You are on page 1of 44

How do experienced users evaluate hybrid financial instruments?

Shana Clor-Proell
s.proell@tcu.edu
Department of Accounting
Neeley School of Business
Texas Christian University
Fort Worth, TX 76129
Tel: +1 (817) 257-7148

Lisa Koonce
Lisa.Koonce@mccombs.utexas.edu
Department of Accounting
McCombs School of Business
The University of Texas at Austin
Austin, TX 78712
Tel: +1 (512) 471-5576

Brian White
Brian.White@mccombs.utexas.edu
Department of Accounting
McCombs School of Business
The University of Texas at Austin
Austin, TX 78712
Tel: +1 (512) 471-5619

May 11, 2016

We thank Zheng Leitter and Ben Van Landuyt for their research assistance. We thank the M.J.
Neeley School of Business and the Deloitte Foundation for their generous financial support, and
the Alumni Office at the McCombs School of Business for help in recruiting participants. We
thank Jeremy Bentley, Nicole Cade, Scott Emett, Jeff Hales, Eric Hirst, Pat Hopkins, Karim
Jamal, Ross Jennings, Steve Kachelmeier, Justin Leiby, Bob Libby, Mark Nelson, Mark Peecher,
Ray Pfeiffer, Chad Proell, Kristi Rennekamp, Mary Stanford, Shankar Venkataraman, an
anonymous reviewer, and workshop participants at the following universities: Alberta, Cornell,
DePaul, Georgia Tech, Illinois, Massachusetts Amherst, Texas Christian, and Texas for helpful
comments.

Electronic copy available at: http://ssrn.com/abstract=2552779


How do experienced users evaluate hybrid financial instruments?
ABSTRACT

Hybrid financial instruments contain features of both liabilities and equity. Standard setters
continue to struggle with getting the classification right for these complex instruments. In this
paper, we experimentally test whether the features of hybrid instruments affect the credit-related
judgments of experienced finance professionals, even when the hybrid instruments are already
classified as liabilities or equity. Our results suggest that getting the classification right is not of
primary importance for these experienced users, as they largely rely on the underlying features of
the instrument to make their judgments. A second experiment shows that experienced users
reliance on features generalizes to several features that often characterize hybrid instruments.
However, we also find that experienced users vary in their beliefs about which individual
features are most important in distinguishing between liabilities and equity. Together, our results
highlight the importance of effective disclosure of hybrid instruments features.

JEL Codes: G23; M41; M48; M49

Keywords: hybrid financial instruments; accounting classification; disclosure

Data availability: Contact the authors

Electronic copy available at: http://ssrn.com/abstract=2552779


1. Introduction

Hybrid financial instruments contain features of both liabilities and equity (FASB [2003,

2007]). For example, preferred stock typically has a fixed dividend payment, which is a liability

feature, but it can also have voting privileges, which is an equity feature.1 Some studies suggest

that financial statement users would be best served by determining the appropriate classification

for these instruments (e.g., Barth et al. [1998], Barth et al. [2013]); that is, are they liabilities,

equity, or some combination of both? Fewer studies suggest that disclosure of the underlying

features of these instruments may be central to allowing users to evaluate them (e.g., Kimmel

and Warfield [1995]). In this study, we investigate the extent to which experienced finance

professionals, who are making credit-related judgments and decisions, rely on the underlying

features of a hybrid instrument once it has been classified as a liability or equity.

Investigating this issue is important for several reasons. First, the use of financial instruments

with features of both liabilities and equity is widespread. Indeed, the last two decades have

witnessed a large increase in the number and type of hybrid instruments used for financing, such

as convertible debt (e.g., BNP Paribas [2013], Dutordoir et al. [2014], Bolger [2015]), and for

compensation (e.g., employee stock options, stock appreciation rights payable in stock). Standard

setters have long struggled with the appropriate accounting for these financial instruments,

debating whether they should get the classification issue right or whether disclosure, given a

particular classification, is the best solution (Ryan et al. [2001], Botosan et al. [2006]). Thus,

shedding light on whether users rely on classification and/or disclosure of hybrid instruments

features provides scholarly input to standard setters.

1
Ryan et al. [2001, 388] distinguish between hybrid and compound financial instruments. While hybrid instruments
have characteristics of liabilities and equity, they do not have distinct components that are straight debt or common
equity. Compound instruments, in contrast, consist of multiple components, with at least one being a liability and
one being equity. For ease of exposition in this paper, we refer to all financing instruments that have characteristics
of liabilities and equity as hybrid instruments.

1
Second, prior research has shown the importance of hybrid instruments classification

(Hopkins [1996], Shah [1996], Frischmann and Warfield [1999]), but has not systematically

tested whether users differentiate hybrid instruments based on their features once they have been

classified as liabilities or equity. Although some research has documented that characteristics of

the firm may cause users to treat financial instruments differently than their classification would

suggest (Linsmeier et al. [2008], Cheng et al. [2003]), it has not shown that users do so on the

basis of instruments features. If users do, in fact, distinguish hybrid instruments based on more

than just their classificationthat is, also based on their featuresthis would suggest that, while

classification may be important, disclosure about features is also significant and, therefore,

should be a consideration for standard setters as they are now re-examining the effectiveness of

financial statement disclosures (FASB [2012, 2015a]).

To examine these issues, we conduct two experiments. In Experiment One, experienced

finance professionals in the role of potential lenders evaluate how hybrid instruments that differ

in both classification and features affect a companys creditworthiness and their willingness to

lend to that company. Experiment Two tests the generality of the findings in Experiment One

and also examines whether experienced finance professionals agree on the relative importance of

several common features of hybrid instruments.

We draw on categorization theory from psychology to develop predictions about how the

features and classification of hybrid instruments will affect the judgments and decisions of those

with substantial domain-specific experience. This theory argues that category representations

held in memory are fundamental to peoples understanding of the world (Loken et al. [2008]).

That is, the category to which items are assigned is important in how items are evaluated.

However, items within a particular category are not assumed to be identical, and may differ in

2
their specific features. Such variation in features is potentially important as it may change how

an item is evaluated despite its classification.

Drawing on this theory, we posit that experienced finance professionals will rely on the

underlying features of a hybrid instrument when evaluating the creditworthiness of a firm, even

when the instrument is classified as a liability or as equity. The basis for this prediction lies in

the fact that experienced professionals frequently encounter these instruments and understand the

wide variation in their underlying features that can occur, even for instruments that are classified

similarly. Thus, experienced finance professionals, even if they are also influenced by the

classification of hybrid instrumentsas documented in an equity setting by Hopkins [1996]

are likely to rely on the features of the hybrid instruments.

Experiment One results are supportive of this prediction, in that experienced users

judgments of the companys creditworthiness and their lending decisions are influenced by the

features of the hybrid instruments. Specifically, variation in the instruments features

significantly affects participants judgments of the issuing companys vulnerability to default and

balance sheet strength, as well as their willingness to lend additional capital to the company.

Interestingly, we find little evidence that our participants creditworthiness judgments and

lending decisions are influenced by the hybrid instruments classification. Of the three

measuresvulnerability to default, balance sheet strength, and willingness to lendonly judged

balance sheet strength is affected by the hybrid instruments classification. A follow-up study

confirms our suspicion that this effect arises because the balance sheet strength measure largely

captures the balance sheet presentation effects of issuing hybrid instruments.

Because we manipulate only one featurematurity of the hybrid instrumentin Experiment

One, we conduct a second experiment to more thoroughly examine how differences in several

3
common features of hybrid instruments affect finance professionals judgments. In addition to

maturity, the instruments in Experiment Two reflect variation in priority in liquidation, voting

rights, settlement in cash versus stock, and dependence on profitability for payments to holders

(FASB [1985, 2003, 2007], Ryan [2007], Sangiuolo et al. [2010]). Results show that on average,

no single feature dominates conceptions of what constitutes a liability or equity instrument,

suggesting that the results of Experiment One would generalize to features other than maturity.

Interestingly, though, experienced users vary significantly in their beliefs about which features

are most important in distinguishing between liabilities and equity, thereby underscoring the

importance of disclosing information about instruments features, as there may not be one binary

classification that all individuals would agree with.

Our study offers insights for both standard setters and researchers. Our study is informative

to standard setters as they continue deliberations surrounding vexatious issues related to

classification of financial instruments with characteristics of both liabilities and equity (FASB

[2015b], IASB [2015]). Our main findingthat experienced users rely primarily on features, and

not classification, in judging the economic consequences of hybrid instrumentssuggests that

the burden of getting the classification right may not be as important as is generally thought. In

other words, resolving the long-standing issue of how a financial instrument with characteristics

of liabilities and equity should be classified may not be as key as resolving how to best disclose

information about the instruments underlying features. Although our study is set within the

domain of instruments with characteristics of liabilities and equity, the insights from our study

may also extend to other areas where classification and disclosure are important, such as

classification and disclosure for instruments with characteristics of liabilities and assets (e.g.,

4
callable bonds) and/or for items within the cash flow or performance statements (e.g., operating

versus investing activities).

For researchers, we extend the accounting literature by showing that experienced finance

professionals not only evaluate hybrid instruments based on their classification (cf., Hopkins

[1996]), but also on their underlying features, even when the instruments are already classified as

liabilities or equity. While Hopkins [1996] showed that experienced equity analysts judgments

were affected by the classification of a hybrid instrument, the underlying features of the hybrid

instrument were held constant in his study; therefore, it was not possible to discern whether

participants would have also made discriminations based on the features of the instrument. The

fact that classification effects are largely not present in our study is interesting, and could be

attributed to the passage of time since Hopkins [1996], or the different contexts of the two

studies. Hopkins [1996] was set within an equity valuation setting, while ours is set within a

credit/lending setting. Thus, the differential influence of classification in the two studies may be

attributable to the fact that equity and credit analysis are commonly regarded as very different

(Penman [2013, 682]). While our results are suggestive of the possibility that the use of

classification information may be context dependent, future research will be required to

rigorously explore this idea.

Our study is also related to, but distinctly different from, previous archival research (Kimmel

and Warfield [1995], Barth et al. [1998], Barth et al. [2013]). The primary aim of these studies is

to provide a market-based test of whether particular instruments (e.g., redeemable preferred

stock, convertible debt, employee stock options) should be classified as liabilities or equity. In

contrast, our purpose is not to address whether hybrid instruments should be classified as

liabilities or equity. Rather, we investigate the extent to which experienced finance professionals

5
rely on the underlying features of a hybrid instrument, once it has been classified as a liability or

equity.

The remainder of the paper is organized as follows. Section 2 provides background on hybrid

instruments and discusses categorization theory. Sections 3 and 4 discuss Experiments One and

Two, respectively. Section 5 concludes.

2. Background and Theory

2.1 HYBRID INSTRUMENTS

Hybrid instruments are challenging from a classification (i.e., liability versus equity)

standpoint because, by definition, they have one or more features that are associated with equity

and one or more features that are associated with liabilities. For example, a convertible bond has

a written call option (on the companys equity) embedded with the underlying bond. While the

written call option is often considered an equity item, the underlying bond is clearly a liability.

As another example, mandatorily redeemable preferred stock (MRPS) has liability features in

that it typically has a fixed payout (like interest) and must be redeemed, most often for cash, on a

specific date or following a designated event. However, it also has equity features as it is

preferred stock and pays dividends (Sangiuolo et al. [2010]).

Importantly, hybrid instruments can affect how the issuer of the instrument is judged in terms

of its perceived creditworthiness (Bianco and Deng [2007]). In this study, we make the

assumption that as liabilities increase as part of a companys capital structure, users are likely to

judge the company as less creditworthy (e.g., Brealey and Myers [2000], Wahlen et al. [2014]).

While credit analysis can be a complex process, we view this simplifying assumption as

reasonable for most firms.

6
2.2 THEORY

2.2.1 Categorization Theory. The extensive research on categorization in psychology has

examined two fundamental questions. First, it has examined how individuals initially group or

classify novel people, objects, events, and experiences into existing categories. Second, and

relevant to our setting, categorization research has examined how individuals use prior

knowledge of category membership to form judgments about novel category members. That is,

once an item has been classified into an existing category, information about that category is

applied to the novel item.

Because not all members of a particular category are identical, though, variation in features

also can be important in judging an item. In a consumer context, for example, the category

vacuum cleaners contains uprights and handhelds, plug-in and battery-powered models, and

models with varying numbers of attachments. Although all of these are vacuum cleaners and

share features common to most vacuum cleaners (e.g., suction), they vary in their specific

features. Thus, the theory indicates that judgments about a new vacuum cleaner can be based on

both its membership in the category of vacuum cleaners and the particular features associated

with the specific vacuum cleaner at hand (Meyvis and Janiszewski [2004]).

This line of research also has shown that what we know about categorization can be specific

to context and to the experiences of individuals (Ahn et al. [2005], Honeck et al. [1987], Proffitt

et al. [2000]). For example, experience in a particular domain has been found to affect how items

are categorized and evaluated once in a category (Loken et al. [2008]). In the consumer-

marketing domain, individuals with high product experience are more likely to make distinctions

among products (Johnson and Mervis [1997]). That is, they are not only able to think more

abstractly on a conceptual basis, but they also make finer distinctions among items than less

7
experienced individuals. As a consequence of this, experts have been found to be relatively more

flexible in their categorizations and, thus, can change the way they think about category

members depending on the specifics of the context (Cowley and Mitchell [2003]).

2.2.2 Categorization Theory in Accounting. Applying these ideas to an accounting context,

the theory suggests that both the financial statement classification and underlying features of a

hybrid instrument can affect experienced users judgments. Consistent with the notion that

classification affects judgments, prior research by Hopkins [1996] shows that the category to

which a novel hybrid instrument is assigned (i.e., liabilities versus equity) affects the valuation

judgments of buy-side equity analysts. His study participants had previously acquired knowledge

about the categories into which financial statement items are classified; i.e., assets, liabilities and

equity. As a result, when they encountered a novel financial instrument (i.e., MRPS) that had

been classified into one of these categories, they used their knowledge about categories to

understand how to evaluate the hybrid instrument for stock valuation purposes.

Hopkins [1996] did not vary the features associated with the hybrid instrument in his study.

This aspect of his study is important, as categorization theory suggests that the particular features

of a hybrid instrument could also affect experienced users judgments. That is, as described in

the vacuum cleaner example above, category membership provides some information, but also

allows for variation among category members that could be important to users. Thus, it remains

an unanswered question as to the extent to which users in the financial reporting domain will rely

on features of hybrid instruments once they have been classified in the financial statements.

We posit that because experienced professionals have seen the variation in features typical of

hybrid instruments, they should appreciate the variation in instruments that are eventually

classified in the financial statements in either the liability or equity category. As previously

8
noted, not all financial instruments have features that are typical of liabilities (i.e., fixed maturity

date, fixed payment, etc.) nor do they necessarily have features that are typical of equity (i.e.,

discretionary payment, voting rights, etc.). Thus, hybrid instruments are not typical of either the

liability or equity category, and so the placement of the instrument into either category is not

perfect. Indeed, because the definitions of the liability and equity categories within financial

reporting are general, the mapping between a particular financial statement item (like a hybrid

financial instrument) and a given category is not straightforward.2 Given this knowledge of the

finer distinctions among items that are placed into the liability and equity categories, the

categorization theory discussed earlier suggests that experienced users are likely to look to

variation in features when evaluating a hybrid instrument and unlikely to rely solely on the

category in which the item is placed. This prediction is summarized below.

HYPOTHESIS: When judging the creditworthiness of, and making a lending decision for, a
company issuing a hybrid financial instrument, experienced finance
professionals will rely on the underlying features of the instrument, even when
the instrument is classified as a liability or as equity.

Note that the underlying psychology theory suggests that both classification and features can

affect experienced users judgments. Importantly, though, the theory does not indicate that

experienced uses reliance on features would depend on whether the hybrid financial instrument

was classified as a liability versus equity. That is, while our hypothesis predicts a main effect of

features and allows for potentially observing a main effect of classification, we do not expect to

2
The formal definition of a liability describes three essential features: (a) it embodies a present duty or
responsibility to one or more other entities that entails settlement by probable future transfer or use of assets at a
specified or determinable date, on occurrence of a specified event, or on demand, (b) the duty or responsibility
obligates a particular entity, leaving it little or no discretion to avoid the future sacrifice, and (c) the transaction or
other event obligating the entity has already happened (FASB [1985, paragraph 36]). Equity is defined as the
residual interest in the assets of an entity that remains after deducting its liabilities; the definition of equity also
refers to several classes of equitywith different degrees of risk stemming from different rights (FASB [1985,
paragraph 62]).

9
observe an interaction between features and classification. Below, we describe the experiment

used to test our predictions.

3. Experiment One

3.1 DESIGN OVERVIEW AND PARTICIPANTS

Experiment One has a 2 2 between-participants design in which experienced participants

assume that they are following a company that has recently issued hybrid financial instruments.

We manipulate the features of the hybrid instruments (2 liability & 3 equity vs. 3 liability & 2

equity) and the classification of the hybrid instruments on the balance sheet (liability vs. equity).

Participants evaluate how the issuance affects the companys creditworthiness and indicate their

willingness to lend additional funds to the company.

To recruit participants, we draw on a business-school alumni database. We request

participation in the study if the graduate has had five years of current work experience in

banking, finance, and related fields. To encourage broad participation and avoid selection bias,

the request indicates only that alumni will participate in a research study involving a short

case study. That is, the request does not mention financial instruments or hybrids. From the set

of volunteers, we randomly select 300 to receive a hard copy of the experimental instrument in

the mail. Of these, 162 individuals (54 percent) return a completed instrument to us.3 Our

participants have an average of 20 years of work experience, with 49 percent having banking

experience and 38 percent having corporate finance experience. Forty-four percent indicate that

they have made lending decisions in the past. All participants are randomly assigned to one of

the four experimental conditions.4

3
Our response rate of 54 percent is consistent with response rates reported in prior studies using experienced
participants. For example, Clor-Proell and Nelson [2007] report a response rate of 64 percent, Clor-Proell [2009]
reports a response rate of 30 percent, and Elliott, Krische and Peecher [2010] report a response rate of 31 percent.
4
Because our access to experienced professionals occurred through a mail-out experiment, our results could be

10
3.2 CASE MATERIALS, MANIPULATIONS, AND DEPENDENT VARIABLES

Our experimental materials are adapted from Hopkins [1996]. Participants receive

background information about a company that has just issued a press release announcing a $7

million security issuance. The press release indicates that proceeds from the security issuance

will be used to expand domestic operations, and it lists the features of the instruments (see

Appendix, Panels A and B). There are five features in total, and we hold constant four of these

features across conditions. Two of the four features that are held constant suggest that the

instruments are similar to traditional liabilities in that the holders receive fixed periodic

payments, and these payments accumulate interest if not paid on a timely basis. The two

remaining features that are held constant suggest that the instruments are similar to traditional

equity in that the holders receive additional payments if yearly profit targets are met, and the

holders have voting rights. The information about the final feature is manipulated between

participants. Specifically, in the 3 liability/2 equity conditions, this feature indicates that the

instruments will mature in 2020 (i.e., liability-like); in the 2 liability/3 equity conditions, this

feature indicates that the instruments have no maturity date (i.e., equity-like).

Following the press release, the materials present summary income statement and balance

sheet data for the company. This information does not include the effects of the new issuance,

and is held constant across conditions. The size of the issuance is material to the financial

statements, representing 5.4 percent of total assets. As dictated by the treatment condition, the

materials indicate either that the company will account for the new instruments as equity or as a

liability. We also provide participants with the journal entry that will be used to record the

subject to response bias if people who are more comfortable analyzing hybrid instruments are more likely to
respond. To address this possibility, we compare the demographics of early responders to late responders
(Oppenheim [1992], Wallace and Mellor [1988]), and find that the two groups are reasonably homogeneous.
Second, we cross our independent variables with proxies for comfort with analyzing hybrids (i.e., response time,
experience, and familiarity with SFAS 150), and find no systematic interaction of our participants comfort level
with features or classification. Based on these results, we believe it is unlikely that response bias affected our results.

11
issuance (with the credit to the entry being to a liability account or to an equity account) and we

ask participants to recast the companys balance sheet (on a new balance sheet that has blank

amounts for the liabilities and equity sections) to reflect the new offering.

Following this information are two questions about the effect of the issuance on the

companys creditworthiness. The first question, based on Hales et al. [2011], asks participants to

judge how the issuance of the $7 million in securities affects the companys vulnerability to

default. Participants respond to this question on a 101-point scale with endpoints of 0 (100),

labeled as much less vulnerable (much more vulnerable). The second question asks

participants to judge how the issuance of the $7 million in securities affects the strength of the

companys balance sheet. Participants respond on a 101-point scale with endpoints of 0 (100),

labeled as creates much weaker balance sheet (creates much stronger balance sheet).

Participants then indicate how willing they would be to lend additional money to the

company. For this question, participants first select one of two options, indicating that they either

would or would not be willing to lend the company additional funds. After selecting one of the

options, they then complete a corresponding scale that measures the strength of their preference

for lending or not lending using a 101-point scale, where 0 (100) indicates a very weak (strong)

preference. Using these responses and scales, we construct a continuous measure for

participants lending decisions by coding the decision to lend (not lend) as +1 (1) and then

multiplying by the related preference score. This measure ranges from 100 to 100, where a

value of 100 indicates a very strong preference against lending, and a value of 100 indicates a

very strong preference in favor of lending. The final section of the materials includes debriefing

questions and manipulation check questions.

12
Two comments are in order about the design of our experiment and our dependent measures.

First, the between-participants design greatly reduces hypothesis guessing, as participants only

see one of the four experimental conditions. Second, because there is no normatively correct

answer in our study, we use our dependent measures to examine the relative levels of our

manipulated variables, rather than the absolute levels (Libby, Bloomfield, and Nelson [2002]).

3.3 MANIPULATION CHECKS

To confirm that our features manipulation was successful, we ask participants to indicate

whether the instruments were described as maturing in 2020, or as having no maturity date. One

hundred (ninety-eight) percent of participants in the 3 liability/2 equity (2 liability/3 equity)

conditions correctly indicate that the instruments do (do not) have a maturity date. We also ask

participants to indicate how the company was planning to account for the instruments. Eighty-six

(seventy-eight) percent of participants in the equity (liability) conditions correctly indicate that

the company was planning to account for the instruments as equity (a liability). Thus, we

conclude that both manipulations were successful, and include all participants in the results

reported below.

3.4 RESULTS

We predicted that when judging the creditworthiness of, and rendering a lending decision for,

a company issuing a hybrid financial instrument, experienced users would rely on the underlying

features of the instrument, even when the instrument is classified as a liability or as equity. To

test this hypothesis, we rely on the two creditworthiness judgmentsnamely, the vulnerability to

default measure and the balance sheet strength measureas well as the lending decision.5

3.4.1 Vulnerability to Default and Balance Sheet Strength. The Cronbach alpha coefficient,

which measures the consistency with which the vulnerability to default and balance sheet
5
All reported p-values are one-tailed when associated with directional predictions.

13
strength measures capture the same underlying construct, is 0.52, suggesting that each measure

captures a relatively unique aspect of the creditworthiness judgment (e.g., Nunnally [1978],

Peterson [1994]). We therefore analyze the effect of our independent variables on each measure

separately. For each measure, we conduct a 2 2 analysis of variance (ANOVA) with two

independent variablesthe classification of the instrument (liability or equity), and the

underlying features of the instrument (3 liability/2 equity or 2 liability/3 equity). The results are

reported in Table 1, and the pattern of cell means is depicted in Figure 1.

-------------------------------------------------
Insert Table 1 and Figure 1 about here
-------------------------------------------------

Starting with the vulnerability to default measure, Panel A of Table 1 shows descriptive

statistics (see also Panel A of Figure 1), while Panel B reports the overall ANOVA. We find a

significant main effect of features (F1,158 = 3.14, p = 0.04), as predicted. On average, participants

judged the companys vulnerability to default as greater when the instrument had more liability

features and fewer equity features (mean of 49.85) compared to when it had fewer liability

features and more equity features (mean of 45.00). This result provides support for our prediction

that experienced finance professionals rely on a hybrid instruments features when judging its

effects on the issuing firms creditworthiness. We do not observe a significant effect of

classification on participants vulnerability to default judgments (F1,158 = 0.07, p = 0.40), nor do

we observe as significant features classification interaction (F1,158 = 1.66, p = 0.20).

Turning to the balance sheet strength measure, Panel A of Table 1 shows descriptive

statistics (see also Panel B of Figure 1), while Panel B of Table 1 reports the overall ANOVA.

As predicted, we find a significant main effect of features on balance sheet strength judgments

(F1,158 = 13.64, p < 0.01). Specifically, participants judged the companys balance sheet to be

14
stronger when the instrument had fewer liability features and more equity features (mean of

57.55) compared to when it had more liability features and fewer equity features (mean of

48.11). We also observe a significant main effect of classification on participants judgments of

balance sheet strength (F1,158 = 4.18, p = 0.02); participants judge the balance sheet to be

stronger when the hybrid instrument is classified as equity (mean of 55.49) compared to when it

is classified as a liability (mean of 50.28). The features classification interaction is not

significant (F1,158 = 0.17, p = 0.68).

While both of these measures provide support for our hypothesis that experienced users will

rely on hybrid instruments features in judging the creditworthiness of the issuer, we note that

classification influences participants judgments of balance sheet strength, but not their

judgments of vulnerability to default. As evidenced by the relatively low Cronbach alpha

coefficient reported above, these two measures clearly capture different fundamental constructs.

One possibility is that the vulnerability to default measure largely captures participants beliefs

about the economic impact of issuing hybrid instruments, whereas the balance sheet strength

measure primarily captures beliefs about how the instruments will affect the appearance of the

balance sheet.6 We address this possibility with a follow-up study described in section 3.5.

3.4.2 Lending Decision. We next turn to participants willingness to lend additional funds to

the company. Recall that our lending measure captures their dichotomous decision about whether

they would lend combined with the strength of their preference for that decision. It ranges from

100 to 100, with 100 indicating a strong preference against lending, and 100 indicating a strong

preference in favor of lending. Panel A of Table 2 shows descriptive statistics for the lending

decision (see also Figure 2), and Panel B reports the overall ANOVA.7 We observe a significant

6
We thank an anonymous reviewer for suggesting this explanation.
7
Two participants did not respond to the lending decision. Thus, results for this measure are based on 160

15
main effect of features (F1,156 = 2.20, p = 0.07). On average, participants expressed a stronger

willingness to lend additional funds to the company when the instrument had fewer liability

features and more equity features (mean of 22.59) compared to when it had more liability

features and fewer equity features (mean of 8.67). This result provides additional support for our

hypothesis. We also note that classification does not significantly affect participants lending

decisions. Specifically, the main effect of classification is insignificant (F1,156 = 0.61, p = 0.22),

as is the features classification interaction (F1,156 = 0.19, p = 0.67). These results are consistent

with the results for our vulnerability to default measure, and suggest that features play a larger

role than classification in determining lending decisions.8

-------------------------------------------------
Insert Table 2 and Figure 2 about here
-------------------------------------------------

As additional evidence of our general conclusion that features have a larger impact than does

classification, we also asked participants two questionsone about how important specific

features/details were to their lending decision and another about how important balance sheet

classification was to their lending decision. They responded to both questions on 101-point

scales with endpoints of 0 (100) indicating not at all important (extremely important). They

rated the importance of features at 72.22 (s.d. 20.50) and the importance of classification at 51.54

(s.d. 30.84). These responses are statistically different (t161 = 7.26, p < 0.01), indicating that our

experienced participants consider features to be more important than classification in lending

decisions.9

observations.
8
Follow-up tests indicate that all simple main effect tests are directionally consistent with our predictions. For
equity classification, features significantly affect balance sheet strength (p = 0.01) and lending decisions (p = 0.09)
but do not significantly affect vulnerability to default (p = 0.37). For liability classification, features significantly
affect vulnerability to default (p = 0.02) and balance sheet strength (p < 0.01), but do not significantly affect lending
decisions (p = 0.23).
9
We find additional support for the importance of features from the recasting task (where participants were asked to

16
3.5 FOLLOW-UP STUDY

Recall that our vulnerability to default dependent measure showed only an effect of features

while the balance sheet strength dependent measure showed both a features and classification

effect. We conjectured that these measures might capture different constructsnamely, that the

vulnerability to default measure primarily captures beliefs about the economic impact of issuing

hybrid instruments, whereas the balance sheet strength measure largely captures beliefs about

how the instruments affect the appearance of the balance sheet. To provide empirical evidence

for this conjecture, we conducted a follow-up study with experienced professionals, randomly

selecting 100 from the 300 to whom we originally sent the Experiment One materials. We mailed

materials to these 100, and received 50 completed responses. As expected, participants in this

follow-up study have similar characteristics to those who participated in Experiment One.10 They

have an average of 23 years of work experience, with 44 percent having banking experience and

35 percent having corporate finance experience.

Participants in the study are asked to assume that a firm has recently issued a financial

instrument that has characteristics of both debt and equity. They are told that the instrument has

both an economic impact on the firm and an effect on the presentation of the firms financial

statements. Participants then evaluate two statements about the issuance, with one statement

referencing the firms vulnerability to default and the other statement referencing the firms

balance sheet strength. Participants indicate whether each statement refers to the economic

impact versus the presentational impact of the new financial instrument. Participants responded

recast the balance sheet for the effects of the new security). Specifically, participants were more likely to reclassify the
instruments on the recasted balance sheet when the companys balance sheet classification did not match the
instruments features (e.g., liability classification and 2 liability/3 equity features) than when they did match (e.g.,
liability classification and 3 liability/2 equity features) (reclassification of no-match is 43.2 percent and match is 17.3
percent; 2 = 12.22, p < 0.01).
10
Because responses to both Experiment One and the follow-up study are anonymous, we cannot tell whether a
given participant participated in Experiment One, the follow-up study, or both.

17
on 101-point scales with end points labeled 50 (Refers exclusively to the economic impact)

and 50 (Refers exclusively to the presentational impact). See also Panel C of the Appendix.

Results indicate that the mean response to the vulnerability to default statement is statistically

lower (mean = 27.84; s.d. = 18.79) than the mean response to the balance sheet strength

statement (mean = 6.52; s.d. = 31.98) (t49 = 6.75, p < 0.01), indicating that participants view

these questions as capturing fundamentally different constructs. Follow-up tests confirm our

earlier conjecture that vulnerability to default primarily captures economic impact, as the mean

of 27.84 is significantly lower than the midpoint of zero (t49 = 10.47, p < 0.01). Also as we

conjectured, the balance sheet strength statement largely captures presentation, as the mean of

6.52 is significantly greater than zero (t49 = 1.44, p = 0.08). These results confirm our suspicion

as to why in our main experiment, the balance sheet strength measure shows a classification

main effect, while the vulnerability to default measure does not.11

3.6 SUMMARY OF EXPERIMENT ONE RESULTS

The overall pattern of results from Experiment One supports the prediction that experienced

finance professionals rely on the underlying features of hybrid instruments, even when the

instruments are classified as liabilities or equity. Although participants do consider classification

in judging balance sheet strength, this appears to be explained by the balance sheet presentation

effects of issuing hybrid instruments (rather than by their economic effect), as we observe no

significant effect of classification on judgments of vulnerability to default or willingness to lend.

11
Other possible, but unlikely, explanations for the lack of a classification effect for the vulnerability measure and
lending decision are a lack of statistical power or lack of participant attention. Both of these seem unlikely given that
the classification variable was significant for the balance sheet strength dependent variable. Further, power concerns
are unlikely as our sample size is quite large for an experiment. Lack of attention seems unlikely given that most
participants passed the manipulation checks.

18
4. Experiment Two

The purpose of Experiment Two is two-fold. First, we test the generality of the results in

Experiment One by investigating whether experienced users distinguish between liabilities and

equity based on features other than the maturity feature that we investigated in the first

experiment. Second, if our results show that participants do in fact make distinctions based on

these other features, we then examine the level of agreement about individual features. If

experienced finance professionals exhibit disagreement about the relative importance of

individual features, then it may be difficult to mandate a classification with which all users

would agree. Thus, Experiment Two addresses this possibility.

4.1 DESIGN OVERVIEW AND PARTICIPANTS

To identify the features, in addition to maturity, that are most commonly associated with the

liability versus equity classification, we reviewed the conceptual framework (FASB [1985]),

financial accounting standards and exposure documents (e.g., FASB [2003, 2007]), materials and

discussions at past FASB/IASB issues conferences (particularly 2004 and 2006), and the

academic and practitioner literature (e.g., Ryan [2007], Sangiuolo et al. [2010]). The most

common features are:12

1. Maturity: traditional liabilities have a fixed maturity date; common equity is perpetual.

2. Priority in liquidation: traditional liability holders have priority over equity holders in

liquidation; common equity holders do not.

3. Voting rights: traditional liabilities do not confer voting rights; common equity typically

does convey such rights.

12
While other features are sometimes mentioned as distinguishing liabilities and equity (e.g., time to maturity,
transferability, uncertainty about existence, difference in rights under liquidation and non-liquidation), these five
were the most frequently mentioned during our review.

19
4. Settlement in cash or common stock: traditional liabilities are usually settled in cash;

although common equity is perpetual, the ability to settle in common stock is sometimes

considered to make a hybrid instrument more equity-like.

5. Dependence on profitability: interest payments on traditional liabilities do not depend on

the issuing company generating profits; dividend payments to equity holders must come

from retained earnings.

Relying on these findings, we ask participants in Experiment Two to evaluate multiple hybrid

instruments that differ with respect to these five features. We manipulate each feature so that it is

either representative of a liability or equity. For example, we manipulate the voting feature so

that holders do not have voting rights (which is arguably similar to a liability) or have voting

rights (which is arguably similar to equity).

The experiment utilizes a 25 half-fractional factorial design. Half-fractional factorial designs

are useful when investigating the effects of a relatively large number of independent variables

(Kuehl [2000]). In this study, we are interested in the effects of five independent variables that, if

fully crossed, would create 32 separate hybrid instruments. If presented between-participants,

these 32 instruments would require a participant pool that is larger than we could realistically

expect to obtain. In contrast, if presented within-participants, requiring individuals to evaluate all

32 instruments would take a substantial time commitment from our participants that could

threaten our response rate. To balance these concerns, we use a half-fractional factorial design,

which requires participants to evaluate only 16 of the 32 possible instruments on a within-

participants basis. Importantly, we carefully select the 16 instruments so that all main effects and

two-way interactions are fully interpretable in what is termed a Resolution V design.13

13
In Resolution V designs, the three- and four-factor interactions cannot be independently estimated (see Kuehl
[2000] for additional information about fractional factorial designs). However, these higher order interactions are

20
We recruit experienced participants from the same alumni database, and using the same

selection criteria, as we used in Experiment One. We mailed the experimental instrument to 100

individuals that did not participate in Experiment One, 56 of whom returned a completed

instrument to us. Our participants are similar to those that participated in Experiment One. They

have an average of 23 years of work experience, with 45 percent having banking experience and

34 percent having corporate finance experience. Forty-five percent of participants indicate that

they have made lending decisions in the past.

4.2 CASE MATERIALS AND DEPENDENT MEASURES

Participants are told that financial instruments can have features that vary along a liability-

equity continuum. We ask participants to evaluate instruments that differ with respect to the five

features described earlier. Each of these features can take one of two possible values for each

instrument; Figure 3 provides an example of an instrument that we used in the experiment. We

refer to the hybrid instruments as Security #1, Security #2, etc., so that we do not change any

other aspect of the instrument, such as the name normally assigned to the instrument.

---------------------------------
Insert Figure 3 about here
---------------------------------

Participants rate each of the 16 hybrid instruments on separate 101-point scales to indicate

where the instrument falls along the liability-equity continuum. Scale endpoints of 50 (50) are

labeled exactly like a bond liability (exactly like common stock). The scale midpoint of 0 is

labeled equally like a bond liability and common stock.

generally negligible (Louviere [1988], Slovic [1969]). For an example of a similar design in the accounting
literature, see Ashton [1974]. To address any potential order effects, the 16 instruments appear in one of four
separate orders. Order is not a significant variable in any of our analyses, so we do not discuss it further.

21
4.3 RESULTS

4.3.1 Are Other Features Used? To determine if experienced users make distinctions for the

other features that were not manipulated in Experiment One, we use the participants responses

to estimate a repeated-measures ANOVA with the five features as independent variables. Panel

A of Table 3 presents descriptive statistics for each of the 16 instruments, while Panel B reports

the ANOVA. We find that each of the five features significantly affects experienced participants

judgments (all F1,55 > 41.44, all p < 0.01).14 These results indicate that experienced finance

professionals do make distinctions based on differences in all five features.

---------------------------------
Insert Table 3 about here
---------------------------------

Further examination of the mean judgments by instrument (Panel A) shows that each feature

contributes independently and monotonically to experienced users similarity judgments. That is,

an increase in liability features increases the judged similarity to a typical liability item, and an

increase in equity features increases judged similarity to a typical equity item (r893 = 0.64, p <

0.01). For example, focusing on the number of equity features, we observe that the average

judgment is 21.35 when the hybrid instrument contains only one equity feature, 11.25 when the

hybrid instrument contains three equity features, and 40.14 when the hybrid instrument contains

five equity features.

Based on these results, we conclude that our generalizations from Experiment One are

unlikely to be limited by our use of the maturity feature in that experiment. Indeed, these

14
Given that our focus is on whether other features are used, this conclusion is based on the main effect results. As
shown in Panel B of Table 3, we also observe that three (of the ten possible) two-factor interactionsthe priority
settlement (p = 0.05), priority by profit (p = 0.06), and settlement profit (p = 0.05) interactionsare significant.
Although statistically significant, these interactions seem to have relatively small effect sizes. Taking the settlement
by profit interaction as an example, we find that when settlement is in cash, varying the profit feature results in a 17-
point difference in judgment; when settlement is in common stock, varying the profit feature results in a 14-point
difference in judgment. Because of these small effect sizes for the interactions, we do not further explore them.

22
Experiment Two results show that, on average, each of the five features is viewed as important to

the liability-equity distinction.

4.3.2 Are All Features Treated Similarly? Given that experienced users do distinguish among

other features, our second set of tests then examines the extent to which all features are treated

similarly. To conduct these analyses, we compute the effect of each feature for each individual

participant based on his/her responses to the 16 scenarios.15 That is, we calculate how much an

individuals mean response changes when we manipulate, for example, the voting feature from

holders do not have voting rights (which is arguably similar to a liability) to holders have

voting rights (which is arguably similar to equity). Scores for changing individual features from

liability-like to equity-like range from 22.25 to 74.88 on the 101-point scale, suggesting wide

variation in individual responses. We then classify responses into five response buckets based on

the magnitude of change.

We first sort these data by feature. As seen in Figure 4, we observe responses in at least four

buckets for all five features. For two featuresmaturity and prioritymost participants increase

their judgments by either zero to 20 points or 21 to 40 points, with an equal number in each of

these response buckets (both p > 0.39). However, for the three other featuresvoting,

settlement, and profitwe observe that most participants increase their judgment by a smaller

amount. Specifically, for these latter features, most increase their judgments by zero to 20 points

rather than by 21 to 40 points (all p < 0.01). This analysis indicates that all features are not

evaluated similarly.

15
This analysis also allows us to rule out demand effects, as participants who are merely reacting to the level of the
manipulation (e.g., positive responses for equity features and negative responses for liability features) would likely
do so in a similar fashion for each of the five features we study. We also note that a between-subjects design, in
which each participant evaluated a single hybrid instrument with a particular combination of features, would
preclude this type of analysis because we would be unable to measure the effect of each feature for individual
participants. That is, a between-participants design would allow only an on-average analysis.

23
---------------------------------
Insert Figure 4 about here
---------------------------------

With this same data, we now sort responses by person (not tabulated). This analysis indicates

that there is variation in how individuals judge the features. Specifically, 73 percent of

participants changed their judgments by more than 20 points, a relatively large amount, in

response to the manipulation of just one or two features. At the same time, many participants

also reacted weakly to some features. Specifically, 54 percent of participants changed their

judgment by less than five points, a relatively small amount, in response to at least one feature.

Again, this indicates there is significant variation in users opinions about the relative importance

of individual features.

Based on these results, we conclude that experienced users do not treat all features similarly,

and that there is no broad agreement on their relative importance. While all five features have a

significant effect on average across all participants, there is in fact much more nuance associated

with individual responses. These results lend further support to our conclusion from Experiment

One that features matter to experienced users. Further, these results highlight the difficulty

associated with establishing a classification rule that all users would agree with, and underscore

the importance of effective disclosure of all features associated with hybrid financial

instruments.

5. Discussion

Standard setters have long grappled with how to account for hybrid instruments within the

financial statements and related disclosures. Much of the deliberation has focused on the proper

classification for these instruments, and prior research has documented the importance of

classification (Hopkins [1996], Shah [1996], Frischmann and Warfield [1999]). Prior research

24
has also addressed accounting for hybrid instruments by trying to determine the appropriate

classification for particular instruments (Kimmel and Warfield [1995], Barth et al. [1998], Barth

et al. [2013]), and has suggested that enhanced disclosure of hybrid instruments features may

help resolve the problem (Kimmel and Warfield [1995]). Our research addresses this issue.

Drawing on categorization theory, we predict and find that experienced finance professionals

are likely to rely on the features of a hybrid instrument, even when the instrument is classified as

a liability or as equity. Examining reliance on features more closely, we find that various features

of hybrid instruments have independent and incremental effects on creditworthiness judgments,

on average. However, there is significant variation in the judged importance of each feature.

These results suggest that it would be difficult for standard setters to mandate classification

criteria that all experienced users would agree with. Overall, our results illustrate the importance

of effective disclosure of features.

Our findings are important to standard setters as they consider how to classify and disclose

information about hybrid instruments in the financial statements (FASB [2003, 2007, 2012,

2015c]). Specifically, our study indicates that a strong focus on effective disclosure of hybrid

instruments features is warranted, particularly as the FASB has indicated that it is concerned

with providing useful information to reasonably sophisticated users of financial reports (FASB

2015d), such as the experienced finance professionals who participated in our study.

Interestingly, we also find that classification effects are largely absent from our results,

suggesting that classification is not of primary importance to experienced users making credit-

related judgments (as in this context, experienced users rely on features regardless of how the

hybrid instruments are classified). However, our results, in conjunction with those of Hopkins

[1996], indicate that meeting the needs of diverse user groups may require accounting standards

25
that address classification and disclosure. Thus, determining how to best disclose information

about the underlying features of a hybrid instrument may be at least as important as determining

the instruments proper classification as either a liability or equity item.

Our experiments include a number of design choices that create opportunities for future

research. First, we examine the effects of classification and features in the context of hybrid

instruments with characteristics of liabilities and equity. However, future research could apply

categorization theory in the context of other financial statement classifications, including assets,

revenues, and expenses; or indeed the operating, investing and financing categories on the

statement of cash flows. Such research could investigate, for example, the relative importance of

differences in features for items in these categories or for different types of users.

Second, in contrast to Hopkins [1996] who investigated the effect of hybrid instrument

classification on equity analysts stock price judgments, we examine the effects of classification

and features on creditor judgments. While creditor judgments represent an important, yet

relatively under-researched setting (e.g., Trotman et al. [2011]), this design choice also makes it

difficult to reconcile our results to those of Hopkins [1996]. Accordingly, future research could

examine whether classification is more central in valuation settings than it is in credit and

lending contexts. Key differences between valuation and credit analysis (Penman [2013]) could

explain why the classification effects obtained by Hopkins [1996] are largely absent in our study.

For example, Penman [2013, 683] notes that equity analysis often requires extensive balance

sheet reformulation whereas credit analysis does not. As a result, equity analysts may be more

sensitive to balance sheet classifications than are credit analysts.

Third, in practice, firms may engage in strategic classification shifting (e.g., McVay [2006],

Fan et al. [2010]). In our context, this might take the form of strategically classifying financial

26
instruments as either liabilities or equity, or structuring the features of an instrument to achieve a

desired classification. As such, future research could examine how this type of strategic behavior

affects the effects of classification and underlying features that we document here. In short, there

are numerous opportunities for additional research on this important topic.

27
APPENDIX
Excerpts from Experiment One Materials

Panel A: Press release provided to participants in the 3 liability & 2 equity feature
conditions

Bransford Announces Financing Plan


By a WALL STREET JOURNAL STAFF REPORTER

SOUTH CREEK, Wis. In an unexpected announcement, Bransford


Sensors, Inc. said it will issue $7.0 million of preference securities.

The new offering will help Bransford, a manufacturer of flow and


measurement instruments, expand domestic operations and establish
manufacturing and distribution facilities in Canada. Earlier this
month, Branfords board of directors announced general expansion
plans but did not take a position on the projects financing.

Bransford said it will issue 1.4 million securities for $5 each.

The securities will be redeemed at a liquidation value of $5 each in


2020.

Holders of the securities will be entitled to quarterly preferential


dividend distributions, over the rights of common shareholders, at
a rate of 7.5% per annum.

If any 7.5% per annum payment is not paid on a timely basis,


holders of the securities will still be entitled to receive those
payments along with an interest accumulation on the past-due
payment.

In addition to quarterly preferential dividend distributions, holders


of the securities will be entitled to additional dividend payments
(beyond the 7.5% per annum) to the extent that firm meets certain
yearly profit targets.

Holders of the securities will have the right to vote, along with the
common shareholders, at a rate of one-quarter vote per security
held.

Bransfords common stock closed yesterday at $38.25, unchanged, in


NASDAQ composite trading. The Company made the
announcement after the close of yesterdays trading.

28
Panel B: Press release provided to participants in the 2 liability & 3 equity feature
conditions

Bransford Announces Financing Plan


By a WALL STREET JOURNAL STAFF REPORTER

SOUTH CREEK, Wis. In an unexpected announcement, Bransford


Sensors, Inc. said it will issue $7.0 million of preference securities.

The new offering will help Bransford, a manufacturer of flow and


measurement instruments, expand domestic operations and establish
manufacturing and distribution facilities in Canada. Earlier this
month, Branfords board of directors announced general expansion
plans but did not take a position on the projects financing.

Bransford said it will issue 1.4 million securities for $5 each.

The securities will have no maturity date and will continue to pay
dividends indefinitely.

Holders of the securities will be entitled to quarterly preferential


dividend distributions, over the rights of common shareholders, at
a rate of 7.5% per annum.

If any 7.5% per annum payment is not paid on a timely basis,


holders of the securities will still be entitled to receive those
payments along with an interest accumulation on the past-due
payment.

In addition to quarterly preferential dividend distributions, holders


of the securities will be entitled to additional dividend payments
(beyond the 7.5% per annum) to the extent that firm meets certain
yearly profit targets.

Holders of the securities will have the right to vote, along with the
common shareholders, at a rate of one-quarter vote per security
held.

Bransfords common stock closed yesterday at $38.25, unchanged, in


NASDAQ composite trading. The Company made the
announcement after the close of yesterdays trading.

29
Panel C: Materials used in follow-up study

Assume that a firm has recently issued a financial instrument that has characteristics of both debt
and equity. The instrument has an economic impact on the firm in that it affects cash flows, etc.
The financial instrument also has an effect on the presentation of the firms financial statements
in that U.S. financial reporting standards will require the firm to classify the instrument as either
a liability or as equity.

If a press release describing the issuance of the financial instrument made the following
statements, to what extent would you interpret these statements as referring to the economic
impact versus the presentational impact of the issuance?

1. The issuance of the financial instrument affects the firms vulnerability to default.

-50 -40 -30 -20 -10 0 10 20 30 40 50


Refers Refers
exclusively to exclusively to
the economic the presentational
impact impact

2. The issuance of the financial instrument affects the strength of the firms balance sheet.

-50 -40 -30 -20 -10 0 10 20 30 40 50


Refers Refers
exclusively to exclusively to
the economic the presentational
impact impact

30
REFERENCES

AHN, W., R. GOLDSTONE, B. LOVE, A. MARKMAN, and P. WOLFF. Categorization


Inside and Outside the Laboratory: Essays in Honor of Douglas L. Medin. Washington,
DC: American Psychological Association, (2005).

ASHTON, R. H. An Experimental Study of Internal Control Judgments. Journal of Accounting


Research 12 (1974): 143-157.

BARTH, M. E., W. R. LANDSMAN, and R. J. RENDLEMAN JR. Option Pricing-Based Bond


Value Estimates and A Fundamental Components Approach to Account for Corporate
Debt. The Accounting Review 73 (1998): 73-102.

BARTH, M. E., L. D. HODDER, and S.R STUBBEN. Financial Reporting for Employee Stock
Options: Liabilities or Equity. Review of Accounting Studies 18 (2013): 642-682.

BIANCO, D., and Z. DENG. Convertible Securities Analysis. UBS Investment Research
March 1, 2007, UBS Securities LLP.

BNP PARIBAS. The Return of the Corporate Hybrid. (2013). http://www.bnpparibas-


ip.com/publications/documents/other/IP/documentlist/newsletters/in-touch/20130305_in-
touch-233_eng_eng.pdf

BOLGER, A. Bond Issuers Increasingly Turn to Hybrids. Financial Times (2015).


http://www.ft.com/cms/s/0/98ea6f3c-bd0a-11e4-b523-00144feab7de.html

BOTOSAN, C., L. KOONCE, S. RYAN, M. STONE, and J. WAHLEN Accounting for


Liabilities: Conceptual Issues, Standard Setting, and Evidence from Academic Research.
Accounting Horizons 19(3) (2006): 159-186.

BREALEY, R. A., and S. C. MYERS. Principles of Corporate Finance. Boston, MA: McGraw-
Hill, (2000).

CHENG, Q., P. FRISCHMANN, and T. D. WARFIELD. The Market Perception of Corporate


Claims. Research in Accounting Regulation 16 (2003): 328.

CLOR-PROELL, S.M. The Effects of Expected and Actual Accounting Choices on Judgments
and Decisions. The Accounting Review 84(5) (2009): 1465-1493.

CLOR-PROELL, S. M., and M. W. NELSON. Accounting Standards, Implementation


Guidance, and Example-Based Reasoning. Journal of Accounting Research 45(4)
(2007): 699-730.

COWLEY, E., and A. MITCHELL. The Moderating Effect of Product Knowledge on the
Learning and Organization of Product Information. Journal of Consumer Research 30
(2003): 443-454.

31
DUTORDOIR, M., C. LEWIS, J. SEWARD, and C. VELD. What We Do and Do Not Know
About Convertible Bond Financing. Journal of Corporate Finance 24 (2014): 3-20.

ELLIOTT, W. B., S. D. KRISCHE, and M. E. PEECHER. Expected Mispricing: The Joint


Influence of Accounting Transparency and Investor Base. Journal of Accounting
Research 48(2) (2010): 343-381.

FAN, Y., A. BARUA, W. M. CREADY, and W. B. THOMAS. Managing Earnings Using


Classification Shifting: Evidence from Quarterly Special Items. The Accounting Review
85 (2010): 1303-1323.

FINANCIAL ACCOUNTING STANDARDS BOARD. Elements of financial statements. FASB


Statement of Financial Accounting Concepts No. 6. Norwalk, CT: FASB, (1985).

FINANCIAL ACCOUNTING STANDARDS BOARD. Statement of Financial Accounting


Standards No. 150: Accounting for Certain Financial Instruments with Characteristics of
both Liabilities and Equity. Norwalk, CT: FASB, (2003).

FINANCIAL ACCOUNTING STANDARDS BOARD. Preliminary Views: Financial


Instruments with Characteristics of Equity. Norwalk, CT: FASB, (2007).

FINANCIAL ACCOUNTING STANDARDS BOARD. Invitation to Comment: Disclosure


Framework. Norwalk, CT: FASB, (2012).

FINANCIAL ACCOUNTING STANDARDS BOARD. Disclosure Framework.


http://www.fasb.org/jsp/FASB/FASBContent_C/ProjectUpdatePage&cid=117615634489
4. (2015a).

FINANCIAL ACCOUNTING STANDARDS BOARD. Technical Agenda Overview,


http://www.fasb.org/cs/ContentServer?c=Document_C&pagename=FASB%2FDocument
_C%2FDocumentPage&cid=1176164187985. (2015b).

FINANCIAL ACCOUNTING STANDARDS BOARD. Accounting for Financial Instruments


Classification and Measurement-Joint Project of the FASB and IASB.
http://www.fasb.org/jsp/FASB/FASBContent_C/ProjectUpdatePage&cid=117615926771
8. (2015c).

FINANCIAL ACCOUNTING STANDARDS BOARD. Obtaining the Investors Perspectives,


http://www.fasb.org/jsp/FASB/Page/SectionPage&cid=1176156545230. (2015d).

FRISCHMANN, P., and T. WARFIELD. Multiple Motivations and Effects: The Case of Trust
Preferred Stock. Issues in Accounting Education 14 (1999): 269-284.

HALES, J., S. VENKATARAMAN, and T. J. WILKS. Accounting for Lease Renewal Options:
The Informational Effects of Unit of Account Choices. The Accounting Review 87(1)
(2011): 173-197.

32
HONECK, R. P., M. FIRMENT, and T. J. S. CASE. Expertise and Categorization. Bulletin of
the Psychonomic Society 25(6) (1987): 431-434.

HOPKINS, P. E. The Effect of Financial Statement Classification of Hybrid Financial


Instruments on Financial Analysts Stock Price Judgments. Journal of Accounting
Research 34 (Supplement) (1996): 33-50.

INTERNATIONAL ACCOUNTING STANDARDS BOARD. Classification and Measurement


of Share-based Payment Transactions. http://www.ifrs.org/Current-Projects/IASB-
Projects/IFRS-2-Clarifications-Classification-and-Measurement/ED-November-
2014/Pages/Exposure-Draft-and-Comment-letters.aspx. (2015).

JOHNSON, K. E., and C. B. MERVIS. Effects of Varying Levels of Expertise on the Basic
Level of Categorization. Journal of Experimental Psychology 126(3) (1997): 248-277.

KIMMEL, P., and T. D. WARFIELD. The Usefulness of Hybrid Security Classifications:


Evidence from Redeemable Preferred Stock. The Accounting Review 70(1) (1995): 151-
167.

KUEHL, R. O. Design of Experiments: Statistical Principles of Research Design and Analysis


(2nd edition). New York, NY: Duxbury Press, (2000).

LIBBY, R., R. BLOOMFIELD, and M. NELSON. Experimental Research in Financial


Accounting. Accounting, Organizations and Society 27 (2002): 775-810.

LINSMEIER, T., C. SHAKESPEARE, and T. SOUGIANNIS. Liability/Equity Classification


Decisions and Shareholder Value. Working paper, University of Illinois (2008).

LOKEN, B., L. W. BARSALOU, and C. JOINER. Categorization Theory and Research in


Consumer Psychology: Category Representation and Category-Based Inference. In
Handbook of Consumer Psychology, edited by P. H. Curtis, M. H. Paul, & R. K. Frank.
New York, NY: Taylor & Francis, (2008).

LOUVIERE, J. J. Analyzing Decision Making: Metric Conjoint Analysis. Sage University


Paper Series on Quantitative Applications in the Social Sciences, 67. Newbury Park, CA:
Sage Publications, Inc, (1988).

MCVAY, S. E. Earnings Management Using Classification Shifting: An Examination of Core


Earnings and Special Items. The Accounting Review 81 (2006): 501-531.

MEYVIS, T., and C. JANISZEWSKI. When Are Broader Brands Stronger Brands? An
Accessibility Perspective on the Success of Brand Extensions. Journal of Consumer
Research 31(2) (2004): 346-357.

NUNNALLY, J. C. Psychometric Theory, 2nd ed, New York: McGraw-Hill, (1978).

33
OPPENHEIM, A. N. Questionnaire Design, Interviewing, and Attitude Measurement. New
York, NY: Pinter Publishers, (1992).

PENMAN, S. H. Financial Statement Analysis and Security Valuation, 5th ed., New York:
McGraw Hill, (2013).

PETERSON, R. A. A Meta-Analysis of Cronbachs Coefficient Alpha. Journal of Consumer


Research 21 (1994): 381-391.

PROFFITT, J. B., J. D. COLEY, and D. L. MEDIN. Expertise and Category-Based Induction.


Journal of Experimental Psychology: Learning, Memory, and Cognition 26(4) (2000):
811-828.

RYAN, S. G., R. H. HERZ, T. E. IANNACONI, L. A. MAINES, K. PALEPU, C. M.


SCHRAND, D. J. SKINNER, and L. VINCENT. Evaluation of the FASBs Proposed
Accounting for Financial Instruments with Characteristics of Liabilities, Equity, or
Both. Accounting Horizons 15(4) (2001): 387-400.

RYAN, S. G. Financial Instruments & Institutions: Accounting and Disclosure Rules. Hoboken,
NJ: John Wiley & Sons, Inc, (2007).

SANGIUOLO, R., S. A. TAUB, and L.F. SEIDMAN. Financial Instruments: A Comprehensive


Guide to Accounting and Reporting. Chicago, IL: CCH, (2010).

SHAH, A. Creative Compliance in Financial Reporting. Accounting, Organizations and


Society 21 (1996): 23-39.

SLOVIC, P. Analyzing the Expert Judge: A Descriptive Study of a Stockbrokers Decision


Process. Journal of Applied Psychology 53(4) (1969): 255-263.

TROTMAN, K. T., H. C. TAN, and N. ANG. Fifty-year overview of judgment and decision-
making research in accounting. Accounting and Finance 51(1) (2011): 278-360.

WAHLEN, J. M., S. P. BAGINSKI, and M. T. BRADSHAW. Financial Reporting, Financial


Statement Analysis, and Valuation: A Strategic Perspective. Mason, OH: South-Western
Cengage Learning, (2014).

WALLACE, R., and C. MELLOR. Nonresponse Bias in Mail Accounting Surveys: A


Pedagogical Note. British Accounting Review 20(2) (1988): 131-139.

34
Panel A: Vulnerability to default

60

55

50

45

40

35

30
2 Liab / 3 Equity 3 Liab / 2 Equity

Liability Equity

Panel B: Balance sheet strength

70

65

60

55

50

45

40
2 Liab / 3 Equity 3 Liab / 2 Equity

Liability Equity

FIG 1.---Experiment One Creditworthiness Judgments. This figure shows the pattern of cell means for judgments
of vulnerability to default (Panel A) and balance sheet strength (Panel B). Descriptive statistics and statistical tests
for these measures are presented in Table 1.

35
40

30

20

10

-10

-20
2 Liab / 3 Equity 3 Liab / 2 Equity
Liability Equity

FIG 2.---Experiment One Lending decision. This figure shows the pattern of cell means for participants
willingness to lend. Descriptive statistics and statistical tests are presented in Table 2.

36
Security # 1
1. The security does not have a fixed maturity date.

2. The holders of the security have priority over common stockholders in a liquidation.

3. The holders of the security do not have voting rights.

4. If repaid, the security will be settled in cash.

5. The payments to security holders do not depend on the firm making profits.

Where does Security # 1 fall along the liability-equity continuum?

-50 -40 -30 -20 -10 0 10 20 30 40 50


Exactly like Equally like a Exactly like
a bond bond liability & common
liability common stock stock

FIG 3.--- Example of a hybrid instrument used in Experiment Two. This figure provides an example of one of the 16
instruments that participants evaluated in Experiment Two. The five features for each instrument were either more
liability-like or more equity-like as determined by the experimental design.

37
70.0%

60.0%

50.0%

40.0%

30.0%

20.0%

10.0%

0.0%
Maturity Priority Vo6ng Se;lement Prot

Decreases (up to -25) Increases (0-20) Increases (21-40) Increases (41-60) Increases (61-80)

FIG. 4.---Feature effect size by participant. This figure reports the effect of each feature manipulation on individual
participants overall judgments. We compute the main effect of each feature for each participant and then group
participants into buckets based on the size and direction of the effect. For example, when the maturity feature is
more equity-like rather than more liability-like, this causes 50 percent of our participants to increase their judgment
by up to 20 points on a 101-point scale in which the endpoints of 50 (50) are labeled exactly like a bond liability
(exactly like common stock).

38
TABLE 1
Experiment One Results Creditworthiness Judgments

Panel A: Descriptive statistics mean (standard deviation)


Vulnerability to Default Balance Sheet Strength
Classification Classification
Row Row
Features Liability Equity Liability Equity
Means Means
42.88 47.12 45.00 55.46 59.63 57.55
2 Liability/3 (19.06) (17.54) (18.32) (15.21) (16.30) (15.80)
Equity n=41 n=41 n=82 n=41 n=41 n=82

51.25 48.45 49.85 44.98 51.25 48.11


3 Liability/2 (15.72) (17.14) (16.40) (17.29) (16.19) (16.94)
Equity n=40 n=40 n=80 n=40 n=40 n=80
47.01 47.78 50.28 55.49
Column
(17.88) (17.25) (17.01) (16.69)
Means
n=81 n=81 n=81 n=81

Panel B: Analysis of variance results

Vulnerability to Default Balance Sheet Strength


F- p- F- p-
Source df M.S. Statistic value df M.S. Statistic value
Features 1 952.52 3.14 0.04a 1 3605.71 13.64 <0.01a
Classification 1 21.11 0.07 0.40a 1 1104.60 4.18 0.02a
Features Classification 1 502.29 1.66 0.20 1 44.83 0.17 0.68
Error 158 303.33 158 264.33

Note: This table reports the results of Experiment One, which investigates how the features and classification of a
newly-issued hybrid financial instrument affect experienced participants assessments of the companys
creditworthiness. We manipulate the features of a hybrid instrument between-participants at two levels, with the five
features of the instrument either indicating that the instrument is more equity-like (2 Liability/3 Equity condition) or
that the instrument is more liability-like (3 Liability/2 Equity condition). Classification is also manipulated between-
participants at two levels with the hybrid instruments either being classified as a liability or as equity. Participants
judge the companys vulnerability to default on a 101-point scale with endpoints of 0 (100), labeled as much less
vulnerable (much more vulnerable). Participants judge the effect of the issuance on the balance sheet on a 101-
point scale with endpoints of 0 (100), labeled as creates much weaker balance sheet (creates much stronger
balance sheet). Panel A reports the descriptive statistics for the vulnerability to default and balance sheet strength
measures, while Panel B reports the ANOVA results for these measures. a One-tailed equivalent given our
directional predictions.

39
TABLE 2
Experiment One Results Lending Decision

Panel A: Descriptive statistics mean (standard deviation)


Lending Decision
Classification
Features Liability Equity Row Means
16.95 28.38 22.59
(60.88) (58.99) (59.85)
2 Liability/3 Equity
n=41 n=40 n=81

7.05 10.33 8.67


(61.56) (56.33) (58.68)
3 Liability/2 Equity
n=40 n=39 n=79
12.06 19.47
Column Means (61.04) (58.03)
n=81 n=79

Panel B: Analysis of variance results


Lending Decision

Source df M.S. F-Statistic p-value


Features 1 7805.61 2.20 0.07a
Classification 1 2162.32 0.61 0.22a
Features Classification 1 662.46 0.19 0.67
Error 156 3540.52

Note: This table reports the results of Experiment One, which investigates how the features and classification of a
newly-issued hybrid financial instrument affect experienced participants willingness to lend to the company. We
manipulate the features of a hybrid instrument between-participants at two levels, with the five features of the
instrument either indicating that the instrument is more equity-like (2 Liability/3 Equity condition) or that the
instrument is more liability-like (3 Liability/2 Equity condition). Classification is also manipulated between-
participants at two levels with the hybrid instruments either being classified as a liability or as equity. Participants
indicate whether they are willing to lend to the company and then indicate the strength of their preference for the
chosen alternative on a 101-point scale with endpoints of 0 (100), labeled as very weak preference (very strong
preference). We convert this to a 201-point scale in which larger positive (negative) numbers indicate a stronger
preference for lending (not lending). Panel A reports the descriptive statistics for the lending decision, while Panel B
reports the ANOVA results for this measure. a One-tailed equivalent given our directional predictions.

40
TABLE 3
Experiment Two Results

Panel A: Descriptive statistics

Instrument Mean
Number Maturity Priority Voting Settle Profit (std dev.)
-21.18
1 E L L L L
(23.08)
-14.43
2 L E L L L
(23.46)
-26.69
3 L L E L L
(20.77)
-21.66
5 L L L E L
(22.99)
-22.89
9 L L L L E
(20.76)
13.36
4 E E E L L
(25.61)
14.04
6 E E L E L
(22.59)
15.57
10 E E L L E
(23.07)
8.21
7 E L E E L
(21.61)
14.43
11 E L E L E
(21.18)
12.57
13 E L L E E
(16.81)
7.70
14 L E L E E
(22.98)
9.30
8 L E E E L
(18.07)
13.34
12 L E E L E
(19.64)
3.85
15 L L E E E
(23.49)
40.14
16 E E E E E
(11.85)

41
Panel B: Repeated measures analysis of variance results

Two-tailed
Source Num df / Den df F-value p-value
Maturity 1/55 105.03 <0.01
Priority 1/55 120.39 <0.01
Voting 1/55 86.20 <0.01
Settlement 1/55 41.44 <0.01
Profit 1/55 97.05 <0.01
Maturity Priority 1/55 2.56 0.12
Maturity Voting 1/55 0.18 0.67
Maturity Settlement 1/55 0.09 0.77
Maturity Profit 1/55 2.35 0.13
Priority Voting 1/55 0.00 1.00
Priority Settlement 1/55 3.87 0.05
Priority Profit 1/55 3.80 0.06
Voting Settlement 1/55 0.95 0.33
Voting Profit 1/55 2.48 0.12
Settlement Profit 1/55 4.18 0.05

Note: This table reports the results of Experiment Two, which examines how the specific features of hybrid
instruments affect experienced users judgments of the instruments similarity to traditional liabilities or traditional
equity. Participants in Experiment Two evaluate 16 hybrid instruments that differ with respect to five separate
features. For each instrument, participants indicate on a 101-point scale whether the instrument is more like a
liability or more like equity. Scale endpoints of -50 (50) are labeled exactly like a bond liability (exactly like
common stock). Panel A presents the descriptive statistics and Panel B reported the results of a repeated measures
ANOVA.

42