at the market (atm) offerings matthew t. billett*...

36
Electronic copy available at: http://ssrn.com/abstract=2178052 1 At The Market (ATM) Offerings Matthew T. Billett* Ioannis V. Floros** Jon A. Garfinkel*** November 19, 2012 ABSTRACT We study a new form of equity offering that emerged in 2008: “At -the-market” (ATM) offerings. Their use has increased significantly, and in 2011 total announced ATM issuance plans were nearly 10% that of traditional SEOs. The key differentiating characteristics of ATMs are their “best-efforts” issuance approach and the ability of firms to “dribble-out” equity issuance at opportune times, often over many months. We document some notable differences in characteristics between ATM and SEO firms. Moreover, ATM issuers appear to experience lower explicit issuance costs (fees), less negative event returns and less negative long-run returns, compared to standard SEO issuers. Finally, we find that firms adjust their dribble-out to reflect time-varying stock return characteristics. Keywords: At-the-market offerings, ATMs, Private equity placements, equity issuances, secondary offerings, seasoned equity offerings, SEOs, external finance, best efforts offerings, issuance costs JEL Classification: G32, G34, G38 We thank Jeffrey Emrich and Dongping Xie for excellent research assistance. All errors are our own. *Kelley School of Business, Indiana University. [email protected] ** Iowa State University and the U.S. Securities and Exchange Commission, [email protected] *** Tippie College of Business, University of Iowa. [email protected]

Upload: others

Post on 05-Aug-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

Electronic copy available at: http://ssrn.com/abstract=2178052

1

At The Market (ATM) Offerings

Matthew T. Billett*

Ioannis V. Floros**

Jon A. Garfinkel***

November 19, 2012

ABSTRACT

We study a new form of equity offering that emerged in 2008: “At-the-market” (ATM) offerings. Their

use has increased significantly, and in 2011 total announced ATM issuance plans were nearly 10% that

of traditional SEOs. The key differentiating characteristics of ATMs are their “best-efforts” issuance

approach and the ability of firms to “dribble-out” equity issuance at opportune times, often over many

months. We document some notable differences in characteristics between ATM and SEO firms.

Moreover, ATM issuers appear to experience lower explicit issuance costs (fees), less negative event

returns and less negative long-run returns, compared to standard SEO issuers. Finally, we find that firms

adjust their dribble-out to reflect time-varying stock return characteristics.

Keywords: At-the-market offerings, ATMs, Private equity placements, equity issuances, secondary offerings, seasoned equity offerings, SEOs, external finance, best efforts offerings, issuance costs

JEL Classification: G32, G34, G38

We thank Jeffrey Emrich and Dongping Xie for excellent research assistance. All errors are our own.

*Kelley School of Business, Indiana University. [email protected] ** Iowa State University and the U.S. Securities and Exchange Commission, [email protected] *** Tippie College of Business, University of Iowa. [email protected]

Page 2: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

Electronic copy available at: http://ssrn.com/abstract=2178052

2

Equity issuance in the U.S. has traditionally followed the firm commitment process (Eckbo,

Masulis, Norli (2007)). Arguments for the dominance of the underwritten offering approach include the

certification that investment banks provide on issuers of uncertain quality, as well as the provision of

liquidity through a network of investors (Booth and Smith (1986), Beatty and Ritter (1986), and others).

However, firms’ needs for certification and/or a sufficient mass of investors to absorb particular issue

sizes, likely vary over time and by the anticipated use of proceeds. For example, an IPO almost certainly

requires more certification than an SEO (Smith (1986)); some SEOs require less certification than others

(Bayless and Chaplinsky (1996)); and larger projects obviously require more capital which augurs for an

SEO.1 Given this variation, it is reasonable to expect that even within the set of firms seeking additional

equity capital, there will be different levels of certification and different quantities of capital needed.

Nevertheless, until recently there appears to be at best marginal variation in firms’ approaches to raising

equity through follow-on issues to the public.2

However, regulatory changes in 2005 and 2008 (Securities Offering Regulation, SOR, and

amendments to forms S-3 and F-3) opened the door to a new form of follow-on equity offering: “at-the-

market” (ATM) issues. These are best-efforts offerings with an additional wrinkle – firms may issue

equity via ATMs in a “dribble-out” fashion where they sell shares directly into the secondary market

over months or even years. The firm pulls these shares “off the shelf” in any quantity they wish,3 at any

time during the “life” of the ATM program.4 They may issue smaller quantities or larger quantities, but

these issues are always sold directly into the secondary market (using a placement agent strictly as a

1 Eckbo, Masulis and Norli (2007) note that average IPO size is 21% smaller than average SEO issuance amounts.

2 The growth in PIPEs is an exception to this. However, our focus is on issues to public investors as they likely face

different information problems than private ones. 3 There is a cap, which we discuss below, but it appears to be far less binding than prior to 2008.

4 Firms either specify an ATM program’s life at announcement, or they may issue the ATM shares up until the shelf

(that the ATM shares were registered under) expires. Below, we analyze firms’ take-down behavior of share issuance over the ATM program lives (section V).

Page 3: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

3

broker) at prevailing market prices. These transactions therefore forego certification and they rely on

the existing stock market demand for the firm’s shares. This paper analyzes the ATM issuance approach

and market.

To facilitate our analysis, we compare ATMs with SEOs. We begin with a basic description of the

anatomies of the two markets, including their size and growth since ATMs appeared. ATMs are

frequently used with 307 announced for over $45 billion during the period 2008-2011. This equates to

roughly 10% of the amount raised using traditional SEOs over the same period. More importantly, we

find that ATM issuance activity is increasing, from total proceeds of roughly $2 billion in 2008 to nearly

$10 billion in 2011. Total SEO issuance activity over the same window declined from approximately

$120 billion in 2008 to $101.6 billion in 2011. Thus the ATM market size has grown in relative

importance compared to SEO issuance activity.5 Importantly, we continue to see growth in this market

during 2012. Through the first ten months of this year, total dollars announced under ATMs (there were

108 events) is already near $12 billion, suggesting a significant rise in activity from last year. We discuss

this below.

Given the size and growth of this market, it is important to understand the economic

motivations of participants. Thus we focus on the key differences between ATMs and SEOs. First, as

noted above ATMs are best efforts offerings while SEOs are traditionally firm commitment offerings.

This implies critical differences in certification that would likely be rather costly for the firms that choose

the ATM route (had they instead done an SEO). Second, the difference in placement venues, with ATMs

going directly into the secondary market and SEOs into the primary, suggests market liquidity will be

particularly important for firms that choose the ATM approach. Third, the single block of immediate

5 From roughly 1% of the amount of SEO proceeds in 2008 to nearly 10% in 2011. A significant portion of ATM

issuance activity (and SEO activity) involves REITs and regulated firms. When we separate out these firms, the remaining ATMs show a similar (roughly five-fold) increase in activity, from $0.5 billion to $2.3 billion from 2008 to 2011. For SEOs, the subset of unregulated corporations issues amounts of $9.3 billion in 2008 and $16.6 billion in 2011. Among unregulated firms, ATM use relative to SEO use has grown over time (from 5% to 14%).

Page 4: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

4

funds generated by an SEO may be less necessary for firms choosing an ATM, where the shares are

dribbled out over several months. Thus, the timing of capital needs is likely different for firms that

choose different issue approaches. Overall, these differences suggest that variation in the costs and

benefits (both explicit and implicit) of issuance under the two approaches will lead firms to select one

approach over the other.

Our main analysis of differential characteristics between ATM and SEO firms is built around the

three differences noted above.6 We begin with a comparison of firm and issue characteristics across the

ATM and SEO samples. The lack of certification suggests ATM firms either perceive little need for, or

would face higher costs of certification. Consistent with the latter, we find that ATM issuers tend to be

smaller with more R&D, and lower sales and profitability ratios. Second, given ATMs are directly sold

into the secondary market we might expect ATM firms to have greater stock liquidity. Indeed,

controlling for other factors, larger turnover associates more with ATM firms than SEO firms, which is

surprising in light of their smaller size. On the other hand, their percentage bid-ask spreads are also

higher, which may be partially driven by their smaller size. Third we investigate differences in

investment behavior. Given SEOs raise a large block of capital all at once, one might expect SEO issuers

to exhibit different financing timing needs. Prior to the financing announcement we see that capital

expenditures, as a fraction of assets, are similar for ATM and SEO firms. In contrast, R&D appears to be a

larger ongoing expense for ATM firms, perhaps suggesting that the dribble-out nature of ATMs better

suits the more staged nature of R&D expenditures.

Given these economic differences in the two samples of firms, ATMs and SEOs likely associate

with different costs (explicit and implicit) of issuance. We address this in several ways. We investigate

explicit fees, event returns, and stock returns preceding and following the events. ATMs show lower

fees. Depending on the ATM sample analyzed (we discuss our subsampling below), explicit fees for

6 The below analyses focus on the sample of unregulated firms, consistent with the common approach among SEO

studies (see e.g. Eckbo and Masulis (1995), Cornett and Tehranian (1995), and Loughran and Ritter (1997)).

Page 5: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

5

ATMs can be lower by anywhere from 25% to over 50%.7 ATMs also have less negative abnormal event

returns than SEOs. The announcement of the ATM program is accompanied by significantly negative

abnormal returns, -3.7%. By comparison, we find that issuance of SEOs8 over this same time period are

accompanied by significantly negative abnormal returns of -4.2% on average. Given firms have the

choice of issuance method, the ATM and SEO wealth effects likely reflect the selection in firm types and

circumstances as noted above. When we match ATMs with SEOs based on firm characteristics, the event

abnormal returns are much closer.

Our comparison of runups prior to ATMs and SEOs is predicated on two factors. One is the

simple enduring regularity that SEOs see sizeable runup in stock price prior to the announcement (see

e.g. Loughran and Ritter (1995)). The other factor is the dribble-out option which may affect firms’

incentives to issue following a stock price runup. We find that average runup prior to SEOs is

approximately 25% (41%) over the 3 (12) months prior to SEO issuance. By contrast, ATMs’ average

runup is 7% (21%) over the 3 (12) months prior to their program announcements. This difference is

rather telling about the market timing hypothesis. Firms conducting ATMs may feel less need to do so at

the end of a large runup, if they still have the option to issue again (quickly), should the stock price rise

further. Moreover, if the runup is driven by large changes in investment opportunities that require

immediate and large capital infusions, then SEOs may naturally follow runups. Similarly if ATMs tend to

be used for ongoing funding purposes for pre-existing investment opportunities, then they may not be

preceded by large runups.

7 We are comparing agent cash fees for ATMs with gross spreads for underwritten SEOs.

8 We estimate issue date rather than filing date abnormal returns because of changes in information content of the

two types of dates, likely engendered by the SOR of December 2005. We discuss this further below.

Page 6: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

6

When we investigate the long-run stock performance (LRP) of ATM and SEO issuers, we find the

typical result for SEOs – they underperform peers significantly over the year following the event.9 By

contrast, we see no evidence of underperformance of ATM issuers relative to peers.

We make several contributions to the literature on equity issuance. We provide the first

academic study of ATMs. Given its size relative to the SEO market and given the number and different

types of firms using it, the ATM issuance process is a viable and increasingly popular approach to raising

equity capital. Second, the growth of this market suggests the importance of ATMs is likely to increase

over time. ATM growth far outstrips growth in other equity capital raising approaches, even including

the PIPEs market, since 2008. Third, and perhaps most important, we provide a detailed examination

of actual issuance behavior for ATMs. Since this is the only equity issuance technique that allows for

“dribbling-out” of shares, a description of actual issuance activity is informative.

Finally, the change in regulations in 2005 (SOR) which allowed for immediate issue of shares off

the shelf, opened the door for firms to issue shares under favorable market conditions (a key motive

behind dribbling out shares). The 2008 amendments to Form S-3 issuance likely had a hand in

broadening the set of issuers using ATMs, also speaking to the importance of regulation for capital

acquisition. Thus our research speaks to the importance of regulatory policy for the critical corporate

activity of capital raising.

The remainder of our paper is organized as follows. Section I describes the changes in

regulation (in 2005 and 2008), and how they encouraged ATM issuance activity. Section II describes our

data. Section III highlights differences in issuer and issue characteristics between ATM firms and SEO

firms that issued during the same time period. Section IV presents results from our analysis of returns

around the ATM and SEO events, as well as runups and long-run performance on either side. Section V

9 See for example, Spiess and Affleck-Graves (1999). Eckbo, Masulis and Norli (2007) suggest a resolution of the

puzzle. Aggarwal (2000) provides an alternative explanation of long-run underperformance, but for IPOs.

Page 7: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

7

contains information on firms’ take-down activity over the life of the ATM. Section VI concludes with a

particular eye towards future work in the area of ATMs as a viable capital source and its drivers.

I Regulatory Reform

The Securities Offering Reform (SOR) policy was passed by the SEC on December 1, 2005. There

were several broad motives to the reform: to allow more disclosure prior to follow-on equity offerings

and to reduce asymmetric information problems that impede capital formation;10 to define a new

category of issuer – a “well-known seasoned issuer” (WKSI); and to provide more timely information to

investors without mandating delays in the offering process. The latter opened the door to offering

securities from the shelf very quickly after the firm made a decision to do so. In particular, SOR

eliminated the prohibition against immediate takedowns off delayed shelf registration statements.

To understand the importance of the latter for ATMs, we highlight similarities between them

and open-market share repurchases (OMRs). They are in some respects mirror images of each other.

We know that firms which perceive their shares to be undervalued may choose to repurchase them

(Vermaelen (1981) and many others). OMR programs allow firms to move into the market and

repurchase any number of shares at any time at prevailing market prices. They also do not require the

firm to execute any repurchases at all. Thus there are several options inherent in OMR programs: the

timing option, the size option, and the execution option. ATMs offer the same options but with respect

to equity issuance. The firm chooses if and when to move into the market and sell shares at prevailing

market prices, and they decide on quantity. The removal of delays from decision to issuance enhances

the timing option’s value.

However, SOR did not appear to catalyze ATM issuance activity. One possible explanation for

this was that eligible firms did not perceive the value of dribbling out shares, as exceeding the implicit

10

Clinton, White and Woidtke (2011) and Shroff, Sun, White and Zhang (2012).

Page 8: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

8

and explicit costs of foregoing the (underwriter) benefits of firm commitment offerings. For large firms,

this may be due to limitations on the size of issuance off the shelf. The SEC specifically noted concerns

with limiting the utility of shelf by constraining the size of issuance off it. For smaller firms, the door

may not even have been open to issue shares off the shelf, because of binding rules on issuer

characteristics.

Both of these concerns were addressed in January 2008 with revisions to requirements

governing issuance via forms S-3 and F-3. For our purposes, the key revisions encouraging ATM activity

broadened the set of companies eligible to issue securities off the shelf, and increased the allowable size

of issuances. Regarding the former, the SEC removed the “public float” restriction to defining WKSI

companies, as long as the issuers met other eligibility conditions for the use of Form S-3. This had the

net effect of allowing companies with less than $75 million in public float to issue via the shelf.

Commenters on the SEC’s proposed policy welcomed expansion of Form S-3 eligibility, noting potential

enhancement to smaller companies’ access to capital.

The SEC further amended regulations that had previously restricted the value of securities that

could be sold in an ATM to 20% of the issuer’s public float. The new policy allows for fully one third of

public float to be issued within any 12 month period. It is this latter change that opened the door to

larger issues (or more total issuance activity over the course of an ATM program) by qualified firms

(WKSIs).

In sum, the 2005 SOR removed wait times for securities issuance off the shelf. The lack of wait-

time from decision to execution of issuance is likely important to realization of one of the purported

benefits of ATMs – allowing firms to issue shares under “favorable” market conditions. The 2008

changes to forms S-3 and F-3 increased both the breadth of companies eligible to issue securities and

the allowed issuance amount relative to the firm’s public float. Given fixed costs to issuance, the

increase in allowable issue size allowed firms to amortize that cost over larger (total over time) capital

Page 9: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

9

raises. It also likely made ATM issuance more attractive simply because the amount of capital able to be

raised became significant. Combined with the wider set of eligible firms, total ATM activity would be

expected to rise.

II Data

A. Sources

Our study is built around two datasets of follow-on equity issuance: secondary equity public

offerings (SEOs) data and At-the-Market offerings (ATMs). The sample window spans the period of

1/1/2008 to 12/31/2011. Our SEOs are drawn from the Securities Data Corporation (SDC) database (U.S.

common stock issuances). Our sample of ATMs is primarily hand-collected, but also draws from

DealFlow Media’s (DFM) PrivateRaise database in 2011 to cross-check our hand-collection. Our SEO

sample begins with 703 completed common stock SEOs and our initial ATMs sample includes 307 closed

equity agreements.

For our hand-collection of ATM data, we use the Knowledge Mosaic platform to search all 8-K

and 6-K filings searching for the following keywords: "at-the-market", "at the market", "controlled

equity offering", "sales agency agreement", "distribution agreement". Also, we search for "ordinary

brokers" information in 8-Ks and 6-Ks that are non-registration statements (in order to avoid getting all

424B2 filings). We account for any reinstitutions or amendments of earlier ATM programs (we find 12 of

them in total).11

Our ATMs data includes the following fields: name of the issuer, the closing date, the placement

status, the planned issuance amount (available for 234 of the total number of 307 ATMs), the closing

and current trading symbol, the closing trading platform, SIC code, the issuer's country and state, the

11

An amendment is considered a revisiting of an initial equity sales agreement whereby the agreement's details change slightly. A reinstitution is a past agreement that is re-activated, usually due to the shelf-registration window expiring.

Page 10: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

10

closing market capitalization, the market price at closing, the planned use of proceeds, the roster of

placement agents together with the agent fees charged. Finally, we obtain the commitment period

within which the issuing company commits itself to dribble-out all (or some) of the ATM shares

(available for 59 ATM transactions). All of these ATM specifics are available in Item 1.01 of the

respective 8-K/6-K filing. Using factiva.com announcements, we gather the ATMs’ announcement dates

(we find that only 6 ATMs have announcement dates preceding the closing date).

Our SEOs sample is drawn from the SDC database and includes only common stock offerings.

Specifically, the filtering criteria we employ are the following (in parentheses we offer the available

number of observations after imposing each criterion): a) all follow-on offerings with an issue date

within 1/1/2008-12/31/2011 (2,688), b) that are of a firm commitment type (1,874), c) with no rights

issues (1,874), d) with issuer SIC code not including regulated industries or the financial industry (1,105),

e) where the issuer is traded on any of the main U.S. stock exchanges (NYSE, AMEX, NASDAQ, NASDAQ

SmallCap) (983), f) with no unit issues (983), g) with no closed-end funds (983), h) with no LBO or RLBO

firms (983), i) with no limited partnerships (901), j) with no simultaneous international offerings (883),

and finally k) with an offering price exceeding $ 5 (703). These screens are broadly consistent with the

extant literature studying SEO activity.

Requiring CRSP and Compustat data trims our ATMs sample to 300 and our SEO sample to 499

observations. To facilitate comparisons of ATMs with SEOs, we create two subsamples of our 300 ATMs:

those with no REITs issuers (189 observations) and the subset of these with no regulated and no (other)

financial firms (133 observations). We do this to better understand ATM activity by these very different

types of firms.

Page 11: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

11

B. Market Differences

Table I presents important information differentiating ATMs from SEOs. Panel A highlights

procedural differences between the two. ATM offering shares are sold strictly into the secondary

market. A placement agent is chosen by the firm and essentially acts as a broker of the shares in sales

on the open secondary market. By contrast, SEO shares are sold in primary market transactions. The

primary market sale of shares typically involves firm commitment, another distinct difference from

ATMs. Finally, ATM issuance programs may be executed over time with only a fraction of shares sold

during each visit (by the placement agent) to the secondary market. By contrast, (firm commitment)

SEOs involve the issuance of shares in a single transaction. These three differences between ATM

issuance and SEOs present new opportunities to companies seeking additional equity capital.

Evidence consistent with this is seen in Panel B of Table I. Beginning with 2008 (the start of ATM

market activity), we compare issuance activity for SEOs with ATMs. We start with ATMs. In both

frequency and proceeds, ATM activity is clearly gaining rapidly in importance. From 2008 through 2011,

announced ATM issuance programs grew from 26 programs to 115 programs. Moreover, total proceeds

grew by 412% over the same period.

As noted above, regulated firms and especially REITs comprise significant portions of ATM

activity. Focusing our attention on ATMs outside of these groups, we still see dramatic rises in activity.

2008 saw only eight ATMs, but this rose monotonically to 56 programs in 2011. Total proceeds too grew

from one-half billion dollars in 2008 to over $2.3 billion in 2011. The year 2009 saw a particular spike in

issuance proceeds, despite fewer ATMs than in the next two years, with total proceeds of over $3.3

billion. However, much of this was concentrated in a few very large programs ($1.65 billion across three

programs and $2.8 billion allocated across nine ATMs).

SEO activity shows a different pattern over time. Across all SEOs (including all industries),

issuance actually fell from $120 billion in 2008 to $101 billion in 2011. Within the subsample of non-

Page 12: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

12

regulated firm SEOs, proceeds rose moderately from $9.3 billion in 2008 to $16.6 billion in 2011. Taken

together with the growth in ATM proceeds, the relative importance of ATMs to SEOs in capital

acquisition is clearly rising. As noted earlier, across all issuer types, it rises from 1% to nearly 10% and

among non-REIT non-regulated firms from 5% to almost 14%. These numbers highlight our expectation

that ATMs will continue to be a viable equity issuance technique.

Confirming preliminary evidence of this prediction is seen in 2012 ATM issuance activity.

Though not tabled, we note here that through the first ten months of 2012 total ATM programs

announced equal 108 – nearly the total in 2011. Moreover, total planned proceeds are almost $12

billion. This exceeds 2011 total proceeds by over $2 billion, again indicating a continued high rate of

growth in ATM usage by firms. These numbers strongly suggest the viability of ATMs as a permanent

fixture in the US equity issuance landscape.

III Issuer and Issue Characteristics of ATMs and SEOs

A. Issuer Characteristics

Table II presents information on ATM issuer characteristics, ex-ante the announcement of the

ATM program. For comparison, it also presents SEO issuer characteristics prior to SEO issuance.

Following the extant literature, we focus on the subsamples of non-REIT non-regulated firms. We

examine asset size, leverage, R&D expenditures, market to book, cash holdings, cash burn, sales,

income, and capital expenditures. Our variables are defined as follows:

TA: Total assets. Lvg: The sum of short-term and long-term debt, all divided by total assets. R&D: R&D expenditures divided by total assets. If R&D is missing, we set it equal to zero. M/B: Market value of equity divided by book value of equity. Cash: Cash plus cash equivalents, all divided by total assets.

Page 13: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

13

Burn: Cash burn rate is the absolute value of operating income before depreciation divided by the sum of cash and cash equivalents. When the income number is positive, burn is set equal to zero. This follows Chaplinsky and Haushalter (2010).

Sales: Revenues divided by total assets. EBITDA: EBITDA divided by total assets. CAPEX: Capital expenditures divided by total assets.

Ex-ante, there are important differences in firm characteristics for the ATM vs. SEO samples.

We discuss means first. ATM firms are smaller than SEO firms in terms of TA ($1.2 billion versus $3.3

billion). They show higher average expenditures on R&D relative to assets (37% versus 9%). They have

lower sales relative to assets (42% vs. 79%) and lower EBITDA to assets (-46% vs. 1%). They have more

volatile stock (σret of 6.1% vs. 4.8%). They are a bit younger (11.4 vs. 12.7 years listed on CRSP)12 and

they carry more cash relative to assets (38% vs. 25%). In short, ATM firms have many of the classic

markings of higher growth firms relative to SEO firms. Confirming evidence of this is seen in their M/B

equity average value of 5.1 relative to SEO firms’ average M/B equity of 3.26. Overall, these differences

imply the selection of high-certification-cost firms into ATMs.

On the other hand, there are notable similarities between our samples of ATM and SEO firms.

Average LVG is very close (50% vs. 48%), as is CAPEX relative to assets (7.6% vs. 8%). These results are

inconsistent with the view that ATM firms are more growth oriented than SEO firms. Finally, the cash

burn rate appears higher on average for SEO firms (5.66) relative to ATM firms (3.0). However, this may

be due the nature of the burn rate construction (which follows Chaplinsky and Haushalter (2010)).

Because it is set to zero for positive operating income firms, only negative operating income contributes

to the burn rate. For larger firms that lose money, they are likely not in a growth phase, whereas

12

Though this difference is not significant. Moreover, the median ages indicate ATM firms are significantly older.

Page 14: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

14

smaller (younger) firms that lose money are more likely to be outspending their revenues because they

are growing quickly.13

Median figures tell a similar story to the means, with a few exceptions. Notably, M/B equity

now looks similar for ATM (2.6) and SEO (2.25) firms. There appear to be similar growth rates for the

middle firms in each issuance group. Taken together with the similar LVG and (mean) CAPEX across

ATM and SEO firms, this suggests variability in the factors that cause firms to choose one issuance

approach over another.

B. Issue Characteristics

We also examine characteristics of the ATM issue and some pre-event (but near the event date)

characteristics for ATM firms and SEO firms.14 We characterize all of these items as “issue

characteristics” because the calculations are always near the event date.15 The extant literature on SEOs

typically focuses on the window [-25, -3] preceding the SEO when studying firm liquidity indicators. We

define our variables of analysis as follows:

% spread: bid-ask spread divided by stock price [averaged over -25, -3] Turnover: volume divided by shares outstanding, per day [averaged over -25, -3] N_trade: number of trades per day [averaged over -25, -3] Proceeds: dollar proceeds planned for issue (ATM) or actually issued (SEO) Procs/mvequ: proceeds divided by market value of equity (on event day) %Fees(sprd): for SEOs, gross spread over proceeds; for ATMs, agent cash fees over planned proceeds

13

Cash burn is higher in the median for ATM firms (.55) than for SEO firms (0). 14

As with our analysis of issuer characteristics, we focus on the samples of non-REIT, non-regulated firms. 15

It’s important to emphasize here that the event date for ATMs is the announcement date of the program, whereas for SEOs it’s the issue date of the shares. The reasons for this are discussed in greater detail below (when we examine event returns), but the short of it is that SOR changed SEO activity such that the “event” is likely to be the issue date, since 2005 (Clinton, White and Woidtke (2011) and Shroff, Sun, White and Zhang (2012)).

Page 15: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

15

σret : Standard deviation of daily stock returns over trading days [-25, -3] where day 0 is the announcement date of ATM program or issuance date of SEO.

Table III presents our results. ATMs have somewhat higher percentage bid-ask spreads on the

event day compared to SEOs (0.008 vs. 0.003). However, this is in part due to the lower price of ATM

stocks relative to SEO stocks ($8.37 vs. $22.68).16 We see similar turnover for ATMs relative to SEOs (2%

of shares outstanding for each sample’s mean) as well as similar numbers of trades (3,963 vs. 3,916). In

light of the fact that ATM firms are smaller (Table II), the similar trading activity suggests ATM events

associate with better liquidity (notwithstanding the higher percentage spreads).

We also examine issuance proceeds (announced plans for ATMs and actual issue amounts for

SEOs). ATM firms announce smaller planned proceeds than SEOs ($69 million vs. $231.5 million),

consistent with ATM firms requiring smaller blocks of capital at once (and over the life of the program).

Similarly, ATM proceeds are smaller amounts relative to market value of equity as well (23% vs. 32%).

Finally, Table III offers information on explicit costs of issuance to ATMs and SEOs. For ATMs,

where there is no firm commitment by the underwriter, the costs are “cash agent fees”. These are paid

to the selling agent who essentially acts as broker of the shares on the open market. Agent cash fees

average 3.37% of planned proceeds (total committed amount). The median is 3%. SEOs are

underwritten and therefore involve more investment bank risk. Gross spreads average 4.59% of

proceeds. If we hypothesize an issue amount of $100 million (somewhat between ATM average and

SEO average proceeds), the extra 1.22% of fees translates into an extra $1.22 million dollars out of the

firm’s pocket.17 This lower direct expense to ATM issuance is one potential driver of the popularity of

this issuance approach.18

16

ATM firms have lower raw bid-ask spreads when compared to SEO firms (.02 vs. .04 in the mean). 17

This ignores potential costs of lacking a firm commitment, but it is illustrative. 18

Notably, when we examine fees for the full samples of ATMs (including regulated and financial firms), they are even smaller. Mean (median) ATM fees equal 2.57% (2%). This differs markedly from the spreads on SEOs for the full sample: the mean (median) SEO spread equals 4.69% (5.08%).

Page 16: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

16

C. Logit explaining choice between ATM and SEO

Given that both firm and issue ex-ante characteristics correlate with follow-on equity issuance

approach, we estimate a logit with variables from both characteristic groups. The dependent variable

equals one if the firm does an ATM and equals zero for an SEO. Table IV presents our results. They are

mostly consistent with the inferences from univariate comparisons. Ex-ante, ATM firms are smaller with

higher percentage spreads but also higher turnover. They have more volatile stock returns and less

positive runup prior to the event. They are less profitable.

Notably, the univariate evidence of more R&D among ATM firms does not survive the inclusion

of all characteristics. Further investigation reveals that the inclusion of the profitability measure as a

regressor associates with a much smaller coefficient on R&D. The two are highly correlated. Also we

see that ATM firms are associated with lower tobin’s q than SEO firms. Thus, the characterization of

ATM firms as more growth oriented would be premature. Consistent with this, ATM firms more often

pay a dividend (the coefficient on the dividend dummy is significantly positive in the logit).

Taken together, the logit results suggest a nuanced view of the commonalities among ATM

firms/events vs. SEO firms/events. The motives for conducting an ATM discussed above receive some

support from the data. Given the ability of firms to dribble-out shares at opportune times in the market

through an ATM, more volatility suggests a higher value to this timing option. This may also explain the

negative coefficient on runup. Too, the greater turnover among ATM firms suggests better liquidity,

which is consistent with ATMs’ reliance on market conditions for placement; though the larger

coefficient on percentage spread gives us some pause in this conclusion. ATM firms also more often

have paid a dividend. They may be less financially constrained and therefore feel less need to raise large

bulbs of capital at once. Finally, despite the more common dividend payment, the smaller size and

lower profits of ATM firms suggest higher certification costs should they instead choose to do an SEO.

This may also help explain their choice.

Page 17: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

17

On the other hand, ATM firms indeed have higher percentage spreads which may inhibit

liquidity. Moreover, they have lower tobin’s Q, inconsistent with the idea that ATM firms are more

growth oriented. However, it is consistent with the idea that greater growth opportunities imply greater

investment opportunities (requiring more capital) to be taken, which augurs for an SEO. Overall, there

appear to be multiple considerations that influence the selection of either an ATM or an SEO as follow-

on equity offering approach.

IV. Returns Before, Around and After ATMs and SEOs

This section explores event returns, runups prior to them, and long-run ex-post returns for the

samples of non-REIT, non-regulated ATMs and SEOs. The events we study for ATMs are the

announcements of the issue plan. We cannot observe the various “take-down” dates as they are not

reported (even ex-post). In section V, we explore take-down behavior for ATMs, but we can only obtain

information on a quarterly basis.19 For SEOs, the event is the actual issue date, rather than the

announcement of putting shares on the shelf for eventual SEO issuance (as firms may do). As noted in

footnote 14, the passage of SOR changed the information environment and markets now appear to

revise their priors about firm valuation when the firm actually issues shares.

A. Event Returns

Table V, Panel A presents abnormal event returns for both ATMs and SEOs. We estimate the

abnormal returns using both a market model and simple market-adjusted returns. The event window is

[-1,1] and for the market model the estimation window is [-268, -16]. We are able to obtain stock return

data for 132 ATM events. Mean abnormal returns under both methodologies are very similar at -3.74%

and -3.72% respectively for market model and market-adjusted.

19

Firms are required to report their take-down activity under the ATM on form 10-Q.

Page 18: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

18

For SEOs, event returns are noticeably more negative, at -4.56% (market model) and -4.14%

(market-adjusted returns). These abnormal returns are significantly worse than ATM event returns, with

z-statistics of -2.6 and -2.3 respectively. Combined with higher explicit issue costs to SEOs, the worse

abnormal returns to SEOs strongly suggest a cost savings motive behind some firms’ use of ATMs.20

However, selection may be driving this difference. The last two lines of Panel A indicate event ARs are

similar for matched (on size and b/m) subsamples of ATMs and SEOs (-2.98% and -2.86% respectively).

B. Runup (pre-event)

Given the sizable differences in event returns for ATMs and SEOs, we investigate one of the

proposed underpinnings of them – stock price runups prior to the events. As Ritter (2003) summarizes,

there is a negative correlation between prior stock price performance and SEO event returns. We

therefore examine stock returns in the several months prior to both ATM and SEO events. We find that

they’re dramatically different.

We calculate Runup as the cumulation of daily market-adjusted returns over various windows

preceding the ATM and SEO events. The windows we analyze are [-252, -60], [-252, -30], [-90, -3], [-60, -

3], and [-30, -3]. We focus our discussion on the [-252, -30] and [-90, -3] windows, but the inference is

the same for all windows. SEOs see far larger Runup prior to their event than ATMs do. Our results are

presented in Table V, Panel B.

For SEOs, their stock price Runup over one year preceding the event is broken down into two

windows: [-252, -30] and [-30, -3]. The returns are respectively 32% and 9%, with the cumulative effect

equaling 42.6%. By comparison, for ATMs the numbers tell a very different story, with a one-year runup

of only 21.3%. Runup is far smaller prior to the event, with most of the difference coming in the 11

20

It’s also worth noting at this point that SEO announcement returns are significantly negative as well (though not as much as the issue date returns). The average abnormal return using the market model is -1.7%. This further reduces shareholder wealth among SEO firms.

Page 19: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

19

months (14% return) preceding the final month (7% return) before announcement. Still, the 7% final

month Runup before ATMs is significantly smaller than the 9% final month Runup before SEOs. Overall,

SEOs show far stronger performance prior to the event than ATMs.

We can interpret the difference in Runups in several ways. First, firms conducting ATMs have a

timing option that SEO issuers do not. The ATM announcement does not require issuance immediately,

so Runup need not be so large as to make it costly to eschew issuance now. An ATM firm may wait until

the stock rises further to issue and take advantage of a higher price. Second, even if the firm issues

under an ATM and then observes a further Runup in stock price, it still has the option to issue again and

take advantage of that further rise in price. Overall, the timing option affords ATM firms the ability to

announce and even execute issues after smaller Runups than in the case of SEOs.

While the above assumes a market-timing element to the firm’s decision, runup may associate with

ATM issuance regardless of active timing. For example, an alternative interpretation is that Runup

results from the market perceiving new investment opportunities for SEO firms to a greater degree than

for ATM firms. Indeed, the negative relationship between Tobin’s q and ATM (vs. SEO) incidence is

consistent with this. Such a situation may arise if a shock to investment opportunity occurs that requires

a large scale immediate investment. This would result in a large stock price response to the shock,

runup, and a likely SEO. If ATMs associate with more ongoing investment, like meeting working capital

needs, continuing R&D, etc., then perhaps a lower runup is expected.

C. Long-run performance

We study the long-run performance of firms’ stock following ATM and SEO events. We use the

BHAR approach and eschew the calendar time approach of Fama and French for two reasons. First, our

limited sample of events would lead to some months of the Fama and French test portfolios containing

Page 20: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

20

very few observations. Second, the short time series of our data would imply a small sample for our

time-series regression under the Fama and French approach.

Loughran and Ritter (2000) advocate for the use of the event-based approach; peer-adjusted

returns to measure long-run performance. We follow the implementation of Spiess and Affleck-Graves

(1999), matching on size (within 10% of market cap) and then book-to-market equity to minimize the

sum of absolute valued differences in both characteristics. We are able to find 358 peers for our SEO

analysis and 57 peers for our ATM calculations.21

Table V, Panel C contains our results. We show estimates of one-year buy-and-hold returns for

event firms and peer firms in addition to the BHARs. For SEOs, the event firms earn 7.18% over the year

following, while their peers see average returns of 17.04%. The difference of -9.86% is significantly

different from zero with a t-statistic of -2.38. The typical result that SEO firms underperform the

benchmark (their peers) is seen in our sample too.

ATMs exhibit event-firm returns of 9.01% over the year following announcement. This is a bit

larger than SEO firms. However, their peers show only 5.065% returns over the same window, with the

net effect being positive BHARs on average (3.94%) for ATM firms. This is not significantly different from

zero, but it highlights the difference between SEO and ATM firms’ long-run performance. Obviously,

some of this is due to selection (the peers’ returns are vastly different across our two samples), but this

does not mean the BHAR differences are unimportant. If the different peers imply different risks, this is

worth noting. ATM firms may be less risky, though some of our firm characteristics suggest otherwise.

The alternative view is more compelling for us. The ATM events may be at different times than SEO

events. This is in line with our Runup results. As we investigate take-down behavior for ATMs, this will

also highlight timing decisions by these issuers.

21

There are three years of data in these tests because we require one year of stock returns ex-post and we do not have CRSP data yet for 2012.

Page 21: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

21

V. Actual Issuance Behavior in ATM Programs

Given ATMs’ flexibility to be executed in a dribble-out fashion, we investigate firms’ actual

takedown behavior under their ATM programs. We collect data on firms’ issuance of equity under the

ATM program from their 10-Q filings (or 10-k in the case of the fiscal year end filing). The 10-Q provides

aggregated (across all issues in the quarter) information on takedown activity. The sample is comprised

of ATM programs of non-REIT firms because of REITs’ different characteristics. We also restrict our

focus to those ATMs that we were able to find price data on from CRSP and that we were able to match

with a comparable SEO firm (in terms of size and b/m equity). This is in deference to the theme of our

paper focusing on differences in issuance approach.

A. Univariate Statistics

We focus on three related measures of takedown activity. They are as follows.

Takedown %age: Equals total number of shares actually issued during the ATM program, divided

by the announced number of shares that the firm planned to issue under the ATM program.

%age active quarters: Equals the number of quarters the firm actually took down equity from the shelf

and issued it, divided by the number of quarters in the announced life of the ATM program. If there is no announced ATM life (commitment period), we assume the life equals the shelf’s life, for the shelf under which the ATM was issued.

Takedown duration: Equals the number of quarters it took the firm to complete the takedown of

shares under the ATM. Even if a firm did not issue shares during a particular quarter during its ATM program, we count that quarter as long as it occurs before the ATM is completed.

Table VI presents means, medians and standard deviations of the above three variables, across

various samples. For the full sample of firms (in Panel A), takedown percentage is slightly less than 50%

(49% mean, 43% median), with substantial variation (41% standard deviation). The typical firm takes

down less than half its announced ATM amount. The typical firm’s percentage of active quarters is very

Page 22: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

22

close to 50%. About a third of firms complete their takedown of the announced quantity of shares in

the ATM, and for them it takes a little over one year to do so.

Panels B and C report statistics for (respectively) the subsamples of firms that announced their

ATM prior to or after the start of calendar year 2010. Given we typically see ATM programs of two or

more years, some ATMs announced after the start of calendar 2010 might be expected to show less

measured issuance activity since we are not completely through 2012 yet.22

However, the evidence suggests the opposite. ATMs announced prior to 2010 show lower

takedown percentages in the mean and median. This suggests more aggressive issuance behavior

(relative to announced intentions) among more recent users of the ATM issuance process. Confirming

evidence is seen in the shorter duration to completion among those ATMs with full takedown of the

announced number of shares. Finally, there is little difference between the percentage of active

quarters across the two sub-samples.

Panel D reports the cumulative takedown percentages, by quarter number following the

announcement. The means suggest an apparent preference to accomplish significant takedown in the

first two quarters (nearly 25%), followed by less activity over the second half of the year. The second

year sees somewhat less activity (15%) than the first. This general downward trend presents in our

regression results next.

B. Regressions

Numerous factors may influence firms’ preferences to execute takedowns in ATM programs.

Our Tobit regressions analyze such tendencies. We investigate the influence of time, prior takedown

activity, stock volatility, and prior stock performance using a methodology similar to Stephens and

22

Also, splitting the sample this way assigns similar calendar time to each window.

Page 23: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

23

Weisbach (1998) who study the timing of actual share repurchases in OMR programs. Table VII presents

our results.

We find several interesting results. First, the later in a “program” a firm is, the less they take

down. This is consistent with the univariate evidence in Table VI (panel D). It suggests a preference for

funds immediately upon setting up the ATM shelf. However, we hasten to emphasize that this

preference is apparently not as strong as with SEOs. We also see that more shares taken down up to

this quarter of the program encourages more shares taken down in the current quarter. Firms that were

more aggressive in their takedowns remain so throughout the program.

In quarters with greater stock return volatility, ATM firms take down less. This suggests firms

pay particular attention to market conditions to execute their ATM issuance efficiently. Consistent with

this, more positive stock returns in the prior quarter encourage larger takedowns in the current quarter.

The insignificant coefficient on current quarter stock returns is perhaps inconsistent with the attention

to market conditions. However, it is not clear when the takedown occurs relative to the stock return

(since we cannot observe any more than the quarter-long takedown behavior). An alternative

interpretation of the runup results is driven by the market feedback hypothesis. The higher returns

indicate greater investment opportunities to the firm, and it issues shares to take advantage of those

opportunities.

Overall, our analysis of takedown behavior suggests a few conclusions. Firms do vary their

takedown behavior in terms of activity levels and prices around that activity. Moreover, the key

difference between ATMs and other best efforts offerings – the ability to dribble-out shares when

market conditions are favorable – appears to hold empirical content.

Page 24: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

24

VI. Conclusions

We study the anatomy of a new approach to offering equity. At-the-money (ATM) offerings

came into fashion starting in 2008, most likely because of regulatory changes that made such offerings

more attractive. Since then, they have grown significantly to comprise a meaningful portion of the

follow-on equity issuance market.

We contrast ATMs with SEOs as they represent two dramatically different equity issuance

approaches. There are many differences in both the firms that opt into the issuance methods and the

characteristics of the issues. ATM firms are smaller and more growth oriented on average, with more

R&D and cash, but lower sales and profits. At the same time, they have similar leverage and CAPEX. A

notable characteristic of the ATM market is the preponderance of REITs in our sample of issuers. It

seems that varying firm characteristics cannot alone explain the choice of issuance approach.

Another potential explanation for firms’ choice of ATMs is the lower fee structure. Because

these offers are not underwritten with an investment bank committed to buying the shares, fees are

lower. Combined with the less negative announcement returns to ATMs compared to event returns for

SEOs, ATMs look less costly. We also find that long-run performance following the two issuance

approaches is different. It’s significantly negative for SEOs and not significantly different from zero

(though apparently positive) for ATMs.

Finally, we examine firms’ takedown behavior in their ATM programs. The key takeaway from

this analysis is that firms appear to use the timing option to dribble-out shares when market conditions

are favorable. Overall, we believe the ATM issuance approach is an important part of the equity

issuance landscape, and is likely to remain so for the foreseeable future.

Many avenues for future research remain. We have not yet analyzed whether variation in

planned use of proceeds is important. Carlson, Fisher and Giammarino (2010) find that this is an

important sorting variable among SEOs. Perhaps firms use ATMs to accomplish different objectives

Page 25: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

25

depending on the planned use of proceeds (operational, investment or financial). Too, the purported

benefit of flexibility to raise equity capital under favorable market conditions begs the question of

whether firms can effectively identify such conditions. Perhaps future research will be able to pinpoint

the days on which firms (their agents) actually executed takedowns and correlate these with market

microstructure measures of market conditions.

Page 26: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

26

References

Aggarwal, R., 2000, Stabilization Activities by Underwriters after IPOs, Journal of Finance 55, 1075-1103.

Bayless, M., Chaplinsky, S., 1996, Is There a Window of Opportunity for Seasoned Equity Issuance?,

Journal of Finance 51, 253-278.

Beatty, R., Ritter, J., 1986. Investment Banking, Reputation, and the Underpricing of Initial Public

Offerings. Journal of Financial Economics 15, 213-232.

Booth, J., Smith, R., 1986, Capital Raising, Underwriting, and the Certification Hypothesis, Journal of Financial Economics 15, 261-281.

Carlson, M., A. Fisher, R. Giammarino, 2010, SEO Risk Dynamics, Review of Financial Studies, 23, 4026-

4077. Chaplinsky, S., Haushalter, D., 2010. Financing Under Extreme Uncertainty: Evidence from Private

Investments in Public Equities, Review of Financial Studies 23, 2789-2820.

Clinton, S., White, J., Woidtke, T., 2011, Is More Information Better? Deregulation of Voluntary

Disclosure Prior to Seasoned Equity Offerings, University of Tennessee Working Paper.

Eckbo, E., Masulis, R., Norli, O., 2007, Security Offerings, Handbook of Corporate Finance: Empirical Corporate Finance, Volume 1, Elsevier/North-Holland, Chapter 6.

Eckbo, E., Masulis, R., Norli, O., 2000, Seasoned Public Offerings: Resolution of the ‘New Issues Puzzle’,

Journal of Financial Economics 56, 251-291. Loughran, T., Ritter, J., 1995, The New Issues Puzzle, Journal of Finance 50, 23-51. Loughran, T., Ritter, J., 2000, Uniformly Least Powerful Tests of Market Efficiency, Journal of Financial

Economics 55, 361-389. Morrison & Foerster LLP, 2012, Frequently Asked Questions About At-the-Market Offerings, White

Paper. Ritter, Jay, 2003, Investment banking and securities issuance, in G. Constantanides et al., eds.: Handbook

of the Economics of Finance (Elsevier Science). Shroff, N., Sun, A., White, H., Zhang, W., 2012, The Impact of the 2005 Securities Offering Reform on

Voluntary Disclosure Before Seasoned Equity Offerings, Working Paper. Smith, C., 1986, Investment Banking and the Capital Acquisition Process, Journal of Financials Economics

15, 3-29. Spiess, K., Affleck-Graves, J., 1999, The Long-Run Performance of Common Stock Following Debt

Offerings, Journal of Financial Economics 54, 45-73.

Page 27: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

27

Stephens, Clifford P., and Michael S. Weisbach, 1998, Actual share reacquisitions in open market repurchase programs, Journal of Finance 53, 313–333.

Vermaelen, T., 1981, Common Stock Repurchases and Market Signalling: An Empirical Study, Journal of

Financial Economics 9, 139-183.

Page 28: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

28

Table I Comparison of ATMs and SEOs

Panel A presents the main characteristics of two types of equity offerings: At-The-Market offerings (ATMs) and common stock Secondary Public offerings (SEOs). Panel B reports the distribution of completed transactions as well as the total gross proceeds amounts raised for all three sub-samples of ATMs (all ATMs, all ATMs with no REITs and all ATMs with no regulated industries). Information on SEOs and PIPEs number of observations spans the time period of 1995 to 2011. The source for SEOs is the Securities Data Corporation (SDC) database, and for ATMs it is both DealFlow Media's PrivateRaise (for the year 2011) and our hand-collection (for the years 2008, 2009 and 2010). The information on ATMs' public announcements preceding the completion date of the offering type "Announced before completion" are hand-gathered from factiva.com, and for SEOs are drawn from Securities Data Corporation (SDC) The information on the: a) selling platform "Sold in the secondary markets", b) the selling pattern "Sold in increments" and c) the underwriting method "Underwritten" are drawn from the related S.E.C. documents.

Panel A Issuance process

Announced before completion

Sold in the secondary market

Sold in increments

Underwritten

At-The-Market Offerings (ATMs)

We find that 95.4% of ATMs programs are announced on (or after) the closing date of the commencement of the program. The dates of the securities' sale are not announced.

Through ATMs, newly-issued shares are sold to the secondary markets.

YES Through ATMs, newly-issued shares are dribbled out into the trading market through a designated broker-dealer at prevailing market prices. There is a placement agent used that acts on a best efforts basis. In the rare case that the placement agent commits to purchase the issuer's securities for its own account with a view to reselling securities, he does not conduct any roadshows or other solicitations. The placement agent is still liable with respect to material misstatements or omissions in the accompanying shelf registration statement.

Secondary Public Offerings (SEOs)

We find that 89.1% of SEO issuances have their filing date preceding the issue date.

Through SEOs, pure primary or combined primary and secondary shares (with the primary shares proportion being at least 50% of the entire offering) are sold to the secondary markets.

NO SEO issuers use underwriters that act on a firm commitment basis.

Page 29: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

29

Panel B Issuance activity

At the market offerings: ATMs Seasoned Equity Offerings: SEOs

Year # of issues Proceeds (in $ billions) # of issues Proceeds (in $ billions)

All 2008 26 1.92 163 119.66

2009 74 26.14 387 138.72

2010 92 7.42 389 90.54

2011 115 9.85 304 101.56

307 45.33 1,243 450.58

Excluding REITs 2008 15 0.93 136 101.22

2009 47 23.34 348 113.66

2010 56 2.49 329 79.78

2011 71 4.52 231 97.72

189 31.28 1,044 392.38

Excluding REITs and regulated firms 2008 8 0.49 76 30.74

2009 27 3.30 218 39.07

2010 47 1.50 234 34.76

2011 56 2.30 176 46.50

138 7.59 704 151.07

Page 30: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

30

Table II

Issuer characteristics

Table presents and compares the mean and the median values of annual financials of all ATMs and all common stock SEOs without REITs, financials and utilities (regulated industries). Financials are drawn from Compustat (Fundamentals Annual) database as of the year before the two samples. Total leverage ratio is both short-term and long-term debt over total assets, R&D ratio is R&D expenses over total assets (if the total assets value is available then R&D expenses are set equal to zero when missing), market-to-book ratio is market value of equity (product of end-of-year shares outstanding by end-of-year closing price) over book value of equity, cash ratio is cash and cash equivalents over total assets, ROA is net income available to common shareholders over total assets, sales ratio is total revenues over total assets, EBITDA ratio is EBITDA over total assets, CAPEX ratio is capital expenditures over total assets and cash burn rate is the absolute value of the ratio of operating income before depreciation over cash and cash equivalents (when the operating income before depreciation is positive then cash burn rate is set to zero). Total assets values are contemporaneous with the other financial values. The last two columns compare mean and median financial values. The Satterthwaite two-sided t-statistic (Wilcoxon two-sided z-statistic) is presented in parenthesis for the mean and the median values, respectively. a,b,c denote the statistical significance at the 10%, 5% and 1% level.

ATM (N = 124) SEO (N = 669) Difference

Variable mean median mean median t-stat z-stat

TA 1220 66 3277 540 -1.65a -9.15c

Leverage 0.50 0.42 0.48 0.45 0.42 1.51

R&D 0.37 0.08 0.09 0.02 2.96c 3.83c

M/B equity 5.10 2.60 3.26 2.25 1.24 0.68

Cash/TA 0.38 0.31 0.25 0.12 4.07c 3.60c

Sales/TA 0.42 0.22 0.79 0.60 -7.14c -6.88c

Cash Burn 3.00 0.55 5.66 0.00 -0.79 10.40c

EBITDA/TA -0.46 -0.19 0.01 0.10 -4.28c -9.22c

CAPEX/TA 0.08 0.03 0.08 0.04 0.03 2.34b

Page 31: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

31

Table III

Issue characteristics

Table reports offerings' specifics, market-related information and fee-related information. Sample is all ATMs and all common stock SEOs without REITs or any other regulated industries. Mean and median values of market-related data refer to percentage spread (bid-ask spread adjusted by the daily closing price), stock turnover (volume divided by shares outstanding), and number of trades. The values are drawn from 23 trading days [-25, -3], where 0 is the offering's closing date. The mean and median values of offerings' specific information are computed as of the offerings' closing date. Mvequ is market value of equity, as of the event date. The fees' information on the SEOs refers to the gross spread percentage (adjusted by the total gross proceeds amount). On the ATMs, the fees' information refers to the agent cash fee. The Satterthwaite two-sided t-statistic (Wilcoxon two-sided z-statistic) is presented in parenthesis for the mean and the median values, respectively. a,b,c denote the statistical significance at the 10%, 5% and 1% level. *fee/spread data is available for 130 ATMs and 476 SEOs.

ATM (N=132) SEO (N=499) Difference

Variable mean median mean median t-stat z-stat

(B-A Spread)/Price 0.008 0.005 0.003 0.001 -5.02c 10.43c

Turnover .02 .007 .02 .008 -2.08b -0.02

Number of Trades 3,963 882 3,916 1,358 0.05 -2.61c

Proceeds ($millions) 69.00 20.00 231.52 103.70 5.50c -8.62c

Proceeds/Mvequ 0.23 0.15 0.32 0.15 -0.51 3.02c

%Fees (spread)* 3.37 3.00 4.59 4.75 -9.31c -9.40c

σret 0.06 0.05 0.05 0.04 1.67a 3.58c

Page 32: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

32

Table IV Logistic Regression Explaining ATM vs. SEO Event

Table reports estimates from logistic regressions explaining incidence of ATM or SEO. Dependent variable equals one if the observation is an ATM, equals zero if it’s an SEO. Size is total assets (TA). RD_ratio equals R&D expenditure divided by TA. If R&D is missing, we set it equal to zero. Capex_ratio equals capital expenditure divided by TA. Cash_acqu_ratio equals cash expenditures on acquisitions divided by TA. Tobin’s q equals market value of equity plus book value of debt, all over TA. %spread equals bid-ask spread scaled by stock price – the median value over the window [-25,-3]. Turnover equals volume divided by shares outstanding – the median value over the window [-25,-3]. σret equals the standard deviation of stock returns over [-25,-3]. Div_dum is a dummy equal to one if the firm pays a dividend, zero otherwise. Proceeds/mvequ equals announced proceeds scaled by market value of equity. Runup equals the stock’s cumulative market-adjusted return over [-252,-3]. External_finc_needs equals EBITDA less change in CAPEX less change in working capital, scaled by TA. Sales ratio equals sales scaled by TA. EBITDA ratio equals EBITDA scaled by TA. Cash ratio equals cash scaled by TA. Leverage equals the sum of long-term debt and short-term debt, scaled by TA. . a,b,c denote the statistical significance at the 10%, 5% and 1% level.

Coefficient Estimate Chi-square stat

Intercept 2.88 7.11c

Size -0.81 26.78c

RD_ratio 0.44 0.21

Capex_ratio -0.38 0.05

Cash_acqu_ratio -4.33 1.45

Tobin’s q -0.26 8.09c

%spread 95.07 12.57c

Turnover 0.03 9.09c

σret 8.01 4.49b

Div_dum 1.76 16.43c

Proceeds/mvequ -0.09 0.36

Runup -1.21 15.29c

External_finc_needs -0.08 1.06

Sales ratio -0.49 2.37

EBITDA ratio -1.75 4.79b

Cash ratio -1.28 2.47

Leverage

-2 Log Likelihood

-0.64

302.80

1.01

Page 33: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

33

Table V

Stock Returns Before, At, and After Event

Panel A reports the mean cumulative abnormal returns (CARs) of all ATMs, all ATMs with no REITs and all ATMs with no REITs and no regulated industries as well as all common stock SEOs with no REITs and no regulated industries. For the ATMs subsamples we use the first public announcement date as our event date whereas for the common stock SEOs we use the issue date as our event date. We report the CARs using the one-factor model and the CRSP equally-weighted index, respectively. Our event window captures the three trading days surrounding t = 0 event day [-1,+1]. The last two lines of Panel A report the mean cumulative abnormal returns (CARs) of all ATMs with no REITs and no regulated industries when they are matched to common stock SEOs with no REITs and no regulated industries using the propensity score matching methodology. With regards to our propensity score matching model, we use the ATMs subsample as our treatment and the SEOs sample as our control sample (making sure that there are no overlaps). We utilize the nearest neighbour matching approach. We are able to match 115 distinct pair of ATMs/SEOs transactions. We allow replacement for the control sample's transactions. Statistical significance is based on the Patell t-test. The number of observations is presented on the second column. Returns are computed over the 3-trading day event window [-1,+1] according to the one-factor market model using the CRSP equal-weighted index as the market proxy. Panel B presents the mean cumulative abnormal returns and the mean calendar-time portfolio regressions returns from randomly selected matched portfolios for all ATMs and all common stock SEOs with no regulated industries and no REITs, respectively. For the ATMs subsamples we use the first public announcement date as our event date whereas for the common stock SEOs we use the issue date as our event date. The number of observations is presented on the first row. Returns are computed over the [-252,-60], [-252,-30], [-90,-3], [-60,-3] and [-30,-3] time windows. Daily returns information is drawn from the CRSP database. Adjustment is performed according to the one-factor market model using the CRSP equal-weighted index as the market proxy. Panel C reports the mean 12-month buy-and-hold raw returns for the ATMs with no REITs and no regulated industries event firms and the common stock SEOs with no REITs and no regulated industries peer firms using daily CRSP pricing information. Peer firms are chosen based on size and book-to-market. Specifically, we make sure that the market capitalization at closing of the ATMs and the common stock SEOs samples do not exhibit a difference of more than 10% and then we choose the peer firms with the lowest sum of both size and book-to-market percentage difference. The t-statistic values are reported in the last row.

Page 34: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

34

Panel A: Event Returns N Market model abnormal rets Market-adj. rets

ATMs 132 -0.0374 -0.0372

SEOs 495 -0.0456 -0.0414

ATMs w/SEO match 117 -0.0298 -0.0287

SEOs w/ATM match 115 -0.0333 -0.0292

note that all of these are significant at the 1% level

Panel B: Runups ATMs (N = 128) SEOs (N = 386)

[-252,-60] 0.1162 0.2297

[-252,-30] 0.1403 0.3174

[-90,-3] 0.0684 0.2473

[-60,-3] 0.0972 0.1806

[-30,-3] 0.0735 0.0905

Panel C: Long run performance (BHRs, BHARs) ATMs (N=57) SEOs (N=358)

BHR_event firm 0.0921 0.0718

BHR_peer -0.0199 0.1704

BHAR 0.1120 (t=1.23)

-0.0986 (t=-2.38)

Page 35: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

35

Table VI

Takedown Behavior of ATMs

Table presents univariate statistics on firms’ takedown behavior of ATM offerings. Data are collected from firms’ 10-Q filings. Takedown percentage equals the number of shares issued relative to the number of shares the firm announced it planned to issue in the original filing. The numerator is the sum of shares issued over the life of the ATM, which equals the planned life announced originally, or if no announcement of planned life was provided, it is the life of the shelf registration that the ATM was filed under. Active quarters percentage equals the number of quarters during the ATM life that the firm actually takes down equity, divided by total number of quarters in the ATM life. Takedown duration equals the number of quarters it takes a firm to complete the takedown of all shares in the ATM, and is calculated only for those firms that complete the ATM (takedown 100% of planned shares). We report statistics for three samples: all firms, firms without explicitly stated commitment periods, and firms with explicitly stated commitment periods.

Panel A: All Firms

Mean Median StdDev N

Takedown Percentage 0.44 0.34 0.41 150

Percentage of Active Quarters 0.50 0.50 0.35 150

Takedown duration (completed only) 3.67 3 2.82 33

Panel B: Pre-2010

Mean Median StdDev N

Takedown Percentage 0.40 0.25 0.41 100

Percentage of Active Quarters 0.53 0.50 0.35 100

Takedown duration (completed only) 4.08 3 3.68 12

Panel C: Post-2010

Mean Median StdDev N

Takedown Percentage 0.53 0.54 0.40 50

Percentage of Active Quarters 0.46 0.50 0.34 50

Takedown duration (completed only) 3.43 3 2.27 21

Panel D: Timing

Q1 Q1-Q2 Q1-Q3 Q1-Q4 Q1-Q5 Q1-Q8

Takedown Percent, mean 0.15 0.23 0.25 0.28 0.34 0.43

Takedown Percent, median 0.017 0.096 0.12 0.15 0.28 0.41

Page 36: At The Market (ATM) Offerings Matthew T. Billett* …schwert.ssb.rochester.edu/f423/BFG_WP12.pdfSEO.1 Given this variation, it is reasonable to expect that even within the set of firms

36

Table VII

Regressions Explaining Takedown Behavior

Table presents estimates from Tobit regression of percentage takedown (one observation per firm/quarter) on the following explanatory variables: quarter_counter is the number of quarters since the ATM announcement; cumulative prior takedown is the total since ATM start of shares taken down, divided by announced number of intended shares to take down.; stock return volatility is the contemporaneous quarter’s standard deviation of daily stock returns; prior quarter CAR is the cumulative abnormal return (market-model based) over the quarter preceding the takedown measure’s quarter; current quarter CAR is the CAR over the contemporaneous quarter; commitment_dum is a dummy equal to one if the firm announces a commitment period for takedown (rather than allowing it to expire at the end of the shelf); announcement CAR is the three-day ATM announcement window CAR. T-statistics are reported underneath coefficients, in parentheses. a,b,c indicates significant at the 10%, 5%, 1% levels.

I

Intercept 0.132 (2.40)b

Quarter Counter -0.050 (-5.59)c

Cumulative Prior Takedown

0.127 (2.94)c

Stock Return Volatility

-1.788 (-1.75)a

Prior Quarter CAR

0.197 (3.46)c

Current Quarter CAR

-0.058 (-0.96)

Commitment_dum

0.004 (0.11)

Announcement CAR

0.463 (1.40)