As a consequence, the price at which the market supports itself after stabilization activities are completed reflects a lower premium to offer price as compared to the issues without large overallocation where the syndicate expects secondary market demand to be more typical.

5.3. Syndicate Trading Profits

Syndicate members are net purchasers of bonds in the secondary market, particularly when issues are overallocated. Since offerings are underpriced these purchases are likely to be completed at prices higher than the offering price, implying that the syndicate may incur trading losses covering their short positions. However, in light of the evidence presented by Ellis, Michaely, and O’Hara (2000) for equity markets, it may also be anticipated that syndicate members act as market makers, from which they would be expected to earn positive trading revenue.

In Table IX, we report on dealers’ secondary market trading profits during the first twenty-one trading days after the offer. Results are reported separately for large overallocated and other issues, and

22 Among control variables, the results in Table VIII indicate that large deals, IPOs and issues by firms that have not issued during the prior two years, high yield, long maturity, low credit rating, and 144A issues are more underpriced.

26 for IG (Panels A and B) and HY (Panels C and D) issues. Results on Panel A pertain to members of the underwriting syndicate while Panel B focuses on non-syndicate dealers. Panel C pertains to the lead-left underwriter, who enters short positions on large overallocated HY issues, while Panel D reports results for all other non-syndicate dealers. For each trading profit measure, we report Wilcoxon p-values for tests of the hypothesis that the distribution of outcomes does not differ across large overallocation and other issues.

The results reported on Table IX indicate that the difference between the weighted average dealer sell price and dealer purchase price over the period (scaled by the offering price), referred to in the table as the weighted-average spread, is positive, as dealers earn gross profits from round-trip trades in the secondary market. Spreads are on average larger for HY bonds than IG bonds. For example, the mean weighted-average spread for syndicate trades in IG issues that are not overallocated is 0.103%, while the mean spread earned by the lead-left underwriter on HY issues that are not overallocated is 0.258%.

Although secondary market trading spreads are positive on average, they are negative for some issues.

The proportion of issues where the syndicate’s weighted-average bid-ask spread during the 21-day interval is negative is higher for offerings with large overallocation (36% for IG issues and 25% for HY issues) relative to offerings without large overallocation (30% for IG issues and 16% for HY issues).

These results are consistent with the reasoning that aftermarket demand for issues with large overallocation is more uncertain, and that price-supporting dealer purchases tend to occur at higher prices than would otherwise be observed.

We assess two components of dealers’ dollar secondary market trading profits. First, we measure round-trip volume for each syndicate group/lead bookrunner as the minimum of their total purchases and total sales over the 21-day trading period, and compute market-making profit as the product of round-trip volume and the weighted-average spread. Mean market-making profits to non-syndicate dealers range from $100,176 to $120,408 (Panels B and D of Table IX) across subsamples, and do not differ meaningfully across issues that are or are not overallocated. Mean market making profits to

27 syndicate/lead left dealers average between $73,000 (IG issues that are not overallocated) to $161,000

(HY issues that are not overallocated).

We compute the second component of dealer trading profits as the minimum of dollar overallocation and net dealer secondary market purchases (to allow for the fact that a few short positions are not fully covered within the 21 days considered) times the difference between the offer price and the dealer’s weighted-average secondary market buy price. We exclude from this measure trades that occur after an overallocated issue is fully covered. If dealers are net sellers or do not trade over this 21-day period, this measure is set to zero. The results reported on Table IX show that that syndicate members incur trading losses on average when they cover short positions on issues with large overallocation. For overallocated IG issues the mean trading loss is $73,128, while for overallocated HY issues the mean trading loss incurred by the lead-left underwriter is $307,985. The 75th percentile loss is $109,380 for IG issues and $416,069 for HY issues.

For large overallocation IG offerings, overall trading profits, inclusive of both short covering and market-making trades, are essentially zero ($1,721) on average, as compared to $73,654 for IG offerings that are not overallocated. For HY issues with large overallocation, the lead left underwriter incurs an average overall trading loss averaging $240,967, as compared to a profit of $161,578 on issues that are not overallocated.

These results verify that the decision to overallocate corporate bond issues leads to non-trivial short-covering costs borne by members of the syndicate. For IG issues, the short covering costs are comparable to the amount of market making revenue earned in the month after the offer, while for HY issues the short covering costs are substantially larger than the first-month market making revenues, on average. A relevant measure from the syndicate’s perspective is the difference in total profits for handling an overallocated issue versus not, which is $71,933 on average for IG issues and $402,545 for

HY issues.

Of course, the syndicate also earns commissions, sometimes termed the “gross spread”. As discussed earlier, while commission data is not reported by SDC for all issues, for IG issues with

28 available data the syndicate earns average commissions of $4.4 million, while for HY issues with available data the commission averages $8.1 million. Hence, while the decision to overallocate issues leads to predictable secondary market trading losses for the syndicate, these appear to be small relative to commissions. A moderate increase in the likelihood that members of the syndicate would be selected to participate in future bond offerings as a result of their stabilization efforts would readily justify the costs.

6. Offering Characteristics and Secondary Market Outcomes

We next turn to an assessment of the determinants of secondary market trading outcomes in the period following the issuance. In Table X, we report the results of multivariate regressions that rely on the same control variables used for prior tables, augmented by an indicator variable for large overallocation issues. We focus on four dependent variables, including the weighted-average spread between the prices at which dealers sell to and the prices at which they buy from customers, the signed net (buy minus sell) position change in institutional-sized ($1 million or more) customer trades, the percentage of trades that are likely instituted by retail traders (those less than or equal to $100,000), and the signed net position change in retail-sized customer trades. The position change variables are each scaled by own trading volume, and all measures are computed over the 21 days from the issue date. We use the full sample of bond issues when studying the bid-ask spread and institutional order flow, but when studying the percentage of retail participation and retail trading volume we study non-144A issues, since retail traders are largely excluded from buying 144A bonds. Results are reported separately for IG

(columns 1, 3, 5, and 7) and HY (columns 2, 4, 6, and 8) issues.

The sample mean of each dependent variable is reported at the top of each panel. The secondary market bid-ask spread averages 9.8 basis points for IG bonds and 21.4 basis points for HY bonds.

Institutions are on average net sellers in both IG (net signed volume is -8% of overall volume) and HY

(net signed volume is -12% of overall volume) issues. Retail-size trades comprise about a quarter of overall activity in both IG and HY non-144A issues, and retail traders are net buyers (net signed volume

29 is 54% of volume in non-144A IG issues and 32% of volume in non-144A HY issues) in the first month after the issue.

Focusing first on multivariate regression results for secondary market spreads, it can be observed that spreads are on average significantly lower (by 2.1 basis points for IG offerings and 3.9 basis points for HY offerings) for issues with large overallocation. The fact that the underwriting syndicate typically engages in net purchases to cover short positions in overallocated deals may encourage other market participants to supply liquidity, in line with the strategic complementarity argument advanced by

Bessembinder, Hao, and Zheng (2020) with regard to designated market makers. Among control variables, it can also be observed that spreads are significantly lower for larger issues and multiple tranche issues, and, perhaps surprisingly, when LIBOR rates are higher, in both IG and HY offerings.

Spreads are wider for IPOs issues and for issuers that have not issued a bond within the prior two years, but the coefficient estimates are statistically significant only for HY issues.

It can be observed in columns 3 and 4 of Table X that the signed institutional position change is lower for issues with large overallocation, in both IG and HY issues. This verifies that institutions, including those who flip their primary allocations, on balance supply the bonds purchased by the underwriting syndicate to cover their short positions in deals with large overallocation. The net institutional position change is not consistently related to other explanatory variables, except for a positive relation with bond maturity.

Focusing on columns 5 and 6 of Table X, it can be observed that non-144A issues with large overallocation involve a significantly higher proportion of retail trading in the aftermarket, for both IG and HY issues. Since primary placements are typically with institutional investors, these results support the reasoning that the syndicates’ aftermarket purchases in issues with large overallocation effectively facilitate a shift in the investor base from institutions toward retail investors. Among control variables, retail size trades account for a higher proportion of trading activity for bonds issued by industrial firms and firms with publicly-listed common stock, and for a lower proportion of trading activity for firms with

30 lower credit ratings, though the final result is of marginal statistical significance in the small (N=395) sample of non-144A HY issues.

The estimates reported in columns (7) and (8) show that the net retail position change in the secondary market is higher for non-144A issues with large overallocation, which also supports the reasoning that overallocation effectively invites retail participation in the aftermarket. The coefficient estimate for IG offers of 0.0524 is statistically significant, while the estimate for HY offers, despite being of similar magnitude (0.0475) is not significant, reflecting the smaller sample size of non-144A HY offers. Retail traders accumulate larger positions in the secondary markets for larger offers in both HY and IG bonds. The net retail position change is negatively and significantly related to recent bond index returns in both IG and HY markets, which may suggest that simple extrapolation of recent price trends affects retail position changes.

On balance, the results reported in Table X support the conclusion that institutions and other dealers are net sellers of overallocated bonds, effectively supplying the bonds that syndicate members use to cover their short positions, and that the underwriting syndicate’s decision to overallocate in the primary market is an important determinants of secondary market outcomes, being associated in particular with narrower bid-ask spreads and higher retail participation.

7. Conclusions

We study the corporate bond issuance process, focusing in particular on the underwriting syndicate structure, the syndicate’s decision to overallocate some issues, the syndicates’ trading activity in the secondary market, and secondary market liquidity. Such linkages are of interest, in part, because members of the underwriting syndicate are also active as secondary market bond dealers. The syndicate’s implicit obligations to stabilize secondary market trading implies that uncertainty regarding secondary market demand may affect the syndicate’s primary market allocation decisions. However, unlike equities, corporate bond offerings typically do not include a “Greenshoe” option, implying that overallocation is

31 potentially costly to the syndicate. Thus, it is of particular interest to assess the nature of overallocation decisions, and their effects on aftermarket trading, underpricing, and liquidity in corporate bond offerings.

Our main findings are as follows. The structure of the syndicate is affected by both the size and the complexity of the issue. Based on information obtained during the book building process, the syndicate overallocates bond offerings that face greater uncertainty regarding secondary market order flow. The incidence and the extent of overallocation is greater for HY offers, which tend to involve more uncertainty. The multivariate evidence suggests that overallocation is observed for deals with greater uncertainty regarding secondary market order flow, and at times when market conditions are uncertain, as evidenced by higher dispersion in bond prices or higher levels and changes in VIX.

With overallocation, the syndicate attains a net short position that effectively hedges the purchases required to provide secondary market support for a weak issue. The syndicate on average incurs losses on the trades undertaken for price stabilization purposes, implying that the syndicate incurs a cost for managing overallocated issues, but these trading losses on short covering trades are small relative to the commissions earned by the syndicate for managing the issue.

We assess relations between overallocation and secondary market trading activity. For IG issues with large overallocation, two thirds of the syndicate’s secondary market volume during the two days after the offering consists of bond purchases, allowing the syndicate to cover on average about 80% of the overallocation. For HY issues the “lead left” bookrunner on average makes secondary market purchases equating to about 65% of the overallocation during the first two days.

Despite the stabilizing effect of short covering purchases, issues with large overallocation ultimately trade at a smaller secondary market premium to offer price; that is, overallocated bonds are less underpriced, which reflects the more modest investor demand that motivated the overallocation. The price stabilizing activities of the syndicate are associated with narrower secondary market bid-ask spreads for issues with large overallocation relative to other issues. Overallocation further effectively facilitates the repositioning of the bond from “flipping” institutional investors into portfolios of longer-horizon retail investors, thus achieving greater ownership dispersion.

32 Our results are predicated on the notion that the underwriting syndicate learns during the bookbuilding process not only about the total quantity demanded (i.e. the degree of oversubscription) but also about characteristics of that demand, including the composition of long-term investors versus potential flippers. While researchers cannot observe the same information, our results support the reasoning that overallocation in the absence of a Greenshoe option comprises an observable indicator that the syndicate anticipates weak or uncertain secondary market demand for an issue.

Individual primary investors do not have access to the same information as the underwriting syndicate, and investing firms have recently pushed for increased transparency in the issuance process.23

Credit Roundtable (2015) expressed the concern that “Once a deal is announced, there is often limited or no contact available with issuer management, making it impossible for investors to ask questions or clarify concerns” and that “there is often inadequate time for analysts to thoroughly review issuer fundamentals, structural features, and covenant protections allowing them to make well-informed decisions.” We find that the primary investors receive larger average allocations, and that allocations are more concentrated, in overallocated as compared to other offers. Since we also show that overallocated offerings are less profitable to primary investors (i.e., are less underpriced), this result implies a potential adverse selection problem faced by institutions who have access to less information than the syndicate during the book building process. Disclosing order book updates in regular intervals during the book building phase might help mitigate information asymmetry for investing firms. The largest bond underwriters have proposed an electronic platform named DirectBooks to partially automate the process.

It will be of interest to assess the extent to which primary investors, underwriters and other market participants benefit from a more transparent and standardized issuance process should the proposed reforms be enacted.

23 See https://www.ifre.com/story/2498707/buyside-body-wants-changes-in-how-us-bonds-are-sold-l8n2fn0el and https://www.bloomberg.com/press-releases/2019-10-11/global-bank-consortium-creates-capital-markets- syndication-platform-directbooks.

33 References Aggarwal, R., 2000. Stabilization activities by underwriters after initial public offerings. Journal of Finance 55, 1075–1103.

Allahrakha, M., J. Cetina, B. Munyan, S.W. Watugala. 2019. The Effects of the Volcker Rule on Corporate Bond Trading: Evidence from the Underwriting Exemption. Working paper. Available at SSRN: https://ssrn.com/abstract=3440712

Auh, J.K., Y.S., Kim, and M. Landon. 2019. Tricks of the Trade? Pre-Issuance Price Maneuvers by Underwriter-Dealers. Working paper. Available at SSRN: https://ssrn.com/abstract=3190476.

Bessembinder, H. J. Hao, and K. Zheng, 2020, Liquidity provision contracts and market quality: Evidence from the New York Stock Exchange, Review of Financial Studies, 44: 44-74.

Bessembinder, H., S. Jacobsen, W. Maxwell and K. Venkataraman, 2018, Capital Commitment and Illiquidity in Corporate Bonds, Journal of Finance 73:1615-1661,

Bessembinder, H., C. Spatt, and K. Venkataraman, 2020, A survey of the microstructure of fixed income markets, Journal of Financial and Quantitative Analysis 55: 1-45.

Brugler, J., C. Comerton-Forde and S. Martin, 2016, Secondary Market Transparency and Corporate Bond Issuing Costs, working paper. University of Melbourne and UNSW Business School.

Booth, J.R., and L. Chua. 1996. Ownership dispersion, costly information and IPO underpricing. Journal of Financial Economics 41: 291-310.

Cai, N., J. Helwege, and A. Warga. 2007. Underpricing in the Corporate Bond Market. Review of Financial Studies 20, 2021-2046.

Corrigan, P.M. 2020. Footloose with Green Shoes: Can Underwriters profit from IPO Underpricing? Yale Journal of Regulation, forthcoming.

The Credit Roundtable, 2015. Corporate bond underwriting and distribution practices, presented to the U.S. Securities and Exchange Commission on 7/1/2015.

Datta, S., M. Iskandar-Datta, and A. Patel. 1997. The Pricing of Initial Public Offers of Corporate Straight Debt, Journal of Finance 52:379–96.

Ederington, L. H. 1974. The Yield Spread on New Issues of Corporate Bonds, Journal of Finance 29:1531–43.

Ellis, K., Michaely, R., and M. O’Hara. 2000. When the underwriter is the market maker: An examination of trading in the IPO aftermarket, Journal of Finance 55:1039-1074.

Ellul, A., and M. Pagano. 2006. IPO Underpricing and After-Market Liquidity. Review of Financial Studies 19:381–421.

Eckbo B., R. Masulis, and O. Norli, 2007, Security Offerings, Handbook of Corporate Finance: Empirical Corporate Finance, Vol. 1, B. E. Eckbo, ed., Chapter 6, pp. 233-373, Elsevier/North-Holland Handbook of Finance Series.

34

Flanagan, T., S. Kedia, A. Zhou. 2019. Secondary Market Liquidity and Primary Market Allocations in Corporate Bonds. Working paper. Available at SSRN: https://ssrn.com/abstract=3449431.

Fishe, R.P., 2002, How stock flippers affect IPO pricing and stabilization, Journal of Financial and Quantitative Analysis, 37: 319:340.

Fung, W. K. H., and A. Rudd. 1986. Pricing New Corporate Bond Issues: An Analysis of Issue Cost and Seasoning Effects. Journal of Finance 3:633–43.

Gande, A., M. Puri, A. Saunders, and I. Walter, 1997, Bank Underwriting of Debt Securities: Modern Evidence, Review of Financial Studies, 10, 1175-1202.

Green, R., B. Hollifield, and N. Schurhoff. 2007. Dealer intermediation and price behavior in the aftermarket for new issues. Journal of Financial Economics 86: 643–682.

Goldstein, M., E. Hotchkiss and S. Nikolova, 2021, Dealer Behavior and the Trading of Newly Issued Corporate Bonds, working paper. Babson College.

Han, S., A. Huang, M. Kalimipalli and K. Wang, 2019, Information and Liquidity of OTC Securities: Evidence from Public Registration of Rule 144A Bonds (2018-08-30). FEDS Working Paper No. 2018- 061. Available at SSRN: https://ssrn.com/abstract=3245055.

Hale, G., and J.A. Santos, 2009. Do banks price their informational monopoly? Journal of Financial Economics 93: 185-206.

Helwege, J., and P. Kleiman. 1998. The Pricing of High-Yield Debt IPOs. Journal of Fixed Income 8:61– 8.

Helwege, J. and L. Wang. 2019. Liquidity and Price Pressure in the Corporate Bond Market: Evidence from Mega-Bonds. Working paper. Available at SSRN: https://ssrn.com/abstract=2801840.

Lindvall, J. 1977. New Issue Corporate Bonds, Seasoned Market Efficiency and Yield Spreads. Journal of Finance 32:1057–67.

Nagler, F., and G. Ottonello, 2017, Structural Changes in Corporate Bond Underpricing. Working paper. Bocconi University.

Nikolova, S., L. Wang, and J. Wu. 2019. Institutional Allocations in the Primary Market for Corporate Bonds. Journal of Financial Economics, Forthcoming.

Ruud, J., 1993, Underwriter price support and the IPO underpricing puzzle, Journal of Financial Economics, 34, 135-151.

Schultz, P., 2012. The market for new issues of municipal bonds: The roles of transparency and limited access to retail investors. Journal of Financial Economics 106: 492-512.

Sorenson, E. H. 1982. On the Seasoning Process of New Bonds: Some are More Seasoned than Others. Journal of Financial and Quantitative Analysis 42:195–208.

35 Wang, L., 2020, Lifting the Veil: The Price Formation of Corporate Bond Offerings, Journal of Financial Economics, forthcoming.

Wasserfallen, W., and D. Wydler. 1988. Underpricing of Newly Issued Bonds: Evidence from the Swiss Capital Market. Journal of Finance 43:1177–91.

Weinstein, M. I. 1978. The Seasoning Process of New Corporate Bond Issues. Journal of Finance 33:1343–54.

36 Figure 1 Distribution of Overallocation

This figure shows a histogram of overallocation for investment grade, high yield, small issue, and large issue categories. Small (large) issues are defined as issues less (greater) than $500 million. Green dot pattern bars denote investment grade (IG) small issues, green solid bars denote investment grade large issues, red striped pattern bars denote high yield (HY) small issues, and red pattern bars denote high yield large issues.

80%

70%

60%

50%

40%

30%

20%

10%

0% 100% - 101% 102% - 104% 105% - %107 108% - 115%

IG Small IG Large HY Small HY Large

37 Figure 2 Underwriter Secondary Market Trading Activity

These figures report the speed for which the overallocation is covered and the underwriter position change in the days surrounding the offering. We report mean statistics from the offering date (day 1) to twenty-one days following the offering date (day 21). The shaded portion represents the 95% confidence interval. Panel (a) shows the syndicates’ net signed secondary market position change for the investment grade large overallocation sample and Panel (b) shows the lead underwriter’s (defined as the dealer who is the P1 allocator) net signed position change for the large overallocation high yield sample. Panel (c) is based on the large overallocation sample and shows the mean percent of the overallocation offset each day for the investment grade and high yield sample. Panel (d) is based on the large overallocation sample and shows the percentage of issues with overallocations that are fully offset each day for the investment grade and high yield sample.

A. Syndicate signed position change (IG) B. Lead left signed position change (HY)

C. % of overallocation offset D. % With fully offset overallocation

38 Figure 3 Distribution of Initial Returns

These figures present histograms of one, two, and five-day returns (the weighted average price on day n relative to the offering price) for both investment grade and high yield issues. These figures are based on the sample with large overallocation. The solid blue bar denotes returns that exceed zero, the patterned red bar denotes returns that are less than zero.

A. Investment Grade B. High Yield

39 Figure 4 Return Price Paths

These figures show the relative price paths for the first 21 days following the offering (day 1). Relative price path is the weighted average raw price on day n scaled by the offering price. 95% confidence intervals are represented with shaded areas. The blue solid line represents the sample average relative prices for the large overallocation sample and the red dashed line represents the sample average relative prices for the non-overallocation sample.

A. Investment Grade B. High Yield

40 Table I Sample Overview Panel A summarizes the sample construction. Corporate bond trade data are from TRACE, bond descriptive data are from the Mergent Fixed Income Securities Database (FISD), and issue and underwriter data are from Thompson Reuters Securities Data Company (SDC). The sample period is March 2010 to March 2018. We match FISD to SDC data with the same CUSIP, offering amount, and if both offering and maturity dates are within 3 days. We match to TRACE data based on CUSIP. Our sample consists of 5,573 bond issues between 2010 and 2018. Panel B provides descriptive statistics for the full sample, investment grade issues, and high yield issues. We report median values (except percent Panel A: Sample Construction # Issues Corporate bonds in FISD 14,409 Match to SDC 10,756 Match to TRACE 9,958 Retain all trades between offering date + 1 year, with a trade size < issue size 9,950 Exclude issues without at least one P1 trade and one S1 trade 9,849 Exclude variable rate issues 9,156 Exclude issues by financial firms 6,925 Exclude Yankee issues 6,072 Exclude issues without P1 trades on offering date (-1,+1) 6,063 Exclude issues with non-FINRA member bookrunners 5,990 Exclude issues without P1 trades by Top 34 dealers 5,971 Exclude issues with overallotment > 1.15 5,922 Exclude issues with overallotment < .95 5,573 Panel B: Sample Description Investment Full Sample High Yield Grade # Observations 5,573 3,611 1,962 Issue Size ($ in 000,000) 500 500 450 # Tranches 1 2 1 % > 1 Tranche 49% 69% 14% % Tranche Size > $1B 43% 57% 19% Maturity 8 10 7 % Maturity < 5 years 21% 14% 35% % Maturity >=5 and < 10 years 31% 23% 47% % Maturity >=10 and < 30 years 33% 41% 18% % Maturity >= 30 years 15% 23% 0% % First issue reported in FISD 11% 4% 23% % 144A 35% 11% 80%

41

Table II Syndicate Structure This table reports median syndicate structure and fee statistics. Manager, bookrunner, and underwriting commitment data are from Thompson Reuters Securities Data Company (SDC). Data for the underwriting commitment by each bookrunner and manager are available for 52% of issues in the full sample and 71% and 16% in the investment grade and high yield samples, respectively. Variable definitions are provided in Appendix I. Issue size and underwriting commitment numbers are in millions. Investment Full High Yield Grade Sample Sample Sample (N=5,573) (N=1,962) (N=3,611) Issue Size 500 500 450 # Managers 10 11 7 # Bookrunners 4 4 4 % Bookrunners in Top 10 33% 29% 50% % Deals w/ at least 1 top10 BR 98.1% 98.8% 96.9% Syndicate Member Underwriting Commitment 50 50 50 Bookrunner Member Underwriting Commitment 96 98 83 Bookrunner Underwriting Commitment/Issue Size 17% 18% 17% Largest Bookrunner Underwriting 22% 22% 30% Commitment/Issue Size % Total Offering Underwritten by Bookrunners 80% 80% 83%

42

Table III Primary Market Activities This table provides descriptive statistics on primary market activities related to pricing and allocation decisions. Panel A reports median pricing and primary market allocation statistics based on the primary placement transactions flagged as “P1” customer and interdealer trades in TRACE. Panel B reports median (unless otherwise denoted) overallocation statistics. To compute the overallocation we do the following: 1) retain all P1 buy trades (dealer sells to customer or another dealer), 2) exclude P1 interdealer trades between syndicate members, 3) divide these P1 trades by the offering amount. Panel C reports statistics for "large" overallocation issues, where large refers to overallocation greater than or equal to 102%. Statistics are reported for both investment grade and high yield samples. Variable definitions are provided in Appendix I. Investment Grade High Yield Sample Sample (N=3,611) (N=1,962) Panel A: Pricing and Primary Market Allocation Statistics Offering Spread over Benchmark Treasury 122 443 # Primary Trades 91 91 Trade Size 6,743,705 5,815,690 Trade Size / Offering Amount 1.1% 1.2% Largest Primary Trade 49,883,500 45,000,000 Largest Primary Trade / Offering Amount 8.4% 10.0% Primary Herfindahl 0.038 0.040 Panel B: Overallocation Statistics % Overallocation 100.5% 104.5% % Overallocation (Mean) 101.4% 104.9% $ Overallocation 2,500,000 19,719,500 $ Overallocation (Mean) 10,145,435 27,335,649 Large Overallocation Indicator (> =102%) 30% 77% Panel C: Large Overallocation Statistics $ Overallocation 20,000,000 26,345,000 $ Overallocation / Trading Volume (Offering Day) 33% 35% $ Overallocation / Trading Volume (Offering Day+1) 27% 37%

43

Table IV Primary Market Activities: Large Overallocation This table provides descriptive statistics on primary market activities related to pricing and allocation decisions for both the large overallocation and non-large overallocation samples. Issues are included in the large overallocation sample if the overallocation is greater than or equal to 102%. This table reports median values unless otherwise denoted. We report Wilcoxon p -Values based on tests of differences between the large overallocation and non- large overallocation samples. For investment grade issues, gross spread data is available for 79% of non-overallocated and 89% of large overallocated issues. For high yield issues, gross spread data is available for 11% of non-overallocated and 22% of large overallocated issues. Variable definitions are provided in Appendix I. Non-Large Large Wilcoxon p - Overallocation Overallocation Value Panel A: Investment Grade Sample (N=3,611) Issue Size ($ in 000,000) 500 550 <.0001 Offering Spread over Benchmark Treasury 120 128 <.0001 Gross Spread 0.65% 0.65% <.0001 # Primary Trades 90 93 0.0108 Trade Size 6,566,201 7,186,505 <.0001 Largest Primary Trade 41,692,070 69,011,600 <.0001 Primary Herfindahl 0.037 0.041 <.0001 Panel B: High Yield Sample (N=1,962) Issue Size ($ in 000,000) 400 450 <.0001 Offering Spread over Benchmark Treasury 463 435 <.0001 Gross Spread 1.42% 1.50% <.0001 # Primary Trades 90 92 0.4200 Trade Size 4,819,301 6,115,802 <.0001 Largest Primary Trade 33,000,000 50,000,000 <.0001 Primary Herfindahl 0.036 0.041 0.0040

44 Table V Determinants of Syndicate Structure and Primary Market Outcomes This table reports regressions of syndicate structure and primary market outcomes on offering and syndicate characteristics and market conditions. Columns (1), (3), and (5) report regression results for the investment grade sample and Columns (2), (4), and (6) report regression results for the high yield sample. % Overallocation is the sum of dealer sell quantities across primary placement (“P1”) transactions scaled by the offering amount. Ln(Spread) is the natural log of the offering spread over the benchmark treasury. Variable definitions are provided in Appendix I. Results in Columns (1)-(2) and (5)-(6) are based on OLS regression framework. Results in Columns (3)-(4) are based on a Tobit regression framework. Dependent variable averages are shown above the regression results. Standard errors are estimated using the Huber-White sandwich estimator. ***, **, and * stand for statistical significance at the 1%, 5%, and 10% level, respectively. (1) (2) (3) (4) (5) (6) # Bookrunners Overallocation % Ln(Spread) IG HY IG HY IG HY Dependent Variable Average 4.9 4.5 101.4 104.9 4.8 6.1 Ln(Offering Amount) 1.137*** 1.347*** 0.121** 0.408*** 0.082*** -0.035** (0.000) (0.000) (0.049) (0.005) (0.000) (0.010) Maturity 0.004 -0.033* 0.014*** 0.011 0.016*** -0.005** (0.271) (0.055) (0.000) (0.632) (0.000) (0.020) Industrial -0.258** -0.393* 0.164* 0.421* 0.105*** -0.003 (0.019) (0.061) (0.084) (0.081) (0.000) (0.875) Multiple Tranche 0.920*** 0.287 -0.245*** 0.094 -0.142*** -0.002 (0.000) (0.129) (0.002) (0.690) (0.000) (0.923) 144A Indicator -0.498*** -0.906*** -0.508*** -0.876*** 0.210*** 0.083*** (0.001) (0.000) (0.000) (0.000) (0.000) (0.000) First issue reported in FISD Indicator -0.075 -0.848*** -0.429** -0.694*** 0.057 0.103*** (0.729) (0.000) (0.020) (0.002) (0.104) (0.000) First Issue w/in 2 Years Indicator -0.392*** -0.389*** -0.167** -0.397** 0.064*** 0.052*** (0.000) (0.003) (0.017) (0.029) (0.000) (0.002) Public Stock Indicator -0.576*** -0.634*** 0.013 -0.205 -0.090*** -0.039** (0.000) (0.000) (0.872) (0.228) (0.000) (0.011) Credit Rating 0.388*** -0.071 0.052 0.002 0.395*** 0.257*** (0.000) (0.408) (0.294) (0.987) (0.000) (0.000) # Bookrunners 0.054*** 0.125*** 0.020*** -0.007** (0.000) (0.000) (0.000) (0.047) S.D. of Offering Date Prices for Issues (q-1) -5.306*** -4.145** 6.310*** 8.365*** 0.249 0.693*** (0.000) (0.016) (0.000) (0.001) (0.238) (0.002) % Chg. Ave. Daily VIX (t-1, t-5 to t) 0.054* -0.035 0.075*** 0.242*** 0.004 0.009* (0.078) (0.333) (0.000) (0.000) (0.248) (0.074) Ave. Daily VIX (t-1 to t-5) -0.022*** -0.059*** 0.033*** 0.035 0.019*** 0.023*** (0.001) (0.000) (0.000) (0.104) (0.000) (0.000) Ave. Corp Bond Index Return (t-1 to t-5) -87.843*** 45.772 -37.262 -139.501* 3.059 8.611 (0.008) (0.349) (0.213) (0.057) (0.547) (0.176) # Corporate Bond Offerings (t-1 to t-5) 0.009 -0.030* 0.001 0.062*** 0.004*** 0.002 (0.117) (0.055) (0.909) (0.009) (0.000) (0.443) Ave. 3-Month Libor (t-1 to t-5) 0.260** 0.476*** 0.010 -0.320 -0.119*** -0.058*** (0.016) (0.010) (0.916) (0.167) (0.000) (0.005) Constant -10.678*** -9.086*** 97.419*** 96.114*** 1.758*** 4.528*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Observations 3,611 1,962 3,611 1,962 3,538 1,752 Adjusted R2 0.198 0.200 0.524 0.423 45 Table VI Secondary Market: Underwriter Trading Activity This table reports OLS regressions of dealer secondary market trading activity on offering and syndicate characteristics and market conditions. We focus on secondary market trading activity over the first 2 days following the issue date where Day 1 is the offering date. We report regression results for investment grade (IG) and high yield (HY) samples. In columns (1)-(2), the dependent variable is the syndicate/lead underwriter net signed secondary market position change excluding the initial overallocation (in 000s). In columns (3)-(4), the dependent variable is the percent of total syndicate/lead underwriter dealer volume that is secondary market purchases from customers or other dealers. The Syndicate/Lead indicator refers to the full syndicate for investment grade issues and the lead underwriter (defined as the P1 allocator) for high yield issues. Large Overallocation is an indicator variable for issues that are overallocated by at least 102%. Variable definitions are provided in Appendix I. Dependent variable averages are shown above the regression results. Standard errors are clustered at the issue level. ***, **, and * stand for statistical significance at the 1%, 5%, and 10% level, respectively. (1) (2) (3) (4) Net Signed Position Change: Days 1-2 % Volume Buys: Days 1-2 IG HY IG HY Dependent Variable Average 12,654 11,322 59.5 58.1 Syndicate/Lead Underwriter Indicator 3,719*** 4,721*** 6.63*** 6.24*** (0.000) (0.000) (0.000) (0.000) Large Overallocation Indicator -1,340** -2,391*** -0.44 -0.16 (0.017) (0.000) (0.156) (0.707) Syndicate/Lead x Large Overallocation 11,932*** 18,865*** 7.61*** 10.94*** (0.000) (0.000) (0.000) (0.000) Ln(Offering Amount) 11,135*** 9,552*** -1.40*** -1.74*** (0.000) (0.000) (0.000) (0.000) Maturity 138*** 146 0.04** -0.01 (0.000) (0.239) (0.014) (0.786) Industrial 1,423*** 383 -1.47*** -0.30 (0.009) (0.726) (0.003) (0.572) Multiple Tranche 470 -319 0.19 0.74* (0.298) (0.762) (0.617) (0.075) 144A Indicator 1,974*** 1,552 1.14* 0.16 (0.009) (0.143) (0.054) (0.706) First issue reported in FISD Indicator 283 -430 -0.43 -1.66*** (0.788) (0.775) (0.679) (0.001) First Issue w/in 2 Years Indicator 152 -1,185* -0.39 -1.69*** (0.736) (0.086) (0.240) (0.000) Public Stock Indicator 129 883 -0.29 -0.05 (0.797) (0.178) (0.466) (0.879) Credit Rating -444 95 -1.02*** 0.45* (0.210) (0.911) (0.000) (0.075) # Bookrunners -60 -195 -0.11* 0.00 (0.598) (0.204) -0.095 (0.992) S.D. of Offering Date Prices for Issues (q-1) 51,573*** 23,776** 1.85 6.92 (0.000) (0.020) (0.697) (0.179) % Chg. Ave. Daily VIX (t-1, t-5 to t) -379*** -515*** -0.19** -0.06 (0.000) (0.005) (0.011) (0.660) Ave. Daily VIX (t-1 to t-5) -243*** -115 -0.19*** 0.07 (0.000) (0.159) (0.000) (0.166) Ave. Corp Bond Index Return (t-1 to t-5) 2,207 535,992** -182.64 -4.58 (0.991) (0.037) (0.206) (0.978) # Corporate Bond Offerings (t-1 to t-5) -89*** 239 0.04* 0.09* (0.003) (0.139) (0.097) (0.069) Ave. 3-Month Libor (t-1 to t-5) 96 1,004 -0.33 1.28** (0.886) (0.363) (0.545) (0.013) Constant -145,217*** -130,919*** 80.74*** 66.04*** (0.000) (0.000) (0.000) (0.000) Observations 7,114 3,872 7,013 3,794 Adjusted R2 0.232 0.187 0.160 0.404 46 Table VII Secondary Market: Price Statistics This table reports underpricing and return statistics for investment grade and high yield issues and for large overallocation issues (issues that are overallocated by at least 102%) and non-large overallocation issues. Raw returns are computed as the return of the weighted-average price on day n ( expressed as the sum of the flat price and accrued interest) relative to the offering price; we then subtract the cumulative index return over n days. For investment grade bonds we use the BofA Merrill Lynch U.S. Corporate Total Return Index and for high yield bonds we use the BofA Merrill Lynch U.S. High Yield Total Return Index. Underpricing measures have been winsorized at the 1% and 99% levels. Negative Return refers to issues with underpricing on day n less than zero. We report Wilcoxon p -Values based on tests of the difference between underpricing (% Negative Return) for investment grade and high yield and for non-overallocated and overallocated issues. Full Sample Investment Grade High Yield Diff.

Skew- Kurt- Skew- Kurt- Wilcoxon Mean SD 10% 90% Mean SD 10% 90% ness osis ness osis p -Values

Underpricing-Day 1 0.35 0.46 -0.12 0.92 1.43 3.78 1.01 0.75 0.14 2.11 0.60 -0.25 <.0001 Underpricing-Day 2 0.44 0.61 -0.20 1.20 1.21 2.90 1.10 0.96 -0.01 2.42 0.45 -0.30 <.0001 Underpricing-Day 5 0.55 0.89 -0.34 1.63 1.05 2.45 1.26 1.24 -0.25 2.98 0.19 -0.35 <.0001 Underpricing-Day 21 0.86 1.70 -0.85 2.96 1.01 2.01 1.67 2.12 -1.04 4.43 -0.01 -0.18 <.0001 % Negative Day 1 Return 19.3% 5.6% <.0001 % Negative Day 5 Return 24.0% 15.3% <.0001 Investment Grade Non-Large Overallocation Large Overallocation Diff. Underpricing-Day 1 0.38 0.48 -0.11 0.97 1.47 3.75 0.29 0.41 -0.15 0.79 1.21 3.12 <.0001 Underpricing-Day 2 0.48 0.61 -0.17 1.24 1.18 2.69 0.37 0.59 -0.26 1.10 1.31 3.57 <.0001 Underpricing-Day 5 0.58 0.87 -0.29 1.63 1.15 2.70 0.48 0.91 -0.49 1.61 0.89 2.00 0.0022 Underpricing-Day 21 0.86 1.62 -0.76 2.86 1.09 2.33 0.87 1.85 -1.10 3.19 0.87 1.46 0.8033 % Negative Day 1 Return 17.5% 23.1% 0.0001 % Negative Day 5 Return 23.0% 26.3% 0.0490 High Yield Non-Large Overallocation Large Overallocation Diff. Underpricing-Day 1 1.36 0.82 0.28 2.57 0.15 -0.87 0.91 0.70 0.12 1.90 0.69 0.08 <.0001 Underpricing-Day 2 1.49 1.02 0.19 2.94 0.04 -0.48 0.99 0.91 -0.05 2.27 0.54 -0.11 <.0001 Underpricing-Day 5 1.62 1.25 0.05 3.44 -0.05 -0.23 1.16 1.22 -0.32 2.91 0.25 -0.31 <.0001 Underpricing-Day 21 2.17 1.95 -0.36 4.48 -0.26 0.25 1.54 2.14 -1.17 4.42 0.07 -0.21 <.0001 % Negative Day 1 Return 3.0% 6.4% 0.0084 % Negative Day 5 Return 9.9% 16.8% 0.0015

47 Table VIII Secondary Market: Underpricing This table reports OLS regressions of underpricing on high yield and large overallocation indicator variables, offering and syndicate characteristics, primary allocation measures, and market conditions. Our underpricing measure is computed over the following days relative to the offering date: 1 (offering date), 2, 5, and 21. Underpricing measures have been winsorized at the 1% and 99% levels. Large Overallocation is an indicator variable for issues that are overallocated by at least 102%. All variable definitions are provided in Appendix I. Dependent variable averages are shown above the regression results. Standard errors are estimated using the Huber-White sandwich estimator. ***, **, and * stand for statistical significance at the 1%, 5%, and 10% level, respectively. (1) (2) (3) (4) Underpricing-Day 1 Underpricing-Day 2 Underpricing-Day 5 Underpricing-Day 21 Dependent Variable Average 0.58 0.67 0.80 1.13 High Yield Indicator 0.401*** 0.529*** 0.777*** 1.188*** (0.000) (0.000) (0.000) (0.000) Large Overallocation Indicator -0.215*** -0.237*** -0.250*** -0.297*** (0.000) (0.000) (0.000) (0.000) Ln(Offering Amount) 0.044*** 0.062*** 0.047 0.211*** (0.004) (0.002) (0.119) (0.000) Maturity 0.014*** 0.014*** 0.017*** 0.041*** (0.000) (0.000) (0.000) (0.000) Industrial 0.072*** 0.016 -0.015 -0.036 (0.000) (0.566) (0.719) (0.651) Multiple Tranche -0.020 -0.010 0.026 0.038 (0.265) (0.679) (0.491) (0.602) 144A Indicator 0.176*** 0.202*** 0.134*** 0.243** (0.000) (0.000) (0.007) (0.012) First issue reported in FISD Indicator 0.331*** 0.380*** 0.470*** 0.452*** (0.000) (0.000) (0.000) (0.000) First Issue w/in 2 Years Indicator 0.114*** 0.138*** 0.160*** 0.199*** (0.000) (0.000) (0.000) (0.001) Public Stock Indicator 0.016 0.020 -0.050 -0.133** (0.375) (0.381) (0.149) (0.046) Credit Rating 0.071*** 0.039*** 0.026 0.135*** (0.000) (0.009) (0.255) (0.002) # Bookrunners -0.009*** -0.007 -0.007 -0.021* (0.005) (0.124) (0.266) (0.098) S.D. of Offering Date Prices for Issues (q-1) 0.918*** 0.501 -0.307 -3.985*** (0.000) (0.115) (0.514) (0.000) % Chg. Ave. Daily VIX (t-1, t-5 to t) -0.017*** -0.001 0.018** 0.052*** (0.000) (0.825) (0.047) (0.005) Ave. Daily VIX (t-1 to t-5) 0.004** 0.009*** 0.029*** 0.086*** (0.018) (0.000) (0.000) (0.000) Ave. Corp Bond Index Return (t-1 to t-5) 5.629 5.217 3.091 -6.171 (0.409) (0.571) (0.814) (0.804) # Corporate Bond Offerings (t-1 to t-5) -0.002 -0.002 0.000 0.011*** (0.168) (0.196) (0.879) (0.008) Ave. 3-Month Libor (t-1 to t-5) -0.067*** -0.107*** -0.143*** -0.476*** (0.002) (0.000) (0.001) (0.000) Constant -0.752*** -0.730*** -0.531 -3.064*** (0.000) (0.009) (0.214) (0.000) Observations 5,300 5,317 4,539 3,924 Adjusted R2 0.323 0.220 0.160 0.134 48 Table IX Dealer Profit Analysis This table reports dealer spreads and profits for non-large overallocation and large overallocation issues (issues that are overallocated by at least 102%). Panels A and B provide statistics for the investment grade sample and Panel C and D provide statistics for the high yield sample. Panel A provides statistics for the members of the underwriting syndicate and Panel C provides statistics for the lead underwriter (defined as the dealer who is the P1 allocator). Panels B and D provide statistics for dealers that are not part of the underwriting syndicate. Statistics are computed using trade data between the offering date and 21 days subsequent to the offering date. We report Wilcoxon p -Values based on tests of the difference between the non-large overallocated and large overallocated issues with exception to the % Negative Spread for which we compute a t -test. Non-Large Overallocation Large Overallocation Difference Mean Q1 Median Q3 Mean Q1 Median Q3 p -Value Panel A: Investment Grade - Syndicate Profit W.A. Spread / Offer Price 0.103% -0.020% 0.077% 0.213% 0.093% -0.070% 0.077% 0.255% 0.2868 % Negative Spread 30% 36% 0.0015 Profit/Loss from Round-Trip Volume 73,654 (5,909) 26,197 108,680 74,849 (34,918) 32,705 165,130 0.7299 Profit/Loss from Covering Overallotment (73,128) (109,330) (21,241) 4,186 <.0001 Total Profit 73,654 (5,909) 26,197 108,680 1,721 (90,695) 3,068 111,870 <.0001 Panel B: Investment Grade - Non-Syndicate Profit W.A. Spread / Offer Price 0.126% 0.024% 0.097% 0.199% 0.124% 0.006% 0.104% 0.217% 0.8533 % Negative Spread 20% 24% 0.0203 Profit/Loss from Round-Trip Volume 120,408 6,655 57,597 176,854 122,514 2,766 75,847 217,026 0.0379 Panel C: High Yield - Lead Bookrunner Profit W.A. Spread / Offer Price 0.258% 0.139% 0.281% 0.419% 0.154% 0.002% 0.185% 0.354% <.0001 % Negative Spread 16% 25% <.0001 Profit/Loss from Round-Trip Volume 161,578 26,731 124,866 253,636 67,018 0 60,927 154,957 <.0001 Profit/Loss from Covering Overallotment (307,985) (416,069) (161,922) (51,629) <.0001 Total Profit 161,578 26,731 124,866 253,636 (240,967) (320,646) (102,462) 9,636 <.0001 Panel D: High Yield - Non-Syndicate Profit W.A. Spread / Offer Price 0.198% 0.091% 0.174% 0.244% 0.147% 0.075% 0.149% 0.223% 0.0117 % Negative Spread 14% 12% 0.2884 Profit/Loss from Round-Trip Volume 100,176 20,271 88,138 168,173 109,459 31,016 88,296 180,354 0.2100

49 Table X Overallocation and Secondary Market Outcomes This table reports OLS regressions of secondary market outcomes on overallocation, offering and syndicate characteristics, and market conditions for both investment grade (IG) and high yield (HY) samples. Our measures are computed over the first 21 days following the issue date. The results in Columns (1)-(4) are based on the full sample of issues. Columns (5)-(8), which focus on retail-sized trades, are based on only public bond issues and exclude 144A issues. Large Overallocation is an indicator variable for issues that are overallocated by at least 102%. In Columns (1)-(2), we compute the W.A. Spread, or difference between the volume-weighted average dealer customer sell price and buy price scaled by the offering price. 'Retail' is defined as customer trades less than or equal to $100,000 and 'Institutional' is defined as customer trades of at least $1 million. We compute the percentage of all trades that are retail-sized trades (Columns (7)-(8)) and the signed net order flow scaled by volume for both retail (Columns (5)-(6)) and institutional-sized (Columns (3)-(4)) trades. All other variable definitions are provided in Appendix I. Dependent variable averages are shown above the regression results. Standard errors are estimated using the Huber-White sandwich estimator. ***, **, and * stand for statistical significance at the 1%, 5%, and 10% level, respectively. (1) (2) (3) (4) (5) (6) (7) (8) W.A. Spread (bp) Order Flow-Institutions % Trades Retail (%) Net Order Flow-Retail IG HY IG HY IG HY IG HY Dependent Variable Average 9.8 21.4 -0.08 -0.12 24.6 26.0 0.54 0.32 Large Overallocation Indicator -2.106*** -3.878** -0.0702*** -0.1103*** 3.242*** 10.852*** 0.0524*** 0.0475 (0.001) (0.024) (0.000) (0.000) (0.000) (0.000) (0.001) (0.565) Ln(Offering Amount) -2.749*** -7.461*** 0.0097* 0.0510*** 3.344*** 2.420 0.0547*** 0.2432*** (0.000) (0.000) (0.100) (0.000) (0.000) (0.184) (0.000) (0.000) Maturity 0.243*** 0.070 0.0010*** 0.0025*** -0.746*** -1.938*** -0.0128*** -0.0192 (0.000) (0.717) (0.000) (0.007) (0.000) (0.000) (0.000) (0.165) Industrial 0.045 -2.046 0.0107 0.0061 3.824*** 10.499*** 0.0786*** 0.1147 (0.955) (0.386) (0.195) (0.581) (0.000) (0.000) (0.001) (0.231) Multiple Tranche -1.898*** -4.201*** -0.0027 -0.0146* 0.460 0.725 0.0225 0.1660** (0.007) (0.005) (0.670) (0.078) (0.512) (0.773) (0.241) (0.017) 144A Indicator 1.375 0.157 -0.0152 0.0006 (0.203) (0.916) (0.129) (0.944) First issue reported in FISD Indicator 1.578 7.011*** -0.0012 0.0160 -4.711** -2.277 0.0196 -0.1426 (0.446) (0.000) (0.951) (0.111) (0.011) (0.596) (0.735) (0.279) First Issue w/in 2 Years Indicator 0.477 4.999*** 0.0041 0.0281*** 0.267 -2.048 0.0133 -0.0884 (0.386) (0.000) (0.428) (0.000) (0.640) (0.322) (0.400) (0.135) Public Stock Indicator 1.082* -1.888 -0.0064 0.0098 2.214*** 5.612*** 0.0168 0.1072* (0.082) (0.107) (0.287) (0.174) (0.002) (0.003) (0.368) (0.079) Credit Rating -0.048 -0.452 0.0037 -0.0072 -1.430*** -2.814 -0.0437*** 0.0430 (0.902) (0.702) (0.350) (0.199) (0.000) (0.115) (0.000) (0.377) # Bookrunners 0.027 -0.399* -0.0010 -0.0014 -0.148 -1.090*** -0.0018 -0.0191 (0.808) (0.061) (0.307) (0.369) (0.244) (0.003) (0.585) (0.146) S.D. of Offering Date Prices for Issues (q-1) 0.924 8.108 -0.2050** -0.1188 30.838*** 43.443 0.7125*** 0.0100 (0.926) (0.612) (0.017) (0.297) (0.001) (0.133) (0.001) (0.991) % Chg. Ave. Daily VIX (t-1, t-5 to t) -0.350** -0.433 0.0024* -0.0051** -0.432*** 0.374 -0.0047 0.0514*** (0.035) (0.228) (0.069) (0.035) (0.001) (0.573) (0.181) (0.005) Ave. Daily VIX (t-1 to t-5) 0.016 -0.339** 0.0017*** 0.0010 -0.042 -0.619** 0.0018 -0.0057 (0.834) (0.027) (0.005) (0.282) (0.489) (0.011) (0.242) (0.482) Ave. Corp Bond Index Return (t-1 to t-5) 204.771 85.728 1.6668 6.6310** 25.197 -2,681.270*** -12.2562* -64.5569** (0.387) (0.891) (0.504) (0.036) (0.921) (0.003) (0.063) (0.021) # Corporate Bond Offerings (t-1 to t-5) 0.042 -0.022 -0.0001 0.0005 0.014 0.213 -0.0008 0.0086 (0.267) (0.905) (0.824) (0.662) (0.743) (0.413) (0.455) (0.286) Ave. 3-Month Libor (t-1 to t-5) -1.913** -4.512*** -0.0187** -0.0211* 1.608* -2.211 0.0675*** 0.0178 (0.017) (0.007) (0.014) (0.057) (0.057) (0.472) (0.003) (0.826) Constant 44.547*** 135.477*** -0.1817** -0.6367*** -15.959** 6.611 -0.1507 -3.0802*** (0.000) (0.000) (0.037) (0.000) (0.037) (0.815) (0.442) (0.000) Observations 3,498 1,886 3,557 1,936 3,181 395 3,188 395 Adjusted R-squared 0.0441 0.0566 0.0463 0.119 0.254 0.156 0.127 0.0983 50 Appendix I Variable Definitions Variable Definition 144A Indicator Indicator variable for non-public 144A offerings. Ave. 3-Month Libor (t-1 to t-5) Average 3-Month Libor in the week prior to the offering date. Ave. Corp Bond Index Return (t-1 to t-5) Average BofA Merrill Lynch U.S. Corporate Total Return Index return in the week prior to the offering. Ave. Daily VIX (t-1 to t-5) Average VIX level in the week prior to the offering. Ave. Stock Market Index Return (t-1 to t-5) Average S&P 500 stock index return in the week prior to the offering. Bookrunners in Top 10 Of the 34 large dealers for which we can link syndicate names in SDC to TRACE transactions, this is the top 10 bookrunners by issue volume over the full 2010 to 2018 sample period. Based on SDC data, the portion of the offering amount allocated to all bookrunners. This variable Bookrunner Member Underwriting is reported both in dollar amounts for each bookrunner (in millions) and a percent allocation Commitment which is the bookrunner allocation amount scaled by the sum of all bookrunner allocation amounts. Credit Rating Credit rating data is from FISD and range from 1 (highest credit quality issuers) to 7 (lowest credit quality issuers). We use Moody ratings, except when the Moody rating is unavailable (i.e., set to 99 or reported after the issue date); in these instances we use the Standard & Poor rating. For issuers without credit rating data, we assign investment grade issues a 4 and high yield issues a 7.

First issue reported in FISD Indicator Indicator variable set to 1 if first issue by issuer (based on issuer_id) in FISD database. Indicator variable set to 1 if the issuer has not completed a corporate bond offering within two First Issue w/in 2 Years Indicator years of the current offering. Gross Spread The underwriting commission; the difference between the price that the issuer receives for its securities and the price that investors pay for them. Data is obtained from Mergent FISD. The gross spread is reported both in dollar amounts and as a percentage of the offering amount. Industrial Indicator variable for industrial industry issues based on Mergent FISD Industry Group. Issues 100% Offset This measure is computed only for issues with positive overallocation. We calculate the signed net order flow of the syndicate for investment grade (lead left for high yield) issues, which is the difference between secondary market purchases and sales during the period. The percent offset is the signed order flow divided by the dollar overallocation if net buys > 0, otherwise the percent offset is set to zero. If the syndicate/lead left does not trade during the period, percent offset is set to zero. Percent offset is bound between 0 and 1. Percent offset remains at 100% once the overallocation is fully offset (covered) during a period. Large Overallocation Indicator variable for overallocation greater than or equal to 102%. Based on SDC data, the portion of the offering amount allocated to the largest bookrunner in Largest Bookrunner Underwriting terms of allocation amount. This variable is scaled by the sum of all bookrunner allocation Commitment amounts. Multiple Tranche Indicator variable if the issue is part of a larger tranched issue. Negative Day Return Negative Day Return is set to 1 when underpricing < 0, otherwise 0. Net Signed Position Change We calculate the signed net order flow of the syndicate for investment grade (lead left for high yield) issues, which is the difference between secondary market purchases and sales during the period. This measure does not incorporate the overallocation (the underwriter's short after the issue).

51 Variable Definition Number of Bookrunners Number of bookrunners for the issue from SDC. Number of Corporate Bond Offerings Number of corporate bond offerings in the week prior to the issue. Totals are based on (t-1 to t-5) investment grade offerings for investment grade issues and high yield offerings for high yield issues. Number of Managers Number of managers for the issue from SDC. Number Primary Trades Number of primary placement (“P1”) transactions in the enhanced TRACE data for each issue. Offering Spread over Benchmark Treasury Yield spread over a comparable maturity treasury at offering. Data obtained from SDC. Overallocation We first sum dealer sell quantities (dealer sells to a customer or to another dealer) across primary placement (“P1”) transactions in the enhanced TRACE data; P1 interdealer trades between syndicate members are excluded. Dollar overallocation is the difference between the sum of the dealer sell quantities and the offering amount. Percent overallocation is dollar overallocation scaled by the offering amount. Percent Volume Buys Percent of dealer volume that is secondary market purchases from customers or in the interdealer market. 2 Primary Herfindahl For each P1 trade i , the sum of each (P1 tradei/total P1 trades) . Profit/Loss from Covering Overallocation If the dollar overallocation is positive and syndicate/lead dealers are net buyers, the profit/loss from covering the overallocation is: minimum (dollar overallocation, net order flow) * (offer price – weighted average buy price). Once the overallocation has been fully covered, this measure is set to zero. If dealers are net sellers, this measure is set to zero. Profit/Loss from Round-Trip Volume Round-trip volume is the minimum of total purchases and total sales over a period. Trading profits/losses from round-trip volume is the weighted average spread for the syndicate/lead left multiplied by round-trip volume. Public Stock Indicator Indicator variable for issuers with publicly traded stock. S.D. of Offering Date Prices for Issues (q-1) Median standard deviation of secondary market prices on the offering date and day after the offering date for offerings in the previous quarter. This measure is based on investment grade offerings for investment grade issues and high yield offerings for high yield issues. Based on SDC data, the portion of the offering amount allocated to bookrunners and managers. Syndicate Member Underwriting This variable is reported both in dollar amounts for each bookrunner/manager (in millions) and a Commitment percent allocation which is the bookrunner or manager allocation amount scaled by issue size. Based on SDC, the portion of the offering amount allocated to all bookrunners relative to the Total Offering Underwritten by Bookrunners total allocated to bookrunners and all managers. Total Profit Total profit is the profit/loss from round-trip trading + profit/loss from covering overallocation.

Underpricing Raw returns are computed as the return of the weighted-average price on day n (expressed as the sum of the flat price and accrued interest) relative to the offering price; we then subtract the cumulative index return over n days. For investment grade bonds we use the BofA Merrill Lynch U.S. Corporate Total Return Index and for high yield bonds we use the BofA Merrill Lynch U.S. High Yield Total Return Index. We compute underpricing for day n = 1, 2, 5, and 21.

W.A. Spread / Offer Price Difference between the volume-weighted average dealer sell price and the volume-weighted average dealer buy price scaled by the offering price.

52