Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
ESSAYS ON ACCOUNTING INFORMATION SYSTEMS
AND ACCOUNTING CHOICE
By Jongkyum Kim
A dissertation submitted to the
Graduate School-Newark
Rutgers, The State University of New Jersey
in partial fulfillment of requirements
for the degree of
Doctor of Philosophy
Ph.D. in Management
Written under the direction of
Professor Bharat Sarath
and approved by
Dr. Bharat Sarath
Dr. Miklos A. Vasarhelyi
Dr. Divya Anantharaman
Dr. Steven B. Lilien
Newark, New Jersey
October 2016
© [2016]
Jongkyum Kim
ALL RIGHTS RESERVED
ii
ABSTRACT OF THE DISSERTATION
Two Essays on Accounting Information Systems and Accounting Choices
By Jongkyum Kim
Dissertation Director: Professor Bharat Sarath
The first essay investigates whether the implementation of enterprise resource
planning (ERP) systems affects the audit repot lag, the time period between a company’s
fiscal year end and the date of the audit report. Prior research has shown that the
implementation of ERP systems can significantly affect a firm’s business operations and
processes. However, scant research has been conducted on the relationship between ERP
implementation and the timeliness of external audits, such as audit report lags. While some
of the alleged benefits of ERP are closely related to removing impediments contributing to
audit report lags, others argue that the complex mechanisms of ERP systems create greater
complexity for control and audit. The test results indicate that overall a firm’s ERP
implementation is negatively associated with audit report lag. However, this negative
association is significant only at fourth and fifth years after initial ERP implementation.
These results imply that the use of ERP systems by client firms may help decrease the audit
report lag, but it takes time for the full impact of the firms’ accounting systems to be
realized.
The second essay examines the consequences resulting from the adoption of the
fair value option (FVO) of SFAS 159 and explore the intentions of firms that adopt FVO.
The primary objective of SFAS 159 is to mitigate the earnings volatility. However, there
are also arguments implying that FVO could in fact increase earnings volatility. Thus, I
iii
first examine whether adopting FVO helps improve earnings volatility. Also, prior studies
report the negative relationship between earnings volatility and the accuracy of earnings
prediction. Therefore, the second hypothesis tests whether adopting FVO has an impact on
the analysts’ forecast accuracy. The test results indicate that there is a positive association
between FVO and earnings volatility and the relation can be amplified depending on the
size of gains or losses resulting from the FVO. Also, the results indicate that the analysts’
forecast error of FVO adopter is greater than that of non-FVO adopters. I test two more
hypotheses to explore the intention of firms’ decision to adopt FVO: opportunistic intention
and informative intention. However, I do not find results which confirm my hypotheses
except for limited evidence for the opportunistic intention.
iv
DEDICATION
To my parents, my brother and sister, my wife, and my sons
v
ACKNOWLEDGEMENTS
I would never have been able to finish my dissertation without the guidance of my
committee members, assistance from friends, and support from my family.
I would like to express my deepest gratitude to my advisor, Dr. Bharat Sarath, for
his enormous and continuous support and guidance. Dr Sarath has been such a great mentor
not only as an academic advisor but also as a true teacher for my life. The conversations
with him have been a good opportunity to broaden my horizons on different cultures and
our society. I am very fortunate to have worked under the supervision of an outstanding
scholar and a great person like him.
I am also very thankful to Dr. Miklos Vasarhelyi for his warm advices and
continuous support throughout the Ph.D. program. I sincerely admire his passion for
research and students, and the lessons from him will be very important guidance for my
future career as a college professor.
I also would like to thank Dr. Divya Anantharaman for her insightful comments
and support. Whenever I stop by her office for advices, she has always welcomed me with
smile and given me very helpful advices. I deeply appreciate her help and kindness. I also
thank Dr Steven Lilien for serving on my dissertation committee and for his valuable
comments and suggestions for my proposal and defense.
I am also very thankful to my friends and colleagues at Rutgers Business School,
Seokyoon, Changhee, Kyunghee, Karina, Vicky, Li He, Yubin, Hyunsang, and … Sunjung.
Most importantly, I cannot imagine that I could have accomplished this work
without the love and support of my family. I would like to express my warmest appreciation
to my parents, sister, brother, and wife, Jiyoung.
vi
Table of Contents
ABSTRACT OF THE DISSERTATION.................................................................... ii
DEDICATION ............................................................................................................. iv
ACKNOWLEDGEMENTS ........................................................................................ v
Table of Contents ........................................................................................................ vi
List of Tables ............................................................................................................. viii
Chapter 1: The Impact of Enterprise Resource Planning (ERP) Systems on the Audit
Report Lag ............................................................................................... 1
I. Introduction ............................................................................................................. 1
II. Literature Review and Research Question ............................................................. 6
III. Research Design ................................................................................................. 12
IV. Results ................................................................................................................. 19
V. Analyses with ERP data of the 2000s ................................................................... 23
VI. Conclusion .......................................................................................................... 26
References ................................................................................................................ 31
Tables for Chapter 1 ................................................................................................. 36
Chapter 2: The Fair Value Option (FVO) under SFAS 159: Effectiveness of FVO and
Intention of FVO Adopters .................................................................. 49
I. Introduction ........................................................................................................... 49
II. Literature Review on Fair Value Accounting ....................................................... 51
III. SFAS 159 ............................................................................................................ 62
IV. Hypothesis Development .................................................................................... 72
vii
V. Research Design ................................................................................................... 78
VI. Results ................................................................................................................ 84
VII. Conclusion, Limitations, and Extensions .......................................................... 90
References ................................................................................................................ 93
Tables for Chapter 2 ................................................................................................. 97
Appendix A for Chapter 2 ....................................................................................... 120
viii
List of Tables
Table 1.1 Determinants of audit report lag ................................................................. 36
Table 1.2 Tangible and intangible benefits of ERP systems ........................................ 37
Table 1.3 Sample selection and identification of ERP firms ....................................... 38
Table 1.4 Distribution of firm years for audit report lags ............................................ 39
Table 1.5 Variable definitions ..................................................................................... 40
Table 1.6 Descriptive statistics .................................................................................... 41
Table 1.7 Multivariate test results (Model 1) ............................................................... 45
Table 1.8 Multivariate test results (Model 2) ............................................................... 46
Table 1.9 Multivariate test results ................................................................................ 47
Table 1.10 Multivariate test results .............................................................................. 48
Table 2.1 Distribution of FVO firms by FVO adoption year and industry .................. 97
Table 2.2 Variable definitions ..................................................................................... 98
Table 2.3 Descriptive statistics (for H1) .................................................................... 100
Table 2.4 OLS regression analyses of earnings volatility on adoption of FVO (full
sample) ...................................................................................................... 103
Table 2.5 OLS regression analyses of earnings volatility on adoption of FVO (financial
firms only) ................................................................................................. 104
Table 2.6 OLS regression analyses of earnings volatility on adoption of FVO (non-
financial firms only) .................................................................................. 105
Table 2.7 Descriptive statistics (for H2) .................................................................... 106
Table 2.8 OLS regression analyses of analysts’ forecast error on adoption of FVO 108
Table 2.9 Descriptive statistics (for H3) .................................................................... 109
ix
Table 2.10 OLS regression analyses of financing activity on adoption of FVO ....... 111
Table 2.11 Logistic regression analyses of financing activity on adoption of FVO .. 112
Table 2.12 Gains or losses?: number of firms which reposrt gains or losses resulting from
FVO in the first FVO adoption year .......................................................... 113
Table 2.13 Logistic regression analyses of financing activity on adoption of FVO with
only firms which adopted FVO in 2008 (financial vs non-financial firms
seperately) ................................................................................................. 114
Table 2.14 Logistic regression analyses of financing activity on adoption of FVO with
only firms which adopted FVO in 2008 (financial vs non-financial firms
together) ..................................................................................................... 115
Table 2.15 Descriptive statistics (for H4) .................................................................. 116
Table 2.16 OLS regression analyses of exotic features of bonds on adoption of FVO
(Model 4-2a) .............................................................................................. 118
Table 2.17 Logistic regression analyses of exotic features of bonds on adoption of FVO
................................................................................................................... 119
1
Chapter 1: The Impact of Enterprise Resource Planning (ERP) Systems on the Audit
Report Lag
I. INTRODUCTION
The primary purpose of this study is to examine the relationship between the
implementation of enterprise resource planning (ERP) systems and the timeliness of
audited financial statements. Specifically, this study investigates whether the
implementation of ERP systems affects the audit report lag, the time period between a
company’s fiscal year end and the date of the audit report.
It is argued that an ERP system is the most important development in the corporate
use of information technology in the 1990s (Davenport 1998). An ERP system is a
packaged business software system that enables a company to manage resources (material,
human, financial, etc.) more efficiently and effectively by providing an integrated solution
for the organization’s information-processing needs (Nah et al. 2001). Kumar and
Hillegersberg (2000) define ERP systems as information system packages that integrate
information and information-based processes within and across functional areas in an
organization. As a result, ERP systems collect and distribute information more timely and
thus help managers improve their ability to process and analyze information (Hitt et al.
2002). Therefore, ERP systems can radically change the way in which accounting
information is processed, prepared, audited, and disseminated (Brazel and Dang 2008).
The alleged benefits and widespread usage of ERP systems have motivated many
empirical research inquiries over a variety of issues such as success factors of ERP, post-
implementation firm performance, and market reaction. However, a relatively small
amount of research has been conducted on how ERP systems affect the external audit.
2
Because an ERP system is typically the most important investment in information
technology of corporations, this study can provide useful insight on how information
technology (ERP) of a client firm affects the efficiency of external audit, at least to the
extent that audit report lag is a good proxy for audit efficiency.
Our research is motivated by several factors concerning recent audit and IT
environment of corporations. First, the usage of ERP systems is now widespread. ERP
software packages have become popular especially for both large and medium-sized
organizations to overcome the limitations of fragmented and incompatible legacy systems
(Robey et al. 2002). Thus, it is important to thoroughly examine the impacts of ERP
systems. However, scant research has been conducted on the relationship between ERP
implementation and the timeliness of external audits, such as audit report lags. An
interesting aspect of ERP is that even though managers decide to implement ERP systems
for their own operational purposes, ERP implementation has also changed audit
environment affecting the efficiency of audit work. According to Behn et al. (2006),
auditors regard inefficient client closing process and consolidation process as important
impediments to reduce the audit report lag. Also, they point out that poor integration of
different systems within the same company is the biggest system impediment to eliminate
the audit lag. However, some prior studies which investigate the characteristics or benefits
of ERP imply that those problems can be improved by implementing ERP systems
(O’Leary 2004). Thus, it will be interesting to examine the relationship between ERP
implementation and the audit report lag.
Second, there are contradicting implications over how advancements in
technology or ERP systems affect audit report lag. Issuing new corporate filing
3
requirements in 2002 which reduce the number of days allowed for filings of Forms 10-K
and 10-Q (SEC 2002), the Securities and Exchange Commission (SEC) argued that
advancements in information technology have improved the ability of companies to capture,
process, and disseminate their financial information. On the other hand, some studies
suggest that ERP systems can deteriorate the internal control mechanism of corporations.
For example, Lightle and Vallario (2003) point out that integrated ERP systems provide a
single point of control segmentation for segregation of duties, but also provide
opportunities for inappropriately configured access privileges that violate internal control
guidelines. As poorer internal controls might be positively associated with audit delay
(Ashton et al. 1987), this control risk can be a factor which leads to longer audit delay.
Thus, whether ERP systems implementation affects audit delay positively or negatively is
a matter of empirical investigation.
Third, there has been a growing interest in continuous auditing (CA) over the past
twenty years, and CA is increasingly under consideration as a tool to enhance the external
audit (Kuhn and Sutton 2010). However, theoretically CA can be realized by reducing time
to audit by taking advantage of more advanced IT systems, and it is expected that the
movement to CA model will be evolutionary rather than revolutionary (Behn et al. 2006).
Even though this paper neither directly relates ERP systems to CA nor argues that ERP
systems will realize CA, we think that ERP systems are critical investments of firms in
their IT systems, and thus the development and pervasive use of ERP systems can provide
the significant infrastructure necessary for the evolution of the assurance function from the
traditional audit to CA (Kuhn and Sutton 2010). In this regard, examining the relationship
between ERP implementation and the audit report lag can add empirical evidence on
4
whether ERP systems or advances in information technology of client firms can actually
help reduce the time spent by external auditors to complete audit works and furthermore
whether they can make a contribution to the movement toward CA.
The results of this study suggest that ERP implementation is negatively associated
with audit report lag. The test results of the first model indicate that an increase in audit
report lags of post-ERP implementation period for ERP firms is significantly smaller than
that of matched firms. On the other hand, it is not likely that ERP systems are associated
with audit report lags immediately after those systems are implemented. According to the
results of the second model, only the fourth and fifth years after ERP implementation show
significant negative coefficients. This result may suggest that it takes time for ERP systems
to be fully utilized and to make a significant impact on the firm’s accounting systems.
This study contributes to the research in both the accounting information systems
and auditing literatures which examine the effects of ERP systems implementation on a
firm’s information environment. What is interesting in our findings is that implementing
ERP systems can have unintended effects. Managers decide to implement ERP systems in
order to improve operational performances of their firms. However, this paper indicates
that implementing ERP systems can affect not only various measures of firms’
performances but also the external audit which is generally regarded as something
independent of client firms’ decisions. This is the first paper to show empirical evidence
that investments in information technology of client firms can make an impact on the
efficiency of external audit even though the investment is motivated by the management’s
operational purposes. This finding also contributes to auditing literature in that it identifies
an additional determinant of the audit report lag, which is the quality of IT systems of client
5
firms. Identifying the determinants of audit delay is meaningful because understanding the
determinants of audit delay helps us to better explain the auditing processes or behaviors
of auditors. For example, given the results of this study, we may conjecture that external
auditors rely at least partially on the clients’ IT systems when they collect and process data
necessary to carry out their audit tasks. Although Brazel and Dang (2008) report a negative
association between ERP implementation and earnings announcement lags, the earnings
announcement lag cannot properly reflect the audit delay because it is influenced by
management’s incentives (Bamber et al. 1993). Thus, our study more directly examines the
association between ERP implementation and audit delay. In addition, our study indirectly
supports the argument that implementation of ERP systems is positively associated with
the effectiveness of internal controls. From the result that ERP implementation is associated
with shorter audit report lags, we may conjecture that as ERP systems get effectively
utilized external auditors are likely to regard ERP firms as having stronger internal controls
than non-ERP firms. That is because weak internal control indicates high control risk which
may force auditors to conduct more strict audits in order to decrease detection risk,
ultimately leading to an increased audit work.
The remainder of this paper is organized as follows: Section II discusses prior
research and develops our research question; Section III describes data collection process
and research methodology; Section IV presents empirical results; Section V discusses the
analyses with the 2000s’ ERP data; and Section VI concludes by discussing implications,
limitations, and further research.
6
II. LITERATURE REVIEW AND RESEARCH QUESTION
Audit Report Lag
Relevance, along with reliability, is a primary attribute that makes accounting
information useful for decision making (FASB 1980) and one of the critical ingredients of
relevance is timeliness. Being delayed in releasing financial statements affects the
timeliness of information provided (Ashton et al. 1989). Audit delay, which can be
measured by audit report lag, is likely to increase the level of uncertainty associated with
decisions for which the financial statements provide information, and thus can ultimately
impair the usefulness of these reports (Givoly and Palmon 1982). In general, the concept
of timeliness in financial reporting can be defined by two dimensions: the frequency of
reporting, which is the length of reporting period, and the reporting lag (Davies and
Whittred 1980). This study uses the latter one to examine whether ERP implementations
affect the timeliness of financial reports. Thus, the audit report lag in this study is defined
as the time period between a firm’s fiscal year end and the audit report date. For the audit
report dates, we have used the dates when auditors sign on the independent auditors’ reports
of 10-Ks.
There are numerous studies which investigate the determinants of the audit report
lag. One stream of the research on the determinants of the audit report lag is related to
auditors. Using Australian data, Whittred (1980) examines the effect of qualified audit
reports on the timeliness of annual reports. The results show that qualifications are
positively associated with delay in releasing preliminary profits and final annual reports.
They also indicate that the more serious the qualification is, the greater the audit report lag
is. He argues that this result is due to the increase in time spent for the year-end audit and
7
auditor-client negotiations. Newton and Ashton (1989) examine the relationship between
audit structure and audit delay for Canadian Big Eight firms, and find that auditors using
structured audit approaches tend to have greater audit report lag than auditors using
unstructured approaches. Knechel and Payne (2001) relate audit report lag to incremental
audit hours, resource allocation of audit team effort, and provision of non-audit services.
Their results indicate that the presence of controversial tax issues and the use of less
experienced audit staff are positively correlated with audit report lag.
Another research stream on the determinants of the audit report lag is related to
the characteristics of client firms. Using the U.S. firm data, Givoly and Palmon (1982)
show that the reporting lag of individual firms seems to be more closely related to industry
patterns and tradition rather than to firm attributes such as size and complexity of
operations. Using data from an international accounting firm, Ashton et al. (1987)
investigate the relationship between the audit report lag and fourteen different firm
attributes. Among these attributes, five variables show significant relationship with the
report lag: total revenue representing firm size, quality of internal control, public company,
operation complexity, and relative amount of interim work. On the other hand, Ashton et
al. (1989) examine cross-sectional variability in audit delay with a sample of Canadian
firms. They test eight variables and the results indicate that firm size (total asset), industry
(financial service), negative net income, and the existence of extraordinary items are
significantly related to audit delay. Bamber et al. (1993) try to propose a more
comprehensive model of audit report lag and emphasize the importance of the amount of
audit work required, the level of resources expended to complete the audit, and the audit
firm technology. The results show that the explanatory power of their model is almost three
8
times greater than that of previous models.
Besides these studies, Krishnan and Yang (2009) explore the trends in audit report
lags and earning announcement lags, and examine the quality of reporting for the period
from 2001 to 2006. The results show that although the filing lags decreased after 2003
when the new corporate filing requirements became effective, audit lags increased
substantially following the new rules particularly in 2004 and 2005 with the introduction
of SOX Section 404. On the other hand, the long audit report lags do not seem to be
associated with lower quality of reporting, measured by absolute value of discretionary
accruals and quality of accruals. Table 1.1 exhibits the determinants of the audit report lag
which have been identified as significant in some of the prior studies.
[Insert Table 1.1 about here]
Enterprise Resource Planning (ERP) Systems
Empirical research related to ERP systems can be broadly divided into four
streams. First, several studies explore the impacts of ERP systems on corporations. Deloitte
Consulting’s (1998) study suggests the rationale and specific benefits of implementing
ERP, and O’Leary (2004) identifies tangible and intangible benefits. Table 1.2 presents
parts of benefits reported on Deloitte Consulting (1998) and O’Leary (2004). However,
Granlund and Malmi (2002) argue that ERP systems have led to relatively small changes
in management accounting and control procedures.
[Insert Table 1.2 about here]
Second, a significant amount of ERP research has identified the factors necessary
for a successful ERP implementation and an effective ongoing usage. Through an extensive
9
literature review, Nah et al. (2001) identify eleven critical factors for the success of ERP
projects; among those are the teamwork and composition of ERP project team, top
management support, business process reengineering, and effective communication.
Bradley (2008) examines success factors using the classical management theory. The
author suggests that a proper project manager, personnel training, and the presence of a
champion are important to implement ERP systems successfully. In addition, the
organizational culture and top management leadership, including strategic and tactical
conducts, appear to be significantly associated with successful ERP implementation (Ke
and Wei 2008).
Third, a stream of research investigates the relationship between ERP
implementation and post-implementation firm performance. Research findings show
mixed results on this association. For example, Poston and Grabski (2001) find no positive
relation between ERP implementation and overall post-implementation financial
performance, but Hunton et al. (2003) find that although the financial performance of ERP
adopters does not change much, non ERP adopters tend to decrease compared to the
performance of ERP adopters. Nicolaou (2004) indicates that ERP implementation has a
positive association with a long-term financial performance and this relation is stronger
when implementation characteristics are controlled. Nicolaou and Bhattacharya (2006;
2008) also show that both the timing of ERP customizations and carrying of such
modifications in conformity with the adopting organizations’ strategic objectives increase
post-implementation operational performance.
Finally, prior studies have found that in general the market reacts positively to
announcements of ERP implementation. Hayes et al. (2001) have found the overall positive
10
reaction to initial ERP announcements measured by standard cumulative abnormal returns.
This positive reaction seems to be more significant when firms are small and healthy and
when the ERP vendors are large. Hunton et al. (2002) show that analysts react to ERP
implementation plans by revising their earnings forecasts positively. Consistent with Hayes
et al. (2001), this tendency is stronger when firms are financially healthy.
Research Question
Key characteristics of ERP systems are integration, standardization, routinization,
and centralization. These attributes allow managers to access more comprehensive
information on a near real-time basis so that they can make better decisions. However,
implementation of ERP systems affects not only internal information processes but also
external audit environment. Using ERP systems, firms can integrate data from different
departments or business segments, standardize data, improve financial control, and reduce
financial closing cycles and data entry mistakes (O’Leary 2004). These benefits have
potential to remove the main impediments recognized by auditors for the audit delay, such
as poor integration of different systems, poor standardization of data, poor financial control,
and inefficient financial closing (Behn et al. 2006). Thus, the alleged benefits of ERP
systems can imply that the audit report lags are likely to decrease by implementing ERP.
In addition, Morris (2011) finds that ERP firms are less likely to report internal
control weaknesses under SOX Section 404. This may imply that external auditors are more
likely to regard ERP firms as having stronger internal controls compared to non-ERP firms.
According to the audit risk model suggested by the AU Section 312 (2007), audit risk
consists of inherent risk, control risk, and detection risk. Thus, if control risk is low,
11
auditors can increase detection risk, maintaining the same level of overall audit risk. Then,
ceteris paribus, auditors can carry out audit tasks less strictly, and thus the amount of end-
year audit work will decrease. As a result, the audit report lag is likely to decrease.
Therefore, we might conjecture that ERP implementation can lead to shorter audit report
lag. In fact, Ettredge et al. (2006) report an association between material weakness in
internal control and audit delay. With introduction of SOX section 404, attestation of
corporate internal control by auditors and managers became mandatory. Using data on
internal control assessments over financial reporting, Ettredge et al. (2006) find that the
presence of material internal control weakness is associated with longer audit delay. Also,
Pae and Yoo (2001) presents a model which shows that when the auditor’s legal liability is
large the client firm underinvests in the internal control system, and thus the auditor
overinvests effort, leading to an decrease in audit efficiency.
However, some others are against this argument. Grabski et al. (2011) say that the
hidden and complex mechanisms of ERP systems create greater complexity for control and
audit (Grabski et al. 2011). Lightle and Vallario (2003) point out that integrated ERP
systems provide a single point of control segmentation for segregation of duties, but also
provide opportunities for inappropriately configured access privileges to violated internal
control guidelines. These studies imply that ERP implementation can rather increase the
overall control risk. In fact, we have witnessed that several firms had experienced serious
problems in the process of implementing advanced information systems. For example,
building and implementing a new system known as ‘Pulse’, Oxford Health Plans
experienced critical problems which led to nearly out of control on its various business
processes (Hammonds and Jackson 1997). Thus, if we apply the same logic of the audit
12
risk model mentioned earlier, it might be expected that ERP implementation is associated
with longer audit report lag.
It is, therefore, an empirical question whether ERP implementation affects audit
report lag positively or negatively. This paper seeks to provide empirical evidence on this
issue. Thus, the research question examined in this study is:
RQ: Is there an association between ERP implementation and audit report lag?
III. RESEARCH DESIGN
Sample Selection
We used the original Nicolaou (2004) data set of ERP and non-ERP firms in order
to examine our research question. For collecting ERP data, a two-phase process was
employed to identify the sample of ERP firms. First, ERP firms were identified by
extracting ERP implementation announcements from the Lexis-Nexis Academic Universe
(News) Wire Service Reports for the period of January 1, 1990 through December 31, 1998.
Following Hayes et al. (2001), we employed a keyword search method using a combination
of the search terms "implement," "convert," and "contract" with the name of each of the
following ERP vendors: Adage, BAAN, Epicor, GEAC Smartstream, Great Plains,
Hyperion, Intentia Intemational, JBA Intemational, JD Edwards, Lawson, Oracle
Financials, Peoplesoft, QAD, SAP, SSA, or SCT. This key word search method yielded an
initial sample of 1,825 announcements. After eliminating announcements which are not
related to ERP implementation, a total number of 332 announcements remained. Second,
the Global Disclosure database was searched for any mention of ERP system
13
implementation in annual reports and SEC filings. The search term of “enterprise resource
planning” was applied to the full text of 10-Ks and 10-Qs for the time period from January
1, 1990 to February 28, 1999. The initial search identified 1,453 documents. Subsequently,
each of these documents was examined individually. Finally, this process resulted in 131
valid ERP implementation related disclosures. These 131 disclosures are all from firms that
are not included in the original Lexis-Nexis Newswires search. Thus, a total number of 463
firms were identified to have implemented ERP systems in the sample period through our
two-phase process. It is also important to notice that all of these announcements or SEC
filings are related to initial implementations of ERP systems, thus providing important
information on ERP adopters.
A subsequent cross-match with the “Global Vantage Key” (GVKey) file from
Research Insight database resulted in the elimination of 142 firms. Furthermore, 67 firms
were eliminated because financial data for these firms were not available in Compustat
database. In addition, five foreign firms were excluded, along with two other firms that
reported discontinued ERP projects. After these elimination processes, we had the final
sample of 247 firms. The paired t-test results indicate that the two groups from different
data sources are not significantly different in terms of size, measured by total assets and
net sales. Panel A of Table 1.3 presents the distribution of the final sample of 247 firms by
year and source, while Panel B presents the distribution of ERP firms by industry.
[Insert Table 1.3 about here]
Identification of ERP Adoption and ERP Completion Year
ERP project completion time is defined as the system go-live date, the time point
14
at which the system gets ready to use. Of the 247 firms included in the final sample, only
105 firms disclosed both the start and completion years of the ERP implementation. For
the 142 sample firms that did not disclose a completion year, it was assumed that their
expected completion time was equal to the mean expected completion time of 9.92 months,
which is based on reporting by 72 firms from the original Lexis-Nexis newswires sample
and 89 firms from the original Global Disclosure database sample that reported expected
completion times. The year of completion was coded as the time period ‘t0’ to indicate that
this is the starting year of ERP use.
Matching Procedure
Following Barber and Lyon (1996), each ERP firm was matched with a control
group company on both industry and size at the year preceding the ERP adoption year (time
t-1). Firms were first matched by the four-digit Standard Industrial Classification (SIC)
code, and then matched by total assets. There were some instances where it was necessary
to go beyond the four-digit SIC code. For such cases, firms were matched using a three-
digit code. Also, an inability to find proper U.S. matches for certain unique firms resulted
in five lost firms; an example of one such firm is Boeing. As a result, the sample which has
been actually used for subsequent analyses consisted of only 242 firms.
In order to validate the soundness of the matching procedure, we performed a
search of the Lexis-Nexis Newswires and the websites of the major ERP vendors. Through
this procedure, we could find that 12 firms which were originally included in the non-ERP
matched group had been actually implemented ERP systems. These 12 firms were
substituted with other non-ERP firms. Furthermore, a graduate assistant called the IT
15
director or other person in charge of system implementation to make sure that the firms
selected as non-ERP matched firms are actually not ERP adopters. This procedure
eliminated 38 firms which either did not respond or responded that they had implemented
an ERP system in the recent past. These firms were also substituted with others.
Audit Report Lag Data
The data for audit report lags were manually collected with all available 10-Ks of
ERP firms and matched firms for the time period from January 1, 1994 to December 31,
2006. Because 10-Ks before 1994 were not available and the time period of our ERP
sample data is up to 1998, the time period of 1994 to 2006 was used. Initially, 4,330 firm
years were collected: 1,927 firm years for ERP firms and 2,403 firm years for matched
firms. Eliminating the firm years whose control variables used in our models were missing
resulted in the decrease of our sample size to 3,710 firm years. In addition, we deleted firm
years which are more than five years earlier or more than five years later than the ERP
completion year. This elimination gave us a final sample of 3,225 firm years. The final
numbers of firm years for ERP firms and for non-ERP firms are 1,616 and 1,609,
respectively. In case of ERP firms, the number of firm years for pre-ERP period is 863 and
that of post-ERP period is 753. As to non-ERP firms, the number of firm years for pre-ERP
period is 673 and that of post-ERP period is 936. Panel A and B of table 1.4 describes the
distribution of firm years for audit report lag data.
[Insert Table 1.4 about here]
Model Specification
16
We have tested the association between ERP implementation and audit report lag
by estimating log-normal duration models. The standard log-normal model is based on the
assumption that the durations follow a log-normal distribution with density function:
𝑓(𝑡) =1
𝑏𝑡𝜙 (
log(𝑡)−𝑎
𝑏) , 𝑏 > 0
where 𝜙 and Φ denote the standard normal density and distribution functions respectively:
𝜙(𝑡) =1
√2𝜋exp (−
𝑡2
2) and Φ(t) = ∫ 𝜙(𝜏)𝑑𝜏
t
0.
The survivor function and transition rate functions are
𝐺(t) = 1 − Φ(log(𝑡)−𝑎
𝑏)
𝑟(𝑡) =1
𝑏𝑡
𝜙(𝑧𝑡)
1−Φ(𝑧𝑡) with 𝑧𝑡 =
log(𝑡)−𝑎
𝑏
where a=Aα and b=exp(Bβ). A and B denote covariate vectors introduced into the model.
It is assumed that the first component of each of the covariate (row) vectors A and B is a
constant equal to one. The associated coefficient vectors α and β are the model parameters
to be estimated (Blossfeld et al. 2007).
There are several reasons why we chose a log-normal model. First, a simple OLS
model is not appropriate for this study since the audit report lag is positively skewed
(Krishnan and Yang 2009), which is a violation of one of the OLS assumptions for
hypothesis testing. Second, transformation of data can be avoided if we use a log-normal
model. Some prior studies, such as Krishnan and Yang (2011) and Ashton et al. (1987,
1989), use the logarithmic transformation to address the normality issue. However, we try
to avoid this transformation since transformation of data can be a distortion of data. Third,
a log-normal model is a member of the class of accelerated failure time models which can
be interpreted in terms of time duration. Audit report lag is time duration measured by
17
number of days. Thus, it appears to be appropriate for this type of study. Lastly, the hazard
ratio graph of our data has the property that the transition rate initially increases with time
to the maximum and then decreases, which is consistent with the shape of log-normal
transition rate graph. A log-normal model is widely used when the transition rate at first
increases and then, after reaching a maximum, decreases (Blossfeld et al. 2007). For
example, Levinthal and Fichman (1988) examine auditor-client relationships to explore the
dynamics of interorganizational relations by estimating a log-normal model. They find that
interorganizaitonal attachments have positive duration dependence in the initial years of
attachment and negative duration dependence for longer durations. In particular, the results
show that the hazard rate of dissolution in the early stages of attachments increases with
time and after this honeymoon period the rate declines as the tenure of the attachment
increases. Bucklin and Sismeiro (2003) have also used the log-normal model to explore the
browsing behavior of visitors to a Web site. This study focuses on two basic elements of
browsing behavior: a user’s decision to continue browsing or to exit the site and the length
of time spent for each page. They specify a log-normal model for examining page-view
durations and the results indicate that the browsing patterns of the Web site users are
consistent with learning effects, site stickiness, time constraints, and cost-benefit trade-
offs.1
We have two models to provide empirical evidence on our research question. In
the first model, the independent variable of our interest is AfterERP which is an interaction
term of two indicator variables, ERP and After. The variable ERP is an indicator variable
1 Log-normal models have been also employed for studies in other disciplines such as medicine,
food science, and zoology (Strum et al. 2000; McCready et al. 2000; Gamel and McLean 1994;
Hough et al. 2003; Aragao et al. 2007; Tokeshi 1990; Tolkamp and Kyriazakis 1999).
18
which takes 1 if a firm-year is for an ERP firm and 0 otherwise. The variable After is also
an indicator variable which takes 1 if a firm-year belongs to post-ERP period and 0
otherwise. Thus, the variable AfterERP takes 1 only if a firm year is one of the post-ERP
firm years for ERP firms. Consequently, by using an interaction term, we can examine the
difference of differences, and the coefficient of AfterERP shows how much audit report
lags of ERP firms change compared to the change in audit report lags of non-ERP firms.
However, it is not likely that ERP systems make a significant impact on the firms
right away as soon as it becomes implemented. The results of Nicolaou (2004) suggest that
it takes time for ERP systems to significantly affect the accounting systems of companies.
Thus, in the second model, we divide the variable ‘AfterERP’ into five post-ERP years
(from AfterERP1 to AfterERP5) so that we can examine how long it takes for ERP systems
to get associated with the audit report lag.
In addition, both models include various control variables which seem to have
significant impact on audit report lag in prior research. Givoly and Palmon (1982) report
that report lag appears to be associated with industry patterns and tradition, and other
studies indicate that different industries are a significant factor which affects audit report
lag (Ashton et al. 1989; Newton and Ashton 1989; Bamber et al. 1993; Krishnan and Yang
2009). Thus, following Krishnan and Yang (2009), we included four indicator variables to
control different practices and accounting rules across industries: HighLit which represents
high litigation industries, HighGrowth which represents high growth industries, Financial
which represents financial services industry, and HighTech which represents high tech
industries. Prior studies found accounting or firm complexity to be a significant
determinant for audit report lag (Ashton et al. 1987; Ashton et al. 1989; Newton and Ashton
19
1989; Bamber et al. 1993; Krishnan and Yang 2009). Thus, to control the different degree
of complexity in the firm’s accounting and operations, we included variables which
represent the number of business segments (NumSegment), the existence of extraordinary
items (ExtraItem), and the existence of foreign operations (ForeignOper). We also included
Size and Auditor to control different firm size and audit firm effect. In addition, we
controlled different fiscal year end by including BusyEnd that is an indicator variable which
distinguish firms whose fiscal year end is December or January from firms whose fiscal
year end is the other months. Furthermore, financial distress seems to affect the audit report
lag (Ashton et al. 1989; Bamber et al. 1993; Krishnan and Yang 2009). Financial distress
is proxied by variables representing negative earnings (Loss) and debt to total asset ratio
(Leverage). Audit opinion is also found to be a significant determinant in several prior
studies (Bamber et al. 1993 and Krishnan and Yang 2009). We have captured audit opinion
with the variable of AuditOpi.2 Finally, we controlled the time trend of audit report lags by
including year dummies. Table 1.5 summarizes the definitions of variables used in this
study.
[Insert Table 1.5 about here]
IV. RESULTS
Descriptive Statistics
Table 1.6 reports the descriptive statistics, where Panel A, B, and C display the
2 The indicator variable ‘AuditOpi’ takes 1 if the firm receives an opinion other than clean opinion,
0 otherwise. We used ‘AUOP’ (Auditor Opinion) item in Compustat for this variable. When
constructing this variable, we coded only ‘unqualified opinion’ as clean opinion and ‘unqualified
opinion with additional language’ was not coded as clean opinion because the additional language
can affect the time that auditors need to spend to complete the audit.
20
distributional properties of the variables used in this study. Panel A shows the descriptive
statistics of all samples including both ERP firms and non-ERP firms. The mean of audit
report lags is 46.2 days. This figure is similar to means of lags of year 2001 or year 2002
of Krishnan and Yang (2009) data and about 22 days shorter than one reported in Knechel
and Payne (2001) that use data of year 1991. Also, the descriptive statistics indicate that
about sixty percent of firm years have fiscal year end month of December or January and
more than ninety percent of firm years were audited by one of big six auditors. Also, less
than thirty percent of firm years received other than clean opinion from auditors. Panel B
exhibits the descriptive statistics of ERP firms versus non-ERP firms. The descriptive
statistics indicate that ERP firms are bigger than non-ERP firms, more likely to have
foreign operations and have fewer business segments. The audit report lags of two samples
do not seem to be different in average. Panel C presents descriptive statistics of pre- and
post- ERP implementation periods. Numerous variables of post-ERP implementation
period are significantly bigger than those of pre-ERP implementation period: number of
segments, incidence of extraordinary items, size, incidence of negative earnings, firms
audited by big six auditors, firms receiving other than a clean opinion, and firms operating
abroad. Also, the mean of audit report lags of post-ERP implementation period is
significantly longer than that of pre-ERP implementation period. Panel D reports the pair-
wise Pearson correlations and Spearman correlations. The table indicates that audit report
lag is negatively correlated with high litigation industry, high tech industry, size, big six
auditors, and foreign operations, and positively correlated with net loss and modified
opinions.
[Insert Table 1.6 about here]
21
Multivariate Results
We examine how ERP implementation affects the audit report lag by estimating
two log-normal duration models. The test results of the first model are presented in Table
1.7. The results show a negative and significant coefficient for the interaction term,
AfterERP, which indicates that ERP implementation is negatively associated with the
length of audit report lag. The coefficients of other control variables are consistent with
prior studies in general. High litigation industry, high tech industry, and client size seem to
be negatively associated with the audit report lag. It seems that the negative association
between the audit report lag and high litigation and high tech industries is due to the
concern of the market on firms that belong to risky industries. It is expected that being
aware of this concern of the market, those firms try to relieve the market’s anxiety by
releasing annual reports more timely. On the other hand, presence of extraordinary items,
negative earnings, and modified audit opinion are positively associated with the audit
report lag. The dummy variable for year 2004 is positive and strongly significant. This can
be attributed to the SOX Section 404 of reporting requirements which became effective in
2004. The SOX section 404 requires managers and auditors to assess the soundness of
corporate internal control, an attestation which was not mandatory before the regulation.
The introduction of this new regulation probably leads to increased audit hours to be spent
in order to complete the audit. The result for year 2004 dummy is also consistent with
Ettredge et al. (2006) and Krishnan and Yang (2009) that report dramatic increases in audit
report lags in 2004, 19.8 days (from 50.3 to 70.1) and 23.3 days (from 49.3 to 72.6) in
average respectively.
22
[Insert Table 1.7 about here]
Table 1.8 reports the test results for the second model. The coefficients of the post-
ERP 4th and 5th year interaction terms are negative and significant. Even though the
coefficients of the other three post-ERP period terms are not significant, they are all
negative. This result implies that ERP systems may help decrease the audit report lag, but
it takes time for the significant impact on the firms’ accounting systems to be realized. The
results for other control variables are mainly the same as those of the first model.
[Insert Table 1.8 about here]
Additional Tests
There exists possibility that the auditors in the late 1990s might have more
knowledge or better understanding on ERP systems than in the early 1990s because ERP
systems began to be introduced in the early 1990s and as time passes the auditors may get
accustomed to those systems. Thus, we checked whether there is knowledge effect (or
education effect) of auditors by examining whether there is any difference in impacts of
ERP on ARLs between early ERP adopters and late ERP adopters. However, we couldn’t
find any significant differences between two groups (untabulated).3
In addition to two log-normal models, we ran OLS regressions using the natural
logarithm of audit report lag, rather than the audit report lag itself, as a dependent variable.
The results for both models are qualitatively the same as those of log-normal models
(untabulated). The coefficient of the interaction term in the first model is negative and
3 We divided ERP firms into two groups, early adopters and late adopters. However, because the
dividing point is not clear, we examined three pairs of groups: pre and post-1994, 1995, and 1996
pairs. Nevertheless, we couldn’t find significant results for all three pairs.
23
significant at five percent level. The coefficients of the post-ERP 4th and 5th year interaction
terms in the second model are also negative and significant. The results for control
variables are very similar to those of log-normal models.
Furthermore, we examined the VIF values of the independent variables to check
whether there is a serious multicollinearity problem. However, we could not find an
indication of multicollinearity. The VIF values for all variables do not exceed five.
V. ANALYSES WITH ERP DATA OF THE 2000S
ERP data of the 2000s
We tried to collect additional information on the firms that have implemented ERP
systems in the 2000s, from year 2000 to year 2008, in order to check whether our results
hold with more recent data. The data collection method is similar to that employed for
collecting our main ERP data. We used Lexis-Nexis Newswires to search for any
announcements relevant to ERP implementation and searched for any mentions of ERP
implementation in SEC filings. This search process yielded 786 unique firms which had
ERP-related announcements or mentions related to ERP implementation in SEC filings
during the search period. However, 151 firms were eliminated because those firms are not
available in Compustat. In order to collect information on ARLs, we employed the Audit
Analytics database instead of collecting manually. Additional 43 firms were eliminated
which are not available in Audit Analytics or whose data for ARLs are not available in
Audit Analytics. Consequently, we had the final sample of 592 firms. Finally, the ERP
firms were matched by the four-digit SIC code and total assets to form a non-ERP control
group. The firm years were selected as they had been for our main ERP data. These
24
processes gave us a final sample of 8,622 firm years.
Multivariate Results
The models estimated with the recent ERP data are the same as those used for the
main ERP data, and the test results are presented in Table 1.9 and 1.10. We could find
results which are consistent with those from our main ERP data, which makes the reliability
of the results even stronger. Table 9 reports the results of the first model. The results show
that the coefficient of the interaction term, AfterERP, is negative and significant at 1 percent
level. The results for other control variables are similar to those from the main data. In
general, the signs of coefficients are the same, but they tend to be more significant than
those of the previous corresponding model. For example, the coefficients of NumSegment
and BusyEnd which were positive and insignificant are now positive and significant. The
coefficients for year dummies are all significant, but the sign turns to positive from negative
in year 2004, which has a similar implication to that of previous results, the effect of SOX
section 404. One possible reason for the increase in statistical significance is probably due
to the increase in the number of observations estimated.
[Insert Table 1.9 about here]
Table 10 reports the results for the second model using the recent ERP data. The
coefficients of all five interaction terms are negative, but in contrast to the results of the
previous corresponding model, they are all statistically significant. We can find one
possible reason for this result from the difference in (IT) audit environments between 1990s
and 2000s. In the 1990s when ERP systems were first introduced and both implementing
firms and auditors were not familiar with those systems, it might take much more time for
25
firms and auditors to learn about how they work and to assess their reliability. However, in
the 2000s ERP systems are not new anymore to auditors and probably even to
implementing firms. In other words, their knowledge level on ERP systems in the 2000s
may be much higher than that in the 1990s. Thus, we can conjecture that it might not take
much time for auditors to understand and effectively utilize ERP systems for their audit.
Another possible reason can arise from the possibility that those ERP announcements in
the 2000s are not indications of initial ERP implementations, rather just additions of some
more systems to or upgrades of existing ERP systems. This is more likely especially for
the 2000s’ data than 1990s’ data. The period of the 1990s is the booming period of
implementing ERP systems. The popularity of ERP did soar in the early 1990s and firms
began to invest billions in ERP systems (Chen 2001). Thus, it is likely that most major
companies were already using ERP systems in the early 2000s. Therefore, we cannot deny
the possibility that a significant portion of ERP-related announcements in the 2000s can be
related to additions of some more systems to existing ERP systems or upgrades of them. If
this is the case, auditors wouldn’t need to spend time in order to understand those
modifications to the systems as much as when the client firms are first implementing ERP
systems. Thus, it would also take less time for those modifications to make a significant
impact on the firms’ accounting systems.
[Insert Table 1.10 about here]
Limitations of Data
In spite of our efforts to collect more recent ERP data, we should confess that this
recent ERP data can be noisy. As mentioned above, because it is highly likely that most
26
major firms had already implemented ERP systems until the early 2000s, it is more difficult
to find proper matched non-ERP firms in the 2000s than in the 1990s. Also, when collecting
ERP data, it is very difficult to distinguish between initial ERP implementations and
different sorts of system modifications such as upgrades, additions of more systems, or
system integrations of foreign branches. This problem is more pronounced in the 2000s’
data because the 1990s’ data includes only initial implementations (as mentioned in the
sample selection section) and it is more likely that ERP-related announcements in the 2000s
are just additions of some more systems to existing ERP systems or upgrades of them
because of the aforementioned reason.
Despite of these potential disadvantages of the 2000s’ ERP data, we however think
that the data still provide interesting information for a different time period. Also, the
results of this data make our main conclusion of negative association between ERP
implementations and ARLs even stronger in that overall they are quite consistent with those
of the 1990s’ data. Nevertheless, we need to be careful in interpreting the results from the
2000s’ data, especially for the second model, because as mentioned above the data can be
noisy and there is a discrepancy between the results of the second model and prior ERP
studies. Prior studies, in general, report that it takes time for ERP systems to make an
impact on implementing firms.
VI. CONCLUSION
Implications
We investigated whether ERP implementation is associated with the audit report
lag. The audit report lag is affected by numerous factors as shown in prior research. Recent
27
advance in technology has radically changed business operation and processes. Meanwhile,
it has also changed audit environment and can affect the determinants of the audit report
lag. This paper is the first study to show empirical evidence that investment in IT of client
firms can make an impact on the external audit even though the investment is motivated
by managers’ operational purposes. The test results of this study indicate that the ERP
implementation has negative association with the audit report lag. Given that an ERP
system is typically the most important investment in information technology for a company,
this result has an implication that advanced technology of client firms can have a positive
influence on the efficiency of external audit, at least to the extent that audit report lag is a
good proxy for audit efficiency. Also, the test result of our second model suggests that it
takes time for ERP systems to be fully utilized and to make a significant impact on the
accounting systems. Our findings are interesting in that implementing ERP systems can
make unintended effects on the external audit. Companies decide to implement ERP
systems to improve their own operational performances. However, our results indicate that
a decision to implement ERP systems can affect not only various measures of firms’
performances but also the external audit which is generally regarded as something
independent of client firms’ decisions.
We might be able to find possible reasons for these results from the characteristics
of ERP systems. In the accounting aspect, ERP systems can be characterized as integrated
and comprehensive enterprise-wide record keeping systems which embrace most activities
and processes of an organization. Thus, the firms with ERP systems can be equipped with
more efficient accounting closing and consolidation processes. Under the ERP
environment, auditors first need to examine the reliability of systems, which means the
28
adequacy of system design and compliance. Once the results of this examination are
acceptable, auditors can save time spent on collecting data from different departments or
business segments and on confirming management assertions because they can be provided
with comprehensive, integrated, and reliable information much more quickly. Our test
results might indicate that auditors are realizing these benefits of ERP systems.
The SEC mandated accelerated filing requirements for corporate 10-K and 10-Q
filings in 2003 (SEC 2002). The requirements reduced the 10-K filing period for
accelerated filers from 90 days to 75 days, and for large accelerated filers the filing period
was further reduced to 60 days. Concerning this filing environment, some might argue that
auditors might not have as much incentive as before to complete audit works sooner than
required periods and thus people might not be interested in audit delay as much as before.
Those arguments might be true. However, the main finding of this paper is that auditors
are more efficient for ERP firms, and this can mean that even if firms file at the same date
the audit quality will be higher for ERP firms. Also, the audit efficiency is closely related
to auditors’ costs. Thus, the audits can be more profitable for ERP firms as the audit costs
may be lower because of the increased efficiency. Therefore, the finding of this paper still
provides implications on different aspects of audit under the current filing rule. Research
questions related to audit fee or audit quality may be worth examining.
Prior studies focus on effects of ERP systems only on firm performances, such as
operational performances or market performances, which are managers’ intended
(expected) effects. However, this paper implies possibility that implementing advanced IT
systems can have unexpected impacts on different aspects including outsiders of the
implementing firms. Therefore, this study provides an implication that as more advanced
29
and sophisticated IT systems get introduced related parties should endeavor to understand
all potential effects of the systems more comprehensively so that they can maximize the
benefits of the systems.
Limitations
This study has some limitations. First, because ERP related announcements, which
are the source of data for this study, are not detailed enough, we couldn’t take into
consideration different degrees of integration of a firm’s IT systems. If the ERP systems
are implemented only in financials and not integrated with logistics, MRP or CRM, or if
the subsidiaries are not included in the implementation, the benefits of ERP for the external
audit will be limited. However, we couldn’t distinguish different degrees of integration due
to the lack of information. Future research with more complete and thorough information
on the integration of ERP systems might be more convincing.
Additionally, more comprehensive and in-depth research is needed about how
advance in technology of client firms affects the behaviors of external auditors or about
what the true factors are which improve audit efficiency. There is a concern on the change
in audit environment caused by the advance in technology. It is argued that auditors do not
properly recognize the heightened degree of risk in ERP systems regarding internal control,
such as network security, database security, and application security (Hunton et al. 2004).
Hunton et al. (2004) say that auditors do not have enough knowledge on systems and ERP
audits should be performed by a cross-functional team which consists of both auditors and
IS experts. If this argument is true, it could be conjectured that auditors trust IT systems of
client firms without proper evaluation on them and reduce their audit work. Then, this is
30
not the realization of benefits of ERP, but rather they are increasing the efficiency of audit
at the expense of effectiveness of audit. Thus, future research that deals with this issue will
be interesting and meaningful.
Further Research
Morris and Laksmana (2010) indicate that ERP implementation is negatively
associated with earnings management activities. Earnings management by client firms
increases the detection risk of auditors and thus ultimately increases the audit risk. Pratt
and Stice (1994) and Simunic and Stein (1996) suggest that one of the responses of auditors
to risk (litigation risk) is audit fee adjustment. Also, it is obvious that the time which
auditors spend to carry out audit tasks is a critical factor which affects the level of audit
fees. Thus, jointly the results of Morris and Laksmana (2010) and of this study, we may
conjecture that ERP implementation is negatively associated with audit fees. By examining
this additional hypothesis, we may be able to understand more comprehensively how
advances in technology of client firms affect the environment of external audit.
31
REFERENCES
Aragao, G. M. F., M. G. Corradini, M. D. Normand, and M. Peleg. 2007. Evaluation of the
weibull and log normal distribution functions as survival models of Escherichia Coli
under isothermal and non isothermal conditions. International Journal of Food
Microbiology 119: 243-257.
American Institute of Certified Public Accountants (AICPA). 2007. Audit Risk and
Materiality in Conducting an Audit. AU Section 312. New York, NY: AICPA.
Ashton, R. H., J. J. Willingham, and R. K. Elliott. 1987. An empirical analysis of audit
delay. Journal of Accounting Research 25 (Autumn): 275–292.
Ashton, R. H., P. R. Graul, and J. D. Newton. 1989. Audit delay and the timeliness of
corporate reporting. Contemporary Accounting Research 5 (Spring): 657–673.
Bamber, E. M., L. S. Bamber, and M. P. Schoderbek. 1993. Audit structure and other
determinants of audit report lag: An empirical analysis. Auditing: A Journal of
Practice & Theory 12 (Spring): 1–23.
Blossfeld, H., K. Golsch, and G. Rohwer. 2007. Event history analysis with Stata.
Lawrence Erlbaum.
Bradley, J. 2008. Management based critical success factors in the implementation of
enterprise resource planning systems. International Journal of Accounting
Information Systems 9 (3): 175–200.
Brazel, J. F., and L. Dang. 2008. The effect of ERP system implementations on the
management of earnings and earnings release dates. Journal of Information Systems
22 (2): 1-21.
Behn, B. K., D. L. Searcy, and J. B. Woodroof. 2006. A within firm analysis of current and
expected future audit lag determinants. Journal of Information Systems 20 (1): 65-
86.
Bucklin, R. E., and C. Sismeiro. 2003. A model of web site browsing behavior estimated
on clickstream data. Journal of Marketing Research 40 (3): 249-267.
Chen, I. J. 2001. Planning for ERP systems: Analysis and future trend. Business Process
Management Journal 7 (5): 374-386.
Davenport, T. H. 1998. Putting the enterprise into the enterprise system. Harvard Business
Review 76 (July/August): 121-131.
32
Davis, B., and G. P. Whittred. 1980. The association between selected corporate attributes
and timeliness in corporate reporting: further analysis. A Journal of Accounting,
Finance and Business Studies 16 (June): 48-60.
Deloitte Consulting. 1998. ERP’s Second Wave. New York, NY: Deloitte Consulting.
Ettredge, M. L., C. Li, and L. Sun. 2006. The impact of SOX section 404 internal control
quality assessment on audit delay in the SOX era. Auditing: A Journal of Practice &
Theory 25 (2): 1-23.
Financial Accounting Standards Board (FASB). 1980. Qualitative Characteristics of
Accounting Information. FASB Concept No. 2. Norwalk, CT: FASB.
Gamel, J. W., and I. W. McLean. 1994. A stable, multivariate extension of the log-normal
survival model. Computers and Biomedical Research 27: 148-155.
Givoly, D., and D. Palmon. 1982. Timeliness of annual earnings announcements: Some
empirical evidence. The Accounting Review LVII (3): 486-508.
Grabski, S. V., S. A. Leech, and P. J. Schmidt. 2011. A review of ERP research: a future
agenda for accounting information systems. Journal of Information Systems 25 (1):
37-78.
Granlund, M., and T. Malmi. 2002. Moderate impact of ERPs on management accounting:
A lag or permanent outcome? Management Accounting Research 13 (3): 299–321.
Hammonds, K. H., and S. Jackson. 1997. Behind Oxford’s billing nightmare.
BusinessWeek. November 17. Available at:
http://www.businessweek.com/1997/46/b3553148.htm.
Hayes, D. C., J. E. Hunton, and J. L. Reck. 2001. Market reaction to ERP implementation
announcements. Journal of Information Systems 15 (1): 3–18.
Hitt, L. M., D. J. Wu, and X. Zhou. 2002. Investment in enterprise resource planning:
Business impact and productivity measures. Journal of Management Information
Systems 19 (Summer): 71-98.
Hough, G., K. Langohr, G. Gomez, and A. Curia. 2003. Survival analysis applied to
sensory shelf life of foods. Journal of Food Science 68 (1): 359-362.
Hunton, J. E., A. M. Wright, and S. Wright. 2004. Are financial auditors overconfident in
their ability to assess risks associated with enterprise resource planning systems?
Journal of Information Systems 1 (2): 7–28.
33
Hunton, J. E., B. Lippincott, and J. L. Reck. 2003. Enterprise resource planning systems:
Comparing firm performance of adopters and non-adopters. International Journal of
Accounting Information Systems 4 (3): 165–184.
Hunton, J. E., R. A. McEwen, and B. Wier. 2002. The reaction of financial analysts to
enterprise resource planning (ERP) implementation plans. Journal of Information
Systems 16 (1): 31–40.
Ke, W., and K. K. Wei. 2008. Organizational culture and leadership in ERP
implementation. Decision Support Systems 45 (2): 208–218.
Khun, J. R., and S. G. Sutton. 2010. Continuous audit in ERP system environments: The
current state and future directions. Journal of Information Systems 24 (1): 91-112.
Knechel, W. R., and J. L. Payne. 2001. Additional evidence on audit report lag. Auditing:
A Journal of Practice & Theory 20 (May): 137–147.
Krishnan, J., and J. S. Yang. 2009. Recent trends in audit report and earnings
announcement lags. Accounting Horizons 23 (3): 265-288.
Kumar, K., and J. V. Hillegersberg. 2000. ERP experiences and evolution.
Communications of the ACM (April): 22–26.
Levinthal, D. A., and M. Fichman. 1988. Dynamics of interorganizational attachments:
Auditor-client relationships. Administrative Science Quarterly 33 (3): 345-369.
Lightle, S. S., and C. W. Vallario. 2003. Segregation of duties in ERP. The Internal Auditor
60 (5): 27-31.
McCready, D. R., J. W. Chapman, W. M. Hanna, H. J. Kahn, D. Murray, E. B. Fish, M. E.
Trudeau, I. L. Andrulis, and H. L. A. Lickley. Factors affecting distant disease-free
survival for primary invasive breast cancer: Use of a log-normal survival model.
Annals of Surgical Oncology 7 (6): 416-426.
Morris, J. J. 2011. The impact of enterprise resource planning (ERP) systems on the
effectiveness of internal controls over financial reporting. Journal of Information
Systems 25 (1): 129-157.
Morris, J. J., and I. Laksmana. 2010. Measuring the impact of enterprise resource planning
(ERP) systems on earnings management. Journal of Emerging Technologies in
Accounting 7: 47-71.
Nah, F., J. Lau, and J. Kuang. 2001. Critical factors for successful implementation of
34
enterprise systems. Business Process Management 7 (3): 285–296.
Newton, J. D., and R. H. Ashton. 1989. The association between audit technology and audit
delay. Auditing: A Journal of Practice and Theory 8 (Supplement): 22-37.
Nicolaou, A. I. 2004. Firm performance effects in relation to the implementation and use
of enterprise resource planning systems. Journal of Information Systems 18 (2):79-
105.
Nicolaou, A. I. and S. Bhattacharya. 2006. Organizational performance effects of ERP
systems usage: The impact of post-implementation changes. International Journal of
Accounting Information Systems 7 (1): 18-35.
Nicolaou, A.I. and S. Bhattacharya. 2008. Sustainability of ERP performance outcomes:
The role of post-implementation review quality. International Journal of Accounting
Information Systems 9 (1): 43-60.
O’Leary, D. E. 2004. Enterprise resource planning (ERP) systems: An empirical analysis
of benefits. Journal of Emerging Technologies in Accounting 1: 63-72.
Pae, S., and S. Yoo. 2001. Strategic interaction in auditing: An analysis of auditors’ legal
liability, internal control system quality, and audit effort. The Accounting Review 76
(3): 333-356.
Poston, R., and S. Grabski. 2001. Financial impacts of enterprise resource planning
implementations. International Journal of Accounting Information Systems 2 (4):
271–294.
Pratt, J., and J. D. Stice. 1994. The effects of client characteristics on auditor litigation risk
judgements, required audit evidence, and recommended audit fee. The Accounting
Review 69 (October): 639-656.
Robey, D., J. W. Ross, and M. C. Boudreau. 2002. Learning to implement enterprise
systems: An exploratory study of the dialectics of change. Journal of Management
Information Systems 19 (1): 17–46.
Securities and Exchange Commission (SEC). 2002. Acceleration of Periodic Report Filing
Dates and Disclosure Concerning Website Access to Reports. Release No. 33-8128.
Washington, D.C.: SEC. Available at: http://www.sec.gov/rules/final/33-8128.htm.
Simunic, D. A., and M. T. Stein. 1990. Audit risk in a client portfolio context.
Contemporary Accounting Research 6 (Spring): 329-343.
35
Strum, D. P., J. H. May, and L. G. Vargas. 2000. Modeling the uncertainty of surgical
procedure times. Anesthesiology 92 (4): 1160-1167.
Tokeshi, M. 1990. Niche apportionment or random assortment: Species abundance patterns
revisited. Journal of Animal Ecology 59 (3): 1129-1146.
Tolkamp, B. J., and I. Kyriazakis. To split behavior into bouts, log-transform the intervals.
Animal Behaviour 57 (4): 807-817.
Whittred, G. P. 1980. Audit qualification and the timeliness of corporate annual reports.
The Accounting Review LV (4): 563-577.
36
Tables for Chapter 1
Table 1.1
Determinants of Audit Report Lag
(1) (2) (3) (4) (5) (6)
Audit
client
related
Client size + - - - -
Net loss + + +
Poor financial condition +
Extraordinary Item + + + +
Operational complexity + +
Public company - -
Magmt incentive to provide timely reports -
Quality of Internal Control -
Industry (financial) - - - -
Auditor
related
Opinion qualification + +
Audit technology + +
Audit team allocation (partner or manager) -
Year end (Dec, or Dec and Jan) - +
Interim work - -
(1): Ashton et al. 1987, data of international accounting firm
(2): Ashton et al. 1989, data of Canadian firms
(3): Newton and Ashton 1989, data of Canadian firms
(4): Bamber et al. 1993, data of U.S. firms
(5): Knechel and Payne 2001, data of international accounting firm
(6): Kirishnan and Yang 2009, data of U.S. firms
37
Table 1.2
Tangible and Intangible Benefits of ERP Systems
Tangible Benefits Intangible Benefits
Inventory Reduction Information Access/Visibility
Personnel Reduction New Improved Processes
Productive Improvements Customer Responsiveness
Order Management Improvements Integration
Financial Close Cycle Reduction Standardization
IT Cost Reduction New Reports/Reporting Capability
Cash Management Improvement Sales Automation
Revenue/Profit Increases Financial Control
Maintenance Reductions No Redundant Data Entry
38
Table 1.3
Sample Selection and Identification of ERP firms
Panel A: Distribution of ERP implementations by year in sample period
Implementation
year Number of valid
ERP anncmts Number of ERP firms with
GVKey identification
Number of ERP firms
with available data in
Compustat
Source: Lexis-Nexis Newswires
98 141 69 52
97 83 55 44
96 43 30 21
95 19 15 11
94 32 19 17
93 3 4 2
92 4 4 4
91 3 2 2
90 4 2 2
Sum 332 200 155
Source: Global Disclosure
98 49 46 39
97 66 62 41
96 12 10 8
90-95 4 3 4
Sum 131 121 92
Grand Sum 463 321 247
Panel B: Distribution of ERP firms by industry
ERP firms represented by the following SIC codes Number of firms
0700 - Agricultural Services 3
1000 - Mining, Construction 5
2000 - Manufacturing 56
3000 - Manufacturing 127
4000 - Transportation, Communications, Utilities 12
5000 - Wholesale and Retail Trade 18
6000 - Finance, Insurance, Real Estate 10
7000 - Services 14
8000 - Health Services 2
9000 - Public Institutions 0
Total 247
39
Table 1.4
Distribution of firm years for audit report lags
Panel A: Distribution of firm years by year
Year Number of firm years
1993 22
1994 161
1995 237
1996 352
1997 425
1998 422
1999 368
2000 340
2001 294
2002 260
2003 219
2004 119
2005 4
2006 2
Sum 3225
Panel B: Distribution of firm years by ERP project completion year
Implementation
Year Number of firm years of audit report lags
ERP Firms Matched Firms
Pre-ERP
Years
-5 44 16
-4 87 45
-3 145 91
-2 196 145
-1 201 180
0* 190 196
Sum 863 673
Post-ERP
Years
+1 182 194
+2 156 194
+3 151 190
+4 139 185
+5 125 173
Sum 753 936
*denotes ERP project completion year.
40
Table 1.5
Variable Definitions
Audi Report Lag = the number of days from the fiscal year-end to the audit report date;
ERP = 1 if the firm is in the ERP firm data, 0 otherwise;
After = 1 if the firm year is an post-ERP implementation year, 0 otherwise;
AfterERP = Interaction term of ERP and After;
After1 to After5 = Indicator variables representing the first, second, third, fourth, and fifth year
after the ERP project completion year, respectively;
AfterERP1 to
AfterERP5 = Interaction terms between ERP and After1 to After5, respectively;
HighLit = 1 if the firm belongs to industries 28, 35, 36, 38, 60, 67, and 73 (two-digit
SIC code), 0 otherwise;
HighGrowth = 1 if the firm belongs to industries 35, 45, 48, 49, 52, 57, 73, 78, and 80 (two-
digit SIC code), 0 otherwise;
Financial = 1 if the firm belongs to industry 60-67 (two-digit SIC code), 0 otherwise;
HighTech = 1 if the firm belongs to industries 283, 284, 357, 366, 367, 371, 382, 384, and
737 (three-digit SIC code), 0 otherwise;
NumSegment = the number of business segments (Compustat historical segments file);
ExtraItem = 1 if the firm reports extraordinary items, 0 otherwise;
Size = natural log of total assets;
BusyEnd = 1 if the firm’s fiscal year ends in December or January, 0 otherwise;
Loss = 1 if the firm reports a loss before extraordinary items, 0 otherwise;
Leverage = debt to total assets ratio;
Auditor = 1 if the firm is audited by a Big 6 auditor, 0 otherwise;
AuditOpi = 1 if the firm receives an opinion other than clean opinion, 0 otherwise
(Compustat item ‘AUOP’);
ForeignOper = 1 if the firm has foreign operations, 0 otherwise; and
GC = 1 if the firm receives a going concern opinion, 0 otherwise (Audit Analytics).
41
Table 1.6
Descriptive Statistics
Panel A: Distributional Properties of Variables (All Sample), N=3225
Variable Mean Median Std Dev 1Q 3Q
Audit report lag 46.244 41 33.002 29 54
ERP 0.501 1 0.5 0 1
After 0.523 1 0.499 0 1
AfterERP 0.233 0 0.423 0 0
After1 0.116 0 0.32 0 0
After2 0.108 0 0.311 0 0
After3 0.105 0 0.307 0 0
After4 0.1 0 0.3 0 0
After5 0.092 0 0.289 0 0
HighLit 0.465 0 0.498 0 1
HighGrowth 0.222 0 0.415 0 0
Financial 0.026 0 0.16 0 0
HighTech 0.34 0 0.473 0 1
NumSegment 2.47 2 1.904 1 4
ExtraItem 0.143 0 0.35 0 0
Size 6.453 6.386 1.948 4.977 7.829
BusyEnd 0.622 1 0.484 0 1
Loss 0.267 0 0.442 0 1
Leverage 0.964 0.557 19.225 0.386 0.715
Auditor 0.908 1 0.287 1 1
AudOpi 0.294 0 0.455 0 1
Foreign 0.584 1 0.492 0 1
(continued on the next page)
42
(continued on the next page)
TABLE 1.6 (continued)
Panel B: Distributional Properties of Variables (ERP firm sample vs Matched firm sample)a
ERP Firm Sample, N=1616 Matched Firm Sample, N=1609
Variable Mean Median Std Dev 1Q 3Q Mean Median Std Dev 1Q 3Q
Audit report lag 44.939 41 29.096 29 54 47.553 41 36.47 30 55
After 0.465 0 0.498 0 1 0.581 1 0.493 0 1
After1 0.112 0 0.316 0 0 0.12 0 0.325 0 0
After2 0.096 0 0.295 0 0 0.12 0 0.325 0 0
After3 0.093 0 0.291 0 0 0.118 0 0.322 0 0
After4 0.086 0 0.28 0 0 0.114 0 0.319 0 0
After5 0.077 0 0.267 0 0 0.107 0 0.309 0 0
HighLit 0.454 0 0.498 0 1 0.476 0 0.499 0 1
HighGrowth 0.219 0 0.414 0 0 0.224 0 0.417 0 0
Financial 0.038 0 0.192 0 0 0.014 0 0.118 0 0
HighTech 0.334 0 0.471 0 1 0.346 0 0.475 0 1
NumSegment 2.401 1 1.897 1 3 2.54 2 1.9089 1 4
ExtraItem 0.136 0 0.343 0 0 0.149 0 0.356 0 0
Size 6.534 6.474 2.019 5.007 7.953 6.372 6.309 1.87 4.916 7.752
BusyEnd 0.624 1 0.484 0 1 0.621 1 0.485 0 1
Loss 0.255 0 0.436 0 1 0.28 0 0.449 0 1
Leverage 0.562 0.562 0.265 0.379 0.718 1.368 0.55 27.215 0.39 0.714
Auditor 0.916 1 0.276 1 1 0.901 1 0.298 1 1
AudOpi 0.3 0 0.458 0 1 0.288 0 0.453 0 1
Foreign 0.621 1 0.485 0 1 0.546 1 0.497 0 1
43
(continued on the next page)
TABLE 1.6 (continued)
Panel C: Distributional Properties of Variables (Pre-ERP vs Post-ERP)a
Pre-ERP Implementation, N=1536 Post-ERP Implementation, N=1689
Variable Mean Median Std Dev 1Q 3Q Mean Median Std Dev 1Q 3Q
Audit report lag 44.033 40 27.962 29 51 48.254 41 36.888 29 58
ERP 0.561 1 0.496 0 1 0.445 0 0.497 0 1
HighLit 0.439 0 0.496 0 1 0.488 0 0.5 0 1
HighGrowth 0.209 0 0.407 0 0 0.233 0 0.423 0 0
Financial 0.033 0 0.18 0 0 0.019 0 0.138 0 0
HighTech 0.317 0 0.465 0 1 0.361 0 0.48 0 1
NumSegment 2.008 1 1.553 1 3 2.891 2 2.088 1 4
ExtraItem 0.107 0 0.309 0 0 0.175 0 0.38 0 0
Size 6.27 6.153 1.866 4.873 7.521 6.62 6.56 2.005 5.094 8.139
BusyEnd 0.608 1 0.488 0 1 0.635 1 0.481 0 1
Loss 0.203 0 0.402 0 0 0.326 0 0.468 0 1
Leverage 0.6 0.555 0.991 0.391 0.708 1.295 0.56 26.548 0.381 0.722
Auditor 0.878 1 0.326 1 1 0.936 1 0.244 1 1
AudOpi 0.234 0 0.423 0 0 0.349 0 0.476 0 1
Foreign 0.55 1 0.497 0 1 0.615 1 0.486 0 1
44
TABLE 1.6 (continued)
Panel D: Pearson Correlations (top) and Spearman Correlation (bottom)b, N=3225
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Audit report lag 1 -0.04 0.06 -0.07 0.02 -0.02 -0.09 -0.02 0.02 -0.23 -0.01 0.20 0.03 -0.08 0.14 -0.13
ERP 2 -0.03 -0.12 -0.02 -0.01 0.08 -0.01 -0.04 -0.02 0.04 0.00 -0.03 -0.02 0.03 0.01 0.08
After 3 0.05 -0.12 0.05 0.03 -0.04 0.05 0.23 0.10 0.09 0.03 0.14 0.02 0.10 0.13 0.07
HighLit 4 -0.15 -0.02 0.05 0.32 -0.15 0.60 -0.11 -0.07 -0.16 -0.12 0.12 -0.02 -0.03 -0.03 0.25
HighGrowth 5 -0.04 -0.01 0.03 0.32 -0.09 0.08 -0.07 0.01 -0.07 -0.07 0.15 -0.01 -0.04 0.01 0.01
Financial 6 -0.01 0.08 -0.04 -0.15 -0.09 -0.12 0.03 -0.01 0.11 0.09 -0.04 0.00 0.05 0.00 -0.10
HighTech 7 -0.19 -0.01 0.05 0.60 0.08 -0.12 -0.13 -0.09 -0.18 -0.08 0.09 -0.02 -0.02 -0.08 0.18
NumSegment 8 0.00 -0.04 0.23 -0.12 -0.08 0.04 -0.16 0.12 0.37 0.08 -0.06 -0.01 0.11 0.10 0.16
ExtraItem 9 0.06 -0.02 0.10 -0.07 0.01 -0.01 -0.09 0.12 0.17 0.07 0.03 -0.01 0.06 0.19 -0.01
Size 10 -0.28 0.03 0.09 -0.15 -0.08 0.10 -0.16 0.35 0.17 0.20 -0.27 -0.07 0.20 0.10 0.29
BusyEnd 11 -0.01 0.00 0.03 -0.12 -0.07 0.09 -0.08 0.08 0.07 0.20 -0.03 0.02 0.02 0.05 0.04
Loss 12 0.20 -0.03 0.14 0.12 0.15 -0.04 0.09 -0.05 0.03 -0.26 -0.03 0.04 0.00 0.09 -0.04
Leverage 13 0.18 0.00 0.01 -0.30 -0.06 0.11 -0.33 0.26 0.20 0.28 0.17 0.17 0.00 0.04 -0.03
Auditor 14 -0.07 0.03 0.10 -0.03 -0.04 0.05 -0.02 0.11 0.06 0.19 0.02 0.00 0.02 -0.02 0.12
AudOpi 15 0.11 0.01 0.13 -0.03 0.01 0.00 -0.08 0.10 0.19 0.11 0.05 0.09 0.21 -0.02 0.04
Foreign 16 -0.16 0.08 0.07 0.25 0.01 -0.10 0.18 0.15 -0.01 0.30 0.04 -0.04 -0.03 0.12 0.04
aBold text indicates significant differences between two subsamples at the 0.05 level or better, one-tailed. Differences in means (medians) are asseseed using a t-test
(Wilcoxon rank sum test).
bBold text indicates significance at the 0.05 level or better, two-tailed.
45
Table 1.7
Multivariate Test Results (Model 1)
Variable Coefficient Std Error z-value p -value
Constant*** 4.192 0.090 46.37 0.000
AfterERP** -0.042 0.019 -2.21 0.027
HighLit** -0.078 0.020 -3.95 0.000
HighGrowth -0.030 0.019 -1.63 0.104
Financial -0.018 0.046 -0.4 0.692
HighTech*** -0.167 0.019 -8.62 0.000
NumSegment 0.005 0.004 1.06 0.291
ExtraItem*** 0.088 0.022 4.03 0.000
Size*** -0.071 0.005 -15.28 0.000
BusyEnd 0.023 0.015 1.52 0.130
Loss*** 0.154 0.018 8.76 0.000
Leverage 0.000 0.000 -0.3 0.763
auditor -0.034 0.026 -1.32 0.188
AudOpi*** 0.117 0.018 6.36 0.000
Foreign -0.026 0.016 -1.57 0.116
d_1994 -0.027 0.088 -0.3 0.763
d_1995 -0.059 0.086 -0.69 0.492
d_1996 -0.039 0.085 -0.46 0.649
d_1997 -0.029 0.085 -0.34 0.737
d_1998 0.023 0.085 0.27 0.791
d_1999 0.025 0.086 0.29 0.773
d_2000 0.046 0.086 0.53 0.594
d_2001 0.009 0.086 0.1 0.921
d_2002 -0.009 0.087 -0.1 0.920
d_2003 0.131 0.087 1.5 0.134
d_2004*** 0.455 0.091 5 0.000
d_2005 0.373 0.302 1.24 0.216
d_2006 0.048 0.418 0.11 0.909
Numer of Obs 3225
LR chi2(27) 852.84
Prob > chi2 0.000
***, **, * denote significant at the 0.01, 0.05, and 0.1 levels, respectively.
46
Table 1.8
Multivariate Test Results (Model 2)
Variable Coefficient Std Error z-value p -value
Constant*** 4.185 0.090 46.3 0.000
AfterERP1 -0.003 0.033 -0.1 0.921
AfterERP2 -0.038 0.036 -1.07 0.283
AfterERP3 -0.030 0.037 -0.83 0.407
AfterERP4 * -0.066 0.039 -1.72 0.085
AfterERP5 ** -0.107 0.043 -2.51 0.012
HighLit*** -0.077 0.020 -3.91 0.000
HighGrowth -0.030 0.019 -1.59 0.113
Financial -0.017 0.046 -0.37 0.715
HighTech*** -0.167 0.019 -8.6 0.000
NumSegment 0.005 0.004 1.07 0.287
ExtraItem*** 0.089 0.022 4.09 0.000
Size*** -0.071 0.005 -15.21 0.000
BusyEnd 0.024 0.015 1.55 0.120
Loss*** 0.153 0.018 8.73 0.000
Leverage 0.000 0.000 -0.3 0.767
Auditor -0.033 0.026 -1.27 0.206
AudOpi*** 0.116 0.018 6.34 0.000
Foreign -0.026 0.016 -1.6 0.111
d_1994 -0.025 0.088 -0.29 0.774
d_1995 -0.058 0.086 -0.67 0.501
d_1996 -0.038 0.085 -0.44 0.656
d_1997 -0.027 0.085 -0.32 0.752
d_1998 0.023 0.085 0.27 0.788
d_1999 0.021 0.086 0.24 0.807
d_2000 0.041 0.086 0.47 0.637
d_2001 0.011 0.087 0.13 0.899
d_2002 -0.001 0.087 -0.01 0.989
d_2003* 0.151 0.088 1.71 0.086
d_2004*** 0.485 0.093 5.24 0.000
d_2005 0.421 0.303 1.39 0.165
d_2006 0.114 0.420 0.27 0.786
Numer of Obs 3225
LR chi2(31) 857.03
Prob > chi2 0.000
***, **, * denote significant at the 0.01, 0.05, and 0.1 levels, respectively.
47
Table 1.9
Multivariate Test Results
Variable Coefficient Std Error z-value p -value
Constant*** 4.227 0.037 114.75 0.000
AfterERP*** -0.054 0.011 -4.87 0.000
HighLit*** -0.039 0.014 -2.88 0.004
HighGrowth 0.001 0.010 0.06 0.953
Financial * -0.056 0.033 -1.72 0.085
HighTech -0.015 0.013 -1.14 0.253
NumSegment*** 0.016 0.002 7.1 0.000
ExtraItem*** 0.094 0.019 4.86 0.000
Size*** -0.040 0.003 -14.58 0.000
BusyEnd *** 0.031 0.009 3.46 0.001
Loss*** 0.092 0.010 9.48 0.000
Leverage 0.001 0.003 0.23 0.821
Auditor** -0.030 0.013 -2.34 0.019
GC*** 0.170 0.025 6.89 0.000
Foreign -0.006 0.010 -0.58 0.564
d_2000*** -0.288 0.039 -7.34 0.000
d_2001*** -0.285 0.038 -7.59 0.000
d_2002*** -0.194 0.037 -5.31 0.000
d_2003*** -0.104 0.035 -2.93 0.003
d_2004*** 0.177 0.035 5.02 0.000
d_2005*** 0.230 0.035 6.52 0.000
d_2006*** 0.246 0.036 6.94 0.000
d_2007*** 0.215 0.036 5.99 0.000
d_2008*** 0.177 0.036 4.89 0.000
d_2009*** 0.175 0.037 4.73 0.000
d_2010** 0.158 0.038 4.2 0.000
d_2011** 0.155 0.038 4.07 0.000
d_2012* 0.078 0.044 1.77 0.076
Number of Obs 8610
LR chi(27) 2011.8
Prob > chi2 0.000
***, **, * denote significant at the 0.01, 0.05, and 0.1 levels, respectively.
48
Table 1.10
Multivariate Test Results
Variable Coefficient Std Error z-value p -value
Constant*** 4.227 0.037 114.75 0.000
AfterERP1*** -0.066 0.019 -3.49 0.000
AfterERP2** -0.044 0.020 -2.22 0.026
AfterERP3*** -0.056 0.020 -2.72 0.007
AfterERP4** -0.046 0.023 -2.01 0.045
AfterERP5* -0.052 0.029 -1.81 0.071
HighLit*** -0.039 0.014 -2.88 0.004
HighGrowth 0.001 0.010 0.05 0.957
Financial * -0.056 0.033 -1.72 0.085
HighTech -0.015 0.013 -1.14 0.253
NumSegment*** 0.016 0.002 7.09 0.000
ExtraItem*** 0.094 0.019 4.87 0.000
Size*** -0.040 0.003 -14.58 0.000
BusyEnd *** 0.032 0.009 3.47 0.001
Loss*** 0.093 0.010 9.48 0.000
Leverage 0.001 0.003 0.22 0.825
Auditor** -0.030 0.013 -2.33 0.020
GC*** 0.170 0.025 6.89 0.000
Foreign -0.006 0.010 -0.58 0.565
d_2000*** -0.288 0.039 -7.34 0.000
d_2001*** -0.285 0.038 -7.58 0.000
d_2002*** -0.194 0.037 -5.3 0.000
d_2003*** -0.103 0.035 -2.93 0.003
d_2004*** 0.177 0.035 5.01 0.000
d_2005*** 0.231 0.035 6.53 0.000
d_2006*** 0.246 0.036 6.92 0.000
d_2007*** 0.215 0.036 6.01 0.000
d_2008*** 0.177 0.036 4.88 0.000
d_2009*** 0.178 0.037 4.79 0.000
d_2010*** 0.154 0.038 4.05 0.000
d_2011*** 0.155 0.039 4.01 0.000
d_2012* 0.075 0.045 1.66 0.097
Number of Obs 8610
LR chi(31) 2012.65
Prob > chi2 0.000
***, **, * denote significant at the 0.01, 0.05, and 0.1 levels, respectively.
49
Chapter 2: The Fair Value Option (FVO) under SFAS 159: Effectiveness of FVO and
Intention of FVO Adopters
I. INTRODUCTION
In recent years, many countries have adopted the International Financial Reporting
Standards (IFRS) whose critical measurement principle is fair value accounting. In addition,
the FASB has issued several accounting standards which help facilitate the movement
toward fair value accounting from historical cost accounting. However, considerable
debate still exists on the advantages and disadvantages of fair values in financial statements.
While proponents claim that fair values help financial statements reflect more relevant
information in a timely manner, opponents argue that fair value accounting violates the
traditional accounting principle of conservatism, which may reduce the reliability of
reported financial information and thus decrease the usefulness of accounting information.
In 2007, the Financial Accounting Standards Board (FASB) issued Statement of
Financial Accounting Standards No. 159 (SFAS 159) – The Fair Value Option (FVO) for
Financial Assets and Financial Liabilities – which provides firms with an option to measure
and report selected financial instruments at fair value. The objectives of this statement are
for mitigating unnecessary earnings’ volatility caused by mixed-measurement accounting,
and for expanding the use of fair value measurement. However, SFAS 159 has drawn
widespread scrutiny and significant criticism from different parties including media,
regulators, and academic researchers. Critics argue that even though FVO may help reduce
the problem of mixed-measurement accounting, it may not help reduce earnings volatility
due to the inherent feature and estimation errors of fair values. In fact, there are several
studies which report evidence that fair value based earnings are more volatile than earnings
50
under the current GAAP (Barth et al. 1995; Hodder et al. 2006). Also, providing managers
with full discretion on selecting financial instruments to which FVO is applied might
induce other problems, such as reduced reliability and comparability. Furthermore, the
option of implementing FVO on an instrument by instrument basis suggested in SFAS 159
can even increase possibility that managers do use FVO not in accordance with the intent
of SFAS 159, and there exist prior studies which report the possibility of managers’
opportunistic election of FVO (Henry 2009; Hsu and Lin 2016).
Thus, the primary purpose of this study is to examine whether the FVO of SFAS
159 has achieved its intended objective which is reducing unnecessary earnings’ volatility.
In addition, this study also tries to explore the motivation of firms’ decision to adopt FVO.
The results indicate that the earnings’ volatility did not improve after adopting FVO. It
seems that there is even a positive association between FVO and earnings volatility and the
relation is being amplified by the size of gains or losses resulting from the FVO. The results
also indicate that the analysts’ forecast error of FVO adopter is greater than that of non-
FVO adopters, which is consistent with negative relation between earnings volatility and
analysts’ forecast accuracy. Given the results of even higher earnings volatility after
adopting FVO, I have tried to explore the intention of firms’ decision to adopt FVO using
two more hypotheses: opportunistic intention and informative intention. However, I could
not find strong results for either hypothesis but there seems to be some limited evidence
for opportunistic intention.
Prior studies in general examines whether FVO adopters are more likely to meet or
beat analysts’ forecast consensus using unrealized gains resulting from elected financial
instruments in order to find evidence of opportunistic election of FVO. This study
51
contributes to FVO-related literature by examining the change in earnings volatility around
adoption of FVO using a sample of all US firms that adopted FVO over a long time period
(eight years). Fiechter (2011) examines whether FVO adoption affects earnings volatility
He examines the effect of FVO of IAS 39, not SFAS 159, on earnings volatility of
international banks over two year time period (using quarterly data). That study suggests
a different type of opportunistic motivations of firms to adopt FVO related to financial
activity. By using a different method of identifying FVO adopters and by extending the
sample period, this study covers a significantly higher number of FVO adopters across
industries, and thus provides more comprehensive evidence of the consequences of FVO
adoption
This paper proceeds as follows. Section II introduces prior research related with
fair value accounting, and Section III discusses SFAS 159 and related research. Section IV
develops hypotheses, and section V describes data collection and research methodology.
Section VI presents test results and section VII offers a summary and conclusion.
II. LITERATURE REVIEW ON FAIR VALUE ACCOUNTING4
Is fair value relevant?
The primary interest of policy makers and academic researchers is in whether fair
value accounting improves information of financial statements that interested users use
compared to the information of historical cost accounting. Thus, a significant number of
prior studies examine the value relevance of fair value accounting. Financial statements
4 The words of fair value, mark-to-market, market value-based, and market value accounting have
been often used as synonyms (Barth 1994).
52
can be considered value-relevant if the information they reflect helps financial statement
users to assess the firm value (Barth and Landsman 1995). Thus, a common way to assess
the value relevance of fair value information recognized or disclosed could be assessing its
incremental association with share prices (Landsman 2007). In general, prior literature on
fair value accounting indicates that fair values provide more relevant information to
financial statement users than the amounts under historical cost accounting.
Barth (1994) examines disclosed fair value estimates of banks’ investment
securities and securities gains and losses. She reports that investment securities’ fair values
provide incremental explanatory power for equity values of banks beyond that of historical
costs, implying that the fair value information disclosed is value relevant. However, the
results for securities gains and losses are mixed. These mixed results on securities gains
and losses seem to be attributable to too much measurement error in the fair value estimates.
Eccher et al. (1996), Nelson (1996), and Barth et al. (1996) examine the value
relevance of fair value estimates required to be disclosed by SFAS No. 107. Using a sample
of U. S. banks, Eccher et al. (1996) investigate the association between share prices and
disclosures of those fair value estimates. They first regress market-to-book (MB) ratios on
the differences between the fair value estimates and their corresponding book values in
order to examine whether the fair value disclosures are value-relevant after controlling
historical cost book values. Similar to Barth (1994), they find that fair values of investment
securities have association with MB ratios. However, they find a weak association for fair
values of net loans, mixed results for fair values of off-balance-sheet instruments, and no
association for fair values of deposit. They also examine whether fair value disclosures
have incremental value relevance over the amounts from traditional historical cost
53
accounting. In order to do so, the study first develop a benchmark model using the
information only form historical cost accounting, and then augment the benchmark model
with fair value disclosures to assess their incremental value relevance. They find that the
R-square increased significantly when the fair value data is included in the model,
indicating that fair value disclosures provide incremental explanatory power in explaining
the variation of MB ratios.
Like Eccher et al. (1996), Nelson (1996) also examines the value relevance of fair
value estimates under SFAS 107 by investigating the association between share prices of
commercial banks and those fair value estimates. She finds an incremental explanatory
power for fair values of investment securities, but no evidence of incremental explanatory
power for fair value estimates of loans, deposits, long-term debt and off-balance sheet
financial instruments. However, it is also found that when additional value relevant
variables, ROE and growth in book value, which can be captured by the accrual accounting
system and are omitted from prior studies on fair value disclosures are incorporated into
the research model, the fair value estimates of investment securities lose their incremental
power to explain share prices.
On the other hand, unlike Eccher et al. (1996) and Nelson (1996), Barth et al. (1996)
find results indicating that differences between disclosed fair values and book values of
securities and fair value estimates of loans and long-term debt provide significant
incremental explanatory power for bank share prices beyond that of book values, but those
of deposits and off-balance sheet items do not. The primary difference between Barth et al.
(1996) and those two studies is that Barth et al. (1996) include additional explanatory
variables for incorporating institutional features of the bank industry: nonperforming loans,
54
interest-sensitive assets and liabilities, a proxy for the intangible asset attributable to core
deposits, net pension assets and nonfinancial assets and liabilities. Focusing on loans’ fair
value estimates, they also find that the coefficient of loans’ fair value is higher for banks
with higher regulatory capital ratios, implying that the market discounts unrealized gain on
loans disclosed by banks which are financially less healthy.
Venkatachalam (1996) investigates the value relevance of disclosed fair values of
banks’ off-balance sheet derivative financial instruments under SFAS No. 119. The
findings indicate that the fair values of off-balance sheet derivatives help explain cross-
sectional variations in bank share prices beyond the contractual (notional) amounts of such
derivatives. He also finds that fair value gains and losses for on-balance sheet financial
instruments are negatively associated with the fair value gains and losses of off-balance
sheet derivatives, which would be evidence that banks use derivatives to reduce their risk
exposures.
Hodder et al. (2006) investigate the properties of net income, comprehensive
income, and a self-constructed measure of full-fair-value (FFV) income for a sample of
U.S. commercial banks. They find that the volatility of FFV income is more than three
times that of comprehensive income and more than five times that of net income even
though average levels of the three income measures are similar in magnitude over the
analyzing period. They also find that the incremental volatility of FFV income is positively
associated with various market-based risks, such as standard deviation in returns, market
model beta, and long-term interest-rate beta, beyond the effects of volatility in net income
and comprehensive income. These findings imply that volatility of FFV income reflects
risk factors which are not captured by volatility of GAAP net income or comprehensive
55
income, and thus fair value accounting might help investors assess the denominator,
discount rate, in a valuation context.
Opponents of fair value accounting express a concern that fair value accounting
will deteriorate reliability of accounting amounts even though it may enhance their
relevance. However, it is in fact very difficult to test separately relevance and reliability of
accounting amounts and value relevance tests are likely to be joint tests of relevance and
reliability. In general, extant literature examines value relevance of accounting amounts by
investigating their associations with equity values to measure how well accounting
amounts reflect information used in financial markets. However, accounting amounts will
be used by market participants only when the amounts are not only relevant but also reliable
enough. An association between equity values and fair value amounts might indicate that
those amounts are relevant and also reliable enough to be used in the market. However,
likewise it is also difficult to declare that a failure to find such an association is attributable
to lack of only relevance or only reliability (Barth et al. 2001). Thus, examining relevance
of information should not be viewed as separate research from study on reliability.
Disclosure vs recognition:
According to Statement of Financial Accounting Concepts No. 5 (1984),
information is relevant when it is capable of making a difference in user decisions, and
reliable when it is representationally faithful, verifiable, and neutral. However, information
provided by financial statements does not have to be new to financial statements users to
be relevant. Rather, an important role of accounting is to summarize or aggregate
information that might be available from other sources (Barth et al. 2001). This would
56
imply that how information is presented or provided might be important. Financial
statements can provide information through recognition or disclosure. There are several
studies which examine and compare fair value estimates recognized and those disclosed.
In their experimental study, Hirst et al. (2004) construct two alternative income
measures for commercial banks, piecemeal (partial)-fair-value (PFV) income measure and
full-fair-value (FFV) income measure, and examine whether risk and value judgements of
bank-specialist analysts are affected by banks’ income measurement method under exposed
and hedged risk conditions of those banks. According to their research setting, while fair
value gains and losses of only investment securities are recognized and those of all other
financial assets or liabilities are disclosed in footnotes under PFV income measurement,
fair value gains and losses of all financial assets or liabilities are recognized under FFV
income measurement. Thy find that analysts’ risk assessments are higher and value
assessments are lower under FFV measurement than under PFV income measurement only
for banks exposed to interest-rate risk, but not for hedged banks. In addition, they also find
that analysts’ risk assessments are higher and value assessments are lower for exposed
banks than for hedged banks only under FFV measurement, but not under PFV income
measurement. These findings indicate that analysts can better differentiate between risky
firms and less risky firms when more fair value changes are recognized than when they are
just disclosed, and under risker conditions, recognizing fair value changes, rather than
disclosing them, may help analysts to get informed and to react to the risk.
Ahmed et al. (2006) examine if share prices are associated with whether the fair
value of derivative financial instruments is recognized or disclosed. Using a sample of
banks that simultaneously hold both recognized and disclosed derivatives prior to SFAS
57
No. 133, they find that while net fair values of recognized derivatives are positively
associated with share prices, coefficients on disclosed derivatives are not significant. Also,
using a sample of banks that have only disclosed derivatives prior to SFAS No. 133, which
are recognized after SFAS No. 133, the study finds no evidence that the fair value of
disclosed derivatives is associated with share prices in the pre-SFAS 133 period, but finds
significantly positive coefficients on derivatives recognized after SFAS 133.
Barth (2004) states that although disclosure is not a substitute for recognition, it can
be complement to recognition. Prior studies seems to indicate that recognized fair value
estimates are better perceived by market participants. However, recognizing those fair
value estimates only may not be enough to provide investors with useful information to
help them make a better and complete decision. There is an inherent risk involved in fair
value estimates which is injected by the inevitable estimation process. Thus, it would be
useful to investors to provide additional information through disclosures about the fair
value estimates recognized such as values at high risk, inherent past volatility of estimates,
and history of realized amounts. Also, how those estimates are presented would also matter.
With the advent of complex financial instruments, recognizing or disclosing those items
makes financial statements to be complex, and thus it gets more difficult to comprehend
the information presented in financial statements. Therefore, disclosing the additional
information in a way or format where information processing costs are lower would be
important for the information to be truly useful to information users.
Earnings management
In such cases where active markets for assets or liabilities do not exist, managers
58
have to rely on their own assumptions for the fair values of the assets or liabilities. This
reality inherently introduces the general problem of information asymmetry between
managers and investors. The information asymmetry can create the moral hazard problem
which may results in manipulation of information provided to the market (Landsman 2007).
There exists prior literature which examines issues of earnings management associated
with fair value accounting. Aboody et al. (2006) examine four key option pricing model
inputs and find that firms understate estimates of option values and thus stock-based
compensation expense (SFAS 123 expense). Specifically, firms tend to adjust expected
option life, expected stock price volatility, and expected dividend yield which are firm-
specific, but not the risk-free interest rate which is less subject to managers’ discretion. The
findings of this study imply that when managers have incentives to manage earnings and
they need to estimate fair values using their own assumptions, they are likely to adjust
those assumptions and thus manage earnings. Dechow et al. (2010) investigate the
securitization gains or losses in the income statement under SFAS 140. They find that
reported securitization gains are larger when pre-securitization earnings are low or below
last year’s level, indicating that managers may take advantage of considerable discretion
over assumptions used in estimating fair values to manage earnings. Also, they find
evidence that those gains are treated as a regular earnings component and thus managers
are rewarded for the gains. In addition, the board appears to be ineffective in monitoring
this earnings management by managers.
In contrast, there also exist prior studies implying that fair value accounting does
not exacerbate or might help mitigate the problem of earnings management. In a study
using Danish banks, Bernard et al. (1995) find no evidence that mark-to-market accounting
59
numbers are managed to avoid regulatory intervention. Their results also indicate that the
reported numbers under Danish mark-to-market accounting system are more reliable
indicators of firm values than those in U.S. system. They point out that the rigid regulatory
intervention policies of Denmark may have contributed to those results. Furthermore, it is
generally argued that the historical cost regime can be inefficient especially in good times
because firms cannot recognize the increase in fair value of assets under historical cost
accounting system.5 Thus, firms have incentives to time asset sales to manage earnings
under the historical cost regime. Proponents of fair value accounting argue that fair value
accounting helps mitigate this problem. The results of Black et al. (1998) are consistent to
this argument. They find evidence that when fair value measurement is used for
recognizing gains or losses in reported income, firms reduce the practice of timing asset
sales for earnings management purposes.
Broader perspectives on fair value accounting
On the other hand, there exist several theoretical studies which imply that we may
need to consider broader economic consequences of fair value accounting when we move
toward the complete fair value accounting regime from historical cost accounting regime.
Plantin et al. (2005, 2008a, 2008b) state that we live in an imperfect world where markets
are not always fully liquid and there are multiple sources of market imperfection. In such
an imperfect world, true market prices which would be available in a perfect world are not
available. Rather, market prices in an imperfect market play a dual role. They are not only
a reflection of the underlying economic fundamentals and actions taken by market
5 Plantin et al. (2008b) state that while the mark-to-market regime leads to inefficient sales in bad
times, the historical cost regime is particularly inefficient in good times.
60
participants, but also an amplifier of those actions. That is, when actions affect prices, the
change in prices also affects actions in turn. If price changes are reflected continuously and
immediately in financial statements under the fair value accounting system, it would be
possible that market participants manage their financial statements more actively and
immediately to maintain their financial objectives, such as leverage ratio.6 This may induce
additional endogenous source of volatility injected by accounting principle. The market
participants’ reaction to the financial statements which are based on fair value can be more
problematic in that it can amplify the financial cycle. As an attempt to maintain their
financial objectives, market participants may need to purchase assets when prices increase
and sell when prices decrease, which results in greater demand during booms and greater
supply during recession (Plantin et al. 2008a). Plantin et al. (2008b) state that accounting
can be seen as a mere detail of measurement leaving the economic fundamentals unaffected
only in the context of completely frictionless competitive markets, implying that
accounting can be relevant only in the imperfect world. The paper argues that we are living
in an imperfect world and shows that accounting amounts matter. It develops a model
which can compare the real effects of a historical cost accounting and fair value accounting,
and theoretically shows that real decisions can be distorted due to different accounting
measurement systems. The main results of the paper indicate that fair value accounting is
more inefficient (lower informational content of prices and suboptimal real decisions) than
historical cost accounting when claims are long-lived, not sufficiently liquid, or senior.7
6 Especially, financial institutions, which are entities that influence financial markets significantly,
tend to adjust their balance sheets more actively in accordance with the price changes than
households (Plantin et al. 2008a) 7 This explains why banking and insurance industries voice strong objections while equity investors
are enthusiastic proponents for fair value accounting. For those financial institutions, a large
proportion of balance sheets consist of items that are of long duration, illiquid, and senior (Plantin
61
Allen and Carletti (2008) develop a model in the context of banking and insurance sector
and illustrate an important disadvantage of mark-to-market accounting. In times of crisis,
prices in illiquid markets may not properly reflect future payoffs, but rather reflect the
amount of liquidity available. It shows that in such circumstances using mark-to-market
accounting can lead to distortions of financial institutions’ portfolio and contract choices,
contagion, and ultimately unnecessary liquidations which would not arise if historical cost
accounting were used. This result implies that fair value accounting may not be beneficial
when markets are not liquid and historical cost accounting may be preferable in such
circumstances. Gorton et al. (2006) examine the issue of mark-to-market in the context of
a security market. Using a theoretical principal-agent setting, they show that mark-to-
market compensation contracts can introduce an externality among traders. The study
indicates that such contracts have feedback effects on asset prices. Similar to Plantin et al.
(2008a, 2008b), this suggests that while prices are dependent on traders’ actions, the prices
also affect the behavior of traders whose compensation is based on asset prices. If this is
the case, prices may not be exogenous indicators of asset value. Consequently, the results
indicate that traders may rationally herd, trading on irrelevant information, which results
in less informative asset prices than they would be without the mark-to-market
compensation scheme.
The issues aforementioned on fair value accounting seem to ultimately stem from
a question about how much we can trust the reliability of fair values estimated. The
proponents of fair value accounting argue that fair value provides more relevant
information and it is also better in terms of corporate governance than historical cost.
et al. 2008a; 2008b).
62
However, this argument may lose its persuasiveness to the extant which the fair values
estimated are noisy. Plantin et al. (2008b) state that the choice between historical cost
accounting and fair value accounting is a choice between obsolete information and
distorted version of current information. While historical cost accounting system ignores
price signals and induces excessive conservatism, fair values available are based on
incomplete information. If we admit that we are living in an imperfect world, we also
cannot help but admit that market prices reflect not only economic fundamentals but also
other sources of market imperfections, which may lead to distortions or transitory
fluctuations of prices. Thus, reflecting immediately the changes in fair value on financial
statements may induce unnecessary volatility of financial statement amounts and may not
be the best way to provide information users with the most relevant information on
underlying economics in order to help them make a better decision. Therefore, it appears
that enhancing reliability of fair value is certainly one of the most important issues to be
addressed before we move to the complete fair value accounting system, and developing a
better measurement basis to mitigate those problems of fair values would be worthwhile.8
III. SFAS 159
Background of SFAS 159
The advent of innovative financial instruments has caused concerns to many,
including accounting profession, regulating bodies, creditors and investors, and preparers
8 For example, Plantin et al. (2008a) suggest using an average value over some interval of time as
the accounting value of an asset and also using an average of past observed discount factors as the
discount factors for illiquid assets. Glover et al. (2005) proposes intertemporal financial statements
which can show facts separately from forecasts by reporting three different columns: Fact, Forecast,
and Total.
63
of financial statements. As a response, the FASB introduced a project on financial
instruments and off-balance-sheet financing in 1986, and issued three Statements of
Financial Accounting Standards (SFAS) in order to deal with complex issues regarding
financial instruments between 1990 and 1994: No. 105, “Disclosure of Information about
Financial Instruments with Off-Balance-Sheet Risk and Financial Instruments with
Concentrations of Credit Risk,” No. 107, “Disclosures about Fair Value of Financial
Instruments,” and No. 119, “Disclosure about Derivative Financial Instruments and Fair
Value of Financial Instruments.” SFAS 107 is the first standard which requires all firms to
disclose the fair value information of all financial instruments, both recognized and off-
balance-sheet, for which it is practical to estimate fair value. SFAS 119 requires disclosures
about fair value estimates for derivative financial instruments. The statement also requires
firms to distinguish financial instruments held for trading from financial instruments held
for purposes other than trading, and to disclose estimated holding gains and losses for
instruments held for trading. However, those three statements are not without criticisms.
The primary problem seems to be that they focus on “disclosure” issues rather than
“recognition” or “measurement” issues. With those three statements, information users are
provided with much more information that is highly technical, and thus more burden would
rest with users to interpret which information is relevant and useful and to understand the
underlying risk of firms using the information disclosed (Li 2006).
Meanwhile, the FASB appears to notice that recognition and measurement issues
should be resolved in order to address the criticisms on those three statements. As an effort,
the FASB issued SFAS No. 115, “Accounting for Certain Investments in Debt and Equity
Securities,” in 1993. The statement requires that all debt and equity securities be classified
64
into one of the three categories: trading, available for sale, and held to maturity. However,
it retains the uses of both fair value measures and historical cost measures. Securities
classified as held to maturity are carried at amortized cost and securities classified as
trading or available for sale are carried at fair values. This statement is criticized in that it
can provide the opportunity for earnings manipulation because categorizing securities or
transferring securities between categories is subjective and unrealized holding gains and
losses from different categories are treated differently. In 1998, the FASB issued SFAS No.
133, “Accounting for Derivative Instruments and Hedging Activities,” as an effort to
achieve the objective of measuring all financial assets and liabilities at fair value. The
statement establishes accounting and reporting standards for derivatives instruments and
hedging activities. It supersedes SFAS No. 105 and No. 119 and amends SFAS No. 107.
The major issue on SFAS 133 is its complexity. Under SFAS 133, the special hedging
accounting is applied only to qualifying items, not all items held for the purpose of hedging.
However, it is so complicated to follow the requirements for qualification that many
instruments end up not being carried at fair value. Thus, it is not uncommon that financial
assets and liabilities are accounted for differently depending on whether they are qualified
for a hedging relationship or not under complex SFAS 133.
In 2006, the FASB issued SFAS 157, “Fair Value Measurements.” The statement is
aimed at increasing the consistency and comparability of fair value measurements and
expanding disclosures about fair value measurements by improving both disclosure and
measurement issues of fair value. More specifically, the statement tries to establish a single
authoritative definition of fair value (“exit-price,” not “entry price”) 9 , establish a
9 SFAS 157 defines fair value as the price that would be received to sell the asset or paid to transfer
the liability (an exit price), not the price that would be paid to acquire the asset or received to
65
framework for measuring fair value, and expand financial statement disclosure
requirements for fair value measurements. SFAS 157 also simplifies and codifies the
related guidance that exists for developing fair value measurements so that it can eliminate
differences between the guidance that have added to the complexity in GAAP. Although
SFAS 157 does not require any new fair value measurements, it clarifies the meaning of
fair value used under the U.S. generally accepted accounting principles (GAAP). Also, the
statement led to several changes in practice related to certain valuation techniques. For
example, with the adoption of SFAS 157, companies are required to maximize the use of
relevant observable inputs and minimize the use of unobservable inputs when they use
valuation techniques for determining fair values. 10 The FASB also emphasizes in the
statement that fair value is a market-based measurement, not an entity-specific
measurement. 11 Prior to the adoption of SFAS 157, entities may have used internal
assumptions that management believed to be better than available observable inputs
(Deloitte 2008). As to the expanded disclosure, SFAS 157 requires firms to make tabular
disclosure for fair value hierarchy by major category of assets and liabilities on a recurring
basis. The fair value hierarchy prioritizes the inputs used to measure fair value into three
broad levels: level 1, level 2 and level 3. The level in the fair value hierarchy within which
assume the liability (an entry price) (FASB 2006). 10 Inputs refer to the assumptions that market participants would use in pricing the asset or liability.
Observable inputs are inputs that reflect the assumptions based on market data obtained from
sources independent of the reporting entity. Unobservable inputs are inputs that reflect the reporting
entity’s own assumptions based on the best information available to the entity in the circumstances
(FASB 2006). 11 The estimated cash flows used for determining fair value are the cash flows that market
participants could obtain from an asset or liability, not estimates of entity-specific cash flows.
Likewise, the discount rate used for fair value is the rate of interest that market participants would
expect to receive for bearing the risks inherent to those cash flows, not the reflection of entity-
specific risk preferences (Barth 2004).
66
each fair value measurement falls is determined on the basis of the lowest level input that
is significant to the fair value measurement.12 Critics of SFAS 157 argue that the exit value
does not reflect properly the value of an asset (or liability) to a company because it fails to
consider how the asset is used within the company and the value is not assessed in
conjunction with other assets. In addition, because in many cases, active markets for
identical assets or liabilities do not exist, many fair value estimates are based on level 2 or
level 3 inputs. Level 2 or 3 inputs are less reliable and difficult to verify (Landsman 2007).
SFAS 159
In January 2006, the Financial Accounting Standard Board (FASB) issued an
exposure draft, The Fair Value Option for Financial Assets and Financial Liabilities. After
receiving responses to the draft from researchers and practitioners, it finally issued FASB
Statement No. 159 (SFAS 159), The Fair Value Option for Financial Assets and Financial
Liabilities – Including an amendment of FASB Statement No. 115, in February 2007. The
statement became effective for fiscal years beginning after November 15, 2007, with early
adoption permitted (FASB 2007).
SFAS 159 affords reporting entities the option to measure certain financial assets
and financial liabilities at fair value, which was not previously available under U.S. GAAP.
Under SFAS 159, companies have the irrevocable option to measure most financial assets
and financial liabilities at fair value (FVO) without any justification and changes in fair
12 Level 1 inputs are quoted prices (unadjusted) in active for identical assets and liabilities. In
general, level 2 inputs are quoted prices in active markets for similar or related assets and liabilities,
or quoted prices in inactive markets for identical assets and liabilities. Level 3 inputs are
unobservable inputs which are based on the reporting entity’s own assumptions. (FASB 2006).
Level 1 inputs are the most preferred.
67
value of those items chosen for FVO are recognized in current-period earnings. In addition,
the difference between beginning book value and fair value of financial instruments
selected for FVO should be reported as a cumulative-effect adjustment (effect of re-
measurement to fair value) to the beginning balance of retained earnings in the first FVO
adoption year. Also, it states that the FVO is applied on an instrument by instrument basis.
Furthermore, because the Financial Accounting Standard Board (FASB) supported a broad
application of the FVO, it decided to keep scope exceptions to a minimum and thus the
FVO can be applied to most financial assets and liabilities. The financial assets and
liabilities excluded from the scope of SFAS 159 include investments in subsidiaries that
would be consolidated, assets and liabilities associated with pension and other
postretirement benefits, financial assets and liabilities recognized under lease contracts,
deposit liabilities of financial institutions, and financial instruments classified as
shareholder’s equity. Appendix A presents examples of FVO adoptions.
The primary objective of SFAS 159 is to mitigate volatility in reported earnings by
reducing artificial earnings volatility caused by measuring related assets and liabilities
differently. Before SFAS 159, firms could offset the change in fair value of financial
instruments against the change in fair value of other derivatives held for the purpose of
hedging for those instruments by following the hedging accounting of SFAS 133. However,
the hedging accounting was too complicated to apply and thus the costs to comply with the
requirements of SFAS 133 were high. Therefore, to the extent that costly hedging
accounting prevents firms from applying hedging accounting, firms would record related
financial assets and liabilities differently and this discrepancy could induce unnecessary
volatility in earnings (Guthrie et al. 2011). However, under SFAS 159 firms don’t have to
68
follow this costly and complex rules for hedging and they now can measure and record
almost all financial instruments at fair value. Thus, SFAS 159 is expected to help entities
to avoid the time, effort, and systems needed to document fair value hedging relationships
and demonstrate their effectiveness to qualify for continued hedge accounting.
Another purpose of SFAS 159 is to expand the use of fair value measurement which is
consistent with the Board’s long-run measurement objective for accounting for financial
instruments. Even though SFAS 159 permits FVO for only financial assets and liabilities,
the Board has a plan to expand FVO even to certain nonfinancial assets and liabilities that
are similar to financial items. Lastly, the statement is also an attempt by the FASB for
international convergence in financial reporting standards. Prior to SFAS 159, the
International Accounting Standards Board (IASB) issued the amended International
Accounting Standard 39 (IAS 39), Financial Instruments: Recognition and Measurement,
which describes conditions under which firms can elect fair value measurement for
financial instruments, which is called fair value option (IASB 2005).13 The purpose of this
amended IAS 39 is to simplify accounting for financial instruments as well as to provide
an opportunity to reduce accounting mismatches (Fiechter 2011).
Issues of SFAS 159
Complexity: The FASB argues that FVO will help firms to reduce artificial
earnings volatility caused by accounting mismatches because they do not have to apply
complex hedge accounting provisions presented in SFAS No. 133. SFAS No. 133 has been
13 While the IASB incorporates some eligibility requirements for application of FVO, SFAS 159
dose not impose any eligibility requirements primarily for the reason that it would reduce the use
of fair value measurement and increase complexity (FASB 2007, paras. A20 and A21).
69
criticized for its complexity and it has not been effective to mitigate earnings volatility
caused by measuring related assets and liabilities differently. However, fair value
measurement standard is also complex. It is complex and difficult to measure fair values
especially in the absence of quoted market prices in an active market. When quoted market
prices are not available, fair value measurement is based on subjective assumptions, and
thus it can provide an opportunity to manipulate earnings (Dechow et al 2010). In fact,
some respondents to the exposure draft for SFAS 159 believe that there is insufficient
guidance for determining fair values and many of the assumptions used in determining fair
values are subjective and complex.
Comparability: The comparability within and between entities may be impaired
because firms are allowed to choose different measurement attributes on a contract-by-
contract basis (FASC 2007; Li 2006). It is possible that entities with similar financial assets
and financial liabilities apply different measurement attributes and different measurement
attributes are applies to similar financial assets and financial liabilities within the same
entity due to the contract-by-contract basis FVO. This mixed-attribute measurement may
deteriorate investors’ ability to comprehend financial information.
Volatility: The primary rationale for SFAS 159 is to reduce the earnings volatility
caused by different accounting rules for financial assets and liabilities. However, critics
claim that the FASB’s argument is counterintuitive in that traditionally, managers have
resisted market value measurement to avoid volatile earnings and auditors have been
hesitant to use fair values to minimize the risk of attesting to amounts that shortly go up or
down (Li 2006).
Instrument by instrument approach: Under SFAS 159, firms are allowed to
70
irrevocably apply FVO by an instrument by instrument basis and managers do not have to
justify their decisions to apply FVO to certain financial assets or liabilities. This
unconstrained contract by contract election of fair value might reduce the reliability and
transparency of financial information because there is high possibility of management
discretion involved. And furthermore, the reduced reliability and transparency may result
in additional information processing costs for information users (FASC 2007). Also, it is
generally believed that fair value measurement is more representative of entities’
underlying economics and provides more relevant information than cost-based measures.
However, the instrument by instrument option will result in the reporting at fair value of
only some financial instruments and this discretionary mixed-attribute measurement may
rather significantly reduce the relevance and comparability of financial reporting. Thus,
critics argue that the statement does not meet the objective of financial reporting because
it may deteriorate usefulness of information and increase costs for users even though it
might help reduce earnings volatility, which is not an objective of financial reporting.
Therefore, they believe that FASB should require fair value accounting for all financial
instruments (FASB 2007).
Interim step toward fair value accounting: As mentioned earlier, one of the
objectives of SFAS 159 is to expand the use of fair value measurement. However, some
critics believe that even though the FASB argues that SFAS 159 is a cost-beneficial interim
step toward measuring all financial instruments at fair value, the statement may rather
result in a further delay in the broader requirement for fair value for financial instruments.
If SFAS 159 produces its intended effect, which is reduced earnings volatility, the preparer
community would have less incentive to support wider adoption of fair value measurement.
71
In other words, if prepares can manage earnings volatility with FVO, they would be less
supportive of a fair value measurement requirement. Thus, this would result in delay or
resistance to requiring fair value measurement (FASB 2007).
Research on FVO
Prior studies which examine the possibility of opportunistic adoption of FVO yield
mixed results. Henry (2009) explores the possibility of managers’ opportunistic usage of
FVO by examining commercial banks which adopted FVO before 2008 (early adopters).
By analyzing the disclosures of banks which rescinded or revised their decision of early
FVO adoption, she finds evidence that their initial adoption of FVO was not in compliance
with the intent of SFAS 159. Also, Hsu and Lin (2016) find that firms with more level 3
assets and liabilities and with weak corporate governance are more likely to recognize
unrealized gains by adopting FVO to meet or beat analysts’ forecast consensus. However,
Guthrie et al. (2011) find no evidence of systematic earnings management through FVO.
They identify FVO adopters among S&P 1500 companies and investigate whether those
firms adopt FVO with opportunistic motivation. In order to find evidence of opportunistic
election of FVO, they examine firms whose reported earnings meet or beat the analysts’
forecast consensus only with unrealized gains from elected financial instruments (current
earnings management) and firms which recognize prior unrecognized losses as cumulative-
effect adjustment to retained earnings to avoid recognizing losses in earnings in later years
(future earnings management). They find evidence of earnings management only from a
small number of early adopters, but no evidence of systematic opportunistic election of
FVO.
On the other hand, using an international sample of 222 banks from 44 countries,
72
Fiechter (2011) examines whether FVO of IAS 39 affects earnings volatility and whether
different levels of regulatory quality of different countries affect the likelihood of adopting
FVO. He hand collects information about the intentions of those banks to adopt FVO from
their disclosures, and finds that the earnings volatility of banks which adopt FVO to
mitigate the mixed measurement problem is lower than that of other banks in the cross-
section during the time period from 2006 to 2007.
IV. HYPOTHESIS DEVELOPMENT
Volatility
Barth (2004) identifies and discusses three additional sources of volatility of
financial statement amounts associated with fair value accounting, compared to historical
cost accounting. The first source of volatility is the estimation error volatility. In many
cases where deep and liquid markets do not exist, firms need to estimate fair values, and
the estimation error is inherent and inevitable in the process of estimation. This estimation
error can increase volatility of recognized fair values. The second is inherent volatility of
fair values. It is widely accepted that fair value amounts would fluctuate more from period
to period than amounts based on historical cost. However, this inherent volatility results
from economic reality that a firm faces, not from accounting measurement. The better
reflection of economic reality is the primary rationale for fair value accounting, and thus
this inherent volatility is not evitable. The last source of volatility identified is mixed-
measurement volatility. Mixed-measurement volatility is purely a result of accounting
measurement problem. Currently, some assets and liabilities are measured at fair value,
some are measured at historical cost (amortized historical cost), some are measured at the
73
lower of cost or market, and some others are not recognized at all. Thus, under this mixed-
measurement accounting, changes in market conditions do not affect the values of assets
and liabilities in the same way, which induces additional volatility of financial statement
amounts. Mixed-measurement volatility is different from the other two in that this is an
artifact of the accounting measurement system and does not reflect a risk that should be
priced. In addition, she states that estimation error impairs reliability of information,
inherent volatility can help increase relevance, and mixed measurement in general impairs
both reliability and relevance.
One of the FASB’s rationales for SFAS 159 is to mitigate the earnings volatility
caused by a problem of different accounting rules for financial assets and financial
liabilities, which is the mixed-measurement volatility. For example, before SFAS 159,
firms could offset the change in value of financial instruments against the change in value
of other financial instruments held for the purpose of hedging by following the hedging
accounting of SFAS 133. However, the hedging accounting was too complicated to apply
and thus the costs to comply with the requirements of SFAS 133 were high. Therefore, to
the extent that costly hedging accounting prevents firms from applying hedging accounting,
firms would record related financial assets and liabilities differently and this discrepancy
could induce unnecessary volatility in earnings (Guthrie et al 2011). However, under SFAS
159 firms don’t have to follow this costly and complex rules for hedging and they now can
measure and record almost all financial instruments at fair value.
On the other hand, because SFAS 159 provides all firms with an option to measure
financial assets and liabilities at fair value, even firms without a problem of unnecessary
earnings volatility caused by the mixed-measurement volatility can decide to record
74
financial instruments at fair value. This means that managers can decide to record financial
instruments at fair value for purposes other than the objective of reducing unnecessary
earnings volatility. In fact, Henry (2009) finds evidence that there were firms which had
decided to implement SFAS 159 opportunistically. In addition, according to SFAS 159, the
FVO can be applied instrument by instrument and thus there is a room for management
discretion in applying SFAS 159. Furthermore, the objective of SFAS 159 is
counterintuitive in that traditionally managers have resisted market values to avoid volatile
earnings and auditors have been hesitant to apply fair value to minimize the risk of attesting
to amounts that might go up or down temporally (Li 2006), which seems to be associated
with estimation error volatility and inherent volatility of fair values. In fact, Barth et al.
(1995) find evidence using a sample of banks that fair value based earnings are more
volatile than historical cost based earnings, and Hodder et al. (2006) reports that their self-
constructed full fair value income is much more volatile than comprehensive income or net
income under the current GAPP. The fair value option of SFAS 159 is primarily intended
to eliminate mixed-measurement volatility for financial instruments. However, increased
use of fair values is likely to increase the other two sources of earnings’ volatility if FVO
is not used in accordance of the intent of SFAS 159. Thus, it would be an interesting
empirical question to examine whether the earnings’ volatility of FVO adopters have in
fact decreased or not in order to examine the effectiveness of SFAS 159. Therefore, I first
examine the following null hypothesis:
H1: The SFAS 159 does not help improve earnings volatility.
Analysts’ forecast accuracy
75
Accurate earnings forecasts are important in that a number of applications of
accounting data, such as in valuation research and practice, require the prediction of
earnings. Reliably projected future earnings are the key for the valuation models and
investments in capital market. However, it is widespread managerial belief that earnings
volatility reduces earnings predictability (Dichev and Tang 2009). In fact, a number of
studies report a negative relationship between earnings volatility and earnings prediction.
For example, Graham et al (2005) state that earnings volatility is associated with more
disagreement among analysts about earnings forecasts. Considering the reported
relationship between earnings volatility and earnings predictability, if the SFAS 159 has an
impact on earnings volatility, we can also reasonably conjecture that it can affect earnings
predictability. However, because the effectiveness of SFAS 159 is not determinable in
advance, I test the following null hypothesis:
H2: The SFAS 159 does not have an impact on analyst forecast accuracy.
Management discretion (Opportunistic decision)
The results for the first hypothesis indicate that SFAS 159 is not effective in
reducing earnings volatility. Then, the next question we might naturally have would be
about why firms have decided to adopt FVO. Thus, I have developed and examined two
more hypotheses to explore the question. Under SFAS 159, firms are allowed to apply FVO
by an instrument on instrument basis without justification for their decisions. This
instrument by instrument option result in reporting at fair value only part of financial
instruments and this discretionary accounting measurement may provide managers with
opportunities for earnings management. The office of the Chief Accountant of the SEC
76
states on its report for a Congressional committee that the benefit of requiring fair value
accounting for all financial instruments is to reduce opportunities for earnings management
and complexity of financial reporting created by the mixed attribute accounting model
(SEC 2005). However, SFAS 159 does not require all financial instruments to be recorded
at fair value and thus does not improve the problem of the mixed attribute accounting, but
rather increases the managements’ discretion in measuring financial instruments. Also, fair
value itself is inherently less reliable especially when measuring fair values of financial
assets or liabilities for which active markets do not exist. In such cases, managers should
estimate fair values with their own assumptions that can be subject to discretion or
manipulation (Landsman 2007). On the other hand, it should be also noted that FVO is
irrevocable and it is very difficult to accurately forecast the changes in values of financial
instruments in the long term. Therefore, it would be more reasonable to conjecture that the
decision for adopting FVO would be for a temporary one-time event in the near future if
the decision is made with an opportunistic intention. One potential candidate could be the
needs for external financing in the near future. For example, if a company has a plan for
external financing in a short period of time, the company may have more incentives to
manage earnings in order to send a more positive signal to the market and to make its
financial position look better at least temporarily before the external financing. Thus, I
conjecture and test the following hypothesis:
H3: Firms are more likely to adopt FVO in the year preceding the year when they
have needs for external financing.
Provision of more relevant information on financial instruments
77
The decision to adopt FVO might be viewed from a perspective of providing better
(more relevant) information on firms’ financial position. It is generally agreed that fair
value accounting provides more relevant information than historical cost accounting. As
the complexity of financial instruments owned by a company increases, the demand from
the market for more relevant information on them may also increase. Thus, firms with more
complex financial instruments have greater needs to provide information on their current
fair values. I try to examine this conjecture using exotic features of bonds issued. While
some bonds are issued without any additional features, other bonds are issued with some
exotic features, such as convertible, redeemable, and asset-backed. It is difficult to assess
exotic features of bonds and to properly reflect their values on financial statements under
historical cost accounting. Therefore, firms with more bonds with exotic features may have
greater motivation to adopt fair value accounting to provide better information on them. In
this regard, I conjecture and test the following hypothesis:
H4-1: Firms which have bonds with more exotic features are more likely to adopt
FVO.
In addition, if the motivation of firms to adopt FVO is for providing more relevant
information on complex financial instruments, it could mean that FVO may expedite
transactions of complex financial instruments. If firms adopt FVO because they care the
needs of investors for relevant information on financial statements, those firms might be
more willing to acquire complex financial instruments than before FVO is available
because the burden to provide relevant information could be lower with FVO. Thus, I also
conjecture and test the following hypothesis:
78
H4-2: FVO adopters are more likely to issue bonds with exotic features after
adopting FVO than firms that do not adopt FVO.
V. RESEARCH DESIGN
Sample selection
In order to identify the firms that have adopted the FVO, I took an advantage of the
item ‘TFVCE’ in COMPUSTAT. The ‘TFVCE’ stands for ‘total fair value change in
earnings’ and it reflects the gains or losses resulting from the change in fair value of
financial assets or liabilities which are included in reported earnings of income statement
only by SFAS 159. I have also cross-checked the reliability of this item by examining actual
10-Ks of ten firms, and the numbers provided by COMOUSTAT were reliable. Thus, I
assume that the first firm-year whose ‘tfvce’ item has a number other than ‘0’ is the first
year for the firm to adopt FVO. Also, the firms that have not had any numbers other than
‘0’ for ‘tfvce’ item for any years are classified as non-FVO firms. Through this process, a
total number of 331 unique firms were identified to adopt FVO. However, the sample size
was reduced to 216 because data for 8 consecutive years surrounding the first FVO
adoption year are not available, which is necessary for computing standard deviations of
earnings, or information for total assets at the year preceding the first FVO adoption year
(time t-1), which is necessary for identifying matched non-FVO firms, is not available for
115 firms. Standard deviation of net income was used as a proxy for earnings’ volatility
and it was computed using four years’ net incomes before and after FVO. In order to
conduct difference in differences analyses, each FVO firm was matched with a non-FVO
firm on industry and size at the year preceding the first FVO adoption year. Firms were
79
first matched by the four-digit Standard Industrial Classification (SIC) code, and then
matched by total assets.14 Also, in the matching procedure, non-FVO firms were allowed
to be selected more than once. The matching procedure resulted in unique 216 FVO firms
and 195 non-FVO firms, but 216 pairs of firms. In addition, 14 firms were lost due to the
unavailability of information for control variables for the first hypothesis testing which is
about volatility of earnings. Therefore, the final sample size includes 202 pairs of FVO
firms and non-FVO firms. Panel A of Table 2.1 presents the distribution of 202 FVO firms
by the first FVO adoption year and Panel B presents the distribution of the firms by industry.
While it seems that there are some early adopters who adopted FVO in 200715, most firms
adopted FVO for the first time in 2008 which is the first year when FVO became available.
Also, many FVO adopters belong to financial industry, and a significant number of
manufacturing firms have also adopted FVO.
[Insert Table 2.1 about here]
Model Specification
I have tested the association between FVO and earnings volatility by estimating
multivariate OLS regressions using difference in differences methodology. Difference in
differences analysis has advantages over single cross-sectional difference analysis or single
time-series difference analysis in that the single cross-sectional difference estimator is
biased if there are any unobservable differences between the treatment and control groups
14 If selected non-FVO firms do not have 8 consecutive years’ data, those firms were replaced with
other non-FVO firms in order to minimize the number of firms lost due to data unavailability. 15 Henry (2009) identifies 35 early adopters (commercial banks), among which 11 firms rescinded
or revised their FVO elections. She identifies early adopters by searching 10-Qs and financial press.
The 202 FVO firms presented in Table 1 are the only firms used for hypothesis testing. Please refer
to Table 12 to see the number of all FVO firms identified.
80
and the single time-series difference estimator is biased if there exist time trends affecting
the outcome variable that are due to forces other than the treatment. The difference in
differences analysis combines those two estimators to avoid the problems of those methods
(Roberts and Whited 2011). The dependent variable, earnings volatility, is measured as four
year moving standard deviation of net income. The independent variable of my interest is
FVO_After which is an interaction term of two indicator variables, FVO and After. The
variable FVO is an indicator variable which differentiates FVO firms form non-FVO firms.
The variable After is also an indicator variable which differentiates pre-FVO period from
post-FVO period. Thus, the variable FVO_After takes ‘1’ if a firm year is one of the post-
FVO period years for FVO firms. Consequently, by using an interaction term, I examine
the difference of differences, and the coefficient of FVO_After indicates how much
standard deviations of net income for FVO firms change compared to those for non-FVO
firms. In addition, I conjecture that the greater the size of gains or losses resulting from
FVO is, the more significantly the earnings’ volatility is affected. Thus, I also include an
independent variable ‘FVO_Size_sum’ which is a four-year moving sum of ‘tfvce’ in the
second model in order to examine the effect of the size of gains or losses resulting from
FVO on the earnings volatility. Also both models include several control variables which
are likely to have impact on earnings volatility. Table 2.2 summarizes the definitions of
variables used in this study. The resulting models are as follows:
Model 1-1:
𝑁𝐼_𝑠𝑡𝑑𝑖,𝑡 = 𝛽0 + 𝛽1𝐹𝑉𝑂𝑖,𝑡 + 𝛽2𝐴𝑓𝑡𝑒𝑟𝑖,𝑡 + 𝛽3𝐴𝐹𝑡𝑒𝑟_𝐹𝑉𝑂𝑖,𝑡 + 𝛽4𝑇𝐴_𝑎𝑣𝑒𝑖,𝑡
+ 𝛽5𝐷𝑒𝑏𝑡_𝑎𝑣𝑒𝑖,𝑡 + 𝛽6𝐵𝑇𝑀_𝑎𝑣𝑒𝑖,𝑡 + 𝛽7𝑅𝑂𝐴_𝑎𝑣𝑒𝑖,𝑡
+ 𝛽8𝑆𝑝𝑒𝑐𝑖𝑎𝑙 _𝑠𝑢𝑚𝑖,𝑡 + 𝛽9𝐿𝑜𝑠𝑠_𝑠𝑢𝑚𝑖,𝑡 + 𝜀𝑖,𝑡
Model 1-2:
81
𝑁𝐼_𝑠𝑡𝑑𝑖,𝑡 = 𝛽0 + 𝛽1𝐹𝑉𝑂𝑖,𝑡 + 𝛽2𝐴𝑓𝑡𝑒𝑟𝑖,𝑡 + 𝛽3𝐴𝐹𝑡𝑒𝑟_𝐹𝑉𝑂𝑖,𝑡 + 𝛽4𝐹𝑉𝑂_𝑆𝑖𝑧𝑒_𝑠𝑢𝑚𝑖,𝑡
+ 𝛽5𝑇𝐴_𝑎𝑣𝑒𝑖,𝑡 + 𝛽6𝐷𝑒𝑏𝑡_𝑎𝑣𝑒𝑖,𝑡 + 𝛽7𝐵𝑇𝑀_𝑎𝑣𝑒𝑖,𝑡 + 𝛽8𝑅𝑂𝐴_𝑎𝑣𝑒𝑖,𝑡
+ 𝛽9𝑆𝑝𝑒𝑐𝑖𝑎𝑙 _𝑠𝑢𝑚𝑖,𝑡 + 𝛽10𝐿𝑜𝑠𝑠_𝑠𝑢𝑚𝑖,𝑡 + 𝜀𝑖,𝑡
[Insert Table 2.2 about here]
In order to test the second hypothesis, I also estimate a multivariate OLS regression.
The information of analyst forecasts were extracted from IBES. For the forecasted EPS, I
used the latest mean forecast for each forecast period end date. The forecast error,
dependent variable (Per_fe), is a forecast error in percentage which is measured by dividing
mean forecast minus actual EPS by actual EPS multiplied by 100. The independent
variables of my interest in the model for examining the association between the FVO and
analyst forecast accuracy are FVO_After and FVO_size. The variable FVO_Size is the
magnitude of gains or losses from FVO, and it can be also regarded as another interaction
term of FVO_After and size of gains or losses from FVO in that only firm years of the
post-FVO period for FVO firms can have numbers for variable FVO_size. The size of gain
or loss used for the second hypothesis is measured by dividing the value of item ‘tfvce’ by
income before extraordinary items multiplied by 100. I also include several control
variables which have been identified in prior studies to have significant impact on analyst
forecast accuracy. Thus, the resulting model is as follows:
Model 2:
𝑃𝑒𝑟_𝑓𝑒𝑖,𝑡 = 𝛽0 + 𝛽1𝐹𝑉𝑂𝑖,𝑡 + 𝛽2𝐴𝑓𝑡𝑒𝑟𝑖,𝑡 + 𝛽3𝐴𝐹𝑡𝑒𝑟_𝐹𝑉𝑂𝑖,𝑡 + 𝛽4𝐹𝑉𝑂_𝑆𝑖𝑧𝑒𝑖,𝑡
+ 𝛽5𝐿𝑜𝑠𝑠𝑖,𝑡 + 𝛽6𝐷𝑒𝑏𝑡𝑖,𝑡 + 𝛽7𝐿𝑁_𝑇𝐴𝑖,𝑡 + 𝛽8𝑁𝑢𝑚𝐸𝑠𝑡𝑖,𝑡 + 𝜀𝑖,𝑡
82
I have constructed the dependent variable, Financing, for the third hypothesis
following Bradshaw et al. (2006). They develop a measure for external financing to
examine relations between external financing and future earnings performance and analysts’
forecasts. The measure for net external financing includes both equity financing and debt
financing. In addition to this variable, I also used two more indicator variables to test the
hypothesis: Financing_d and High_Financing. Financing_d is an indicator variable to
capture whether a firm has a net external financing activity or not, and High_Financing is
also an indicator variable to capture whether the magnitude of external financing is greater
than median within the same year and industry. Using those three dependent variables, I
estimated OLS and logistic regression models to test the hypothesis. The independent
variables of my interest are After_FVO and Lag_FVO_Size. Because I conjecture that
firms may have incentives to adopt FVO for external financing in a better condition in the
near future, I regressed the external financing on lagged variables. The model used for
hypothesis testing is as follows:
Model 3:
𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑛𝑔𝑖,𝑡 = 𝛽0 + 𝛽1𝐹𝑉𝑂𝑖,𝑡 + 𝛽2𝐴𝑓𝑡𝑒𝑟𝑖,𝑡 + 𝛽3𝐴𝐹𝑡𝑒𝑟_𝐹𝑉𝑂𝑖,𝑡
+ 𝛽4𝐿𝑎𝑔_𝐹𝑉𝑂_𝑆𝑖𝑧𝑒𝑖,𝑡 + 𝛽5𝐿𝑎𝑔_𝐿𝑁_𝑇𝐴𝑖,𝑡 + 𝛽6𝐿𝑎𝑔_𝐷𝑒𝑏𝑡𝑖,𝑡
+ 𝛽7𝐿𝑎𝑔_𝑅𝑂𝐴𝑖,𝑡 + 𝛽8𝐿𝑎𝑔_𝑆𝑝𝑒𝑐𝑖𝑎𝑙𝑖,𝑡 + 𝛽9𝐿𝑎𝑔_𝐿𝑜𝑠𝑠𝑖,𝑡
+ 𝛽10𝐿𝑎𝑔_𝐵𝑇𝑀𝑖,𝑡 + 𝜀𝑖,𝑡
The two hypotheses on relevant information were tested using exotic features of
bonds issued. Three variables related with bonds’ exotic features were constructed: Exotic,
Exodic_d, and High_Exotic. The variable Exotic measures how many exotic features are
83
attached to the bonds issued. Exotic_d is an indicator variable which differentiates bonds
with exotic features from bonds without exotic features. High_Exotic is also an indicator
variable which differentiates bonds with more exotic features from bonds with fewer exotic
features. The information for bonds was extracted from Mergent Fixed Income Securities
Database. The model for H4-1 is Model 4-1, where FVO is dependent variable and
variables for exotic features are independent variables. The model for H4-2 is Model 4-2,
where variables for exotic features are dependent variables and FVO and After_FVO are
included as independent variables. While Model 4-1 use only observations of pre-FVO
period, Model 4-2 use observations of both pre-FVO and post-FVO periods. The control
variables included in Model 4-2 are lagged variables because firm features in the previous
year are more likely to affect bond issuance of the current year than firm features of the
current year. The resulting models for last two hypotheses are as follows:
Model 4-1:
𝐹𝑉𝑂𝑖,𝑡 = 𝛽0 + 𝛽1𝐸𝑥𝑜𝑡𝑖𝑐𝑖,𝑡 + 𝛽4𝐿𝑁_𝑇𝐴𝑖,𝑡 + 𝛽5𝐷𝑒𝑏𝑡𝑖,𝑡 + 𝛽6𝑅𝑂𝐴𝑖,𝑡
+ 𝛽7𝑆𝑝𝑒𝑐𝑖𝑎𝑙𝑖,𝑡 + 𝛽8𝐿𝑜𝑠𝑠𝑖,𝑡 + 𝛽9𝐼𝐶𝑊𝑖,𝑡 + 𝛽10𝐵𝑇𝑀𝑖,𝑡 + 𝜀𝑖,𝑡
Model 4-2
𝐸𝑥𝑜𝑡𝑖𝑐𝑖,𝑡 = 𝛽0 + 𝛽1𝐹𝑉𝑂𝑖,𝑡 + 𝛽2𝐴𝑓𝑡𝑒𝑟𝑖,𝑡 + 𝛽3𝐴𝐹𝑡𝑒𝑟_𝐹𝑉𝑂𝑖,𝑡 + 𝛽4𝐿𝑎𝑔_𝐿𝑁_𝑇𝐴𝑖,𝑡
+ 𝛽5𝐿𝑎𝑔_𝐷𝑒𝑏𝑡𝑖,𝑡 + 𝛽6𝐿𝑎𝑔_𝑅𝑂𝐴𝑖,𝑡 + 𝛽7𝐿𝑎𝑔_𝑆𝑝𝑒𝑐𝑖𝑎𝑙𝑖,𝑡
+ 𝛽8𝐿𝑎𝑔_𝐿𝑜𝑠𝑠𝑖,𝑡 + 𝛽9𝐿𝑎𝑔_𝐼𝐶𝑊𝑖,𝑡 + 𝛽10𝐿𝑎𝑔_𝐵𝑇𝑀𝑖,𝑡 + 𝜀𝑖,𝑡
Model 4-2 was also estimated using OLS or logistic regression models in accordance with
the features of dependent variables.
84
VI. RESULTS
1. Results for tests on earnings’ volatility
Descriptive Statistics
Table 2.3 reports the descriptive statistics of the sample of 808 observations, where
Panel A, B, and C display the distributional properties of the variables used for tests on
earnings volatility. Panel A shows the descriptive statistics of all observations including
FVO firms and non-FVO firms. Panel B exhibits the descriptive statistics of FVO firms
versus non-FVO firms. Panel C presents descriptive statistics of pre-FVO and post-FVO
periods. Panel B indicates that the standard deviation of net income of FVO firms is greater
than that of matched non-FVO firms and FVO firms reported losses more frequently than
non-FVO firms during the sample period. Panel C indicates that there is no statistical
difference in standard deviation of net income between pre-FVO period and post-FVO
period. However, firms seem to have reported special items and losses more frequently in
the post-FVO period than in the pre-FVO period. Panel D of Table 3 reports the pair-wise
Pearson correlations and Spearman correlations. The correlation table indicates that
earnings volatility is correlated with firm size, debt ratio, book to market ratio, and
reporting special items.
[Insert Table 2.3 about here]
Multivariate Test Results
The multivariate test results for earnings volatility are presented in Table 2.4. The
coefficient of the interaction term, After_FVO, in the first model is positive, but not
statistically significant. In the second model where the variable ‘FVO_Size_sum’ is added
which is a variable for the size of gains or losses from FVO, the coefficient of
85
FVO_Size_sum is strongly significant and positive while the coefficient of After_FVO is
insignificant. In the both models, the coefficients for firm size and reporting loss are
positive and strongly significant. These results imply that even though the adoption of FVO
may affect the earnings volatility, the impact can be amplified by the size of gains or losses
by FVO.16
[Insert Table 2.4 about here]
I also conducted the same analyses after dividing the sample into financial firms
and non-financial firms. The reason for this analysis is that because unlike non-financial
firms, financial firms’ financial statements are primarily comprised of financial instruments,
the motivation of non-financial firms for adopting FVO could be different from that of
financial firms and thus the effect of FVO on earnings’ volatility could be also different
between financial firms and non-financial firms. Table 2.5 and 2.6 report the test results for
financial firms and non-financial firms respectively. However, the results are qualitatively
the same to those of the full sample and I could not find any significant differences between
the two groups.
[Insert Tables 2.5 and 2.5 about here]
2. Results for tests on forecast accuracy
Descriptive Statistics
Table 2.7 reports the descriptive statistics of the sample of 2,387 firm years, where
Panel A shows the distributional properties of the variables used for tests on forecast
16 I also conducted the same analyses using different non-FVO firms matched by the propensity
score matching method, but the results are qualitatively the same to the results reported in Table 4.
86
accuracy and Panel B reports the pair-wise Pearson correlations and Spearman
correlations.17 Like the sample used for the first hypothesis, the sample for testing on
forecast accuracy includes observations which are not four years earlier or not four years
(including adoption year) later than the FVO adoption year.18 Panel A indicates that the
mean of forecast error is about 20 percent of actual EPS, about 25 percent of sample firm
years reported loss, and the mean of number of forecasts for each firm year is around 8.4.
Panel B indicates that forecast error is positively correlated with reporting loss, debt ratio,
and negative change in EPS, and negatively correlated with number of forecasts.
[Insert Table 2.7 about here]
Multivariate Test Results
The test results for analyst forecast accuracy are presented in Table 2.8. The
variables of my interest are After_FVO and FVO_Size. Similar to the results for earnings’
volatility, the coefficient of FVO-Size is positive and strongly significant while the
coefficient of After_FVO is not significant. Reporting loss, firm size, and negative change
in EPS seem to be positively associated with forecast error, and the number of forecasts is
negatively associated with forecast error. The implication from these results is similar to
that of analyses for earnings’ volatility. That is, rather than FVO adoption itself, the
magnitude of gains or losses from FVO can affect the analysts’ forecast accuracy.
[Insert Table 2.8 about here]
3. Results for tests on financing activity
17 Because data sets used for each hypothesis testing are different in terms of sample period and
types of variables, descriptive statistics of the samples are separately presented for each hypothesis
test. 18 216 pairs of FVO firms and non-FVO firms were used for last three hypotheses testing.
87
Descriptive Statistics
Table 2.9 reports the descriptive statistics of the sample of 1,645 firm years used
for the third hypothesis, where Panel A shows the distributional properties of the variables
used for tests on financing activity and Panel B reports the pair-wise Pearson correlations
and Spearman correlations. The sample size decreased significantly because I deleted firm
years which are more than two years earlier or more than one year later than the FVO
adoption year. Because the third hypothesis is for testing the managers’ opportunistic
motivation of adopting FVO for the near future (the first year after adopting FVO), I have
shorten the sample period. The mean of Financing_d indicates that about 44 percent of firm
years had positive net financing activity. The variables for financing activity appear to be
positively correlated with reporting loss, and negatively correlated with firm size, ROA,
and book to market ratio.
[Insert Table 2.9 about here]
Multivariate Test Results
Table 2.10 presents results of OLS regressions using the variable Financing as a
dependent variable, and Table 2.11 presents results of logistic regressions using the
variables Financing_d and High_Financing. I could not find significant results for
After_FVO and Lag_FVO_Size from OLS models. The logistic regression models also
yield no significant results for those variables. Next, I divided the sample into financial
firms and non-financial firms as I did for volatility analyses, and ran the same regressions
for each of the subsamples. However, no significant results for After_FVO and
Lag_FVO_Size were found, and also no significant differences between financial firms
and non-financial firms were found (untabulated).
88
[Insert Tables 2.10 and 2.11 about here]
However, I was curious about why some firms adopt FVO as soon as it becomes
available while other firms delay adopting FVO. Thus, I took a look at distribution of
number of firms adopting FVO by the first FVO adoption year and checked whether there
is any difference in number of firms reporting gains and losses. As displayed in Table 2.
12, there exists a significant difference between financial firms and non-financial firms that
adopted FVO in 2008, which is the first year when the FVO became effective for regular
adopters. In 2008, the number of non-financial firms reporting gains is significantly higher
than number of firms reporting losses while the statistics is opposite for financial firms.
[Insert Table 2.12 about here]
Therefore, I conducted similar analyses using only firms which adopted FVO in
2008. Table 2.13 reports the results of logistic regressions which use financial firm and
non-financial firm samples separately. Non-financial firm sample yields a positive and
significant coefficient for After_FVO while no significant results were found from the
financial firm sample. Furthermore, using both financial firms and non-financial firms
together, another regression analysis was conducted which includes different indicator
variables: NFFirm, an indicator variable which differentiates financial firms from non-
financial firms, and After_NFFirm, an interaction of NFFirm and After. The results are
presented in Table 2.14. The coefficient of After_NFFrim is positive and significant, which
indicates that financing activity of non-financial firms after adopting FVO is higher than
that of financial firms.
[Insert Tables 2.13 and 2.14 about here]
89
4. Results for tests on exotic features of bonds
Descriptive Statistics
Table 2.15 reports the descriptive statistics of the sample of 410 firm years for
testing last two hypotheses, where Panel A shows the distributional properties of the
variables used for tests on exotic features of bonds and Panel B reports the pair-wise
Pearson correlations and Spearman correlations. In many cases, there are several bond
issues for the same firm and same year. In those cases, only one bond issue was kept and
others were deleted to avoid over-representation of one firm with many bond issues for one
year. In cases where one firm issues both bonds with exotic features and bonds without
exotic features, only one bond issue from bonds with exotic features and another one bond
issue from bonds without exotic features were kept. As described in Table 2.2, the variable
Exotic is a numerical variable which captures the number of exotic features of bonds issued
and the variable Exotic_d is an indicator variable which takes 1 if a bond has at least one
exotic feature and zero otherwise. Panel A indicates that about 75 percent of bond issues in
the sample have at least one exotic feature. Also, Panel B indicates that variables for exotic
features of bonds are positively correlated with reporting loss in the previous year, and
negatively correlated with previous year’s firm size and debt ratio.
[Insert Table 2.15 about here]
Multivariate Test Results
For the hypothesis 4-1, the variables, Exotic, Exotic_d, and High_Exotic, were
included in turn as an independent variable in Model 4-1. However, none of three logistic
regressions yielded significant results (untabulated). As to the hypothesis 4-2, as I did for
financial activity analyses, I also ran both OLS and logistic regressions using those three
90
variables for exotic features as a dependent variable in turn. Table 2.16 reports the results
of an OLS regression and Table 2.17 presents results of logistic regressions. No significant
results for the variable After_FVO were found from all three regression models.
[Insert Tables 2.16 and 2.17 about here]
I also ran the same regressions using all available bond issues without deleting any
observations during the sample period. However, even though the significance increased
in general probably due to the increase in sample size (the sample size was 1,257), I could
not find results which are consistent with the developed hypothesis (untabulated).
VII. CONCLUSION, LIMITATIONS, AND EXTENSIONS
In this study, I have first examined the effectiveness of FVO of SFAS 159 in
mitigating earnings’ volatility and its effect on analysts’ forecast accuracy. The test results
indicate that the earnings’ volatility of FVO adopters have in fact increased and the impact
is amplified in accordance with the size of gains or losses resulting from the FVO. Also,
the test results of the second hypothesis indicate that analysts’ forecast error of FVO
adopters increased more than that of non-FVO adopters, which is consistent with the
positive relation reported in prior studies between earnings’ volatility and analysts’ forecast
error. Given the results for the first two hypotheses, I have also tried to explore the
motivation of firms to adopt FVO. I conjectured two more hypotheses to try to identify
potential motives underlying FVO adoption (i) an opportunistic intention and (ii) an
informative intention. Unfortunately, I could not find results which support either
hypothesis other than some limited evidence for opportunistic intention.
The results of this study cast doubt on the effectiveness of SFAS 159. This study
91
implies that we need to examine more thoroughly whether SFAS 159 provides managers
with excessive discretion without achieving its intended purpose. Granting full discretion
to managers on which financial instrument is reported at fair value and thus whether
changes in fair values of specific financial instruments are included in net income might
deteriorate reliability and comparability of financial statements without lowering earnings’
volatility which is not even a main purpose of financial reporting. This study provides some
limited evidence for managers’ opportunistic motivation of adopting FVO. Further studies
with more refined data and more robust research methodologies which explore the
intentions of FVO adopters would be interesting.
Limitations and Extensions
Testing hypotheses 4-1 and 4-2 using exotic features of bonds issues may not be
appropriate. I hypothesize that firms with more complex financial instruments are more
likely to adopt FVO. However, I didn’t not test the hypothesis using actual features of
financial instruments owned by firms. Rather, I used information on bonds issued for
several years before adopting FVO. The recently issued bonds may not properly represent
the characteristics of all financial instruments held by the firm.
Examining bond covenants can be a possible future research topic. For example,
investigating whether firms adopt FVO in order to reduce the likelihood of violating bond
covenants or whether bond covenants are adjusted to reflect the effects of FVO would be
interesting. In fact, Demerjian et al. (2016) reports evidence that covenant definitions of
private loan contracts are modified to exclude the effects of FVO.
92
Examining the association between adopting FVO and CEO tenure would also be
interesting. As mentioned in the hypothesis development section, if managers adopt FVO
opportunistically, they would do so for a short term effect as it is very difficult to forecast
accurately the fair values of financial instruments in the long term; because CEOs who
are going to retire in the near future may have more incentives to manage short term
performances, those CEOs are more likely to adopt FVO.
Another possibility is that even if we suppose that results were found indicating
that earnings’ volatility decreased after adopting FVO, it would not necessarily mean that
the FVO was used in accordance with the intent of SFAS 159. For example, if firms use
FVO for smoothing earnings rather than for mitigating the problem of mixed-measurement
accounting, earnings’ volatility would also decrease. This practice is obviously not in
accordance with the intent of SFAS 159, and can be regarded as an example of
opportunistic usage of FVO. This implies that when there is too much discretion allowed
or not enough guidance provided for certain accounting standards, it can often be difficult
to assess whether the standards work in compliance with their intended objective. Thus,
given the current movement toward principle-based accounting from rule-based
accounting, it is also an important issue to determine how much additional guidance should
be provided in order to achieve each standard’s intended objective while simultaneously
avoiding the difficulties that arise from the specific guidance that reduce the effectiveness
of principle-based standards.
93
REFERENCES
Aboody, D., M. E. Barth, and R. Kasznik. 2006. Do firms understate stock-based
compensation expense disclosed under SFAS 123? Review of Accounting Studies 11:
429–461.
Ahmed, A. S., E. Kilic, and G. J. Lobo. 2006. Does recognition versus disclosure matter?
Evidence from value-relevance of banks’ recognized and disclosed derivative
financial instruments. The Accounting Review 81 (3): 567–588.
Allen, F., and E. Carletti. 2008. Mark-to-market accounting and liquidity pricing. Journal
of Accounting and Economics 45: 358–378.
American Accounting Association (AAA), Financial Accounting Standards Committee
(FASC). 2007. Response to FASB Exposure Draft, “The Fair Value Option for
Financial Assets and Financial Liabilities, Including an Amendment of FASB
Statement No. 115”. Accounting Horizon 21 (2): 189-200.
Barth, M. E. 1994. Fair value accounting: evidence from investment securities and the
market valuation of banks. The Accounting Review 69 (1): 1–25.
Barth, M. E. 2004. Fair values and financial statement volatility. In The Market Discipline
across Countries and Industries, edited by C. Borio, W. C. Hunter, G. G. Kaufman,
and K. Tsatsaronis, 323-333. Cambridge, MA: MIT Press.
Barth, M. E., W. H. Beaver, and W. R. Landsman.1996. Value-relevance of banks’ fair value
disclosures under SFAS 107. The Accounting Review 71 (4): 513–537.
———, ———, and ———. 2001. The relevance of the value relevance literature for
accounting standard setting: another view. Journal of Accounting and Economics 31:
77–104.
Barth, M. E. and W. R. Landsman. 1995. Fundamental issues related to using fair value
accounting for financial reporting. Accounting Horizons 9 (4): 97–107.
———, ———, and J. M. Wahlen. 1995. Fair value accounting: Effects on banks’ earnings
volatility, regulatory capital, and value of contractual cash flows. Journal of Banking
& Finance 19: 577–605.
Bernard, V. L., R. C. Merton, and K. G. Palepu. 1995. Mark-to-market accounting for banks
and thrifts: lessons from the Danish experience. Journal of Accounting Research 33
(1): 1–32.
94
Black, E. L., K. F. Sellers, and T. S. Manly. 1998. Earnings management using asset sales:
an international study of countries allowing noncurrent asset revaluation. Journal of
Business Finance and Accounting 25 (9/10): 1,287–1,317.
Bradshaw, M. T., S. A. Richardson, and R. G. Sloan. 2006. The relation between corporate
financing activities, analysts’ forecasts and stock returns. Journal of Accounting and
Economics 42: 53–85.
Dechow, P., L. Myers, and C. Shakespeare. 2010. Fair value accounting and gains from
asset securitizations: A convenient earnings management tool with compensation
side-benefits. Journal of Accounting and Economics 49 (1): 2–25.
Demerjian, P. R., J. Donovan, and C. R. Larson. 2016. Fair value accounting and debt
contracting: Evidence from adopting of SFAS 159. Journal of Accounting Research
54 (4): 1041-1076.
Dichev, I. D., and V. W. Tang. 2009. Earnings volatility and earnings predictability. Journal
of Accounting and Economics 47: 160-181.
Eccher, E. A., K. Ramesh, and S. R. Thiagarajan. 1996. Fair value disclosures by bank
holding companies. Journal of Accounting and Economics 22: 79–117.
Fiechter, P. 2011. The effects of the fair value option under IAS 39 on the volatility of bank
earnings. Journal of International Accounting Research 10 (1): 85-108.
Financial Accounting Standards Board (FASB). 1984. Recognition and measurement in
financial statements of business enterprises. Statement of Financial Accounting
Concepts No. 5. Stamford, CT: FASB.
———. 1991. Disclosures about Fair Value of Financial Instruments. Statement of
Financial Accounting Standards No. 107. Norwalk, CT: FASB.
———. 1993. Accounting for Certain Investments in Debt and Equity Securities.
Statement of Financial Accounting Standards No. 115, Norwalk, CT: FASB.
———. 1994. Disclosure about Derivative Financial Instruments and Fair Value of
Financial Instruments. Statement of Financial Accounting Standards No. 119.
Norwalk, CT: FASB.
———. 1998. Accounting for Derivative Instruments and Hedging Activities. Statement
of Financial Accounting Standards No. 133. Norwalk, CT: FASB.
———. 2006. Fair Value Measurements. Statement of financial Accounting Standards No.
95
157. Norwalk, CT: FASB.
———. 2007. The Fair Value Option for Financial Assets and Financial Liabilities-
Including an Amendment of FASB Statement No. 115. Statement of financial
Accounting Standards No. 159. Norwalk, CT: FASB.
Glover, J. C., Y. Ijiri, C. B. Levine, and P. J. Liang. 2005. Separating facts from forecasts
in financial statements. Accounting Horizons 19 (December): 267–282.
Gorton, G., P. He, and L. Huang. 2006. Asset Prices When Agents Are Marked to Market.
NBER Working Paper No. 12075, National Bureau of Economic Research (NBER).
Graham, J. R., C. R. Harvey, and S. Rajgopal. 2005. The economic implications of
corporate financial reporting. Journal of Accounting and Economics 40: 3-73.
Guthrie, K., J. H. Irving, and J. Sokolowsky. 2011. Accounting choice and the fair value
option. Accounting Horizon 25 (3): 487-510.
Henry, E. 2009. Early adoption of SFAS No. 159: Lessons from games (almost) played.
Accounting Horizon 23 (2): 181-199.
Hirst, D. E., P. E. Hopkins, and J. E. Wahlen. 2004. Fair values, income measurement, and
bank analysts’ risk and valuation judgments. The Accounting Review 79 (2): 453–
472.
Hodder, L. D., P. E. Hopkins, and J. M. Wahlen. 2006. Risk-relevance of fair-value income
measures for commercial banks. The Accounting Review 81 (2): 337-375.
Hsu, P., and Y. Lin. 2016. Fair value accounting and earnings management. Eurasian
Journal of Business and Management. 4 (2): 41-54.
International Accounting Standards Board (IASB). 2005. Financial Instruments:
Recognition and Measurement-The Fair Value Option. International Accounting
Standard 39. London, U.K.: IASB.
Landsman W. R. 2007. Is fair value accounting information relevant and reliable? Evidence
from capital market research. Accounting and Business Research Special Issue:
International Accounting Policy Forum: 19-30.
Li, F. 2006. Fair Value Option: A review and assessment. Tennessee CPA Review (October):
4-6.
Nelson, K. K. 1996. Fair value accounting for commercial banks: an empirical analysis of
SFAS No. 107. The Accounting Review 71 (2): 161–182.
96
Plantin, G., H. Sapra, and H. S. Shin. 2005. Marking to market, liquidity and financial
stability. Monetary and Economic Studies (Special Edition/October): 133–155.
———, ———, and ———. 2008a. Fair value accounting and financial stability.
Financial Stability Review 12 (October): 85-94.
———, ———, and ———. 2008b. Marking-to-market: Panacea or Pandora’s Box?
Journal of Accounting Research 46 (2): 435-460.
Roberts M. R. and T. M. Whited. 2011. Endogeneity in empirical corporate finance.
Working paper, University of Rochester.
Securities and Exchange Commission (SEC). 2005. Report and Recommendations
Pursuant to Section 401(c) of the Sarbanes-Oxley Act of 2002 on Arrangements with
Off-Balance Sheet Implications, Special Purpose Entities, and Transparency of
Filings by Issuers. Washington, D.C.: SEC. Available at:
https://www.sec.gov/news/studies/soxoffbalancerpt.pdf.
Venkatachalam, M. 1996. Value-relevance of banks’ derivatives disclosures. Journal of
Accounting and Economics 22: 327–355.
97
Tables for Chapter 2
Table 2.1
Distribution of FVO firms by FVO adoption year and industry
Panel A: Distribution of FVO firms by the first FVO adoption year
Year Number of FVO firms
2007 30
2008 108
2009 35
2010 14
2011 12
2012 3
Sum 202
Panel B: Distribution of FVO firms by industry
FVO firms represented by the following SIC codes Number of firms
01-09: Agriculture, Forestry, Fishing 1
10-14: Mining 2
15-17: Construction 4
20-39: Manufacturing 57
40-49: Transportation, Communications, Utilities 5
52-59: Retail 7
60-67: Finance, Insurance, Real Estate 100
70-89: Services 24
91-99 Public Administration 2
Total 202
98
Table 2.2 Variable Definitions
NI_std = Four year moving standard deviation of net income, winsorized at ± 1;
FVO = 1 if the firm is a FVO firm, 0 otherwise;
After = 1 if the firm year is an post-FVO adoption year, 0 otherwise;
After_FVO = Interaction term of FVO and After;
FVO_Size_sum = Four year moving sum of absolute value of the item 'tfvce'19;
TA_ave = Four year moving average of natural log of total assets;
Debt_ave = Four year moving average of debt to total assets ratio;
BTM_ave = Four year moving average of book to market ratio;
ROA_ave = Four year moving average of return on assets ratio;
Special_sum = Four year moving sum of a dummy variable for special item;
Loss_sum = Four year moving sum of a dummy variable for loss;
Per_FE = Forecast error in percentage, absolute value of mean forecast less actual eps
divided by actual eps times 100, winsorized at ± 1;
FVO_size = Proportion of gain or loss from FVO in earnings (‘tfvce’ divided by income
before extraordinary items times 100);
Loss = 1 if the firm reports a net loss, 0 otherwise;
Special = 1 if the firm reports special items, 0 otherwise;
Debt = Debt to total assets ratio;
LN_TA = Natural log of total assets;
NumEst = Number of forecasts for each firm year.
Per_EPS = Change in eps in percentage, absolute value of current eps less previous year's
eps divided by previous year's eps times 100;
Neg_Ch_EPS = 1 if change in eps is negative, 0 otherwise;
Per_EPS_Neg_Ch = Interaction term of per_eps and neg_ch_eps.
Financing = Net external financing, sum of equity financing and debt financing scaled by
average total assets, winsorized at ± 1
*Net equity financing: sale of common and preferred stock less purchase of
common and preferred stock less cash payments of dividends
*Net debt financing: issuance of long-term debt less payments for long-term
debts reductions plus net changes in current debt;
19 The ‘tfvce’ stands for ‘total fair value change in earnings’, and it reflects the gains or losses
resulting from the change in fair value of financial assets or liabilities which are included in reported
earnings of income statement only by SFAS 159.
99
Financing_d = 1 if variable 'financing' is positive, 0 otherwise;
High_Financing = 1 if 'financing' is greater than median 'financing' within the same year and
industry (1 digit sic code) group, 0 otherwise;
Lag_FVO_Size = Absolute value of the amount of 'tfvce' divided by net income at the end of
the previous year;
ROA = Return of assets
ICW = 1 if the firm is identfied as an internal control weakness firm, 0 otherwise
(Compustat item 'AUOPIC');
BTM = Book to market ratio.
Lag_LN_TA = Natural log of total assets at the end of the previous year;
Lag_Debt = Debt to total assets ratio at the end of the previous year;
Lag_ROA = Return of assets at the end of the previous year;
Lag_Special = 1 if the firm reports special items in the previous year, 0 otherwise;
Lag_Loss = 1 if the firm reports a net loss in the previous year, 0 otherwise;
Lag_ICW = 1 if the firm is identfied as an internal control weakness firm in the previous
year, 0 otherwise (Compustat item 'AUOPIC');
Lag_BTM = Book to market ratio at the end of the previous year.
NFFirm = 1 if the firm is a non-financial firm (firm whose sic code doesn’t start with 6),
0 otherwise;
After_NFFirm = Interaction term of NFFirm and After;
Exotic = Numerical variable for exotic features of bonds issued.
*The variable gets 1 point for each of the following features: asset backed,
convertible, redeemable, puttable, refundable, soft call, enhancement, and
exchangeable.
Exotic_d = 1 for bonds issued with exotic features
High_Exotic = 1 if 'exotic' is greater than median 'exotic' within the same year and industry
(1 digit sic code) group, 0 otherwise;
100
Table 2.3
Descriptive Statistics (for H1)
Panel A: Distributional Properties of Variables (All Sample), N=808 Variable Mean Median Std Dev 1Q 3Q
NI_std
404.488
40.434
1113.580
9.275
228.799
NI1_std
7.514
1.718
20.002
0.366
5.741
FVO
0.500
0.500
0.500
0.000
1.000
After
0.500
0.500
0.500
0.000
1.000
After_FVO
0.250
0.000
0.433
0.000
0.500
FVO_Size_sum
267.513
0.000
2366.390
0.000
0.001
tfvce1_sum
0.637
0.000
1.271
0.000
0.500
tfvce2_sum
51.145
0.000
655.740
0.000
0.003
TA_ave
7.696
7.483
2.758
5.861
9.353
Debt_ave
0.320
0.132
3.135
0.039
0.292
BTM_ave
0.246
0.630
7.963
0.384
0.982
ROA_ave
-0.169
0.008
3.786
-0.011
0.041
Special_sum
2.423
3.000
1.377
1.000
4.000
Loss_sum
1.134
1.000
1.396
0.000
2.000
101
TABLE 2.3 (continued)
Panel B: Distributional Properties of Variables (FVO vs Matched firm sample)a
FVO Firm Sample, N=404
Matched Firm Sample, N=404
Variable
Mean
Median
Std Dev
Mean
Median
Std Dev
NI_std 529.165 42.165 1351.900 279.812 36.662 790.185
NI1_std 7.390 1.930 20.368 7.637 1.468 19.652
After 0.500 0.500 0.501 0.500 0.500 0.501
TA_ave 7.795 7.573 2.805 7.596 7.391 2.709
Debt_ave 0.439 0.139 4.424 0.201 0.122 0.289
BTM_ave -0.257 0.624 11.214 0.750 0.637 0.855
ROA_ave -0.298 0.008 5.346 -0.039 0.009 0.301
Special_sum 2.473 3.000 1.348 2.374 2.000 1.406
Loss_sum 1.243 1.000 1.411 1.025 0.000 1.375
Panel C: Distributional Properties of Variables (Pre-FVO vs Post-FVO)a
Before FVO, N=404 After FVO, N=404
Variable Mean Median Std Dev Mean Median Std Dev
NI_std 330.310 32.292 849.743 478.667 44.828 1322.920
NI1_std 6.824 1.679 17.218 8.204 1.780 22.443
FVO 0.500 0.500 0.501 0.500 0.500 0.501
TA_ave 7.592 7.375 2.715 7.799 7.609 2.799
Debt_ave 0.215 0.137 0.300 0.425 0.128 4.423
BTM_ave 0.584 0.541 0.542 -0.092 0.830 11.245
ROA_ave -0.053 0.010 0.559 -0.284 0.006 5.326
Special_sum 2.208 2.000 1.368 2.639 3.000 1.354
Loss_sum 0.948 0.000 1.344 1.319 1.000 1.425
102
TABLE 2.3 (continued)
Panel D: Pearson Correlation (top) and Spearman Correlation (bottom) b
1 2 3 4 5 6 7 8 9 10 11 12 13 14
NI_std 1
-0.05 0.11 0.07 0.12 0.53 0.17 0.02 0.57 0.00 -0.14 0.02 0.11 -0.01
NI1_std 2 0.03
-0.01 0.03 0.02 -0.02 -0.01 -0.01 -0.42 0.27 -0.20 -0.30 0.01 0.41
FVO 3 0.06 0.05
0.00 0.58 0.11 0.50 0.08 0.04 0.04 -0.06 -0.03 0.04 0.08
After 4 0.08 0.04 0.00
0.58 0.11 0.50 0.08 0.04 0.03 -0.04 -0.03 0.16 0.13
After_FVO 5 0.08 0.06 0.58 0.58
0.20 0.87 0.14 0.04 0.06 -0.10 -0.05 0.11 0.13
FVO_Size_sum 6 0.09 0.06 0.53 0.47 0.85
0.29 0.09 0.24 0.01 -0.10 0.01 0.09 0.03
tfvce1_sum 7 0.07 0.05 0.68 0.44 0.90 0.76
0.19 0.11 0.02 -0.15 0.00 0.11 0.13
tfvce2_sum 8 0.08 0.03 0.49 0.49 0.85 0.74 0.79
0.03 0.01 0.01 0.00 0.00 0.00
TA_ave 9 0.75 -0.58 0.03 0.04 0.04 0.04 0.05 0.06
-0.15 -0.02 0.19 0.19 -0.42
Debt_ave 10 0.26 -0.03 0.03 0.01 0.02 0.01 0.04 0.05 0.25
-0.17 -0.99 -0.03 0.06
BTM_ave 11 0.11 -0.28 0.04 0.27 0.19 0.17 0.16 0.19 0.26 -0.02
0.12 -0.01 -0.11
ROA_ave 12 0.07 -0.20 -0.03 -0.09 -0.05 -0.06 -0.07 -0.07 0.18 -0.13 -0.23
0.04 -0.08
Special_sum 13 0.32 0.12 0.03 0.18 0.12 0.11 0.09 0.12 0.17 0.12 0.05 -0.05
0.07
Loss_sum 14 -0.01 0.57 0.09 0.14 0.13 0.13 0.13 0.10 -0.39 0.08 0.08 -0.75 0.09
a Bold text indicates significant differences between two subsamples at the 0.05 level or better, two-tailed. Differences in means are assessed
using a t-test.
b Bold text indicates significance at the 0.05 level or better, two-tailed.
103
Table 2.4
OLS Regression Analyses of Earnings Volatility on Adoption of the FVO
(full sample)
Variable Model 1-1 Model 1-2
Estimate p-value Estimate p-value
Intercept -1390.86*** 0.0014 -1147.30*** 0.003
FVO 81.61 0.3194 100.04 0.1679
After 348.29* 0.0575 198.38 0.2218
After_FVO 88.98 0.4427 -91.34 0.3765
FVO_Size_sum 0.17*** <.0001
TA_ave 343.90*** <.0001 298.96*** <.0001
Debt_ave -101.82 0.1456 -81.44 0.1883
BTM_ave -12.94*** 0.0012 -9.91*** 0.005
ROA_ave -112.40* 0.0521 -92.25* 0.0716
Special_sum -64.53*** 0.0051 -70.09*** 0.0006
Loss_sum 198.68*** <.0001 164.70*** <.0001
Industry dummies Included Included
Year dummies Included Included
Adj. R-sq 0.4576 0.5756
F-statistic 26.22 40.08
n 808 808
Dependent variable NI_std NI_std
*, **, *** indicate statistical significance at the 10, 5, 1 percent level respectively.
104
Table 2.5
OLS Regression Analyses of Earnings Volatility on Adoption of the FVO
(financial firms only)
Variable Model 1-1 Model 1-2
Estimate p-value Estimate p-value
Intercept -3162.95*** <.0001 -2630.53*** <.0001
FVO 168.02 0.1953 211.50* 0.0599
After 634.51** 0.0234 491.41** 0.0425
After_FVO 160.66 0.3764 -139.34 0.3817
FVO_Size_sum 0.15*** <.0001
TA_ave 392.34*** <.0001 340.93*** <.0001
Debt_ave -79.13 0.7836 -304.97 0.2229
BTM_ave -16.66*** 0.0017 -13.14*** 0.0043
ROA_ave -1389.05 0.4949 -705.66 0.6886
Special_sum -2.88 0.9398 -11.84 0.7195
Loss_sum 168.84*** 0.007 161.03*** 0.0029
Industry dummies Included Included
Year dummies Included Included
Adj. R-sq 0.5505 0.6639
F-statistic 22.25 33.84
n 400 400
Dependent variable NI_std NI_std
*, **, *** indicate statistical significance at the 10, 5, 1 percent level respectively.
105
Table 2.6
OLS Regression Analyses of Earnings Volatility on Adoption of the FVO
(non-financial firms only)
Variable Model 1-1 Model 1-2
Estimate p-value Estimate p-value
Intercept -344.96 0.265 -336.18 0.2502
FVO 46.05 0.4782 42.45 0.4889
After -131.23 0.4682 -149.70 0.3812
After_FVO 0.18 0.9985 -44.72 0.6061
FVO_Size_sum 0.38*** <.0001
TA_ave 104.64*** <.0001 111.12*** <.0001
Debt_ave -25.63 0.6131 -40.06 0.4035
BTM_ave -3.33 0.4527 -5.14 0.2214
ROA_ave -29.41 0.4802 -41.74 0.2894
Special_sum -3.46 0.8611 -18.48 0.3256
Loss_sum 41.54** 0.026 44.69** 0.0113
Industry dummies Included Included
Year dummies Included Included
Adj. R-sq 0.6814
0.7157 F-statistic
23.32
26.62 n
408
408 Dependent variable NI_std NI_std
*, **, *** indicate statistical significance at the 10, 5, 1 percent level respectively.
106
Table 2.7
Descriptive Statistics (for H2)
Panel A: Distributional Properties of Variables (All Sample), N=2387 Variable Mean Median Std Dev 1Q 3Q
Per_FE 20.199 4.393 49.730 1.397 14.286
FVO 0.506 1.000 0.500 0.000 1.000
After 0.530 1.000 0.499 0.000 1.000
After_FVO 0.266 0.000 0.442 0.000 1.000
FVO_size 6.278 0.000 51.300 0.000 0.000
LOSS 0.247 0.000 0.431 0.000 0.000
Special 0.627 1.000 0.484 0.000 1.000
Debt 0.195 0.114 0.236 0.027 0.283
LN_TA 8.245 8.038 2.503 6.428 9.830
NumEst 8.419 6.000 6.852 3.000 13.000
Per_EPS 142.281 33.014 907.202 14.103 82.609
Neg_Ch_EPS 0.389 0.000 0.488 0.000 1.000
Per_EPS_Neg_Ch 66.080 0.000 597.990 0.000 23.767
107
TABLE 2.7 (continued)
Panel B: Pearson Correlation (top) and Spearman Correlation (bottom) a
1 2 3 4 5 6 7 8 9 10 11 12 13
Per_FE 1 0.039 0.107 0.077 0.095 0.239 0.028 0.075 0.002 -0.151 0.026 0.192 0.036
FVO 2 0.039 -0.008 0.595 0.121 0.064 0.030 0.049 0.009 0.049 -0.022 -0.001 0.008
After 3 0.107 -0.008 0.568 0.115 0.121 0.120 -0.016 0.010 0.008 0.068 0.098 0.062
After_FVO 4 0.077 0.595 0.568 0.203 0.125 0.077 0.018 0.003 0.022 0.010 0.040 0.037
FVO_size 5 0.095 0.121 0.115 0.203 0.028 -0.011 0.024 0.051 -0.009 -0.004 0.053 -0.003
LOSS 6 0.239 0.064 0.121 0.125 0.028 0.099 0.147 -0.275 -0.156 0.078 0.242 0.162
Special 7 0.028 0.030 0.120 0.077 -0.011 0.099 0.092 0.080 0.114 0.031 0.033 0.015
Debt 8 0.075 0.049 -0.016 0.018 0.024 0.147 0.092 0.160 -0.065 0.052 0.054 0.047
LN_TA 9 0.002 0.009 0.010 0.003 0.051 -0.275 0.080 0.160 0.260 -0.011 -0.011 -0.004
NumEst 10 -0.151 0.049 0.008 0.022 -0.009 -0.156 0.114 -0.065 0.260 -0.062 -0.134 -0.052
Per_EPS 11 0.026 -0.022 0.068 0.010 -0.004 0.078 0.031 0.052 -0.011 -0.062 0.024 0.650
Neg_Ch_EPS 12 0.192 -0.001 0.098 0.040 0.053 0.242 0.033 0.054 -0.011 -0.134 0.024 0.138
Per_EPS_Neg_Ch 13 0.036 0.008 0.062 0.037 -0.003 0.162 0.015 0.047 -0.004 -0.052 0.650 0.138
a Bold text indicates significance at the 0.05 level or better, two-tailed.
108
Table 2.8
OLS Regression Analyses of Analysts’ Forecast Error on Adoption of the FVO
Variable Model 2-1 Model 2-2
Estimate Pr > |t| Estimate Pr > |t|
Intercept -22.73 0.3685 -20.97 0.4059
FVO 4.94* 0.08 4.95* 0.0789
After 1.04 0.8052 0.58 0.8903
After_FVO -3.72 0.3342 -5.21 0.1785
FVO_size 0.06*** 0.0013
LOSS 22.69*** <.0001 22.69*** <.0001
Special 0.23 0.9146 0.39 0.8536
Debt -0.98 0.8473 -1.32 0.7958
LN_TA 1.99*** 0.0021 1.98*** 0.0021
NumEst -0.99*** <.0001 -0.99*** <.0001
Per_EPS 0.00 0.542 0.00 0.5587
Neg_Ch_EPS 9.92*** <.0001 9.75*** <.0001
Per_EPS_Neg_Ch 0.00 0.1489 0.00 0.1659
Industry dummies Included Included
Year dummies Included Included
Adj. R-sq 0.116 0.1195
F-statistic 7.52 7.61
n 2387 2387
Dependent variable Per_FE Per_FE
*, **, *** indicate statistical significance at the 10, 5, 1 percent level respectively.
109
Table 2.9
Descriptive Statistics (for H3)
Panel A: Distributional Properties of Variables (All Sample), N=1645
Variable Mean Median Std Dev 1Q 3Q
Financing_d 0.4365 0.0000 0.4961 0.0000 1.0000
Financing 0.9747 -0.3117 15.9051 -4.0187 2.8812
High_Financing 0.5058 1.0000 0.5001 0.0000 1.0000
FVO 0.5064 1.0000 0.5001 0.0000 1.0000
After 0.5052 1.0000 0.5001 0.0000 1.0000
After_FVO 0.2547 0.0000 0.4358 0.0000 1.0000
Lag_FVO_Size 0.0585 0.0000 0.6239 0.0000 0.0000
Lag_LN_TA 7.7920 7.5290 2.7112 5.9038 9.6123
Lag_Debt 0.2098 0.1192 0.2639 0.0214 0.2995
Lag_ROA -0.0488 0.0104 0.6806 -0.0034 0.0475
Lag_Special 0.5994 1.0000 0.4902 0.0000 1.0000
Lag_Loss 0.2748 0.0000 0.4465 0.0000 1.0000
Lag_BTM 0.6155 0.5808 1.9854 0.3376 0.8853
110
TABLE 2.9 (continued)
Panel B: Pearson Correlation (top) and Spearman Correlation (bottom) a
1 2 3 4 5 6 7 8 9 10 11 12 13
Financing_d 1 0.537 0.794 0.026 -0.068 -0.031 -0.019 -0.044 0.024 -0.087 -0.028 0.054 -0.042
Financing 2 0.859 0.515 -0.005 -0.063 -0.047 -0.034 -0.138 0.011 -0.149 -0.050 0.157 -0.041
High_Financing 3 0.794 0.818 0.026 0.009 0.023 -0.012 -0.038 -0.009 -0.078 -0.044 0.072 0.001
FVO 4 0.026 -0.005 0.026 -0.004 0.577 0.093 0.024 0.047 -0.032 0.014 0.071 -0.039
After 5 -0.068 -0.056 0.009 -0.004 0.579 0.093 0.012 0.028 -0.013 0.054 0.127 0.015
After_FVO 6 -0.031 -0.042 0.023 0.577 0.579 0.161 0.025 0.043 -0.011 0.039 0.115 -0.036
Lag_FVO_Size 7 -0.070 -0.067 -0.014 0.376 0.376 0.651 -0.018 0.004 0.001 -0.005 0.023 0.022
Lag_LN_TA 8 -0.076 -0.049 -0.054 0.019 0.011 0.023 0.015 0.119 0.221 0.104 -0.294 0.026
Lag_Debt 9 -0.023 -0.062 -0.050 0.025 0.033 0.033 0.040 0.310 -0.169 0.068 0.195 -0.201
Lag_ROA 10 -0.151 -0.220 -0.159 -0.043 -0.122 -0.083 -0.086 0.107 -0.147 0.015 -0.233 0.150
Lag_Special 11 -0.028 -0.054 -0.044 0.014 0.054 0.039 0.042 0.096 0.073 -0.104 0.106 0.017
Lag_Loss 12 0.054 0.076 0.072 0.071 0.127 0.115 0.117 -0.318 0.051 -0.763 0.106 -0.031
Lag_BTM 13 -0.102 -0.046 -0.016 0.020 0.224 0.145 0.146 0.210 0.053 -0.264 0.085 0.073
a Bold text indicates significance at the 0.05 level or better, two-tailed.
111
Table 2.10
OLS Regression Analyses of financing activity on Adoption of the FVO
Variable Model 3-1 Model 3-2
Estimate p-value Estimate p-value
Intercept -3.078 0.598 -3.088 0.597
FVO -0.120 0.909 -0.117 0.912
After 1.063 0.428 1.048 0.434
After_FVO -0.604 0.683 -0.470 0.752
Lag_FVO_Size -0.582 0.337
Lag_LN_TA -0.409* 0.056 -0.418* 0.051
Lag_Debt -3.596** 0.039 -3.589** 0.039
Lag_ROA -2.601*** <.0001 -2.597*** <.0001
Lag_Special -0.882 0.276 -0.887 0.273
Lag_Loss 4.155*** <.0001 4.140*** <.0001
Lag_BTM -0.129 0.512 -0.124 0.528
Industry dummies Included Included
Year dummies Included Included
Adj. R-sq 0.1161 0.1161
F-statistic 5.91 5.8
n 1645 1645
Dependent variable Financing Financing
*, **, *** indicate statistical significance at the 10, 5, 1 percent level respectively.
112
Table 2.11
Logistic Regression Analyses of financing activity on Adoption of the FVO
Variable Model 3-3 Model 3-4
Estimate Pr > ChiSq Estimate Pr > ChiSq
Intercept -148.900 0.964 8.929 0.989
FVO 0.075 0.318 0.043 0.551
After 0.111 0.254 0.009 0.926
After_FVO -0.030 0.781 0.038 0.711
Lag_FVO_Size 0.009 0.916 0.043 0.608
Lag_LN_TA -0.025 0.428 -0.015 0.617
Lag_Debt 0.542* 0.039 0.652** 0.011
Lag_ROA 1.452*** 0.000 1.468*** 0.000
Lag_Special 0.125 0.279 0.189* 0.090
Lag_Loss -0.020 0.902 -0.073 0.635
Lag_BTM 0.005 0.861 -0.020 0.487
Industry dummies Included Included
Year dummies Included Included
LR ch2 (DF: 45) 207.682 113.800
LR Pr>ch2 <.0001 <.0001
n 1645 1645
Dependent variable Financing_d High_Financing
*, **, *** indicate statistical significance at the 10, 5, 1 percent level respectively.
113
Table 2.12
Gains or Losses?: Number of firms which report gains or losses resulting
from FVO in the first FVO adoption year
Year Financial firms Non-financial firms
Gains Losses Gains Losses
2006 0 0 1 0
2007 26 12 4 3
2008 27 44 54 26
2009 9 7 17 20
2010 7 6 5 3
2011 6 4 7 2
2012 5 0 3 4
2013 2 3 3 2
2014 4 2 3 0
2015 5 3 1 1
114
Table 2.13
Logistic Regression Analyses of financing activity on Adoption of the FVO with only
firms which adopted FVO in 2008
(Financial vs non-financial firms separately)
Variable Non-financial firms only VS Financial firms only
Estimate Pr > ChiSq Estimate Pr > ChiSq
Intercept -8.566 0.980 0.213 0.667
FVO -0.232 0.111 0.003 0.985
After -0.109 0.542 0.176 0.327
After_FVO 0.416** 0.039 -0.303 0.138
Lag_LN_TA 0.184** 0.014 -0.018 0.732
Lag_Debt 0.572 0.171 -0.558 0.172
Lag_ROA 0.680 0.187 9.717** 0.011
Lag_Special 0.517** 0.025 -0.064 0.762
Lag_Loss -0.597** 0.033 1.037*** 0.003
Lag_BTM 0.370** 0.045 -0.050 0.297
Industry dummies Included Not included
Year dummies Included Included
LR ch2 (DF: 27, 11) 91.772 19.673
LR Pr>ch2 <.0001 <.0001
n 482 413
Dependent variable High_Financing High_Financing
*, **, *** indicate statistical significance at the 10, 5, 1 percent level respectively.
115
Table 2.14
Logistic Regression Analysis of financing activity on Adoption of the FVO with
only firms which adopted FVO in 2008
(Financial and non-financial firms together)
Variable Estimate Std error Wald Chisq Pr>ChiSq
Intercept -19.848 782.900 0.001 0.980
NFFirm -0.138 0.563 0.060 0.807
After -0.285 0.187 2.312 0.128
After_NFFirm 0.422** 0.206 4.198 0.041
Lag_LN_TA 0.056 0.068 0.673 0.412
Lag_Debt 1.220** 0.523 5.448 0.020
Lag_ROA 1.008 0.661 2.330 0.127
Lag_Special -0.114 0.224 0.261 0.610
Lag_Loss 0.019 0.292 0.004 0.948
Lag_BTM 0.007 0.033 0.047 0.829
Industry dummies Included
Year dummies Included
LR ch2 (DF: 31) 64.483
LR Pr>ch2 <.0004
n 455
Dependent variable High_Financing
*, **, *** indicate statistical significance at the 10, 5, 1 percent level respectively.
116
Table 2.15
Descriptive Statistics (for H4)
Panel A: Distributional Properties of Variables (All Sample), N=410
Variable Mean Median Std Dev 1Q 3Q
Exotic
1.017
1.000
0.829
1.000
1.000
Exotic_d
0.754
1.000
0.431
1.000
1.000
High_Exotic
0.746
1.000
0.436
0.000
1.000
FVO
0.573
1.000
0.495
0.000
1.000
After
0.559
1.000
0.497
0.000
1.000
After_FVO
0.322
0.000
0.468
0.000
1.000
Lag_LN_TA
10.294
10.458
2.634
8.425
12.172
Lag_Debt
0.272
0.207
0.248
0.071
0.370
Lag_ROA
0.002
0.010
0.136
0.002
0.026
Lag_Special
0.722
1.000
0.449
0.000
1.000
Lag_Loss
0.207
0.000
0.406
0.000
0.000
Lag_ICW
0.049
0.000
0.216
0.000
0.000
Lag_BTM
1.055
0.665
4.516
0.429
1.004
117
TABLE 15 (continued)
Panel B: Pearson Correlation (top) and Spearman Correlation (bottom) a
1 2 3 4 5 6 7 8 9 10 11 12 13
Exotic 1
0.703 0.696 0.018 0.054 0.017 -0.332 0.059 -0.159 -0.027 0.200 0.091 0.001
Exotic_d 2 0.837
0.981 -0.070 0.085 -0.006 -0.248 -0.052 -0.044 -0.001 0.111 0.024 0.042
High_Exotic 3 0.827 0.981
-0.072 0.103 0.006 -0.230 -0.044 -0.036 0.001 0.105 0.028 0.043
FVO 4 -0.016 -0.070 -0.072
0.007 0.595 0.173 0.117 0.037 -0.040 0.101 0.035 0.059
After 5 0.064 0.085 0.103 0.007
0.613 0.123 0.041 -0.064 0.216 0.176 0.019 0.073
After_FVO 6 0.002 -0.006 0.006 0.595 0.613
0.227 0.115 0.032 0.043 0.201 0.014 0.103
Lag_LN_TA 7 -0.341 -0.261 -0.243 0.180 0.114 0.220
0.034 0.199 0.136 -0.257 0.022 0.057
Lag_Debt 8 0.002 -0.107 -0.098 0.182 0.084 0.192 0.081
-0.305 0.081 0.179 0.072 -0.046
Lag_ROA 9 -0.077 -0.018 -0.013 -0.151 -0.176 -0.215 -0.127 -0.173
-0.110 -0.450 -0.084 0.279
Lag_Special 10 -0.016 -0.001 0.001 -0.040 0.216 0.043 0.136 0.116 -0.174
0.156 0.090 0.052
Lag_Loss 11 0.183 0.111 0.105 0.101 0.176 0.201 -0.234 0.118 -0.700 0.156
0.080 0.026
Lag_ICW 12 0.076 0.024 0.028 0.035 0.019 0.014 0.020 0.030 -0.054 0.090 0.080
0.209
Lag_BTM 13 -0.025 0.048 0.054 0.069 0.214 0.201 0.230 -0.103 -0.340 0.090 0.177 -0.037
a Bold text indicates significance at the 0.05 level or better, two-tailed.
118
Table 2.16
OLS Regression Analysis of exotic features of bonds on Adoption of the FVO
(Model 4-2a)
Variable Estimate Std Error t-value p -value
Intercept*** 1.934 0.330 5.860 <.0001
FVO 0.107 0.106 1.000 0.318
After 0.002 0.163 0.010 0.989
After_FVO 0.007 0.142 0.050 0.962
Lag_LN_TA -0.026 0.020 -1.260 0.207
Lag_Debt* -0.347 0.182 -1.900 0.058
Lag_ROA** -0.823 0.327 -2.510 0.012
Lag_Special 0.042 0.085 0.500 0.618
Lag_Loss 0.018 0.110 0.170 0.867
Lag_ICW 0.114 0.168 0.680 0.498
Lag_BTM -0.014 0.009 -1.510 0.131
Industry dummies Included
Year dummies Included
Adj. R-sq 0.3474
F-statistic 6.73
n 410
Dependent variable Exotic
*, **, *** indicate statistical significance at the 10, 5, 1 percent level respectively.
119
Table 2.17
Logistic Regression Analyses of exotic features of bonds on Adoption of the
FVO
Variable Model 4-2b Model 4-2c
Estimate Pr > ChiSq Estimate Pr > ChiSq
Intercept -12.417 0.950 -1.408 0.306
FVO 0.141 0.517 0.161 0.446
After -0.002 0.996 -0.042 0.903
After_FVO -0.144 0.639 -0.131 0.663
Lag_LN_TA 0.098 0.239 0.095 0.248
Lag_Debt 1.581 0.052 1.287 0.107
Lag_ROA 4.872 0.489 0.971 0.803
Lag_Special -0.182 0.602 -0.138 0.687
Lag_Loss 0.859 0.148 0.615 0.271
Lag_ICW 0.526 0.476 0.480 0.513
Lag_BTM -0.973 0.031 -0.815 0.052
Industry dummies Included Included
Year dummies Included Included
LR ch2 (DF: 38) 151.044 139.299
LR Pr>ch2 <.0001 <.0001
n 410 410
Dependent variable Exotic_d High_Exotic
*, **, *** indicate statistical significance at the 10, 5, 1 percent level respectively.
120
Appendix A for Chapter 2
Examples of FVO adoptions
SFAS 159 provides a fair value option that allows firms to irrevocably elect fair
value for the initial and subsequent measurement of certain financial assets and liabilities
on a contract-by-contract basis. If the fair value option is elected for an instrument, SFAS
159 specifies that all subsequent changes in fair value for that instrument shall be reported
in the income statement. Also, SFAS 159 requires that the difference between the carrying
value before election of the fair value option and the fair value of these instruments be
recorded as a cumulative-effect adjustment to the opening balance of retained earnings in
the period of adoption. In this Appendix, I present examples of FVO adoptions by several
firms.
Community Central Bank Corp. (Financial firm)
Community central bank adopted FVO in 2007 for various financial instruments
such as investment securities (including agency debentures and short callable bank qualifi
ed tax exempt municipal bonds), interest rate swap hedging securities, Federal home loan
bank advances, subordinated debentures, interest rate swap hedging subordinated debentu
res, and redeemed subordinated debentures. It made an adjustment of $420,000 to the
beginning balance of retained earnings and recognized a gain of $1,392,000 in earnings.
The following statements and notes are excerpts from the 2007 annual report of
Community central bank.
121
“While not required to adopt the new standard until 2008, the Corporation elected
to adopt it in the first quarter of 2007. …. As a result of the Corporation's adoption
s, certain financial instruments were valued at fair value using the fair value optio
n. The cumulative reduction to opening retained earnings from adopting these stan
dards was approximately $420,000. …. The following table shows the balance
sheet effect of the adoption of SFAS 159.” (p. 39)
122
“The table below contains the fair value measurement at December 31, 2007 using
the identified valuations. Additionally, the changes in fair value for the twelve
month period ended December 31, 2007 for items measured at fair value pursuant
to election of the fair value option.” (p. 41)
NorthStar Realty (financial firm)
NorthStar adopted FVO in 2008 for third party available for sale securities,
exchangeable senior notes, bonds payable and liabilities to subsidiary trust issuing
preferred securities. It made an adjustment of $190.766 million to the retained earnings and
recognized a gain of $782.3 million in earnings. The following statements and notes are
excerpts from the 2008 annual report of NorthStar realty.
123
“We adopted SFAS No. 159 effective January 1, 2008 and have elected to fair value
for our third party available for sale securities, our exchangeable senior notes, our
N-Star bonds payable and our liabilities to subsidiary trust issuing preferred
securities.” (p. 33)
124
“Our available for sale securities that serve as collateral for our term debt
transactions, which we have elected the fair value option pursuant to SFAS 159, are
carried at fair value with the net unrealized gains or losses reported as a component
of earnings in the statement of operations.” (p. 34)
“With respect to derivative instruments that have not been designated as hedges, or
are hedges on debt that is remeasured at fair value pursuant to SFAS 159, any net
payments under, or fluctuations in the fair value of, such derivatives are recognized
currently in income. …. In January 2008, we adopted SFAS 159 and elected the fair
value option for our bonds payable and its liability to subsidiary trusts issuing
preferred securities. …. As a result of this election, the interest rate swap
agreements associated with these debt instruments no longer qualify for hedge
accounting, in accordance with SFAS 133 ‘‘Derivatives and Hedging Activity’’,
since the underlying debt is remeasured with changes in the fair value recorded in
earnings. The unrealized gains or losses accumulated in other comprehensive
income, related to these interest rate swaps, will be reclassified into earnings ….”(p.
36)
“Unrealized Gain (Loss) on Investments and Other: Unrealized gain (loss) on
investments and other increased by approximately $756.6 million for the year
ended December 31, 2008 to a gain of $752.3 million, compared to a loss of $4.3
million for the year ended December 31, 2007. The unrealized gain on investments
for the year ended December 31, 2008 consisted primarily of unrealized gains
related to SFAS 159 mark-to-market adjustments of $958.2 million on various N-
Star bonds payable, $103.2 million on exchangeable senior notes and $95.2 million
on liability to subsidiary trusts issuing preferred securities offset partially by
unrealized losses related to SFAS 159 mark-to-market adjustments of $338.0
million on various N-Star available for sale securities, $30.0 million on our
corporate lending joint venture (which is an equity investment that is marked to
market) and $36.3 million on interest rate swaps as a result of these swaps no longer
qualifying for hedge accounting under SFAS 133.” (p. 42)
“The following table presents a summary of existing eligible financial assets and f
inancial liabilities for which the fair value option was elected on January 1, 2008 a
nd the cumulative-effect adjustment to retained earnings recorded in connection w
ith the initial adoption of SFAS 159.” (p. F-25)
125
ADM Tronics (non-financial firm)
ADM adopted FVO in 2008 for investments in associated companies which are
subject to the equity method of accounting. As a result of the FVO adoption, it recognized
a loss of $10,660,000 from changes in the fair value of the investments in Ivivi in 2008 and
made a pre-tax cumulative-effect adjustment of $ 9,220,483 to the beginning balance of
retained earnings. The following statements and notes are excerpts from the 2008 annual
report of ADM Tronics.
“As a result, our investment in Ivivi was reported during the period from October
18, 2006 until March 31, 2008 under the equity method of accounting, whereby we
126
recognized our share of Ivivi’s earnings or losses as they are incurred. Effective
April 1, 2008 (“the Adoption Date”), we have adopted SFAS No. 159 “The Fair
Value Option for Financial Assets and Liabilities” with respect to our investment in
Ivivi, whereby we report our investment in Ivivi at fair value. …. The fair value of
our investment in Ivivi at the adoption date was approximately $11,375,000. The a
doption of SFAS No. 159, with respect to our investment in Ivivi, resulted in the re
cognition of the following:
Pre-tax cumulative-effect adjustment to retained earnings: $ 9,220,483
Deferred tax liability: 2,425,188
Post-tax cumulative-effect adjustment to retained earnings: $ 6,795,295.” (p. 34)
Five star quality care, Inc. (non-financial firm)
Five Star Quality Care adopted FVO in 2008 for a put right for auction rate
securities. No adjustment to retained earnings was made because the put right was obtained
during 2008. Five Star Quality Care recognized a gain of $11,081,000 as a result of the
FVO adoption. The following statement and notes are excerpts from the 2008 annual report
of Five star quality care.
“The put right is recorded at estimated fair value under the Financial Accounting
Standards Board’s, or FASB, No. 159, ‘‘The Fair Value Option of Financial Assets
and Financial Liabilities’’, or SFAS 159. We calculated this value based on the
127
difference between the fair value for the ARS and the par value of the ARS, which
is the amount we expect to receive from UBS beginning June 2010.” (p. 54)
“During the fourth quarter of 2008, we elected the fair value option available under
SFAS 159 solely for valuing the put right received from UBS related to our
investments in ARS.” (p. F-12)
“We have accounted for our put right as a freestanding financial instrument and
elected to record it at its estimated fair market value under the fair value option of
SFAS 159. As a result, we recorded an $11,081 gain related to receipt of this put
right and recorded its value in other long term assets.” (p. F-22)
Discussions
The actual examples of FVO adoptions presented above indicate that firms adopt
FVO for various financial instruments. Some might argue that FVO may not significantly
affect the asset side of balance sheet especially for non-financial firms since investments
in equity securities had been already measured at fair value even before FVO. However,
because FVO can be applied to almost all financial instruments and there are no qualifying
criteria imposed, it is possible that adopting FVO for certain financial instruments, such as
available for sale securities, held to maturity securities, or various liabilities, makes a
significant impact on earnings or retained earnings. The examples presented above confirm
the possibility. Therefore, by adjusting the timing of adopting FVO, firms might try to
affect the current or future earnings.