Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
The Value Relevance of the Fair Value Hierarchy of FAS 157
A Dissertation
Submitted to the Graduate Division
of the
University of Hawai‘i at Mānoa
in Partial Fulfillment of the Requirements for the Degree of
Doctor of Philosophy
in
International Management
May 2012
by
Eric C. Wen
Dissertation Committee:
Hamid Pourjalali, Chairperson
Shirley J. Daniel
Boochun Jung
Qianqiu Liu
Sumner La Croix
ii
All Rights Reserved
iii
Acknowledgements
I owe deepest gratitude to my dissertation chair, Hamid Pourjalali, who encouraged me to
apply to the doctoral program, taught me everything I know about accounting theory,
introduced me to many of his friends, and provided me not only with countless cups of
coffee and quiet places to work, but also with encouragement and sage advice that
shepherded me through the entire process. I also thank the members of my committee,
Shirley Daniel, Boochun Jung, Qianqiu Liu, and Sumner La Croix for their comments,
encouragement, and support throughout the process. I especially thank Shirley for her
support, through the Center for International Business Education and Research (CIBER),
that allowed me to participate in the 2009 Annual Meeting of the American Accounting
Association (AAA) in New York where I heard Wayne Thomas present the work that
would become Song, Thomas, and Yi (2010); Boochun for his careful help with the
statement of some of my hypotheses and for hosting several doctoral candidates to the
most delicious and varied Korean barbecue dinner that I have ever had where we enjoyed
the evening under a huge tree; Qianqiu for leading the finance seminar that introduced
doctoral candidates to several seminal papers and for always welcoming me into his
office with a warm smile, and Sumner for his wealth of insights especially on research
methods and industrial organization, and for his very careful reading of earlier drafts of
this dissertation.
I thank my dear wife, Aida, for her constant support and encouragement with
everything, really. I could write another dissertation, longer than this one, on all of the
iv
adventures we have shared together. I also thank my parents and three daughters1 for
their support, encouragement, and prayers. In addition, I also thank the members of my
church, International Baptist Fellowship, for their prayers and support. Finally, I thank
the International Baptist Ministries for hosting the Thursday Lunch, right across the street
from the Shidler College of Business, which not only fed my body with a balanced hot
meal during the semester, but also provided an open venue for spiritual refreshment
through the community of international scholars and friends who gathered there.
1 At the time of this writing, our son is only saying a handful of recognizable words and wobbling around.
v
Table of Contents
Acknowledgements ............................................................................................................ iii
List of Tables ..................................................................................................................... vi
Table of Figures ................................................................................................................ vii
ABSTRACT ..................................................................................................................... viii
1. Introduction ................................................................................................................. 1
2. Review of Literature .................................................................................................... 4
2.1 Review of Literature on the FVH of FAS 157 ..................................................... 4
2.2 Overview of the Literature on Characteristics of Industries .............................. 10
3. Development of Hypotheses ...................................................................................... 16
H1. Book-to-Market Equity ......................................................................................... 18
H2. Industry Concentration .......................................................................................... 19
H3. Industry Status....................................................................................................... 20
H4. Liquidity ................................................................................................................. 21
H5. Relative Amount ................................................................................................... 21
3.6 Industry Classification ........................................................................................... 22
4. Sample Selection ....................................................................................................... 24
5. Results ....................................................................................................................... 27
5.1 The Value Relevance of the FVA across Industry Sectors ................................ 33
5.2 Results of Tests of H1 ........................................................................................ 36
5.3 Results of Tests of H2 ........................................................................................ 38
5.4 Results of Test of H3 .......................................................................................... 42
5.5 Results of Tests for H4 ....................................................................................... 45
5.6 Results for H5 ..................................................................................................... 48
6 Sensitivity Analysis ................................................................................................... 49
7 Conclusion ................................................................................................................. 52
References ................................................................................................................... 179
Appendix A ..................................................................................................................... 182
Appendix B. Computation of the Herfindahl of Sales ................................................... 187
vi
List of Tables
Table 1. Four (Known) Papers that Examine the Fair Value Hierarchy (FVH) ............... 61
Table 2. The Main Characteristics of the Fair Value Hierarchy (FVH) of FAS 157 ...... 62
Table 3. Datasets of the Four Early Papers ....................................................................... 63
Table 4. Summary of Hypothesis and Results ................................................................. 65
Table 5 Papers examining characteristics of industries in more depth ............................ 70
Table 6 Summary of Main Results of Selected Papers ..................................................... 72
Table 7. Steps Taken to Develop the Sample .................................................................. 76
Table 8. Summary of Selected Variables by GICS Sectors .............................................. 82
Table 9. Results of OLS Regression Using Winsorized Variables .................................. 86
Table 10. Results of Iteratively Reweighted Least Squares (IRLS) Regression ............. 91
Table 11. Results of OLS Regression on Trimmed Sample ............................................ 94
Table 12. Investigating the Effect of BE/ME .................................................................. 98
Table 13. Investigation of the Effect of Industry Concentration (HHI) ........................ 111
Table 14. Three-year Average Herfindahl of Net Sales ................................................ 123
Table 15. Entering, Incumbent, and Exiting Firms ........................................................ 134
Table 16. Investigation of the Effect of Liquidity, Using the Quick Ratio (QR) and
Operating Cash Flow Ratio (CR) .................................................................................... 142
Table 17. Investigating the Value Relevance of the FAS 157 Ratio ............................. 153
Table 18. Sensitivity Analysis ....................................................................................... 155
vii
Table of Figures
Figure 1: Timeline of the Issuance and Effective Date of FAS 157 ................................. 57
Figure 2. Weekly Close of the S&P 500 Index from Jan 1, 2007 to Jan 1, 2012 ............. 58
Figure 3 Sample Disclosures Required by FAS 157 ........................................................ 59
viii
ABSTRACT
Financial Accounting Standard (FAS) 157 requires disclosures of the fair value hierarchy (FVH)
of assets and liabilities for financial reports issued after November 15, 2007. This dissertation
examines the value relevance of the FVH across a set of industry sectors. The results suggest that
the FVH is value relevant for all industry sectors. However, although the degree of value
relevance varies across industry sectors, the relative amount of assets and liabilities measured at
fair value, as a ratio to total assets and liabilities, does not influence the value relevance of the
FVH. This study also examines the influence of several other characteristics of firms on the
value relevance of the FVH across industry sectors. Its main findings suggest that book-to-market
equity and a firm's status (as entering or incumbent) influence the value relevance of the FVH.
The dissertation also finds some evidence that a firm's size and firm's status are related and that
the firm's size may also influence the value relevance of the FVH. This dissertation is unable to
find a clear pattern of the influence of liquidity (as measured by both the quick ratio and cash
flow from operations) on the value relevance of the Level 3 fair value of assets across industry
sectors.
1. Introduction
The purpose of accounting is to provide useful information to decision-makers. The early users
of accounting deemed reporting net income useful information and therefore adopted an
“Income Statement Approach” to financial reporting. However modern standard-setters view
valuation as useful information and advocate a “Balance Sheet Approach.” One of the key
differences between the two approaches is the preference of methodology for measurement of
assets and liabilities. While the traditional “Income Statement Approach” prefers (depreciated)
historical cost, the “Balance Sheet Approach” advocates measurement by fair value (Dichev
2008). The debate over the merits of measurement by historical cost vs. fair value is not new to
accounting. Laux and Leuz (2009) and Penman (2007) provide recent contributions to this long-
standing conceptual debate. Essentially, the debate revolves around the concepts of relevance vs.
reliability: Proponents argue that fair value measurement (FVM) has greater relevance to users
of financial information, while opponents counter that FVM is less reliable, especially when no
market exists.2
Notwithstanding the philosophical debate for or against it, prior to 2008, generally
accepted accounting principles in the United States (U.S. GAAP) included multiple definitions
for FVM that were scattered across various pronouncements. The lack of consistency and
comparability resulting from the multiple definitions of FVM provided challenges to users of
financial statements with FVM. In response to requests from constituents on this issue, the
Financial Accounting Standards Board (FASB) began its standard-setting procedure by publish-
ing an exposure draft on June 23, 2004 that sought to improve guidance on measuring fair value
(FASB 2004). The FASB followed its standard-setting procedure which it completed in the fall
2 When applied to the balance sheet, the term “mixed attribute model” means that certain items are
measured at (depreciated) historical cost while others are measured at market value.
2
of 2006 with the issuance of Financial Accounting Standard (FAS) 1573 that provided a single
definition of fair value (FASB 2006). “This statement defines fair value, establishes a
framework for measuring fair value, and expands disclosures about fair value measurements”
(FAS 157 paragraph 1). Although the FASB subsequently delayed by one year its application to
nonfinancial assets and liabilities measured or disclosed on a non-recurring basis, for all other
assets and liabilities FAS 157 became a part of the U.S. GAAP for fiscal years beginning after
November 15, 2007 (FASB 2008). This timeline is illustrated in Figure 1.
[Insert Figure 1 about here]
[Insert Table 1 about here]
Four early papers, listed in Table 1, study the value relevance and information content of
the main disclosures of FAS 157. The papers, which report generally consistent results, share
two other common features as well: All used disclosures from unaudited interim quarterly
financial reports before the final accounting period of 2008 and all examined firms from a single
industry: financial services. While some of their results may generalize to firms in other
industries, because firms in the financial services have characteristics that differ from firms in
other industries, some results may be unique to firms in financial services. Therefore, which of
the findings of the early papers apply more broadly to firms in other industries and which are
unique to the financial services industries is an open, important, and interesting research
question.
This dissertation will extend the examination of the value relevance of the fair value
hierarchy (FVH) more broadly and beyond firms in financial services. This extension is
important since certain findings of the early papers may be unique to firms in the financial
services. Through FAS 157, the FASB has standardized the measurement and reporting of the
FVH for all firms across all industries. However the market assigns value based on discounted
3 Under the FASB’s new codification, FAS 157 has become Accounting Standard Codification (ASC)
820.
3
expected future cash flows and other characteristics that can be very different between firms and
industries. Therefore, the value relevance of the FVH could vary among industries.
On the other hand, the value relevance of the FVH may be influenced by similar
variables or common characteristics. For example, Goh, Ng, and Ow Yong (2009), one of the
early papers, found evidence that suggests that firms with higher liquidity (banks with higher
Tier 1 capital ratio) have higher estimated coefficients of Level 3 assets. They attributed this to
the fact that FAS 157 defines fair value measurement as an exit price in the asset’s or liability’s
most favorable market, and firms with higher liquidity are more likely to be able to realize it in
its most favorable market. While the early papers used (unaudited) interim reports, this
dissertation examines the value relevance of the (audited) fiscal year-end reports from the first
fiscal year in which FAS 157 became U.S. GAAP. Furthermore, the sensitivity analysis extends
the findings to the second year (2009) after FAS 157 became GAAP. Due to the volatility of
financial markets during this period, this dissertation addresses an interesting research question
by examining whether patterns of value relevance observed by the early papers persisted through
the reporting period of the (audited) year-end data.
It is important to note that, “In 2008 the financial markets froze” (Spiegel 2011). Figure
2 shows the weekly close of the S&P 500 Index, a well-known indicator of the equity markets in
the U.S. The figure shows that the impact of the crisis on the equity markets in the U.S. was
most severe at the end of 2008 and beginning of 2009. For the study of value relevance, this
study will use prices from March 31, 20094. Because prices are thought to be based on expecta-
tions of future performance, and because the equity markets may have begun to recover from the
crisis by that date, prices of equities on this date may have less influence from the crisis.
However, the fact that the accounting data from this study comes from the period of crisis should
be kept in mind, and this could have an influence on the results.
4 Studies using returns of equities that span this period, the end of 2008 and beginning of 2009, may need
to exercise caution.
4
2. Review of Literature
This section provides a review of the literature on the FVH of FAS 157 followed by a review of
the literature on characteristics of industries.
2.1 Review of Literature on the FVH of FAS 157
[Insert Table 2 about here]
One of the expanded disclosures required by FAS 157 is the three-tiered fair value hierarchy
(FVH) based on the type of input to determine the FVM. “The fair value hierarchy gives the
highest priority to quoted prices (unadjusted) in active markets for identical assets or liabilities
(Level 1) and the lowest priority to unobservable inputs (Level 3)” (FAS 157 paragraph 22).
Table 2 displays the main characteristics of the FVH. Several recent papers have taken an early
look at the value relevance of and mitigation of information risk by the FVH5 of FAS 157. This
section describes the main results of four (known) early papers on the value relevance and
information content of the fair value hierarchy (FVH) of FAS 157. Table 1 lists the papers and
the abbreviations that will be used for each paper.
[Insert Table 3 about here]
Panel A of Table 3 provides an overview of the data used in each paper. These early
studies use (unaudited) interim financial reports from 20086 and examine firms in the financial
services industries. Goh, Ng, and Ow Yong (2009), hereinafter GNO, restrict their investigation
to firms with Standard Industrial Classification (SIC) beginning with 60 (Depository institutions)
5 In addition to the FVH, Kolev also uses FAS 157’s Level 3 Reconciliation disclosure. 6 One paper, Riedl and Serafeim (2009), also includes data from 2007.
5
or 61 (Nondepository credit institutions)7. Kolev (2009) focuses on Industry 40 of the Global
Industry Classification Standard (GICS), which are the Financials, but removes GICS Industry
4040 which are real estate firms. Kolev’s study uses members of the “S&P 1500” that is
comprised of the firms in the following three constituent indices: S&P 500, S&P MidCap 400,
and S&P SmallCap 600. Therefore, Kolev’s study does not include the smaller firms that are
not included in the S&P SmallCap 600 but are included in GNO and Song, Thomas, and Yi
(2010), hereinafter STY. Riedl and Serafeim (2011), herein after RS, take a different approach
and concentrate on large firms (total assets in excess of $10 billion) that are in three particular
financial services industries (SIC 6020, 6035, and 6211). RS have the fewest firms in their
initial sample (56). Finally STY draw their sample from the Compustat Bank dataset.
Panel B of Table 3 also shows that all four papers include accounting data from the first
and second quarters of 2008. Kolev uses data from only these two quarters. Both GNO and
STY include data from the third quarter of 2008, while RS augments the two quarters in 2008
with data from all four quarters of 2007. None of the four papers includes data from the fourth
quarter of 2008 or from 2009. Although the main contribution of this project is to understand
the value relevance of the FVH across industries, the fact that the data used is audited year-end
data for the first year in which FAS 157 became GAAP will allow us to see whether the pattern
observed using the earlier interim reports is also present in the more recent year-end reports as
well.
Panel A of Table 3 also summarizes a few other characteristics of the different data sets,
including the number of distinct firms (N), the number of firm-quarters, the method of handling
outliers, and the method of handling the FVH disclosures. In particular, both GNO and Kolev
subtract Level 1(2)[3] liabilities from the corresponding assets and use net Level 1(2)[3] assets
in their estimations. RS and STY use the “raw” information as-is, while in one series of tests,
7 Website of the 1997 Economic Census, http://www.census.gov/epcd/ec97sic/E97SUSH.HTM, accessed
July 15, 2011
6
RS form the “sum ratio” as the sum of Level 1(2)[3] assets and Level 1(2)[3] liabilities divided
by the sum of assets and liabilities. A more detailed review of these papers follows.
[Insert Table 4 about here]
GNO seek to understand whether investors use the information in the FVH, and whether
investors weigh the information in the FVH differently across banks. Panel A of Table 4
summarizes GNO’s hypotheses, main tests, and findings. As shown in the Panel, their first
hypothesis relates to their first goal of seeking to understand whether investors use information
provided in the FVH. Their findings, which support their hypothesis, suggest that the
information provided by the FVH is value relevant. More specifically, their regression analyses
provide monotonically decreasing values of their estimated FVH coefficients. However, GNO’s
results suggest that the coefficients of Level 2 and Level 3 net assets are not statistically
different. They note, however, that “the difference in the coefficients (0.626 versus 0.489)
appear [sic] to be economically significant” (GNO p. 19).
Their second and third hypothesis examine whether investors weigh the information
provided by the FVH differently across banks. They partition the sample with two indicator
variables. One, based on the bank’s Tier 1 Capital Ratio, which is a measure of its financial
strength, has the value 1 when the Capital Ratio is above the median, and 0 otherwise. The
value of the other indicator variable is based on whether the bank’s auditor is a Big Four auditor
or not. Their findings support their hypotheses and suggest that investors do weigh the
information provided by the FVH differently among banks. In particular, they found the
estimated coefficient of the interaction between Tier 1 Capital Ratio and Level 3 net assets is
statistically significant, suggesting that the market perceives banks with higher financial strength,
and liquidity, as more likely to realize its Level 3 net assets as measured and reported. Finally,
GNO find statistically significant coefficients on the interaction between Big Four auditor and
all three net assets, with the coefficients of the interaction between Level 2 and 3 net assets and
7
the Big Four indicator being highly significant. This suggests that the presence of a Big 4
auditor helps reduce information risk associated with Level 2 and Level 3 net assets. This is
summarized in Panel A of Table 4.
Kolev has two hypotheses. His first posits that investors find Level 2 and Level 3
measurements sufficiently reliable to be reflected in firm value while his second is that investors
find Level 2 and Level 3 estimates less reliable than Level 1. Kolev’s analysis is quite technical.
First, he develops a set of controls that include industry indicators within the GICS financials,
company-level credit ratings, and proxies for size, growth, and profitability. Kolev’s first test
suggests that investors find the quantitative FVH disclosures informative. Similar to GNO, the
magnitude of the coefficients of the FVH regressors are monotonically decreasing. Kolev
performs a Wald test to assess equality among the coefficients that suggests in almost every
specification, the coefficient of the Level 3 net assets is different from the coefficients of the
Level 1 and Level 2 net assets. In order to control for the possibility of correlated omitted
variables, Kolev differences the level model and incorporates information from the Level 3
Reconciliation, another new disclosure introduced by FAS 157. Panel B of Figure 3 provides an
example of the Level 3 Reconciliation disclosure. Where statistically significant, the results are
consistent with earlier findings. Panel B of Table 4 summarizes Kolev’s results.
While the papers above examine the value relevance of the FAS 157 FVH disclosures,
RS begin from a different point. They observe that finance theory suggests that information risk,
the uncertainty regarding valuation parameters of an asset, is reflected in a firm’s equity beta and
the information asymmetry component of bid-ask spreads. RS first estimate the firm’s equity
beta using the single-factor CAPM, and then regress this beta on FVH Level data and controls.
RS use the FVH disclosures as reported and in particular do not compute net assets for each
level of the FAS 157 FVH. The coefficients of the fair value of assets increase monotonically
and F-tests suggest that the coefficient of Level 3 assets is significantly higher, suggesting more
8
risk, than Levels 1 and 2. Although the coefficients of liabilities do not exhibit monotonicity,
nevertheless the coefficient of Level 3 liabilities has the largest magnitude, and F-tests suggest
the coefficient is different from the others. In their model for the cost-of-capital, RS include the
FAS 157 FVH disclosures in “absolute form.” The numerator is the sum of the absolute value of
assets and liabilities deflated by the sum of the absolute value of total assets and total liabilities.
Their results suggest that the coefficient of Level 3 fair values is positive and statistically
significantly while coefficients for Level 1 and Level 2 fair values are not statistically significant.
This is summarized in Panel C of Table 4.
STY extract a sample from the Compustat Bank file for the first three quarters of 2008.
They eliminate observations that have studentized residuals greater than two as outliers and
examine the value relevance of the FVH by regressing stock price onto accounting information
including the FAS 157 FVH. They find that the coefficients of FVH decrease monotonically. In
particular, the coefficient of Level 3 assets is positive, and small which suggests “that investtors
place less weight on Level 3 fair value assets relative to Levels 1 and 2” (STY p. 18), and the
coefficient of Level 3 liabilities is significantly less than minus one, which “is consistent with
investors perceiving Level 3 fair value liabilities to be understated” (STY p. 18). Furthermore,
FAS 157 requires disclosure of the type of asset or liability at each level. Vuong’s statistic
applied to the R2 of regressions with and without the additional type-of-asset and liability
information, “suggests that when the Level information is used together with the Type
information, FAS 157 Level information is incrementally value relevant” (STY p. 20).
Lastly, STY’s final tests suggest that corporate governance is value relevant, with the
greatest impact on the valuation of Level 3 fair value assets. They suggest that because
management has the greatest ability to bias the measurement of Level 3 assets, investors “trust”
firms with better corporate governance more than others. They carefully apply techniques
appropriate for scale effects, as described in the recent paper by Barth and Clinch (2009). As a
9
part of their sensitivity analysis, STY also include a dummy variable for high and low Tier 1
capital. Their results are robust with this characteristic of banks and are summarized in Panel D
of Table 4.
Thus, the results from these four early papers, which are largely consistent, suggest that
the disclosures of the FVH are value relevant and contain information, and in particular that the
coefficients of the Level 1 holdings8 are significantly different from the coefficients of the Level
3 holdings for firms in the financial services industries, and especially banks. However, firms in
the financial services industry, and especially banks, are subject to regulations that are not
applied to firms in other industries. For example, regulators from the Federal Deposit Insurance
Corporation (FDIC) routinely visit banks, and if a bank is not maintaining certain regulatory
ratios that are mandated by law, the regulators are required to revoke its charter to operate,
which effectively forces the bank to close. The firms studied by the early papers are regulated,
and some heavily so, suggesting that some of the findings could be unique to regulated firms.
Therefore it is important to examine the value relevance of the FVH of firms in other industries
that are less regulated (e.g., utilities), or unregulated (most other industries).
Accounting literature provides strong evidence that the level of regulation influences
earnings response coefficient (ERC), earnings relationship with stock prices and alignment
between the market value and balance sheet values. Teets (1992) compares the earnings
response coefficient (ERC) of electric utilities and a random sample of non-regulated firms. He
finds that the ERC of electric utilities are smaller on average than non-regulated firms. El-
Gazzar, et al. (2009) examine “the valuation effects of earnings and two non-earnings-based
measurements (book values and operating cash flow) on security prices of airline companies
under two different market structures: regulated and deregulated” (p. 88). Their results suggest
“that nonearnings measures have higher explanatory power of security prices in regulated times
8 Hereinafter the term “holdings” will be used for the collective expression “assets and liabilities.”
10
for the airline firms. In deregulated times, earnings have a stronger relationship with prices” (p.
88). Comparing electric utilities to comparable manufacturing firms, Nwaeze (1998) provides
evidence that the alignment between the market value and the balance sheet book value for
utilities is considerably high. In particular, Nwaeze’s result suggests that the ratio of BE/ME of
electric utilities is close to one.
2.2 Overview of the Literature on Characteristics of Industries
Researchers in accounting and finance have examined the relationship between firms, industries,
earnings, and market valuation. This section provides a brief overview of this rich literature.
The study by King (1966), one of the earliest that examined a firms’ industry effect on capital
markets, uses factor analysis to identify “a smaller set of clusters of security prices changes that
tend to move as homogenous groups” (King p. 139). Using price data from the Center for
Research in Security Prices (CRSP), King studies 63 firms in six two-digit industries9 from May
1927 – December 1960. Based on this sample, the study finds that over half of the variance in
price of a typical stock is explained by elements that affect the entire market and also that on
average over the period, industries influence over 10% of the variance in price. He concludes,
“…the data give remarkable support to the hypothesis stating that the movement of a group of
security price changes can be broken down into market and industry components” (King p. 163).
As a type of “sensitivity analysis,” King (1966) shows remarkable consistency between his
(principal) market factor and the S&P 500. However, in a subsequent methodological study,
Livingston (1977) suggests that the market factor estimated from factor analysis is sensitive to
the set of firms included in the sample, and therefore recommends using a broader index, such as
the S&P 500, in market-based research. Furthermore, he suggests that regression analysis could
be more robust than factor analysis for market-based research.
9 In the paper, King describes the codes as from the SEC. The SEC currently uses SIC codes.
11
In the year following King (1966)’s paper, Brown and Ball (1967) published research
that examines accounting earnings in a manner similar to the way that King had studied security
prices. Using annual earnings information from Compustat, Brown and Ball identify 316 firms
in ten industries based on two-digit industry classification codes10 spanning 1947–1965. Based
on this sample, they report that on average approximately 35-40% of a the variability of a firm’s
annual earnings is due to the variability of all firms while on average 10–15% can be associated
with the industry average. They note, “On the whole, our results parallel those reported by King”
(Brown and Ball 1967, p. 67). In their suggestions for future research, they note that the
classification scheme for industries may be somewhat arbitrary and discuss some possible
alternate methods on which to base different strategies of distinguishing and classifying
industries.
In their review of marked-based accounting research, Lev and Ohlson (1982) note that,
“Existing empirical literature suggests that, at minimum, firms can be usefully grouped into
industries in the construction of valuation models” (p. 308). Reviewing capital markets research
in accounting in the subsequent decade, the 1980s, Bernard (1989) describes several papers that
have concentrated on particular industries, such as oil & gas and financial services, and
encourages more research to understand unique characteristics of specific industries.
In a more recent review article covering the 1990s, Kothari (2001) mentions one paper,
Biddle and Seow (1991), which specifically includes industries in their methodology to estimate
earnings response coefficients. Biddle and Seow group firms into industries because firms
within an industry generally share similar economic, financial characteristics and generally
would make similar choices of accounting methods. They base their 40 industries on the
traditional Standard Industry Classification (SIC) system that will be discussed in more depth
10 In their paper, Brown and Ball describe the codes as Compustat codes. These could be SIC codes.
12
below. Their results suggest that earnings response coefficients vary significantly across
industries.
Kothari (2001) also reviews important papers from finance that impact capital markets
research in accounting. One such paper is Moskowitz and Grinblatt (1999) whose results
suggest that industry is a significant factor in the momentum anomaly. Another such paper is
Fama and French (1997) who, similarly to Biddle and Seow (1991), created their own industry
categories based on the SIC and whose results suggest risk factor loadings on the ratio of book
equity-to-market equity (BE/ME) factor vary substantially across industries. Lang and
Lundholm (1996) results suggest that the earnings of firms within the same industry are
informative to a particular firm.
[Insert Table 5 about here]
Traditionally, many researchers who included industries in their studies have simply
added “industry fixed-effects” to their models. However, recently some researchers have begun
to investigate the properties or characteristics of industries that have measurable economic or
financial effects. Three such papers whose findings could be relevant to this study are shown in
Panel A of Table 5. Panel B of the same table shows a brief comparison of some of the selection
criteria of the samples of each paper. The remainder of this section will discuss some of the
highlights of these papers.
Banko, Conover, and Jensen (2006), hereinafter BCJ, find evidence that suggests the
value effect varies by industry. The value effect refers to the empirical finding that the BE/ME
is significant for returns (Fama and French 1992). BCJ sought to understand the strength of the
value effect both at the firm- and industry-level. They use data from Compustat and CRSP from
1968–2000 and use SIC from CRSP because these SIC reflect the historical classification, while
the SIC from Compustat reflect the most recent classification.
[Insert Table 6 about here]
13
From this sample, BCJ first form 21 industry groups that each have 15 or more firms.
Dropping firms with negative book equity (BE), they compute BE/ME for each firm, rank them
by BE/ME, and form (five) equally-weighted quintiles based on the BE/ME ranking, following
Fama and French (1992). Panel A of Table 6 highlights their main findings. Their table of
summary statistics of BE/ME by industry displays a rather wide variation in average BE/ME
with two industries (apparel and primary metals) in the extreme “value” range ( > 0.9) and three
industries (communications, services, and chemicals) in the extreme “growth” range ( < 0.5). To
examine the relationship between industry and the value effect, they use a generalized least-
squares approach. The results from this preliminary series of models on the entire sample
suggest that the quintile BE/ME is statistically significant. To further examine the prevalence of
the value effect, they separately run regressions for each industry and find that for over half of
the industries (11 of 21) the coefficient of BE/ME is significant, and for an additional fifth of the
industries (4 of 21) is moderately significant. They conclude that the intra-industry variation in
BE/ME is relevant in explaining stock returns. This intra-industry variation could also be
significant in value relevance studies.
By examining the risky cash flows that firms generate by their activities in product
markets and the market’s assignment of value based on those risky cash flows, Hou and
Robinson (2006), hereinafter HR, seek to develop a link between the theories of industrial
organization, e.g. industry concentration, and asset pricing. They develop their main hypothesis
from two different theoretical models. One model is the creative destruction hypothesis of
Joseph Schumpeter (1912) which states that innovation begins in firms outside of an established
industry and can become so successful that the new firms overtake the established ones. One of
the possible consequences of this hypothesis is that concentrated industries have lower returns
because concentrated industries contain more established firms, and with less innovation, and
lower returns. A second theoretical starting-point is the Structure/Conduct/Performance (S/C/P)
14
theory of Joe Bain (1954) which states that a Structure that is particular to an industry, like a
barrier to entry, affect a firm’s Conduct, and a firm’s conduct and decisions affect its
Performance. The presence of a barrier to entry would lead to a concentrated industry.
Protected firms would have less default risk, and since these firms are less risky, their market
return would be lower. From these two starting-points, HR begin to examine the average returns
of industries by concentration.
They develop their sample ordinary common shares11 from NYSE-, AMEX-, and
NASDAQ-listed companies in Compustat and CRSP. They remove regulated industries from
their sample and use the Herfindahl Index as a measure of industry concentration. The
Herfindahl Index is based on a firm’s market share, and they examine three separate ways to
estimate market share: Net Sales, Total Assets, and Book Equity. They observe that these
measures exhibit high correlation and noting that all of the measures are imperfectly correlated
with “true market share,” they use Net Sales as the basis for their Herfindahl Index. Following
Fama and MacBeth (1973), they create quintile portfolios based on their Herfindahl Index and
estimate pooled regressions. The coefficients from their regressions support their theoretical
predictions.
To establish their main results, they simply tabulate summary statistics for their quintile
portfolios. Their results show that in competitive industries (where the Herfindahl Index is low)
the average return is higher, while in concentrated industries (where the Herfindahl Index is high)
the average return is lower. HR suggest that
“It is well understood from industrial organization that the structure of product
markets affects managers' equilibrium operating decisions. If these operating
decisions affect the risk of a firm's cash flows, then these decisions should
impact stock returns. The main finding in this paper is that firms in highly
concentrated industries earn lower returns, even after controlling for size, book-
to-market, momentum, and other known return predictors. Moreover, the
economic magnitude of these effects is large.” (p. 1928).
11 CRSP assigns ordinary common shares a Share Code of 10 or 11.
15
These results suggest that the relationship between a firm’s holdings and its market
value may be influenced by the industry to which it belongs. In particular the coefficients of
Level 1, Level 2, and Level 3 holdings in a value relevance regression may be influenced by the
industry to which the firm belongs, providing a stronger argument that the results of the early
papers on the value relevance of the FVH could be unique to the financial services industries.
MacKay and Phillips (2005), hereinafter MP, begin with a review of theoretical models
of firms in equilibrium and partial-equilibrium. Instead of developing specific tests for the
different models, they identify and explore the common theme of these models: firms within an
industry behave differently. In particular, they investigate three types of decisions that firms
make: financial structure (debt structure), technology, and risk decisions. To examine a firm’s
technology within its industry, MP use Maksimovic and Zechner (1991)’s concept of a “natural
hedge” which measures how a firm’s technology compares to the rest of its industry. Then,
following Williams (1995) and Fries, Miller, and Perraudin (1997), MP develop a second
measure of a firm’s position within its industry that is based on its status as an entering,
incumbent, or exiting firm.
They develop their sample from active and inactive firms in Compustat and CRSP. In
order to compute diversification Herfindahls, they merge in Compustat’s business segment files,
for the years 1981–2000. Because the theoretical methods they use were developed for firms
with classical notions of capital and labor, they limit their sample to manufacturers with SIC
from 2000–3990. Furthermore, since the theories were developed for firms in competitive
industries, they use the Herfindahl-Hirshman Index (HHI)12 which provides a measure of the
level of competition within an industry, and they select unconcentrated firms with HHI < 1000.
For comparison, they also separately study concentrated firms, with HHI > 1800.
12 The HHI is a part of the Census of Manufactures that is conducted every five years that is published by
the U.S. Census Bureau’s Economic Census.
16
Their first analysis, “a table of subsample means” reported in Panel C of Table 6, clearly
shows that within an industry, firms exhibit characteristics based on their position as an entering,
incumbent or exiting firm. For example, while incumbents carry the lowest amount of relative
debt, entering firms typically have slightly higher amounts of debt, and firms exiting have the
highest debt. Their second analysis consists of a series of regression models with the firm
characteristics (financial structure, technology, and risk decisions) as the dependent variable.
They rely on differences in adjusted R2 to establish their result. In their first set of regressions
they include only industry fixed effects, while in their second they add firm fixed effects. They
find that for each dependent variable, the adjusted R2 of the second set of regressions is higher
than the first set. This clearly suggests that industry fixed-effects alone do not explain firm
characteristics.
They report the results of ordinary least squares (OLS) regressions, but to control for firm
fixed effects and address endogeneity bias, following Whited (1992), they estimate a system of
simultaneous equations using the generalized method of moments (GMM) with first-differences
as the instrument variables. The results from these more rigorous econometric techniques
support their main findings. Their finding suggests that a firm’s position within its industry
differentially affect their coefficients in a value relevance model.
3. Development of Hypotheses
The Efficient Market Hypothesis (EMH) states that security prices fully reflect available
information (Fama 1970, Kothari 2001). This simple assertion has been qualified into several
different forms, e.g. the weak, semi-strong, and the strong, and has been the subject of numerous
studies (Fama 1991, Kothari 2001). Because the purpose of accounting is to provide
information that is useful and the EMH asserts that security prices fully reflect information
(including disclosures of the FVH), therefore studies of value relevance build upon these
foundations by examining whether participants in the equity markets, one of the largest types of
17
users, respond to accounting information, using the change in price of an equity issue as a
measure of their response (Barth, Beaver, and Landsman 2001, STY).
Furthermore, the Capital Asset Pricing Model (CAPM) attributes variation in the
expected return of securities to differences due to the riskiness of individual firms (Kothari
2001). Empirical tests of the static-CAPM have shown difficulty finding evidence to support it
(Fama and French 1992). However a modified conditional-form of CAPM has received
empirical support, and a framework for associating risk with returns, the CAPM is very widely
used (Jagannathan and Wang 1996). For the purpose of developing our hypotheses (below), we
will restate the CAPM’s risk-return principle as investors dislike risk and therefore will demand
a premium to assume risk.
Prior to 2008 the generally accepted accounting principles in the United States (U.S.
GAAP) contained multiple definitions of fair value that were scattered across various
pronouncements. In September 2006, after working with constituents over the course of several
years, the Financial Accounting Standards Board (FASB) issued Financial Accounting Standard
(FAS) 157 that addressed this issue. FAS 157, which became effective for fiscal 2008, defines
fair value as an exit price in an orderly transaction between independent (unrelated) and
knowledgeable market participants in the principal or most advantageous market for the asset or
liability, and introduces several new disclosures, one of which is the fair value hierarchy (FVH)
that classifies fair value measurements (FVM) into three categories based on inputs: Level 1
inputs are from quoted prices for identical assets or liabilities that are traded in an active market;
Level 2 inputs are other-than Level 1 inputs but are observable, either directly or indirectly, and
for example could be observable prices for similar assets or liabilities; and Level 3 inputs are
unobservable for the asset or liability.
Prior research examined the value relevance and information content of the FVH of
firms in the financial services industries. One of the main findings of these efforts is that in
18
value relevance models of firms in the financial services industries, the coefficients of the FVH
are monotonically decreasing with the coefficient of Level 1 (3) assets largest (smallest), and
furthermore the coefficient of Level 3 assets is statistically different from the coefficients of
Level 1 and 2 assets. Another significant result is that the coefficient of Level 3 assets is higher
(lower) for banks with higher (lower) Tier 1 capital ratios, which suggests that investors
incorporate either liquidity or financial strength in their assignment of value to banks.
Although FAS 157 requires firms to disclose the information of the FVH in a consistent
manner and the EMH asserts that security prices fully reflect all information, nevertheless the
CAPM predicts that the investors assign values to firms differently based on their perception of
the risks associated with a particular firm. Thus, a tension may exist between the consistent
information firms provide through the disclosures of the FVH vs. the differing valuation the
market assigns them based on their individual riskiness. Because valuation can be different due
to a firm’s unique risk factors, this dissertation investigates the influence of certain risk factors
on the coefficients of the FVH in value relevance models.
The literature review provided theoretical and empirical evidence that market
participants consider the industry to which the firm belongs as well as other risk factors when
setting a value for a firm. In the following paragraphs, I will expand the possible effect of
industry and other risk factors on the value relevance of assets reported using the FVH. A priori,
I expect that the market will penalize (reward) value (growth) firms and more (less) risky
industries with lower (higher) valuation of the assets reported at fair value and expect to observe
the effects of this lower (higher) valuation in the relative magnitudes of the coefficients of Level
1, Level 2, and Level 3 assets.
H1. Book-to-Market Equity
My first hypothesis posits that BE/ME contains significant explanatory power in the
value of the firm. Banko, Conover, and Jensen (2006) examine how stocks are valued and find
19
evidence that valuation varies by industry. They form 21 industry groups for which BE/ME
displays a wide variation with two industries (apparel and primary metals) in the extreme “value”
range (BE/ME is large) and three industries (communications, services, and chemicals) in the
extreme “growth” range (BE/ME is small). To further examine the prevalence of the value effect,
they separately run regressions for each industry and find that for over half of the industries (11
of 21) the coefficient of BE/ME is significant, and for an additional fifth of the industries (4 of
21) the coefficient is moderately significant. They conclude that the intra-industry variation in
BE/ME is relevant in explaining stock returns. Furthermore, they provide empirical evidence
that suggests “value stocks” may exhibit higher default risk which supports the conjecture of
Fama and French (1992). Studies in asset pricing have consistently shown that BE/ME can
significant explain firm value (Banko, et al.), and therefore should be included when examining
value relevance.
H1a: The BE/ME explanatory variable will be significant in a value relevance model.
Although firms report assets at fair value based exit prices, the market could price them
differently based on the riskiness of the firm’s industry. The following hypotheses will test this
proposition.
H1b: The magnitude of the BE/ME coefficient of the FVH in a value relevance
models will be lower (higher) for firms in value (growth) industries.
The next four hypothesis (H2 to H5) concentrate on the coefficient of the FVH for assets
and liabilities and how they are influenced by industry concentration, firm’s position within an
industry, firm’s liquidity, and the magnitude of FMH.
H2. Industry Concentration
Industries can also be different with respect to level of concentration. In an attempt to
link the theories of industrial organization and asset pricing, Hou and Robinson (2006) develop
their propositions from two different models; one that states that innovation begins in firms
20
outside of an established industry and can become so successful that the new firms overtake the
established ones. A second theoretical model states that a structure that is particular to an
industry, like a barrier to entry, affects a firm’s conduct, and a firm’s conduct and decisions
affect its performance. One of the possible consequences of the first model is that concentrated
industries have lower returns because concentrated industries, having more established firms,
have less innovation, and therefore lower returns. The second model also suggests that the
presence a barrier to entry would lead to protected firms and concentrated industry; hence less
default risk and less expected market return. Hou and Robinson (2006) remove regulated
industries from their sample and use the Herfindahl Index as a measure of industry concentration.
The coefficients from their regressions support their theoretical predictions. In competitive
industries (where the Herfindahl Index is low) risk, including default risk, is higher and therefore
coefficients in a value relevance study should be discounted. In concentrated industries (where
the Herfindahl Index is high) risk is lower, and therefore coefficients in a value relevance study
should be higher.
H2: In industries with lower (higher) concentration, the coefficients of the
FVH of Assets will be lower (higher).
H3. Industry Status
MacKay and Phillips (2005) identify and explore why and how firms within an industry
behave differently. In particular, they investigate three types of decisions that firms make:
financial structure (debt structure), technology, and risk decisions. MacKay and Phillips (2005)
develop a second measure of a firm’s position within its industry that is based on its status as an
entering, incumbent, or exiting firm. Their results show that within an industry firms exhibit
characteristics based on their position as an entering, incumbent or exiting firm. For example,
from the perspective of debt, incumbents carry the lowest relative amount of debt. Entering
firms, typically, have higher amounts of debt, and firms exiting have the highest debt. They find
that industry fixed-effects alone do not explain firm characteristics. Since the location of the firm
21
within an industry can also provide clues to its future cash flow, it will affect how a firm is
valued in the capital market. Fries, Miller and Perraudin (1997) also suggest that a firm’s
position within its industry differentially affect their coefficients in a value relevance model. Of
these three categories, incumbents have the least risk while entering and exiting firms have the
most risk. Therefore, incumbents should be subject to less discounting while entering and
exiting firms will be subject to the more discounting.
H3: In a value relevance regression, the coefficients of the FVH of Assets will be
lower (higher) for entering and exiting (incumbent) firms.
H4. Liquidity
Goh, Ng, and Ow Yong (2009) found that the magnitude of the coefficient of Level 3
Assets, the most risky of the FVH, varied with whether the bank’s Tier 1 capital ratio was high
or low. The Tier 1 capital ratio represents the bank’s financial strength, which is a measure of
risk. FAS 157 states that fair value measurements are based on, “The most advantageous market
is the market in which the reporting entity would sell the asset or transfer the liability with the
price that maximizes the amount that would be received for the asset or minimizes the amount
that would be paid to transfer the liability, considering transaction costs in the respective
market(s)” (FASB 2006).
Clearly, firms with the liquidity would be most able to hold any of their assets, including
Level 2 and Level 3 assets, in order to realize them in the “most advantageous market.” I will
use the quick ratio (QR) and operating cash flow ratio (CR) as measures of liquidity that can be
observed across industries and test the following to determine the effect of firms’ liquidity on
FVH valuation:
H4: The interaction between liquidity and FVA3 will be positive.
H5. Relative Amount
Finally, in a survey of perceptions of CFOs toward the adoption of the Fair Value
Option, Daniel, et al. (2010) created the “FAS 157 Ratio” as a measure of the proportion of FAS
22
157 assets and liabilities to the firm’s total assets and liabilities, and show that this measure has
explanatory power. The rationale for the “FAS 157 Ratio” is straightforward. If the value of the
assets measured at fair value is small (large) compared to the total assets of the firm, then we
expect investors to be less worried about the measurements, and therefore we expect the
coefficients of the FVH in a value relevance model to be higher (lower).
H5: The relative amount of assets measured at fair value to total assets
will be value relevant.
3.6 Industry Classification
Early researchers used industry classification systems developed by the U.S government and
noted some caveats. More recently, several researchers have compared those systems with
another commercially-available system. This section discusses some of this work on using and
understanding industry classification systems. Traditionally, papers that have studied industries
in the U.S. have used the Standard Industrial Classification (SIC) which was first published in
1939 by economists and statisticians in the former Bureau of the Budget (now the Office of
Management and Budget, OMB) in order to provide the departments and agencies of the US
government a common definition of industries (Saunders 1999). Although the SIC was initially
developed to classify individual manufacturing facilities (plants), it has subsequently been
applied to classify entire firms (enterprises) (Clarke 1989). The SIC consists of a hierarchical
four-digit code where the left-most digit indicates a broad industry category and subsequent
digits provides more granularity in the classification.
Due to changes in the domestic and global economy since its original publication, the
SIC has undergone a major revision in each decade through 1987, with several minor revisions
occurring between the major ones. Unfortunately, Guenther and Rosman (1994) have noted that
the differences between the SIC in Compustat and the SIC in CRSP could be significant.
23
Around that time, in the early 1990s, the OMB created the Economic Classification
Policy Committee with representatives from all major Federal statistical agencies to examine
various issues with the SIC. In 1993, the ratification of the North American Free Trade
Agreement (NAFTA) created a need for the participating countries (Canada, Mexico, and the
United States) to have a coordinated industry classification system to meet the monitoring
requirements set-forth in the Agreement. At the time, Canada’s classification system dated to
1980, Mexico did not have one, and the US’s SIC was based on the major revision of 1987.
However, the existing industry classification systems were inconsistent with each other.
Therefore, statistical agencies from the three NAFTA countries cooperatively developed the
North American Industry Classification System (NAICS), whose first version was published in
1998 (Saunders 1999).
Since then, the major statistical agencies in all three countries, including the US, have
transitioned to the NAICS, and the US has no further planned updates to the SIC. Krishnan and
Press (2003) provide a thorough comparison between the SIC and the NAICS. Using techniques
from prior accounting researchers, they document that the newer NAICS provides an
improvement over the older SIC in most cases. Therefore, their results suggest the NAICS is
more suited to modern research in accounting. However, Bhoraj, Lee, and Oler (2003)
conducted a more thorough study that compared not only the SIC and NAICS, but also the
industry classification system of Fama and French (1997) and the Global Industry Classification
Standard (GICS)®. The results of their comparative study show that of the four classification
schemes, the GICS is the most suited to financial research. Kile and Phillips (2009) confirmed
the earlier conclusion of Bhoraj, et al. (2003). Werner (2005) performed a detailed study of
industry classification systems and also found that the GICS leads to lower valuation errors. This
24
dissertation will use the GICS® of Standard and Poor (S&P) and Morgan Stanley Capital
International (MSCI) Barra13.
4. Sample Selection
[Insert Table 7 about here]
This section and Panel A of Table 7 describe the steps taken to obtain the sample. The sample
originated as an extract from the Compustat Fundamentals Annual data service, Step 0 in Panel
A, that yielded 6,380 observations of year-end accounting information for fiscal 2008 of active
issues traded on exchanges in the U.S. Panel B of Table 7 shows the Stock Exchange Code from
Compustat. Fiscal 2008 was selected because this is the first year after FAS 157 became
effective. The sensitivity analysis includes data from fiscal 2009. In order to concentrate on
actively-traded firms, the first step removed 753 issues that were not on a major exchange (e.g.
New York Stock Exchange (NYSE), American Stock Exchange (AmEx), or National
Association of Securities Dealers Association Quotations (NASDAQ)). The second step
removed 744 issues whose information Compustat marked as not final data. Step 3 removed
393 firms whose missing information consisted of general accounting information, such as total
assets, total liabilities, net income, minority interest, or common shares outstanding. Step 4
removed 1,892 firms that did not have a complete set of observations of the fair value hierarchy
(FVH), and Step 5 removed four firms without a code from the Global Industry Classification
Standard (GICS®), which is used in this study for industry classifications.
In order to minimize the effect of fluctuations in the equity markets in this study, this
study concentrates on firms whose fiscal year-end is December because these firms release their
(audited) year-end reports towards the end of the first quarter of the subsequent year. Step 6
removed all firms whose fiscal year-end was not December, while Step 7 extracted stock prices
13 Standard & Poor provides information on GICS at the following website
http://www.gics.standardandpoors.com, accessed July 15, 2011.
25
as of March 31, 2009 of active firms in the monthly database of the Center for Research in
Security Prices (CRSP) and merged this extract with the resulting dataset of Step 6. Because
175 firms could not be merged, the resulting sample consisted of 2,108 firms at the end of Step 7.
The next step examined the book equity, BE, computed as the difference of total liabilities, LT,
and minority interest, MIB, from total assets, AT, i.e. BE = AT – LT – MIB. Because investors
may treat firms with negative book equity differently from firms with positive book equity, Step
8 removed 84 firms with negative book equity.
In addition to the six observations of the levels of the FVH, Compustat also provides
two additional data items: the Total Fair Value Assets, TFVA, and Total Fair Value Liabilities,
TFVL. Therefore, one would expect the computed sum of the three disclosures of the levels of
the FVH of assets or liabilities to be equal to their respective total, e.g. letting SA represent the
computed sum of the Level 1, Level 2, and Level 3 disclosures of assets, one would expect this
sum, SA, to be equal to the reported total, TFVA, i.e. SA = TFVA. Similarly letting SL represent
the computed sum of the Level 1, Level 2, and Level 3 disclosures of liabilities, one would
expect SL = TFVL. Step 9 involved computing the two sums for each firm and comparing the
result to its related reported total. The two comparisons each have two possible results: equal or
not equal. Panel C of Table 7 shows the four possible results of the comparisons. Although in
most cases the computed sum did equal its respective reported total as expected, in 290 cases at
least one of the computed sums did not equal its reported total.14 Panel D of Table 7 shows the
distribution of the firms in each of the four cases by two-digit GICS Sector. The column with
the heading “Less” shows the sum of the three cases with at least one inequality by GICS sector.
This column also represents the number of firms dropped in a particular GICS sector. Step 9,
then, which removed the 290 firms where at least one of the computed sums did not equal its
reported total, yielded a sample with 1,734 firms. Within this sample, Financials (GICS Sector
14 One case, that of JP Morgan Chase, was investigated by downloading the firm’s 2008 Annual Report.
Their TFVA and TFVL was the sum of the respective FVH observations less “FIN 39 Netting” (p. 146).
26
40), with 515 firms, has the largest number of firms, followed by Information Technology,
Health Care, and Industrials with 284, 246, and 202 firms respectively. Two sectors, Consumer
Staples and Telecommunications Services, have fewer than 50 firms (39 and 28 respectively).
Lastly, as explained further below, the final step identified and removed154 outliers, that
resulted in a sample size of 1,580 firms.
[Insert Table 8 about here]
Panels A and B of Table 8 briefly examine some characteristics of the data by GICS sector.
Panel A of Table 8 shows the mean and standard deviation of the ratios of Book Equity-to-
Market Equity (BE/ME), Fair Value of Assets to Total Assets (FVA), and Fair Value of
Liabilities to Total Liabilities (FVL). Because the computation of BE/ME requires the market
equity, ME, and Compustat did not report this item for 159 firms in this sample, this column
summarizes observations of 1,575 firms, instead of 1,734. While the mean of BE/ME for the
overall sample is 1.28, the Health Care, Utilities, and Consumer Staples GICS Sectors have the
lowest means of 0.71, 0.82, and 0.87 respectively and the Consumer Discretionary and
Financials GICS Sectors have highest means of 1.77 and 1.74 respectively. The overall mean of
the FVA is 0.18 for the sample. The Health Care, Information Technology, and Financials
Sectors have the highest means of 0.35, 0.25, and 0.22 respectively while the Energy, Materials,
and Utilities Sectors have the lowest mean of 0.05. The FVL ratio is quite low with an overall
mean of 0.03 with the Telecommunication Services and Materials Sectors having the highest
means of 0.07 and 0.06 respectively. Liabilities reported at fair value compose a very small
proportion of total liabilities. Panel B of Table 8 provides more details by showing the mean
and standard deviation of the ratio of the six individual level categories of the FVH scaled by its
appropriate aggregated total, e.g. the first column shows the mean and standard deviation of
Level 1 Assets scaled by Total Assets (FVA1 / AT). The ratio of Level 3 Assets to Total Assets
is extremely low, as are all of the ratios of liabilities. Statistical models involving these
27
explanatory variables could be very sensitive to measurement errors, and therefore the
discussion of results will concentrate on the coefficients of the fair value of assets. Following
STY, who combined the reported Level 1 and Level 2 liabilities into a single variable, FVL12,
this dissertation will also use this sum.
5. Results
This section begins with a brief overview of the results followed by a more detailed
discussion of the specific tests. In summary, the results suggest that the FVH has maintained its
value relevance in this sample and furthermore that the value relevance does in fact vary by
industry. Results from the tests of hypotheses H1 (BE/ME) suggest that the BE/ME is value
relevant. Furthermore, two industry sectors, Financials and Information Technology, strongly
display the predicted pattern of lower (higher) estimated coefficients for value (growth) firms for
all three estimated coefficients of the fair value of assets. However, as summarized in Panel G
of Table 12, three industries, Energy, Industrials, and Health Care, show the opposite pattern,
and therefore these results vary by industry sector. The results of tests of the second hypothesis,
the effect of industry concentration on the value relevance of the FVH, are mixed. One
possibility is the lack of a broad and accurate measure of industry concentration. Although the
Herfindahl-Hirschman Index is an accurate measure of industry concentration, it is not broadly
available for all industries. Following Hou and Robinson (2006), I computed the Herfindahl of
Sales for all firms in a sample, however this measure does not include data for private firms
across each industry, and the lack of results could be attributed to inadequacy of this scale. The
tests of hypothesis H3, on the effect of a firm’s status as an entering or incumbent on the value
relevance of the FVH, contain results for entering and incumbent firms. As the focus of this
dissertation is on the fiscal year 2008 and it is not yet possible to identify exiting firms, my tests
do not contain results for exiting firms. Panel E of Table 15 shows that four industry sectors,
28
Energy, Industrials, Health Care, and Financials, display the expected pattern with the
coefficient of entering firms lower than incumbents. Only one industry sector, Consumer
Discretionary, displays the opposite pattern, while Information Technology shows mixed results.
Therefore, the results suggest that hypothesis H3 is partially supported. Tests of hypothesis H4,
the effect of liquidity on the coefficient of FVA3, uses two different measures of liquidity: the
quick ratio (QR) and operating cash flow (CR). Results using the quick ratio, which is available
for only about one-third of the firms in the sample, are mixed while results using the operating
cash flow, which is available for about two-thirds of the firms in the sample, are significant and
as expected for three industry sectors. These results provide limited support for H3. Finally,
results of hypothesis H5, on the value relevance of the FAS Ratio, the ratio of FAS 157 assets
and liabilities to the firm’s total assets and liabilities, are mixed as shown in Panel A1 of Table
17. The following paragraphs provide a more detailed discussion of the two methods
investigated to mitigate the effect of outliers and to validate the approach selected by
highlighting the general consistency of its results with those of prior literature. Then, sections
5.1 to 5.5 discuss the results of tests of hypothesis.
In order to maintain compatibility with prior literature, the modified Ohlson regression
model employed by STY is also used in this dissertation, and is as follows:
(1)
where each of the explanatory variables is deflated by the number of common shares outstanding.
Following STY, Level 1 and Level 2 liabilities have been combined into a single explanatory
variable because they have similar characteristics. In order to examine the value relevance of
the FVH within an industry sector, separate regression studies are performed for each GICS
Sector. Therefore, for each model, the results present a set of coefficients for each industry
sector. To fully reflect this, each coefficient in Equation (1) could have an additional index to
29
indicate its GICS Sector. However, for clarity in Equation 1, those subscripts have been
dropped.
Empirical researchers have developed several techniques that identify and mitigate the
effect of outliers on the estimated coefficients of regression. Winsorization, one of the simplest
of such techniques, identifies outliers as values above or below a certain percentile, and
mitigates their effect by assigning the value of the low or high percentile to the respective
outliers. This study begins with a conservative 1%, and therefore the variables are Winsorized at
the first- and 99th-percentiles by GICS Sector.15
[Insert Table 9 about here]
Panel A of Table 9 shows the estimated coefficients for the first set of regressions with
Winsorized variables16. The estimated coefficients, including the intercept, are displayed in the
first set of paired columns, while the number of observations in the regression, N, and the
Adjusted R2 are on the far right. The estimated values and their associated significance statistics
are presented in a pair of columns and rows. The header of the first row of a pair contains the
name of the GICS Sector and will show the estimated value followed by a number of stars
indicating the statistical significance as measured by the p-value. Following convention, if the
p-value is greater than 0.1, no star will be present. For p-values between 0.1 and 0.05, one star
will be displayed. For p-values between 0.01 and 0.05, two stars will be displayed, and for p-
values less than 0.01, three stars will be displayed. In the second row of each pair, the numerical
15 Jie (Jay) Cao provides a SAS macro that Winsorizes data, but it writes the Winsorized values into the
original column: http://ihome.cuhk.edu.hk/~b121456/tools/Winsorize_Macro.txt For this dissertation, his
macro was modified to create a new column within the SAS data set that contains the Winsorized values. 16 In order to present the estimated regression coefficients for each study by GICS Sector consistently and
efficiently, a series of SAS macros were developed to collect the coefficients and relevant parameters
from each regression model and present the results.
30
code of the GICS Sector is displayed followed by the letter “t,” which indicates the value is from
the t-test, as shown in Panel A of Table 9.17 The number in parenthesis is the p-value.
One can see that in general, none of the coefficients of FVA3 exhibit significance except
for the regression of Consumer Staples. The magnitude of the coefficient is extremely large and
negative for the 39 observations in this sector. There are four other industries in which the
coefficients of FVA1 and FVA2 exhibit statistical significance: Consumer Discretionary, Health
Care, Financials, and Information Technology. The first two, Consumer Discretionary and
Health Care, exhibit a similar pattern as reported in STY showing a higher coefficient for FVA1
than FVA2. The other two, Financial and Information Technology, exhibit the opposite pattern.
As discussed earlier in the literature review, the early papers, including STY, collected
FVH disclosures exclusively for banks. Therefore, the estimated coefficients from STY’s
regression studies are most comparable to those in the rows labeled “Financials” throughout this
dissertation. However, the GICS Financials in this dissertation is broader than the banking sub-
sector used by STY. Furthermore, this dissertation includes all financial institutions (both banks
and non-bank financial institutions). Also, while STY develop a pooled sample of unaudited
quarterly disclosures from an earlier time period, this study uses audited year-end disclosures
from the subsequent year. The coefficients of the fair value of assets are much smaller than
those reported by STY and the coefficient of Level 3 Assets is not statistically significant (Panel
A of Table 9). Panel B of Table 9 shows the results of a series of F-tests that examine whether
the coefficients for fair value items are significantly different from one and whether two
coefficients of FVH are different from one another. Prior research, e.g. STY, reports that the
coefficients of FVA1 and FVA2 are not significantly different from one and the coefficient of
FVA3 is different from one. As reported, results in Table 9 do not follow these patterns. These
17 Later, some models will be estimated using iteratively reweighted least squares (IRLS). In panels
showing the results of these studies, the letter “c” will follow the numerical code of the GICS sector,
indicating that the values in the row are from a Chi-square distribution.
31
differences may suggest that the Winsorization have not mitigated the effect of outliers on the
coefficients.
In order to help assess the effectiveness of the Winsorization in reducing the influence
of outliers, a comparison was made between the standard deviation and the standardized median
absolute deviation (MAD).18 The comparisons are in the Appendix as Table A1. The
Untreated column compares the standard deviation and MAD of the original sample before
Winsorization and the Winsorized column contains the same statistics, but after Winsorization.
If outliers are not present, the standard deviation and MAD should be similar. By inspection,
one can see that the Winsorization improves the correspondence somewhat. However, even the
Winsorized values could still be influenced by outliers.
This suggests that outliers are present even in the Winsorized dataset. One possible
solution could be to simply increase the percentiles and re-Winsorize at the third- and 97th-
percentiles and investigate the effectiveness of this treatment. However, Winsorization is
sometimes regarded as an arbitrary method and could be the reason that STY did not employ it.
Another technique is M-estimation, a type of iteratively reweighted least squares (IRLS)
algorithm (Huber 1981). IRLS examines the residuals of a regression, computes weights that
minimize the influence of observations with high residuals, and runs another regression. The
algorithm continues to improve weights based on residuals and to, iteratively, run regressions
until a convergence criterion is met. In this case, the convergence criterion is that the
coefficients of the regression converge. The results of the IRLS regression are shown in Panel A
of Table 10.
[Insert Table 10 about here]
M-estimation identifies outliers as observations whose residuals are three times larger
than the scale of the regression, where the median method is used as a robust estimator for the
18 In his discussion of measures of deviation, Huber (1981) provides an illuminating discussion of the
historical development of the median absolute deviation (MAD) and the standard deviation.
32
scale. The tests of significance of the coefficients of M-estimation are conducted using the Chi-
square distribution, unlike OLS which uses the t-distribution. Another difference between OLS
and M-estimation is that while the former allows a much richer set of F-tests, the later only
allows basic tests of coefficients. The results of Panel A of Table 10 show that the magnitudes
of estimated coefficients of the FVH of the financials more closely resemble those of STY.
Furthermore, as one can see, many more coefficients in other industries have become significant.
[Insert Table 11 about here]
STY eliminate outliers identified as observations having a studentized residual larger
than two and run OLS regressions on the (trimmed) sample. For comparison purposes, the
observations that M-estimation identified as outliers19 are deleted from the sample and OLS is
then run on the trimmed sample. Panel A of Table 11 shows that the estimated coefficients
resemble those of the IRLS studies (Panel A of Table 10). The Adjusted R2 of the OLS
regressions on the trimmed sample is much higher than the IRLS regressions on the full sample.
This suggests that its estimated coefficients explain more of the variation of the smaller sample,
which is consistent with the fact that this trimmed sample has had outliers removed.
In spite of the differences between the sample of STY and the one used in Table 11, the
coefficients of the explanatory variables exhibit several noteworthy similarities. For example,
the intercepts and coefficients exhibit strong statistical significance, and the signs of all the
coefficients are as expected. Moreover, the coefficients are similar in magnitude; e.g. the
coefficients related to assets (NFVA, FVA1, FVA2, & FVA3) all range between 0.5 and 1; and
the coefficients related to liabilities (NFVL, FVL12, & FVL3) all range between –2.2 and –0.7.
There are also differences between the sample of STY and the one used in Table 11.
The results of F-tests do not have the same correspondence. In particular, STY observed that
while the coefficients of FVA1 and FVA2 were not statistically different from one, the
19 Panel B of Table 10 shows the distribution by GICS Sector of the 154 outliers that M-estimation
identifies as outliers.
33
coefficient of FVA3 was statistically different from one. As shown in Panel B of Table 11, for
the GICS Financials, all of the coefficients of the FVH of assets, FVA1–FVA3, are statistically
different from one. Furthermore, STY observed that FVA1 was not statistically different from
FVA2 while FVA3 was statistically different from FVA1 and FVA2. The results in Panel B of
Table 11 suggest that none of the coefficients of the FVH of assets are statistically similar.
Overall while not exactly the same, the results of this study using the GICS Financials
are similar to those of STY. This suggests that the general pattern that STY reported for the
financials in the early quarters of 2007 also exists in this sample at the end of FY2008.
5.1 The Value Relevance of the FVA across Industry Sectors
The previous section describes the approach taken to mitigate the effects of outliers and
validates the approach by highlighting the general consistency of the results with those of prior
literature. This section reports the results of regressions testing the modified Ohlson model
across industry sectors.
Panel A of Table 11 reports the statistical significance of all five of the FVH coefficients
(FVA1, FVA2, FVA3, FVL12, and FVL3). As expected, the coefficients of the fair value of
assets, FVA1, FVA2, and FVA3 are nearly all positive while the coefficients of the fair value of
liabilities, FVL12 and FVL3, are nearly all negative. Variation in the significance of the
coefficients across industry sectors supports the notion that the value relevance varies by
industry sector. For example, one industry sector (Telecom Services—with only 26 firms)
exhibits one significant coefficient of the FVH, two sectors (Materials and Utilities) exhibit two
significant coefficients, four (Energy, Industrials, Consumer Discretionary, and Information
Technology) exhibit three significant coefficients, and the remaining three (Consumer Staples,
Health Care, and Financials) exhibit significance in all five coefficients of the FVH.
The coefficients of Consumer Staples are unexpectedly large in magnitude. This
abnormality could be due to the small sample size of this sector (only 36 observations). When
34
statistically significant, the coefficients of FVA1 range from 1.02 (Consumer Discretionary) to
1.42 (Information Technology). One would expect the coefficient of FVA2 to be lower than that
of FVA1 because investors to place highest credibility on Level 1 assets and discount Level 2
(and Level 3) measurements. Where both coefficients of FVA1 and FVA2 exhibit statistical
significance, of the five pairs where both coefficients are statistically significant (ignoring
Consumer Discretionary), only one sector, Industrials, shows this pattern. The other four
industry sectors, Consumer Discretionary, Health Care, Financials, and Information Technology,
show the opposite pattern. There are four sectors with significant coefficients of FVA3:
Financials (0.56***), Energy (0.73**), Information Technology (1.73***), and Health Care
(4.30***). Except for Health Care, the three other coefficients of FVA3 are the lowest of the
significant coefficients of the fair value of assets within their sector. This suggests that of the
Level 1, 2 and 3 disclosures of the fair value of assets, investors discount Level 3 the most. The
significant coefficients of FVL12 ranged from –1.98*** (Utilities) to –0.76** (Materials) as
expected, except for the Health Care sector, where the coefficient of FVL12 was 1.93***. The
sign of this coefficient is completely unexpected, and together with the surprisingly large
coefficient of FVA3 suggests that in this sector, investors may view Level 3 assets and Level 1
and Level 2 liabilities differently. Further study is necessary to address this anomaly.
The coefficient FVA1 (FVL12) is expected to be closer to one (negative 1). This
expectation was tested with F-tests, whose results are reported in Panel B of Table 11. While
many results are significant, they are not consistent. The results indicate that:
the Energy sector has three coefficients of the FVH that exhibit significance: FVA2,
FVA3, and FVL12. The F-tests suggest that the coefficients of FVA2 and FVA3 are
not significantly different from one and from each other, and the coefficient of FVL12
is not significantly different from –1;
35
the Materials sector has two coefficients of the FVH that exhibit significance: FVA2
and FVL12. The coefficient of FVA2 is significantly different from one, and the
coefficient of FVL12 is not significantly different from –1;
the Industrials sector has three coefficients of the FVH that exhibit significance: FVA1,
FVA2, and FVL12. Both coefficients of FVA1 and FVA2 are not significantly
different from one and from each other. The coefficient of FVL12 is not significantly
different from –1.
the Consumer Discretionary sector, the same three coefficients of the FVH are
significant, however the coefficient of FVA1 is not significantly different from one,
while the coefficient of FVA2 is significantly different from one, and the coefficients
are not similar to each other. The coefficient of FVL12 is not significantly different
from –1.
the Health Care sector, all five coefficients of the FVH exhibit significance. The three
coefficients of fair value of assets are significantly different from one. However, the
coefficients of FVA1 and FVA2 are not significantly different from each other, but
they are statistically significantly different from FVA3. The coefficients of FVL12 are
not significantly similar to –1 while the coefficient of FVL3 is similar to –1, and both
coefficients are significantly different from each other.
the Financials has all five coefficients of the FVH exhibiting statistical significance and
different from each other. None of the coefficients of the fair value of assets are similar
to one. The coefficient of FVL12 is statistically similar to –1 while the coefficient of
FVL3 is not.
the Information Technology sector, the three coefficients of the fair value of assets are
statistically significant. The F-tests suggest that the coefficients do not exhibit
statistical differences. However the coefficients of FVA1 and FVA2 are not
36
statistically similar to one, while the coefficient of FVA3 is. In the Telecommunication
Services sector, the only coefficient of the FVH that exhibits significance is that of
FVA2, and F-tests suggest that it is not statistically similar to one.
the Utilities sector has two coefficients of the FVH that exhibit significance. The
coefficient of FVA1 is not statistically different from one while the coefficient of
FVL12 is not significantly different from –1.
We expect to find that the estimated coefficients of FVA1 (and perhaps FVA2) to not be
different from (positive) one. While we find evidence of such a pattern, the results are not
consistent. In addition, we expect the estimated coefficient of FVL12 to not be different from
negative one. We observed this general pattern except for Health Care. The above discussion
suggests that the coefficients of the FVH are value relevant, and the value relevance is different
across industries, and that the results observed for the GICS financials are generally consistent
with prior literature. Next is the examination of the tests of the hypotheses.
5.2 Results of Tests of H1
H1 investigates the additional value relevance of BE/ME in two ways. The first
approach includes the BE/ME ratio into the modified Ohlson model as follows:
PRCit = α0 + α1NFVAit + α2FVA1it + α3FVA2it + α4FVA3it
+ α5NVFLit + α6FVL12it + α7FVL3it + α8 BE/MEit + β1NIit + εit (2)
Hypothesis H1a predicts that the coefficient of BE/ME, a8, will be significant and negative. The
second approach partitions the sample into two 45th–percentiles based on whether the value of
the BE/ME ratio is low or high. The hypothesis predicts that the coefficients of the fair value of
assets, FVA1 through FVA3, of the lower (higher) 45th–percentile, the growth (value) firms, will
be higher (lower) than those in the value (growth) firms. The results of the regression model in
Equation (2) are shown in the two Panels A1 and A2 of Table 12.
37
[Insert Table 12 about here]
The comparison of Panel A1 of Table 12 with Panel A of Table 11 shows that all
coefficients that were statistically significant in model 1 continue to be significant in model 2
with minor shifts in the value of the coefficients. In Panel A1 of Table 12 one can see that the
coefficient of BE/ME is negative for all industries, and when N>40, the coefficient is also
significant. This suggests that BE/ME is value relevant, as hypothesized.
Panel B of Table 12 shows the composition of the two 45th-percentiles based on low or
high BE/ME. Forty-fifth-percentiles were selected in order to have some separation between the
higher and lower subsamples. Some studies partition the sample into quantiles, however, lower
percentiles were not possible due to the small size of the sample. The 45th-percentile, where
BE/ME is low, contains the growth firms in each GICS Sector, while the 45th-percentile where
BE/ME is high holds the value firms. The estimated coefficients for the growth (value) firms,
with BE/ME low (high), are presented in Panel C (Panel E) of Table 12.
H1b predicts the coefficients of the fair value of assets of the growth firms in Panel C
will be higher than those of the value firms in Panel E. To more easily compare the estimated
coefficients of the fair value of assets, Panel G of Table 12 contains corresponding pairs of
coefficients from Panels C and E where both are significant. For the two largest GICS Sectors,
Financials and Information Technology, all three pairs of coefficients of fair value of assets
exhibit the hypothesized pattern, while for three sectors, Energy, Industrials, and Health Care,
some pairs of coefficients exhibit the opposite pattern. While this suggests that H1b is supported
by only two sectors, an inspection of the relative sizes of the FVA and the number of firms in the
sample (see Panel B of Table 8) indicates that the hypothesis has the hypothesis of both largest
industries where the fair value of assets are the largest.
Panels D and F of Table 12 show the F-tests of coefficients for the corresponding
regression. The new results of F-tests for liabilities are even stronger than what was reported in
38
Panel B of Table 11. For example, the coefficient of FVL12 of Industrials was not significantly
different from –1 in the earlier regression, yet in both growth and value regressions, this
coefficient in the corresponding regressions is significant and not similar to negative one. This
suggests that investors value assets of growth and value firms differently in each industry.
Overall, these results suggest that hypothesis H1b may hold for two sectors (Financials
and Information Technology). Furthermore, these results suggest that in some cases the value
relevance of similar information from the disclosures of the FVH can be different for firms in
different industries and growth vs. value firms.
5.3 Results of Tests of H2
The second hypothesis investigates the effect of industry concentration on the value
relevance of the FVH, and will be approached by partitioning the sample into two groups of low
vs. high concentration. H2 states that in industries with higher (lower) concentration, we expect
the coefficient of the FVH Assets to be lower (higher).
Although the Herfindahl-Hirschman Index (HHI) is generally regarded as the most
accurate measure of industry concentration, it is limited both in issuance and coverage. The US
Census Bureau publishes the HHI once every five years as a part of its Economic Census. At the
time the analysis was performed, the most recently available HHI was from the Economic
Census of 2002. In addition, the Census Bureau computes the HHI only for a limited number of
industries. Panel A of Table 13 shows descriptive statistics of the HHI for the 510 firms in this
sample whose industries were included the computation of HHI in Economic Census of 2002.
[Insert Table 13 about here]
Firms in two GICS Sectors, Financials and Utilities, did not have HHI scores, while four
other sectors, Energy, Materials, Consumer Discretionary, and Consumer Staples, had fewer
than 60 firms each. HHI scores for firms in four of the seven GICS sectors show a wide range of
39
values from a low HHI of under twenty-five to a high HHI of over 2,500. In order to increase
the number of firms that can be included in tests of this hypothesis, an alternate measure of
industry concentration, the three-year average Herfindahl of Sales, was subsequently computed
and studied. The results of the studies that used the HHI are presented in the panels of Table 13
while the results based on the alternate measure are presented in panels of Table 14. Lastly, as
an additional investigation, IRLS regression studies with an indicator variable with 0 for firms in
unconcentrated industries and 1 for concentrated industries are reported.
[Insert Table 14 about here]
Panel B of Table 13 shows the low and high 45th percentiles of HHI. The number of
firms in each GICS Sector is quite small. Furthermore, the Health Care sector has an imbalance
with 46 firms with HHI Low and 110 firms with HHI High. Due to the small number of
observations, the regression tests may lack sufficient power.20 The results of OLS regressions on
the subsample with HHI low (high) are presented in Panel C (E) of Table 13 and are followed by
results of corresponding F-tests in Panel D (F). The significant coefficients of fair value of
liabilities in Panel C of Table 13 are positive, as are two of the three significant coefficients in
Panel E of Table 13. The only pair of significant coefficients of the fair value of liabilities is
that of FVL12 of Health Care where both coefficients are positive. These results of the fair
value of liabilities are unexpected and perhaps are a consequence of the small sample size.
According to H2, the coefficients in Panel C with HHI Low are expected to be higher than those
in Panel E with HHI high. Panel G of Table 13 shows the significant pairs of coefficients of fair
value of assets from Panels C and E. Clearly, the results are mixed.
Because the sample was reduced from its original size due to lack of HHI, in order to
explore the influence of outliers, as a robustness check, IRLS regressions were run on both 45th-
percentiles. The results, presented in Panels H and I of Table 13 and summarized in Panel J, are
20 Subsequently, an alternate measure of industry concentration with broader coverage will be computed
and discussed later.
40
consistent with the OLS regressions summarized in Panel G. In particular all pairs that were
significant in OLS Panel G were also significant in the IRLS Panel J. The coefficients between
the OLS and IRLS regressions generally maintained similar magnitude and, more importantly,
maintained the same relationship with each other.21 Thus, because the five pairs of OLS
regressions continue to exhibit the same pattern under the IRLS regressions, we conclude that
those results are not sensitive to outliers. The results are mixed and do not generally support the
hypothesis.22
As mentioned earlier, although the HHI is considered a very accurate measure of
industry concentration, its issuance and coverage are limited. The Herfindahl of Sales is a
measure of industry concentration that has broader coverage than the HHI, however because it is
computed only for public firms, it is not as reliable as the HHI. Appendix B describes the
computation of the three-year average of the Herfindahl of Sales, and Panel A of Table 14 shows
descriptive statistics of the measure which ranged from 0.03 to 0.41, and the standard deviation
across industries ranged from 0.01 to 0.09. Thus, this measure may not provide sufficient
variation.
In an attempt to follow the methodology in Table 13, the sample was split into two 45th
percentiles. However because the Herfindahl of Sales produces insufficient variation, the 45th
percentiles did not achieve sufficient separation, i.e. for seven of the 10 industry sectors, the high
values of the lower 45th percentile were the same as the low values of the higher 45th percentile.
Therefore, smaller percentiles were formed.
Using 40th percentiles reduced the sectors with insufficient variation to half of the total,
while with 35th percentiles, as shown in Panel B of Table 14, only three sectors exhibited
21 In the IRLS regression, a new pair of significant coefficients, FVA3 of Health Care, emerged. 22 The signs of the coefficients of the fair value of liabilities are also mixed. Two additional pairs of
significant coefficients are now present in Panels H and I. However, the coefficients of FVL12 of
Materials have opposite signs. The signs of the coefficients of the fair value of liabilities are also mixed.
Two additional pairs of significant coefficients are now present in Panels H and I. However, the
coefficients of FVL12 of Materials have opposite signs.
41
insufficient variation23: Energy, Materials and Telecommunication Services. Furthermore,
Consumer Staples, with 16 firms in the low 35th percentile, has insufficient data to perform a
regression study. These four industry sectors have been removed from the panels that follow.
Panel C of Table 14 shows the number of firms in each percentile group by industry sector.
Panels D and E of Table 14 show the results of OLS regression studies, while Panel F contains
the pairs of significant coefficients of assets. Similar to Panels G and J of Table 13, the results
are mixed.
As a final series of studies investigating industry concentration, the modified Ohlson
model with H_Ind as an indicator variable with 0 for unconcentrated industries and 1 for
concentrated industries as reported on the three-year average Herfindahl of Sales as shown in
Equation (3) was run:
PRCit = α0 + α1NFVAit + α2FVA1it + α3FVA2it + α4FVA3it
+ α5NVFLit + α6FVL12it + α7FVL3it + α8 H_Indit + β1NIit + εit (3)
This hypothesis suggests that firms with higher concentration should have smaller coefficients
for FVA1 – FVA3 than firms with lower concentration. Consequently, we expect a negative
coefficient for H_Ind. The first regression was for the entire sample. The results are shown in
Panel G of Table 14. The coefficient of H_Ind is not statistically significant. The same
regression model was run with the sample divided according to GICS Sector. The results are
shown in Panels H1 and H2 of Table 14. The results remain mixed with the coefficient of
H_Ind significant and positive in the following sectors: Energy, Financials, and Telecom
Services, but significant and negative in the Industrials and Consumer Staples. Based on the
results of this study, the effect of industry concentration on the value relevance of the FVH is
unclear. The results are mixed which suggests that the influence of concentration does not
23 The maximum value of the low 35th percentile is equal to the minimum value of the high 35th percentile.
42
clearly affect the value relevance of the disclosures of FVH. Lack of support, however, could be
due to insufficiency of the Herfindahl of Sales as an accurate measure of industry concentration.
5.4 Results of Test of H3
Hypothesis H3 investigates how firm’s status (entering, incumbent, exiting) influences
the value relevance of the FVH. As discussed in more details later, because this study uses
financial information for fiscal year ending December of 2008, we are limited to entering and
incumbent firms and we are unable to evaluate the effect of status of exiting firms on value
relevance of the FVH. Obviously, future exiting firms will be some of the current incumbent
firms, and thus it is possible to compare the value relevance of the disclosures of the FVH of
entering vs. the combined incumbent and exiting firms.
This study follows MacKay and Phillips (2005) (MP) and begins by classifying firms as
entering, incumbent, or exiting by examining the presence or absence of firms from the sample
in the decade preceding and year following 2008. A firm is classified as entering if it appears
for the first time in the decade prior to 2008, exiting if it disappears in the year after 2008, and
incumbent if it persists throughout the entire period. Clearly, exiting firms should be identified
using data from the entire decade following 2008 rather than simply the subsequent year.
However, at the time of this writing the data on the decade after 2008 is not yet available, and
therefore this means that this study is unable to distinguish between incumbent and exiting firms.
In what follows, the term “incumbent firms” will be used to refer to the group that includes both
incumbents and exiting firms that are unidentifiable at the time of this writing. We expect that
the coefficients of FVH of assets will be lower (higher) for entering (exiting and incumbents)
firms.
Table 15 provides yearly number of firms in the 2008 sample that entered during the
1990-2009 period. To obtain information in this table, I first identify the “2008 cohort,” and
then query Compustat for the presence this “cohort” from 1999 through 2009. Panel A of Table
43
15 follows this “2008 cohort” from 1999 through 2009. In order to more clearly identify the
entering and exiting firms, Panel B of Table 15 indicates the result of the subtraction of each
column of Panel A from the column of the “2008 cohort.” Thus, the columns of Panel B show
how many firms must be added to the corresponding column of Panel A in order to recover the
2008 cohort. Stated slightly differently, for the decade preceding 2008, the columns of Panel B
show the number of firms that entered a particular GICS Sector of the 2008 cohort in that year,
while the column after 2008 shows the number of firms that exited the cohort in that year.
[Insert Table 15 about here]
Several industries had less than 30 firms entering, and with so few firms in that
subsample, statistical tests have less power. Consequently, Materials, Consumer Staples,
Utilities, and Information Technology were removed from analysis entering firms whose results
are shown in Panel C of Table 15.
In order to minimize the influence of outliers, IRLS regressions are run, and the
estimated coefficients of the entering firms (incumbents) are reported in Panel C (Panel D) of
Table 15. Comparing significant coefficients between entering firms and incumbents, one can
identify eleven pairs of significant coefficients of the FVH of assets that are shown in Panel E of
Table 15. By inspection, one can easily see that eight pairs support the hypothesis. It is
important to note that the Consumer Discretionary sector has only 31 entering firms, and perhaps
due to a small number of firms, this sector could also have been dropped. If this is done, then in
every case, the corresponding coefficient of the incumbents is higher than that of the entering
firms as hypothesized. This suggests that in almost all cases hypothesis H3 is supported and the
value relevance of FMV is higher (lower) for incumbent (entering) firms.24
These results could be biased by the size of the firms. Entering firms are naturally
smaller than incumbents. Panel F of Table 15 shows the results of a one-tailed t-test that
24 Panel F of Table 15 shows the four pairs of coefficients of FVH of liabilities that are significant. It is
interesting to note that in all four cases, the coefficients of incumbents are more negative.
44
compares the means of the logarithm of total assets of the entering to the established firms. The
one-tailed Scatterthwaite t-value for unequal variances, 6.68, is highly significant, which
suggests that the means of the two groups are different. The 95% confidence interval for the
mean of the entering firms is (6.18, 6.48), while the corresponding interval for the established
firms is (6.85, 7.08). The clear separation between the confidence intervals suggests that on
average the entering firms are smaller than the more established firms (incumbent) as expected.
Researchers in finance have consistently identified size as a significant risk factor in the
cross-section of returns of equities (Banz 1981 and Fama and French 1992), and therefore, the
influence of size on the value relevance of the FVH should also be investigated. The modified
Ohlson model was used with two additional variables. One was an indicator variable whose
value, Entri, was set to one for entering firms and zero for established firms, and the second was
the logarithm of total assets log(Ai):
PRCit = α0 + α1NFVAit + α2FVA1it + α3FVA2it + α4FVA3it
+ α5NVFLit + α6FVL12it + α7FVL3it
(4)
+ α8 Entrit + α9 log(Ait) + α10 Entrit ×log(Ait) + β1NIit + εit
The estimated coefficients of four OLS regressions that use these two variables and their
interaction to investigate the influence of the firm’s status, as entering or incumbent, size, and
their interaction are reported in Panels G1 and G2 of Table 15.
The regressions were estimated on the entire sample which consisted of 1,580
observations. All coefficients are highly significant. The first model, which includes the
indicator variable, entering, by itself, shows that the coefficient is highly significant and negative,
which suggests that investors generally discount entering firms compared to incumbents. The
second model, which includes the logarithm of assets by itself, shows that size is significant and
positive. This indicates that size is value relevant, and furthermore that investors view larger
firms more favorably than smaller ones.
45
The third model includes both terms together, and both maintain both their significance
and sign. This suggests that both status and size are value relevant. Finally, the fourth model
includes the interaction of status and size. In this model, the coefficient of the entering indicator
variable changes sign and continues to be significant, while the coefficient of size maintains its
sign and high significance. Interestingly, the coefficient of the interaction term is highly
significant and negative. Overall, these results clearly suggest that size has explanatory power in
addition to a firm’s status as entering or incumbent in a value relevance regression of the FVH.
5.5 Results of Tests for H4
Hypothesis H4 investigates the effect of liquidity on the value relevance of the FVH. H4 states:
The magnitude of the coefficients of FVH will be higher (lower) for firms with higher (lower)
liquidity. The motivation for this hypothesis is GNO, who found a positive interaction between
high and low Tier 1 Capital Ratio and the coefficient of Level 3 fair value assets (FVA3). The
Tier 1 Captial Ratio is only available for banks, and therefore a different measure of liquidity
must be used for non-bank institutions. Two common measures of liquidity are the quick ratio
(QR)
and the operating cash flow ratio (CR)
where ACT = Total Current Assets, INVT = Total Inventories, XPP = Prepaid Expenses, LCT =
Total Current Liabilities, and OANCF = “Operating Activities—Net Cash Flow.” This
dissertation will investigate both measures of liquidity. As summarized in Panel A of Table 16,
of the trimmed sample, 1,061 firms were missing the QR, while only 480 were missing the CR.
46
As anticipated,25 there are almost no firms in the Financials with QR or CR and therefore the
discussion of this hypothesis will not include results from this sector. However, it is important
to recall that GNO, using the Tier 1 Capital Ratio, already have shown liquidity influences the
value relevance of FVA3 for banks. Thus, although results from financials will not be a part of
this dissertation, their effect has already been examined and supported by prior literature.
[Insert Table 16 about here]
Panel B of Table 16, shows some summary statistics for the QR. For all GICS Sectors,
the mean is higher than the median, which indicates that the distribution of QR is skewed
towards higher values. In order to minimize the effect of outliers, IRLS regressions is used on
the modified Ohlson model that includes QR as shown below:
PRCit = α0 + α1NFVAit + α2FVA1it + α3FVA2it + α4FVA3it
+ α5NVFLit + α6FVL12it + α7FVL3it + α8 QRit + β1NIit + εit
(5a)
Panels C1 and C2 show the coefficients estimated by IRLS. Unfortunately, the
estimated coefficient of QR is marginally significant in only two industries, Health Care and
Information Technology. Moreover, for both industries, its sign is negative, which is the
opposite of what is expected. This suggests that investors are penalizing firms for maintaining
liquidity. One possible explanation for this could be that because interest rates are low during
this period, investors believe that firms in these industries should be borrowing to finance capital
investment.
To test H4, the interaction between QR and FVA3 was investigated as shown below:
PRCit = α0 + α1NFVAit + α2FVA1it + α3FVA2it + α4FVA3it + α5NVFLit
+ α6FVL12it + α7FVL3it + α8 QRit + α9QRit ×FVA3it + β1NIit + εit (5b)
25 Thanks to Prof. Roger Debreceny during the defense of the proposal.
47
Panels D1 and D2 of Table 16 show the estimated coefficients using an IRLS regression.
Only two industries, Industrials and Health Care, have significant coefficients, and unfortunately,
the signs are opposite. For the Industrials, the estimated coefficient of the interaction is
significant, has the anticipated sign, and is very large in magnitude. For Health Care, however,
the same estimated coefficient while significant and reasonably large, has the opposite of the
anticipated sign. These conflicting findings necessitate further study of the effect of liquidity as
measured by the quick ratio, QR, on the value relevance of the FVH.
Panel E of Table 16 shows summary statistics for CR. All GICS Sectors have firms
with negative CR, and in one, Health Care, even the median CR is negative. This negative
operating cash flow could be from the influence of the financial crisis on firms in this sample.
Panels F1 and F2 show the IRLS coefficient estimates in the modified Ohlson model
that includes the CR
PRCit = α0 + α1NFVAit + α2FVA1it + α3FVA2it + α4FVA3it
+ α5NVFLit + α6FVL12it + α7FVL3it + α8 CRit + β1NIit + εit .
(5c)
Although the CR is significant in only two GICS Sectors, Consumer Discretionary and
Consumer Staples, it is positive in both cases, which suggests that liquidity itself is value
relevant in those sectors. This result for the CR is in contrast to that of the QR.
The next pair of half-panels, G1 and G2 of Table 16 show the IRLS estimated
coefficients for the modified Ohlson model that includes the interaction term between CR and
FVA3
PRCit = α0 + α1NFVAit + α2FVA1it + α3FVA2it + α4FVA3it + α5NVFLit
+ α6FVL12it + α7FVL3it + α8 CRit + α9 CRit ×FVA3it + β1NIit + εit .
(5d)
48
The coefficients of CR that were significant in the prior model, Consumer Discretionary and
Consumer Staples, remain significant and shift only slightly in magnitude. However the
coefficient of the interaction term in these models are not significant. This suggests that in these
GICS sectors, Consumer Discretionary and Consumer Staples, the value relevance of liquidity
has only a main effect on the equity price.
It is extremely interesting to note that in three GICS Sectors: Health Care, Telecom-
munication Services, and Utilities, the interaction term is significant and positive, as expected.
This suggests that there is empirical support in this study for some industries that liquidity, as
measured by CR, may indeed be value relevant for Level 3 assets. It is interesting to note that
these industries, as is the Financials, are regulated.
It is also important to note that one possible explanation for the null result of H4 is that
the accounting data is from the period of the financial crisis of 2008 (Spiegel 2011). During this
period, many firms faced external liquidity constraints which may weaken the influence of
liquidity on the value relevance of the disclosure of Level 3 assets.
5.6 Results for H5
The final hypothesis investigates whether the relative degree of assets reported at fair
value (to the total assets, FVArat) influences the value relevance of the FVH. A direct
relationship between the magnitude of fair value of reported assets for the test FVH is expected.
Stated differently, H5 hypothesizes that when the “FAS 157 ratio” is lower (higher), we expect
lower (higher) coefficients of the FVH. Summary statistics of the FVArat (assets reported at fair
value divided by total assets) are shown in Panel A of Table 8. To test this hypothesis, FVArat
was included in the modified Ohlson regression model as shown in Equation (5):
PRCit = α0 + α1NFVAit + α2FVA1it + α3FVA2it + α4FVA3it
+ α5NVFLit + α6FVL12it + α7FVL3it + α8FVAratit + β1NIit + εit . (6)
49
Panels A1 and A2 of Table 17 contain the results of the IRLS regression studies, which show
that none of the estimated coefficients of the FVA ratio are significant. Therefore, in this sample,
the FVArat does not contain explanatory power, which suggests that hypothesis H5 is not
supported.
[Insert Table 17 about here]
6 Sensitivity Analysis
To insure that the results are not specific to a single year, this sensitivity analysis tests the value
relevance of the FVH by pooling data from 2009 with 2008. Because of the lack of data on
industry concentration, H2 will not be included in the sensitivity analysis. Because the focus of
H3 is on the status of firms based on the year 2008, it does not make sense to pool the results
with those of the year 2009. Therefore H3 will also not be included in the sensitivity analysis.
Furthermore, the results suggests that H5 is clearly rejected. Therefore, this sensitivity analysis
concentrates on the general value relevance of the FVH across industry sectors and tests of
hypotheses H1 and H4. The new data is pooled year-end annual accounting data from 2008 and
2009, with stock market prices from March 31, 2009 and 2010 respectively.26 Panel A of Table
18 shows the steps taken to develop this sample for this analysis. These steps follow those
presented earlier in Panel A of Table 7, except that firms with negative book equity are not
dropped yet. Part of the sensitivity analysis is to determine whether the IRLS regression will be
able to minimize the effects of this type of outlier. Panel B of Table 18 shows the composition
of this sample by GICS Sector. Panel C1 (C2) provides descriptive statistics of the ratio of fair
26 The main analyses of the dissertation presented in the previous section used the variable, “Common
Shares Outstanding,” CSHO, to deflate items from the consolidated statement of financial position.
However, Compustat no longer provides this variable, and therefore this sensitivity analysis uses
Compustat’s variable “Common Shares Used to Calculate Earnings Per Share – Basic,” CSHPRI, to
deflate appropriate items. Also, while SAS was used for analyses in the previous section, R is used in this
section.
50
value of assets (liabilities) to total assets (liabilities) expressed as a percentage. Panel C2 also
shows the sum of the Level 1 and Level 2 fair value liabilities scaled by total liabilities reported
as FVL12.
Panel D of Table 18 shows the distribution of outliers that are identified as observations
with Studentized Residuals greater than two. Following STY, these are removed. Panel E of
Table 18 shows the sensitivity studies that were performed and the sample sizes for each test.
The sensitivity studies concentrated on the modified Ohlson model across industry sectors,
hypothesis H1 on the value relevance of the BE/ME ratio, and hypothesis H4 on the effect of
liquidity on the value relevance of the fair value of Level 3 assets, FVA3.
Panel F of Table 18 shows the results of the overall test of value relevance on the
trimmed sample across all industry sectors. As before, the FVH continues to be value relevant,
and the value relevance varies by industry sector. The estimated coefficients of the FVH of
Financials and Information Technology are all significant, and the estimated coefficients of the
FVH of the Financials continue to display the pattern described by STY. The estimated
coefficients of the FVH of the Information Technology sector display an unusual pattern,
however. Although the estimated coefficients of fair value assets are all positive, they do not
display a monotonically decreasing pattern, and although the estimated coefficient of FVL12 is
negative, the estimated coefficient of FVL3 is positive and large.
Because the variables of interest, such as the book-to-market equity, quick ratio, and
operating cash flow ratio, are not available for each firm, the regression models that investigate
their effects are run on subsets of the sample. In order to minimize the effects of outliers, IRLS
regressions are run on the subsamples. Panel G of Table 18, the results of H1 on the value
relevance of the BE/ME ratio, are very similar to those of Panel A1 of Table 12 in that all
estimated coefficients of BE/ME are negative, and in all but two industry sectors (Consumer
Discretionary and Staples) are highly significant, as expected.
51
Panel H of Table 18 investigates effect of liquidity on the value relevance of FVA3.
Similar to Panel C1 of Table 16, the estimated coefficients of QR are nearly all not statistically
significant. Furthermore, for one industry sector, Information Technology, the estimated
coefficient is statistically significant and negative. This is the opposite of the expected sign.
Panel I of Table 18 shows the results of the model that includes the interaction between
QR and FVA3. Similar to Panel D1 of Table 16, only one interaction is significant, Consumer
Staples, and large in magnitude. As with the main study, this suggests that the results of this
test, using QR, are not generally conclusive.
Panel J of Table 18 continues the examination of the value relevance of liquidity using
the operating cash flow ratio, CR. All estimated coefficients, except for one are positive as
expected. Furthermore, in five industries the coefficients exhibit statistical significance. This
result is stronger than that of Panel F1 of Table 16 where estimated coefficients of only two
industry sectors, Consumer Discretionary and Staples, are significant.
Panel K of Table 18 shows the results of an IRLS regression including the interaction
between CR and FVA3. No interaction terms are significant. The next two panels, Panels L and
M, continue this study using a trimmed sample where the stockholder’s equity, including
BE/ME, is not negative. Panel L of Table 18 shows the results of an IRLS regression of the
model that includes CR alone with no interaction. The estimated coefficients are similar to those
of Panel J, as expected. Panel M of Table 18 shows the results of an IRLS regression on a
model that includes the interaction of CR and FVA3. Interestingly, the estimated coefficient of
the interaction term for Utilities exhibits strong statistical significance, and its magnitude is
nearly identical to that of the in Panel G1 of Table 16. However, the interaction term in no other
industry sector exhibits statistical significance.
The next series of tests introduces a dummy variable for the year, with the value zero for
2008 and one for 2009, and investigates its interaction with CR and FVA3.
52
PRCit = α0 + α1NFVAit + α2FVA1it + α3FVA2it + α4FVA3it + α5NVFLit+ α6FVL12it
+ α7FVL3it + α8 CRit + α9 Yr09it + α10 CRit ×FVA3it+ α11 CRit ×Yr09it (7)
+ α12 FVA3it × Yrit + α13 CRit ×FVA3it ×Yr09it + β1NIit + εit .
Panel N introduces the year dummy. When the coefficient of this dummy variable is statistically
significant, its estimated value is larger than three, and exhibits strong statistical significance in
six industry sectors. As one can easily expect, this result suggests that prices in the second year,
from March 31, 2010, are higher than prices from the first year, from March 31, 2009. Panel O
investigates the interaction between the year dummy and CR, and shows the interaction exhibits
statistical significance for only one industry sector, Industrials. Panel P investigates interaction
between the year dummy and FVA3, and shows its interaction exhibits statistical significance
for one industry sector, Consumer Staples. Panel Q. investigates the interaction between FVA3
and CR when the year dummy variable is included as a regressor. Again, the estimated
coefficient of the interaction is positive, as expected, and exhibits statistical significance only for
the Utilities sector. Finally, for completeness, Panel R adds the triple interaction between the
year dummy, CR, and FVA3, but the result for the interaction between CR and FVA3 remains
the same.
7 Conclusion
The influence of the crisis of 2008 on this dissertation may have most heavily affected
the null result of H4, which sought to examine the impact of a firm’s liquidity on the value
relevance of the fair value disclosure of its Level 3 assets.
Although not specifically expressed in the form of hypothesis, this dissertation expected
the value relevance of the FVH to be different across different industries. Based on the findings
of prior literature, this dissertation used the Global Industry Classification Standard (GICS®) to
classify firms into industry sectors. To minimize the effect of outliers, the data was Windsorized
at the 1%/99% levels. However, the coefficients of the fair value hierarchy, FVH, of the GICS
53
Financial sector estimated by ordinary least squares (OLS) on the Winsorized sample were
different from what had been reported in the literature. Song, Thomas, and Yi (2010), STY,
observed a monotonically decreasing pattern of coefficients of Level 1 through 3 fair value
assets, FVA1—FVA3. Consequently, I applied iteratively reweighted least squares (IRLS), a
technique from robust statistics, to the sample and found that coefficients of FVA1—FVA3 of
the GICS Financials estimated using this approach more closely displayed the pattern observed
by STY. Finally, in a method similar to STY, I dropped observations that IRLS identified as
outliers due to their large residuals and also found that coefficients of FVA1—FVA3 of the
GICS Financials estimated by OLS on the trimmed sample were again more consistent with
prior literature.
In each of the ten GICS industry sectors, at least one of the five coefficients of the FVH
(FVA1, FVA2, FVA3, FVL12, and FVL3) exhibited statistical significance. In three industry
sectors, Consumer Staples, Heath Care, and Financials, all five coefficients of the FVH exhibited
statistical significance. In four industry sectors, Energy, Industrials, Consumer Discretionary,
and Information Technology, three of the five coefficients exhibited statistical significance. In
two industry sectors, Materials and Utilities, two of the five coefficients exhibited statistical
significance, and in only one industry sector, Telecommunication Services, only one of the five
coefficients exhibited statistical significant. These results strongly suggest that the FVH is value
relevant, and moreover that the value relevance varies by industry sector.
This dissertation also sought to investigate the additional effect of other characteristics on
the value relevance of the FVH. This dissertation’s first hypothesis posits that BE/ME contains
significant explanatory power in the value of the firm. Banko, Conover, and Jensen (2006)
conclude that the intra-industry variation in BE/ME is relevant in explaining stock returns. This
dissertation extends their work by examining the value relevance of BE/ME (vs. return). Using
the modified Ohlson model with an additional term for the BE/ME ratio, I found that the
54
estimated coefficient of the book-to-market equity ratio (BE/ME) exhibits statistical significance
with the expected sign. However, a comparison of estimated coefficients in regressions on value
vs. growth subsets, the results are mixed. In two industries, Financial Services and Information
Technology, all three pairs of coefficients of the fair value of assets, FVA1 through FVA3,
exhibit statistical significance and follow the expected pattern. Three other industries, however,
have one or two pairs of estimated coefficients that show the opposite pattern.
This dissertation’s second hypothesis examines the influence of industry concentration
on the value relevance of the FVH. Based on the work of Hou and Robinson (2006) who find
evidence that suggests that firms in competitive (concentrated) industries have higher (lower)
average returns, this dissertation’s second hypothesis posits that the coefficients of the FVH of
assets will be lower (higher) in industries with higher (lower) concentration. The results are
mixed, possibly because the two available measures of industry concentration each contain
separate inadequacies. The Herfindahl-Hirshman Index (HHI), computed every five years by
the Census Bureau of the US federal government, is considered to be one of the most accurate
measures of industry concentration. However its coverage is not broad. When published, it is
only available for a subset of manufacturing firms. Researchers can use data items such as net
sales to compute their own Herfindahl index of concentration. To do so, they need to have
access to all data items for all firms in an industry. However, there are often firms that are not in
a dataset that are in an industry whose data items should be included in the Herfindahl index.
Therefore, although a Herfindahl index could be computed for each firm in a dataset, it may not
accurately reflect the concentration in a particular industry.
The third hypothesis discusses the influence of a firm’s status as entering or incumbent
on value relevance of the FVH. Mackay and Phillips (2005) compare a firm’s status as entering,
incumbent, and exiting and find that within an industry, a firm’s status affects its valuation.
Mackay and Phillips identify entering (exiting) firms by examining the decade before (after) the
55
year of interest. Because the year studied, 2008, is less than a decade away from the current year,
exiting firms cannot be identified at this time, and therefore the study only compares entering to
incumbent firms. The results are generally consistently and exhibit the expected pattern. As the
literature in finance has shown that size is a significant factor in the cross sectional analysis of
expected stock returns, and there is a possibility that the effect observed could be due to size and
not the firm’s status as entering vs. incumbent, I briefly investigated this, and found evidence
that suggests that size has explanatory power. Furthermore, the firm’s status as entering or
incumbent remains significant even after the introduction of size in the model.
This dissertation’s fourth hypothesis examines the effect of liquidity on the Level 3
assets, FVA3. Goh, Ng, and Ow Yong (2009) report results that suggest that banks with higher
Tier 1 capital ratios have higher estimated coefficients of Level 3 assets. Because the Tier 1
capital ratio is only available for banks, this dissertation uses two alternate measure of liquidity:
the quick ratio (QR) and operating cash flow ratio (CR). The results do not provide strong
evidence suggesting that either measure of liquidity influences value relevance of Level 3 fair
value assets.
The final hypothesis examines whether the relative amount of assets measured at fair
value to total assets influences the value relevance of the FVH. The results consistently reject
this proposition. This suggests that the relative size of the assets measured at fair value does not
influence the value relevance of the FVH.
Prior studies examined the value relevance of the FVH for banks. This study provides
evidence that the value relevance of the FVH can be extended to other industry sectors. The
strong statistical significance of the BE/ME ratio suggests that it could be considered as a control
variable in future studies that use the modified Ohlson model to examine value relevance.
56
Further studies could examine the effect of size27 directly on the value relevance of the
FVH across industry sectors. The hypothesis of the influence of liquidity on the value relevance
of FVA3 was motivated by GNO’s result on the effect of the Tier 1 capital ratio on Level 3
assets of banks. With the lack of results from this study, it is possible that perhaps the analog of
the Tier 1 capital ratio in non-bank firms is not liquidity. Therefore, perhaps some other
measure of the riskiness of firms is more appropriate. Some possible alternatives are the
auditor’s going concern opinion, credit rating, or default probability. In further studies of the
value relevance of the FVH, the influence of these other measures on the value relevance of the
Level 3 assets could be of interest.
27 Thanks to Prof. Qianqiu Liu for this suggestion.
57
Figure 1: Timeline of the Issuance and Effective Date of FAS 157
The timeline shows that the FASB published an exposure draft seeking comments on
measurement of fair value in the summer of 2004. In the fall of 2006, the FASB issued FAS
157 which was to have become effective for fiscal years beginning after November 15, 2007.
For nonfinancial assets and liabilities recognized or disclosed on a nonrecurring basis, however,
in early 2008, the FASB postponed the effective date of FAS 157 to fiscal years beginning after
November 15, 2008. For all other assets and liabilities, FAS 157 became U.S. GAAP when
initially stated.
58
Figure 2. Weekly Close of the S&P 500 Index from Jan 1, 2007 to Jan 1, 2012
Source: MSN Money
This chart shows the weekly close of the S&P 500 Index from January 1, 2007 to January 1,
2012. The prices used in this dissertation are from the CRSP Monthly database on 3/31/2009.
3/30/2009
59
Figure 3 Sample Disclosures Required by FAS 157
Panel A. Example of the Fair Value Hierarchy (FVH) in tabular form
This sample shows a typical disclosure of the Fair Value Hierarchy (FVH), one of several new
disclosures that are required by Financial Accounting Standard (FAS) 157. The table clearly
identifies particular types of assets and liabilities, and also clearly shows the amounts of each
type that the filing entity measured using inputs that FAS 157 defines as Level 1, 2, and 3.
Table 2 summarizes the main characteristics of Levels 1 to 3. This disclosure was made by
Caterpillar, Inc. in its Form 10-K filing for 2008.
60
Figure 3 (cont’d)
Panel B. Example of the Reconciliation of Level 3 Assets and Liabilities.
This sample shows a typical reconciliation of Level 3 Assets and Liabilities, one of several
new disclosures that are required by Financial Accounting Standard (FAS) 157. Beginning
with the ending balance of the prior year of assets and liabilities measured with Level 3 inputs,
this reconciliation summarizes the main changes during the year that resulted in the ending
balance for the period. This disclosure was made by Caterpillar, Inc. in its Form 10-K filing
for 2008.
61
Table 1. Four (Known) Papers that Examine the Fair Value Hierarchy (FVH)
Abbr Author(s) Title SSRN
GNO Goh, Beng Wee
Ng, Jeffrey
Ow Yong, Kevin
Market Pricing of Banks’ Fair Value Assets Reported
Under SFAS 157 during the 2008 Economic Crisis
1335848
Kolev Kolev, Kalin Do Investors Perceive Marking-to-Model as Marking-
to-Myth? Early Evidence from FAS 157 Disclosure
1336368
RS Riedl, Edward J.
Serafeim, George
Information Risk and Fair Value: An Examination of
Equity Betas and Bid-Ask Spreads
(Journal of Accounting Research 49:4,
pp. 1083–1122)
1439851
& JAR 49:4,
pp. 1083–
1122
STY Song, Chang Joon
Thomas, Wayne
Yi, Han
Value Relevance of 157 Fair Value Hierarchy
Information and the Impact of Corporate Governance
Mechanisms
(The Accounting Review 85:4, pp. 1375–1401)
1198143
& Acc. Rev
85:4, pp.
1375-1401.
This table shows the four (known) early papers that studied the value relevance and content of
information of the fair value hierarchy (FVH) and the abbreviations used in this (proposed)
dissertation for them.
62
Table 2. The Main Characteristics of the Fair Value Hierarchy (FVH) of FAS 157
Level 1 Level 2 Level 3
Approach to Measurement of Fair
Value Identical
Instrument
Similar
Instrument Other*
Prices used in Measurement of Fair
Value Observed** Observed** Estimated
* Market (e.g. Broker’s Quotations), Income (e.g. Discounted Cash Flow) or Cost (e.g. to
replicate the service capacity of an asset)
**Quoted prices in active markets.
This table highlights the main characteristics of the fair value hierarchy (FVH) of Financial
Accounting Standard (FAS) 157.
63
Table 3. Datasets of the Four Early Papers
Panel A: Characteristics of Data used
GNO Kolev RS STY
Industry SIC 60 & 61 GICS40,
removed 4040
SIC 6020, 6035,
& 6211
(Compustat
Bank dataset)
Selection NYSE, AMEX,
NASDAQ
S&P 500, (Mar 08)
S&P Mid-Cap 400,
S&P SmallCap 600
12/31/2006:
TA > $10 B
Banks
(implicit)
Time Period 2008 Q1 – Q3 2008 Q1 & Q2 2007 Q1 –
2008 Q2
2008 Q1 – Q3
N 516 (Q1:177, Q2:172) 56 431
Firm-
quarters 1,462 349 148 1,260
Outliers Winsorized IRLS n/a (sensitiv.) Studentized
Levels
1, 2, 3 Net Assets Net Assets As-is &
“Sum Ratio” As-is
Sources Compustat
Quarterly;
CRSP; 10–Q
Compustat Quarterly;
CRSP; 10–Q
10–Q/10–K
(*Compustat N
= 1,045)
Compustat
Bank Qtrly;
CRSP; 10–Q
*In sensitivity analysis.
This table provides a summary the data sets that the early papers used. The papers and
abbreviations follow those defined in Table 1.
64
Table 3 (cont’d)
Panel B: Reporting Periods from which the papers examined the FVH
2007 (all) 2008 Q1 2008 Q2 2008 Q3 2008
Q4**
2009 Q1
GNO
Kolev
RS
STY
This Dissertation
Italics indicate portions related to this proposed dissertation.
** FAS 157 became US GAAP in this quarter (fiscal years ending after November 15, 2008).
This table provides a summary of the reporting periods that the (known) early papers used. The
papers and abbreviations follow those of Table 1.
65
Table 4. Summary of Hypothesis and Results
Panel A: GNO (Goh, Ng, and Ow Yong 2009)
Statement Findings
H1 Lower level FV assets are priced less than higher level FV assets. Supported
H2 The market pricing of FV assets is higher for banks with higher Tier 1 capital.
(liquidity)
γ13 >0*
H3 The market pricing of FV assets is higher for banks that are audited by Big 4 auditors. γ15 >0*
γ16, γ17 >0***
This table shows the main hypotheses and findings of GNO. The explanatory variables relating to the FVH are underlined because GNO use
net assets, or assets minus liabilities, in their models. The coefficients of their first model are monotonically decreasing and are statistically
different from one. However, β6 is not statistically different from β7. They use the model at the bottom of the table to address their next two
hypotheses. They test their second hypothesis using a dummy variable for whether the bank’s Tier 1 capital ratio is higher or lower than the
median, and find a significant interaction with the Level 3 net asset explanatory variable. Lastly the interaction of their Big4 dummy variable
with the FVH net assets suggests that the presence of a Big4 auditor increases investors confidence in the Level 2 and Level 3 net assets.
itit
ititititit
eFVAFVAFVA
NETBEEPSAMEXNYSEPRICE
321 765
43210
4342414
_13_12_11
_14_1]H1[
17161514
131211
1098
BIGFVABIGFVABIGFVABIGNETBE
INDCAPFVAINDCAPFVAINDCAPFVA
INDCAPNETBEBIGINDCAPPRICE it
***
66
Table 4 (cont’d)
Panel B: Kolev (2009)
Statement Findings
H1 Investors find mark-to-model [Levels 2 & 3] FV estimates sufficiently reliable to
be reflected in firm value.
Supported
H2 Investors perceive mark-to-model [Levels 2 & 3] estimates as less reliable than
mark-to-market FV.
Supported
Level
Difference
This table shows the main hypotheses and results of Kolev. As with GNO, in the previous panel, the explanatory variables are underlined to
indicate that Kolev also used net assets, or assets minus liabilities, in his model. To test his hypotheses, which is supported, he forms a
regression model with a complete set of controls. In order to control for the possibility of correlated omitted variables, he forms a differences
model and uses the Level 3 reconciliation table, an additional disclosure required by FAS 157, in the differences model. The differences model
supports his hypotheses.
ityprofitabil & growth, size,for proxies rating, credit ,indicatorsindustry Controls
ControlsNetBVELevelLevelLevelPrice
~
321 1321
ControlsNetBVEAddLGainL
LevelLevelLevelPrice
i154
321
33
321
67
Table 4 (cont’d)
Panel C: RS (Riedl and Serafeim 2011)
Statement Result
H1 The association between a bank’s equity beta and its financial assets is increasing
in the uncertainty about the parameters of the payoff distribution of those assets,
as measured by the Level 1, 2, or 3 designations.
Supported
H2 The information asymmetry component of the bid-ask spread is increasing in the
illiquidity of the bank’s financial assets, as measured by the Level 1, 2, or 3
designations.
Some supp:
γ7 significant
Sensitivity analysis (A. smaller banks): consistent with H1; not consistent with H2.
RS concentrate on the information content of the FVH of FAS 157. The dependent variable in their first model is the CAPM equity beta while
in their second they choose the information asymmetry component of bid-ask spreads. The results of their first test strongly support their
hypothesis, while their results of their second test generally support their hypothesis. Because their main sample is small, in their sensitivity
analyses, they use a larger and broader set of firms, and find evidence consistent with their first hypothesis, but not their second.
ididid
ititititititit
RETVWRET
LeverageOAFVAFVAFVAadjBeta
_
321_
10
54321
itit
itititit
FVFVFV
LogFollLogRiskLogTurnPriceLogLogSpread
321 765
43210
68
Table 4 (cont’d)
Panel D: STY (Song, Thomas, and Yi 2010 The Accounting Review)
Statement Result
H1 FV measurements under FAS 157’s hierarchy are value relevant. Supported
H2 FV measurements under FAS 157’s hierarchy are incrementally value relevant to
asset/liability type (Type information).
Supported
(Vuong 2&3)
H3 The value relevance of FV measurements under FAS 157’s hierarchy is less
evident for firms with weaker corporate governance. (Managers may “abuse”
private information.)
Generally
Supported
itititit
ititititit
NIFVLFVLNFVL
FVAFVAFVANFVAPRC
312
321
765
43210
1;1~
1;1~, 43
3
1
3
1
3
1
10
543210
(69.6%) ...
%)8.67( ...
C
CC
B
BB
A
AAtit
itititittit
ADEVLOANINVESCNFVAPRC
AOTHERADEVLOANINVSECNFVAPRC
*****
432
0.76 0.14 0.20
...321]1H[ GOVGOVFVAGOVFVAGOVFVAPRC itGitGitGit
69
Table 4 Panel D (cont’d)
STY’s first results very clearly suggest that the FVH exhibits value relevance. The numerical values of their coefficients exhibit a monotonic
pattern, and the coefficient of the Level 3 assets is statistically different from the coefficients of the Level 1 and Level 2 assets. Their second
hypothesis involves the value relevance of the expanded disclosures for the type of asset or liability reported at fair value. Their test involves a
pair of models: one with aggregated explanatory variables and one with detailed explanatory variables for several types of assets or liabilities.
They perform Vuong’s test which suggested that the models are different, and therefore support this hypothesis. Their final hypothesis
involves the effect of corporate governance on the value relevance of the FVH. Using a principal-component factor analysis, they compute an
index of corporate governance. The test of their hypothesis suggests that strong corporate governance, as measured by their index, has the most
effect on the coefficient of Level 3 assets.
70
Table 5 Papers examining characteristics of industries in more depth
Panel A: Identification of the papers
Abv Authors Title Year Journal
BCJ Banko, John C.
Conover, C. Mitchell
Jensen, Gerald R.
The Relationship between the Value Effect and
Industry Affiliation
2006 JoB
HR Hou, Kewei
Robinson, David T.
Industry Concentration and Average Stock Returns 2006 JF
MP Mackay, Peter
Phillips, Gordon M.
How Does Industry Affect Firm Financial
Structure?
2005 RFS
This table shows three papers that studied characteristics of industries or characteristics within industries, and the abbreviations used in this
dissertation for them.
JoB stands for Journal of Business, JF stands for Journal of Finance, and RFS stands for Review of Financial studies.
71
Table 5 (cont’d)
Panel B: Highlights of Samples Used (all draw from Compustat & CRSP)
Abbr Period Industry SIC Concentration
BCJ 1968 – 2000 15 or more Two-digit No
HR 1963 – 2001 No Regulated Three-digit Yes
MP 1981 – 2000 Manufacturing 2000 – 3990 Yes
This table provides a summary the data sets that the early papers used. The papers and abbreviations follow those in Panel A.
72
Table 6 Summary of Main Results of Selected Papers
Panel A: BCJ (Banko, Conover, and Jensen 2006 Journal of Business)
Statement Findings
H0 First look: Examine distribution of BE/ME Value (> 0.9): 2
Table with summary statistics of BE/ME (# Industries = 21) Growth (< 0.5): 3
H1 Understand the relationship between VE & Industry B1** (all models)
Rpt = α0 + B1 (Quintile BE/MEpt) + B2 (Industry BE/MEpt)
+ B3 MEpt+ B4 βp + εpt (whole sample)
B2 (self**; w/ >0.1)
B3** & B4**
H2 Understand the prevalence of VE by industry VE varies by Ind.
Rpt = α0 + B1 (BE/MEpt) + B2 MEpt+ B3 βp + εpt (by industry)
[ Also examined temporal variation of VE – exhibited ]
B1: 11**, 4*, (6)
H3 Understand risk (distress, earn. uncert., leverage) & VE Exhibit correlation
25 ports: two-step sort (1. industry BE/ME, 2. firm BE/ME)
73
Table 6 Panel A (cont’d)
BCJ examine the variation of BE/ME by industry. Their study includes 21 industries with 15 or more firms (in each industry). Industries (and
firms) are considered “value” industries when BE/ME is large (approaches 1), while the same are considered “growth” industries when BE/ME
is small (closer to 0). Their table of summary statistics of BE/ME by industry exhibits variation.
For each year in their study, they rank firms by BE/ME and form quintiles. Their pooled regressions suggest that quintile BE/ME is
significant. To further examine the prevalence of the value effect (VE), they separately run regressions for each industry and find that for over
half of the industries, the coefficient of BE/ME is significant, and for an additional fifth of industries is moderately significant. They conclude
that intra-industry variation in BE/ME is relevant in explaining stock returns.
74
Table 6 (cont’d)
Panel B: HR (Hou and Robinson 2006 The Journal of Finance)
Statement Findings
H0 Table: Quintiles of H(Sales) + size, R&D, profitability, risk; &
Fama-Macbeth (1973) regressions on firm (portfolio) characteristics
Support
Concentrated: protected, profit-rich, low R&D.
H1 (Table) Sort portfolios by H(Sales), and examine average return
Competitive Industries (H ~ 0): Average Return is HIGHER
Concentrated Industries (H ~ 1): Average Return is lower
HR seek to link theories of industrial organization and asset pricing. They examined that concentrated industries have less innovation and other
lower risk characteristics, and therefore have lower returns. They discuss the Herfindahl Index that provides a measure industry concentration,
and after computing and comparing several types of Herfindahl Indices, decide to base their Index on sales, H(Sales). They perform Fama-
MacBeth (1973) regressions and results support their hypothesis. To establish their main result, they form and sort portfolios by H(Sales) and
find that in competitive industries the average return is higher, while in concentrated industries, the average return is lower. These findings
validate their interpretation of theories from industrial organization.
75
Table 6 (cont’d)
Panel C: MP (MacKay and Phillips 2005 Review of Financial Studies)
Statement Findings
H1 {Entering, Incumbent, Exiters} make different choices Supported
Table of subsample means (based on MP Table 2):
Entering Incumbents Exiters
Debt Mid low HIGH
Tech Mid HIGH low
Risk n/a low HIGH
H2 {Debt, Tech, Risk} vary within an industry. Supported
{Debt, Tech, Risk} = {Industry Effects} + 2:{Firm Fixed Effects} R2 < 2:{R
2}
H3 Investigate with robust econometric methods Supported
OLS and GMM regressions
MP review several theoretical models of firms in equilibrium and partial equilibrium. Instead of developing tests for particular theories, they
attempt to explore the common theme of the models; that firms within industries behave differently. They identify firms as entering,
incumbents, and exiting firms and find that their characteristics differ, as predicted. Next, they perform pairs of regression models, the first
with industry effects alone and the second with both industry effects and firm fixed effects. They note that the R2 of the second (full) model is
higher than the first (industry only) and conclude that the industry information contains additional explanatory power, which supports their
hypothesis. Finally, they perform a more robust econometric analysis based on GMM regressions which supports their hypothesis.
76
Table 7. Steps Taken to Develop the Sample
Panel A. Description of Action Taken and Number of Listings at Each Step
Step Description Change Firms
Remaining
0 Extract from Compustat for fiscal year 2008 with:
11 <= Stock Exchange Code (EXCHG) <= 14
6,380 6,380
1 Less: Not Major Market (NYSE, AMEX, or NASDAQ) (753) 5,627
2 Less: Not Final Data (keep UPD = 3 only) (744) 4,883
3 Less: Missing AT, LT, NI, MIB, or CSHO (393) 4,490
4 Less: Missing complete set of FVH observations (1,892) 2,598
5 Less: Firms missing a GICS code (SIC & NAICS present) (4) 2,594
6 Less: Firms with fiscal year-end not December (311) 2,283
7 Extract from CRSP Monthly 3/31/2009 with PRC > 0.0 and
EXCHG <=3 returned 5,792 items. Merge by (TIC=Ticker; or
CUSIP8=CUSIP or CUSIP8=NCUSIP) yielded 2,105 matches.
(175) 2,108
8 Less: Firms with negative book equity, BE < 0, where
BE = AT – LT – MIB
(84) 2,024
9 Less: Firms where the computed sum of Level 1, Level 2, and Level
3 Assets or Liabilities did not equal the respective observation of
the Total from Compustat.
(290) 1,734
10 Less: Outliers (154) 1,580
This panel shows the steps taken to assemble the dataset, the number gained (lost) at each step, and the total number of firms at each step. UPD
is Compustat’s field that indicates the status of the data. UPD = 3 means the data is final. The next panel shows Stock Exchange Codes
(EXCHG). Other abbreviations are expanded below:
77
Table 7. Panel A (cont’d)
AT = Assets, Total
LT = Liabilities, Total
NI = Net Income
MIB = Minority Interest, from the Balance Sheet
CSHO = Common Shares Outstanding
BE = Book Equity
PRC = the price of one share of the firm’s stock
FVH = fair value hierarchy
CUSIP = Committee on Uniform Securities Identification Procedures
SIC = Standard Industrial Classification
NAICS = North American Industry Classification System
GICS® = Global Industry Classification Standard
78
Table 7. (cont’d)
Panel B. Stock Exchange Codes
Stock Exchange Code Name of Exchange
11 New York Stock Exchange
12 Amex
13 OTC Bulletin Board
14 NASDAQ-NMS Stock Market
This panel shows the numeric value of Compustat’s Stock Exchange Code (EXCHG) variable and the name of the stock exchange.
79
Table 7. (cont’d)
Panel C. (Step 9) Description of the Four Cases based on Equality or Inequality
SA, Sum of assets reported at fair value, computed as: SA = FVA1 + FVA2 + FVA3
TFVA = Observed Total, Fair Value of Assets (as reported by Compustat from the financial report)
SL, Sum of liabilities reported at fair value, computed as: SL = FVL1 + FVL2 + FVL3
TFVL = Observed Total, Fair Value of Liabilities (as reported by Compustat from the financial report)
In addition to reporting the disclosures of six FVH levels as individual data items, e.g. FVA1, FVA2, FVA3, FVL1, FVL2, and FVL3,
Compustat also provides the Total Fair Value Assets (TFVA) and Total Fair Value Liabilities (TFVL) as separate data items. Provided that
additional accounting standards do not apply, we expect the Total Fair Value Assets to be the sum of FVA1, FVA2, and FVA3, and similarly
for the Total Fair Value Liabilities. However, there were several cases where the sum of the individual FVH items did not equal the respective
total. This Panel shows the four possible cases. The next panel provides the number of firms in each case by Global Industry Classification
Standard (GICS®) sector.
Case SA TFVA SL TFVL
Assets Not Equal ≠ =
Liabilities Not Equal = ≠
Both Not Equal ≠ ≠
Both Equal = =
80
Table 7 (cont’d)
Panel D. Detail by GICS Sector of the Number of firms with Equality or Inequality of the Computed Sum vs. Observed Total
GICS Sectors Step 8
Assets
Not Equal
Liabilities
Not Equal
Both
Not Equal Less
Step 9
(Both Equal)
10 – Energy 157 6 5 5 (16) 141
15 – Materials 84 4 5 2 (11) 73
20 – Industrials 218 13 2 1 (16) 202
25 - Consumer Discretionary 171 13 4 1 (18) 153
30 - Consumer Staples 44 4 0 1 (5) 39
35 - Health Care 286 40 0 0 (40) 246
40 – Financials 644 100 11 18 (129) 515
45 - Information Technology 320 32 2 2 (36) 284
50 - Telecommunication Services 30 2 0 0 (2) 28
55 – Utilities 70 1 3 13 (17) 53
Total 2,024 215 32 43 (290) 1,734
81
Table 7 Panel D. (cont’d)
This panel shows the number of firms, by GICS Sector, where the sum of the individual FVH items, such as FVA1, FVA2, and FVA3, did not
equal the respective data item of its total, e.g. Total Fair Value Assets, and respectively for liabilities. It shows the number of firms trimmed in
Step 9 of Panel A by GICS Sector. The previous panel provides additional clarification of the meaning of the column headings.
82
Table 8. Summary of Selected Variables by GICS Sectors
Panel A. Selected Ratios
BE-to-ME Ratio FVA Ratio FVL Ratio
GICS Sectors N Mean Std N Mean Std N Mean Std
10 - Energy 140 1.40 1.36 141 0.05 0.07 141 0.05 0.12
15 - Materials 69 1.19 1.96 73 0.05 0.13 73 0.06 0.13
20 - Industrials 199 1.17 1.22 202 0.06 0.10 202 0.04 0.09
25 - Consumer Discretionary 147 1.77 3.01 153 0.07 0.13 153 0.02 0.05
30 - Consumer Staples 38 0.87 0.97 39 0.07 0.13 39 0.03 0.08
35 - Health Care 231 0.71 0.67 246 0.35 0.33 246 0.03 0.11
40 - Financials 412 1.74 1.80 515 0.22 0.24 515 0.03 0.11
45 - Information Technology 262 1.01 0.87 284 0.25 0.23 284 0.02 0.09
50 - Telecommunication Services 25 0.96 1.37 28 0.13 0.24 28 0.07 0.20
55 - Utilities 52 0.82 0.42 53 0.05 0.08 53 0.04 0.08
Sample 1575 1.28 1.62 1734 0.18 0.24 1734 0.03 0.10
BE-to-ME is Book Equity to Market Equity. Book Equity is computed as BE = AT – LT – MIB, where AT = Assets-Total, LT = Liabilities-
Total, and MIB = Minority Interest-Balance Sheet.
83
Table 8 Panel A (cont’d)
This table provides summary statistics by GICS Sector of three ratios at Step 9 of the data selection process. Not every firm had a Book
Equity-to-Market Equity (BE-to-ME) Ratio available. The BE-to-ME is important in hypothesis H1, and the table shows that there appears to
be sufficient variation in the sample to test my first hypothesis. Similarly, the summary statistics of the FVA and FVL Ratios suggest that there
appears to be sufficient variation in the sample for regression studies.
84
Table 8 (cont’d)
Panel B. Mean and Standard Deviation of the Fair Value Hierarchy Observations scaled by Respective Total
FVA1
AT
FVA2
AT
FVA3
AT
FVL1
LT
FVL2
LT
FVL12
LT
FVL3
LT
GIC Sectors N mean std mean std mean std mean std mean std mean std mean std
Energy 141 0.01 0.03 0.03 0.05 0.01 0.04 0.01 0.07 0.03 0.09 0.04 0.12 0.01 0.02
Materials 73 0.03 0.09 0.01 0.02 0.02 0.08 0.01 0.06 0.04 0.12 0.05 0.13 0.00 0.01
Industrials 202 0.04 0.08 0.01 0.03 0.01 0.03 0.00 0.03 0.02 0.07 0.03 0.07 0.01 0.06
Cnsmr Discret. 153 0.04 0.11 0.02 0.05 0.01 0.04 0.00 0.03 0.01 0.04 0.02 0.05 0.00 0.01
Cnsmr Staples 39 0.06 0.13 0.01 0.03 0.00 0.01 0.00 0.02 0.02 0.06 0.03 0.08 0.00 0.01
Health Care 246 0.21 0.27 0.12 0.21 0.02 0.06 0.01 0.06 0.02 0.09 0.02 0.10 0.01 0.04
Financials 515 0.03 0.09 0.16 0.19 0.03 0.13 0.00 0.04 0.01 0.08 0.02 0.09 0.01 0.07
Info Tech 284 0.16 0.19 0.07 0.13 0.02 0.06 0.01 0.04 0.01 0.06 0.02 0.08 0.00 0.04
Telecom Svcs. 28 0.12 0.24 0.01 0.02 0.01 0.02 0.00 0.00 0.03 0.12 0.03 0.12 0.03 0.16
Utilities 53 0.03 0.05 0.02 0.04 0.00 0.01 0.01 0.04 0.03 0.05 0.04 0.08 0.00 0.01
Sample 1734 0.08 0.16 0.08 0.15 0.02 0.08 0.01 0.04 0.02 0.08 0.02 0.09 0.01 0.05
85
Table 8 Panel B (cont’d)
This table provides summary statistics by GICS Sector of each of the levels of the FVH, and the sum of Level 1 and Level 2 liabilities scaled by
total assets or liabilities. FVAn (FVLn) represents Level n fair value assets (liabilities) while FVL12 represents the sum of Level 1 and Level 2
liabilities. AT (LT) means total assets (liabilities). The variable created by the sum of Level 1 and Level 2 Liabilities has no GICS Sector with
a mean of zero. For this statistical reason, and to maintain consistency with prior literature, this combined variable will be used in regression
studies. The Level 3 Liabilities has several GICS Sectors where the mean is zero. However, Level 3 is deemed to be quite different from Level
1 and Level 2. For this theoretical reason, and to maintain consistency with prior literature, this variable will be used by itself and not combined
with another variable.
86
Table 9. Results of OLS Regression Using Winsorized Variables
Panel A. Estimated Coefficients
GICS Sector Int NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 NI N
Adj
R2
Energy 5.27 *** 0.44 *** 2.66 ** 0.46 0.03 -0.36 ** -0.65 * 2.14 1.05 *** 141 0.48
3.39 (<.01) 4.55 (<.01) 2.42 (0.01) 0.69 (0.49) 0.03 (0.97) -2.42 (0.01) -1.71 (0.09) 1.07 (0.28) 4.86 (<.01)
Materials 10.29 *** 0.37 0.49 1.65 -0.53 -0.50 0.55 -17.42 3.68 *** 73 0.48
3.23 (<.01) 1.35 (0.18) 0.26 (0.79) 0.43 (0.66) -0.14 (0.88) -1.45 (0.15) 0.71 (0.48) -0.50 (0.61) 6.93 (<.01)
Industrials 4.38 *** 0.51 *** 1.04 *** 0.73 -1.10 -0.41 *** -1.19 ** -2.39 1.75 *** 202 0.48
3.67 (<.01) 5.68 (<.01) 3.12 (<.01) 1.09 (0.27) -0.84 (0.40) -3.71 (<.01) -2.37 (0.01) -1.12 (0.26) 6.91 (<.01)
Cnsmr
Discret.
5.14 *** 0.70 *** 6.27 *** 1.17 * -3.09 -0.64 *** -4.81 *** -18.03 0.82 *** 153 0.46
2.90 (<.01) 4.37 (<.01) 6.45 (<.01) 1.84 (0.06) -1.14 (0.25) -3.00 (<.01) -7.20 (<.01) -0.28 (0.77) 3.80 (<.01)
Cnsmr
Staples
1.22 0.65 ** 4.20 *** 48.31 *** -223.55 *** -0.37 -1.40 -113.63 ** 3.84 *** 39 0.82
0.44 (0.66) 2.68 (0.01) 2.75 (<.01) 5.07 (<.01) -4.71 (<.01) -1.34 (0.18) -1.44 (0.15) -2.18 (0.03) 3.32 (<.01)
Health Care 6.05 *** 1.12 *** 1.41 *** 0.92 *** 2.71 -1.26 *** 0.53 -7.86 *** 2.31 *** 246 0.52
5.89 (<.01) 7.17 (<.01) 2.97 (<.01) 3.87 (<.01) 1.47 (0.14) -5.76 (<.01) 0.65 (0.51) -3.18 (<.01) 5.19 (<.01)
Financials 7.46 *** 0.17 *** 0.20 *** 0.25 *** 0.04 -0.16 *** -0.75 *** -0.68 *** 1.40 *** 515 0.40
10.80 (<.01) 7.45 (<.01) 6.07 (<.01) 10.61 (<.01) 0.30 (0.76) -7.16 (<.01) -4.13 (<.01) -3.03 (<.01) 9.14 (<.01)
Info Tech 2.15 * 0.78 *** 1.45 *** 2.88 *** 1.72 -0.50 * -0.50 9.85 2.24 *** 284 0.33
1.72 (0.08) 4.22 (<.01) 3.16 (<.01) 4.94 (<.01) 1.15 (0.25) -1.96 (0.05) -0.28 (0.78) 0.24 (0.80) 5.18 (<.01)
87
Table 9 Panel A (cont’d)
GICS Sector Int NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 NI N
Adj
R2
Telecom
Svcs.
5.73 * 0.09 0.88 3.80 2.14 0.15 -0.45 -0.33 3.23 ** 28 0.45
1.83 (0.08) 0.21 (0.83) 0.85 (0.40) 1.55 (0.13) 0.33 (0.74) 0.29 (0.77) -0.23 (0.82) -0.10 (0.92) 2.22 (0.03)
Utilities 5.35 * 0.20 0.75 -0.28 0.89 0.04 -1.27 * -2.37 2.76 ** 53 0.58
1.79 (0.08) 0.84 (0.40) 1.23 (0.22) -0.37 (0.71) 0.31 (0.75) 0.15 (0.88) -1.91 (0.06) -1.28 (0.20) 2.44 (0.01)
Total 1734
88
Table 9 (cont’d)
Panel B. Results of F-Tests of the Coefficients
GICS Sector NFVA=1 FVA1=1 FVA2=1 FVA3=1 NFVL=-1 FVL12=-1
Energy 32.79 <.01*** 2.28 0.13 0.67 0.41 1.11 0.29 17.84 <.01*** 0.88 0.35
Materials 5.31 0.02** 0.07 0.79 0.03 0.86 0.17 0.68 2.08 0.15 4.02 0.04**
Industrials 30.03 <.01*** 0.02 0.89 0.17 0.68 2.58 0.11 28.66 <.01*** 0.14 0.70
Cnsmr Discret. 3.36 0.06* 29.39 <.01*** 0.07 0.79 2.27 0.13 2.74 0.10 32.53 <.01***
Cnsmr Staples 2.01 0.16 4.39 0.04** 24.61 <.01*** 22.41 <.01*** 5.27 0.02** 0.17 0.68
Health Care 0.57 0.45 0.73 0.39 0.10 0.75 0.86 0.35 1.41 0.23 3.48 0.06*
Financials 1,340.26 <.01*** 574.59 <.01*** 1,015.34 <.01*** 66.73 <.01*** 1,324.59 <.01*** 1.82 0.17
Info Tech 1.37 0.24 0.96 0.32 10.40 <.01*** 0.23 0.63 3.93 0.04** 0.08 0.77
Telecom Svcs. 4.13 0.05* 0.01 0.90 1.30 0.26 0.03 0.86 4.84 0.04** 0.08 0.78
Utilities 11.83 <.01*** 0.17 0.68 2.88 0.09* 0.00 0.96 12.43 <.01*** 0.17 0.68
89
Table 9 Panel B (cont’d)
GICS Sector FVL3=-1 FVA1=FVA2 FVA1=FVA3 FVA2=FVA3 FVL12=FVL3
Energy 2.48 0.11 3.16 0.07* 3.56 0.06* 0.15 0.69 1.80 0.18
Materials 0.23 0.63 0.07 0.79 0.06 0.81 0.18 0.67 0.27 0.60
Industrials 0.42 0.51 0.15 0.70 2.31 0.12 1.86 0.17 0.30 0.58
Cnsmr Discret. 0.07 0.78 15.68 <.01*** 10.56 <.01*** 2.32 0.13 0.04 0.83
Cnsmr Staples 4.67 0.03** 23.37 <.01*** 22.07 <.01*** 24.90 <.01*** 4.61 0.04**
Health Care 7.72 <.01*** 0.88 0.34 0.41 0.52 0.85 0.35 11.36 <.01***
Financials 1.97 0.16 1.28 0.25 1.74 0.18 3.31 0.06* 0.05 0.81
Info Tech 0.07 0.79 3.61 0.05* 0.03 0.86 0.52 0.47 0.06 0.80
Telecom Svcs. 0.04 0.84 0.99 0.33 0.03 0.85 0.08 0.77 0.00 0.97
Utilities 0.55 0.46 1.20 0.28 0.00 0.96 0.14 0.70 0.28 0.59
90
Table 9 (cont’d)
Panel A
This table presents the coefficients of a modified Ohlson model estimated by OLS regression with
Winsorization at the 1- and 99-percentiles. The modified Ohlson model, Equation (1), us repeated
here:
PRCit = α0 + α1NFVAit + α2FVA1it + α3FVA2it + α4FVA3it
+ α5NVFLit + α6FVL12it + α7FVL3it + β1NIit + εit
PRC is the price of one share of the firm’s stock. NFVA (NFVL) means non-Fair Value Assets
(Liabilities) and is computed as the total assets (liabilities) less the sum of Level 1, 2, and 3 assets
(liabilities). FVA1, FVA2, and FVA3 are the disclosures of Level 1, Level 2, and Level 3 fair value
assets respectively. FVL12 is the sum of the disclosures of Level 1 and Level 2 liabilities, while
FVL3 is the disclosures of Level 3 fair value liabilities. NI is net income. The explanatory
variables have been deflated by the number of common shares outstanding. Panel A shows the
estimated coefficients of the explanatory variables, the number of observations in the regression,
and its adjusted R2. Each coefficient and its related measures of statistical significance are
presented in a pair of columns and a pair of rows. The estimates of the coefficients of each
explanatory variable are in columns headed by the name of the explanatory variable and in rows
labeled with the name of the GICS® sector. To the immediate right of a coefficient, under a column
with no heading, are a number of stars that visually indicate its statistical significance. Following
convention, *** indicates a p-value < 0.01, ** indicates a p-value between 0.01 and 0.05, *
indicates a p-value between 0.05 and 0.1, and no star indicates a p-value greater than 0.1. Rows
directly beneath those with the name of a GICS Sector are blank. The number is the t-value. The
number in parenthesis to the immediate right of t-value is the p-value. In cases where the p-value is
less than 0.01, “<.01” is displayed instead of the actual p-value. The right-most columns, headed
by N and Adj R2, respectively show the number of observations in each sample for a particular
GICS sector and the adjusted R2 for that study.
Panel B
This panel shows the results of F-tests that examine various properties of and relationships between
coefficients within a regression model.
91
91
Table 10. Results of Iteratively Reweighted Least Squares (IRLS) Regression
Panel A. Estimated Coefficients
GICS Sector Icept NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 NI N
Adj
R2
Energy 0.18 1.00 *** -0.35 1.06 *** 0.72 * -1.04 *** -1.01 *** -0.22 0.60 *** 141 0.31
0.03 (0.86) 174.02 (<.01) 0.68 (0.40) 12.91 (<.01) 2.87 (0.09) 76.99 (<.01) 41.44 (<.01) 0.03 (0.86) 17.98 (<.01)
Materials -0.07 1.26 *** 0.21 6.46 *** 2.30 -1.27 *** -0.82 ** -3.43 2.14 *** 73 0.46
0.00 (0.96) 104.66 (<.01) 0.06 (0.80) 13.87 (<.01) 1.83 (0.17) 66.08 (<.01) 5.61 (0.01) 0.05 (0.82) 79.81 (<.01)
Industrials 2.32 *** 0.69 *** 1.10 *** 0.55 ** 0.40 -0.63 *** -1.21 *** -2.06 1.24 *** 202 0.33
6.96 (<.01) 157.85 (<.01) 17.51 (<.01) 4.70 (0.03) 0.64 (0.42) 81.75 (<.01) 9.51 (<.01) 1.58 (0.20) 31.88 (<.01)
Cnsmr Discret. 3.95 *** 0.57 *** 1.17 *** 2.25 *** 0.17 -0.56 *** -1.09 *** -1.72 0.41 *** 153 0.24
29.63 (<.01) 321.64 (<.01) 8.06 (<.01) 46.20 (<.01) 0.10 (0.75) 120.45 (<.01) 10.51 (<.01) 0.11 (0.74) 18.89 (<.01)
Cnsmr Staples 0.74 0.78 *** 4.85 *** 53.54 *** -
249.68
*** -0.52 * -1.54 -
122.64
** 3.27 *** 39 0.54
0.06 (0.80) 9.19 (<.01) 8.97 (<.01) 28.05 (<.01) 24.65 (<.01) 3.14 (0.07) 2.25 (0.13) 4.93 (0.02) 7.11 (<.01)
Health Care 1.37 *** 1.28 *** 1.42 *** 1.37 *** 4.25 *** -1.47 *** 2.01 *** -1.18 ** 0.00 246 0.37
7.18 (<.01) 340.25 (<.01) 52.35 (<.01) 132.64 (<.01) 32.34 (<.01) 270.10 (<.01) 30.66 (<.01) 6.18 (0.01) 0.00 (0.98)
92
92
Table 10 Panel A (cont’d)
GICS Sector Icept NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 NI N
Adj
R2
Financials 1.49 *** 0.73 *** 0.73 *** 0.81 *** 0.58 *** -0.76 *** -1.11 *** -0.80 *** 0.75 *** 515 0.29
14.13 (<.01) 3,173.68 (<.01) 2,083.66 (<.01) 4,067.69 (<.01) 77.52 (<.01) 3,081.00 (<.01) 3,560.26 (<.01) 48.18 (<.01) 111.29 (<.01)
Info Tech 1.70 *** 0.60 *** 1.38 *** 1.79 *** 1.23 *** -0.45 *** -0.07 2.82 1.14 *** 284 0.37
16.94 (<.01) 109.75 (<.01) 101.47 (<.01) 127.57 (<.01) 20.68 (<.01) 38.60 (<.01) 0.24 (0.62) 0.66 (0.41) 68.72 (<.01)
Telecom Svcs. 4.77 0.04 0.84 4.94 ** 4.05 0.14 0.01 0.86 3.66 *** 28 0.42
2.46 (0.11) 0.01 (0.93) 0.72 (0.39) 4.31 (0.03) 0.40 (0.52) 0.07 (0.78) 0.00 (0.99) 0.07 (0.79) 6.69 (<.01)
Utilities 6.28 * 0.29 1.18 * 0.18 -0.10 -0.12 -1.94 *** -1.54 2.65 ** 53 0.42
3.84 (0.05) 1.36 (0.24) 3.26 (0.07) 0.05 (0.82) 0.00 (0.97) 0.14 (0.71) 7.41 (<.01) 0.60 (0.43) 4.81 (0.02)
Total 1734
This panel shows the coefficients of the same modified Ohlson model from the previous table estimated by an Iteratively Reweighted
Least Squares (IRLS) regression, a technique from robust statistics. IRLS is a type of weighted least squares regression where the
observations with large residuals are weighted less in subsequent regressions. The weights are readjusted at each step until a convergence
criterion is met. IRLS can identify outliers by observations with high residuals. In the rows under the name of GICS Sector, the number
is the value of the Chi-square statistic and the number in parenthesis is its p-value.
93
Table 10 (cont’d)
Panel B. Distribution of Outliers Identified by IRLS
GICS Sector Step 9 Outliers Trimmed
Energy 141 16 125
Materials 73 11 62
Industrials 202 8 194
Cnsmr Discret. 153 14 139
Cnsmr Staples 39 3 36
Health Care 246 36 210
Financials 515 35 480
Info Tech 284 28 256
Telecom Svcs. 28 2 26
Utilities 53 1 52
Total 1734 154 1580
This panel shows the distribution by GICS Sector of outliers identified by IRLS. These outliers
were subsequently dropped in Step 10 of the sample selection process described in Panel A of
Table 7.
94
Table 11. Results of OLS Regression on Trimmed Sample
Panel A. Estimated Coefficients
GICS Sector Int NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 NI N
Adj
R2
Energy 0.00 1.06 *** -0.37 1.22 *** 0.73 ** -1.12 *** -1.12 *** -0.65 0.58 *** 125 0.76
0.00 (0.99) 12.84 (<.01) -0.96 (0.33) 4.58 (<.01) 2.04 (0.04) -8.85 (<.01) -7.74 (<.01) -0.58 (0.56) 4.66 (<.01)
Materials -
0.20
1.21 *** 0.15 6.26 *** 2.25 -1.19 *** -0.76 ** -3.03 2.10 *** 62 0.83
-
0.16
(0.87) 9.29 (<.01) 0.19 (0.84) 3.61 (<.01) 1.46 (0.14) -7.32 (<.01) -2.24 (0.02) -0.24 (0.81) 6.10 (<.01)
Industrials 2.67 *** 0.69 *** 1.14 *** 0.64 *** 0.33 -0.65 *** -1.28 *** -2.03 1.13 *** 194 0.98
3.26 (<.01) 13.52 (<.01) 4.68 (<.01) 2.69 (<.01) 0.70 (0.48) -9.97 (<.01) -3.54 (<.01) -1.34 (0.18) 5.38 (<.01)
Cnsmr
Discret.
4.61 *** 0.50 *** 1.02 * 2.40 *** 0.08 -0.49 *** -0.96 *** -2.44 0.34 *** 139 0.53
5.38 (<.01) 6.19 (<.01) 1.93 (0.05) 6.80 (<.01) 0.16 (0.87) -5.00 (<.01) -2.63 (<.01) -0.50 (0.61) 3.79 (<.01)
Cnsmr Staples 1.06 0.77 *** 4.84 *** 56.04 *** -
254.42
*** -0.51 ** -1.47 * -
135.86
*** 3.09 *** 36 0.90
0.49 (0.62) 3.90 (<.01) 4.08 (<.01) 7.38 (<.01) -6.85 (<.01) -2.28 (0.03) -1.91 (0.06) -3.34 (<.01) 3.37 (<.01)
Health Care 1.53 *** 1.30 *** 1.31 *** 1.42 *** 4.30 *** -1.51 *** 1.93 *** -1.22 *** -
0.08
210 0.77
3.06 (<.01) 17.53 (<.01) 6.98 (<.01) 11.99 (<.01) 5.73 (<.01) -15.86 (<.01) 5.37 (<.01) -2.74 (<.01) -
0.39
(0.69)
95
Table 11 Panel A (cont’d)
GICS Sector Int NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 NI N
Adj
R2
Financials 1.56 *** 0.73 *** 0.73 *** 0.82 *** 0.56 *** -0.77 *** -1.07 *** -0.78 *** 0.68 *** 480 0.87
3.70 (<.01) 40.05 (<.01) 29.49 (<.01) 47.47 (<.01) 8.29 (<.01) -41.02 (<.01) -10.17 (<.01) -7.07 (<.01) 7.39 (<.01)
Info Tech 1.80 *** 0.61 *** 1.42 *** 1.73 *** 1.23 *** -0.47 *** -0.04 2.55 1.12 *** 256 0.72
4.76 (<.01) 10.70 (<.01) 10.66 (<.01) 9.75 (<.01) 4.76 (<.01) -6.00 (<.01) -0.29 (0.77) 0.83 (0.40) 8.23 (<.01)
Telecom
Svcs.
5.63 ** -0.13 0.77 6.43 *** 7.93 0.19 0.14 2.77 4.34 *** 26 0.69
2.48 (0.02) -0.39 (0.69) 1.03 (0.31) 3.40 (<.01) 1.62 (0.12) 0.51 (0.62) 0.10 (0.92) 1.09 (0.29) 3.97 (<.01)
Utilities 6.33 ** 0.27 1.22 ** 0.26 0.10 -0.09 -1.98 *** -1.67 2.54 ** 52 0.62
2.24 (0.03) 1.23 (0.22) 2.03 (0.04) 0.35 (0.72) 0.04 (0.97) -0.32 (0.74) -2.92 (<.01) -0.95 (0.34) 2.39 (0.02)
Total 1580
Observations with large residuals were identified by a process described in Appendix A and removed to produce the trimmed sample. This
panel shows the estimated coefficients of the same modified Ohlson model as used in Table 9, however in this table the OLS regressions were
run on the trimmed sample. The format of presentation is the same in previous tables.
96
Table 11 (cont’d)
Panel B. Results of F-Tests
GICS Sector NFVA=1 FVA1=1 FVA2=1 FVA3=1 NFVL= –1 FVL12= –1
Energy 0.56 0.45 12.50 <.01*** 0.69 0.40 0.58 0.44 0.90 0.34 0.72 0.39
Materials 2.61 0.11 1.23 0.27 9.21 <.01*** 0.66 0.42 1.40 0.24 0.52 0.47
Industrials 35.49 <.01*** 0.32 0.57 2.34 0.12 2.06 0.15 29.07 <.01*** 0.60 0.43
Cnsmr Discret. 38.57 <.01*** 0.00 0.97 15.77 <.01*** 3.26 0.07* 27.20 <.01*** 0.01 0.92
Cnsmr Staples 1.33 0.25 10.49 <.01*** 52.49 <.01*** 47.32 <.01*** 4.92 0.03** 0.37 0.54
Health Care 16.49 <.01*** 2.77 0.09* 12.45 <.01*** 19.35 <.01*** 28.32 <.01*** 66.58 <.01***
Financials 221.35 <.01*** 116.14 <.01*** 112.57 <.01*** 41.10 <.01*** 157.90 <.01*** 0.42 0.51
Info Tech 46.90 <.01*** 9.91 <.01*** 16.91 <.01*** 0.77 0.37 45.21 <.01*** 44.14 <.01***
Telecom Svcs. 12.07 <.01*** 0.10 0.75 8.25 0.01** 2.00 0.17 9.81 <.01*** 0.65 0.43
Utilities 10.84 <.01*** 0.13 0.71 1.02 0.31 0.11 0.74 10.39 <.01*** 2.09 0.15
97
Table 11 Panel B (cont’d)
GICS Sector FVL3= –1 FVA1=FVA2 FVA1=FVA3 FVA2=FVA3 FVL12=FVL3
Energy 0.10 0.75 12.27 <.01*** 4.69 0.03** 1.46 0.22 0.18 0.67
Materials 0.03 0.87 10.46 <.01*** 1.59 0.21 3.24 0.07* 0.03 0.85
Industrials 0.46 0.49 1.41 0.23 2.20 0.13 0.37 0.54 0.23 0.62
Cnsmr Discret. 0.09 0.76 3.99 0.04** 1.67 0.19 13.90 <.01*** 0.09 0.76
Cnsmr Staples 11.02 <.01*** 49.92 <.01*** 46.68 <.01*** 52.54 <.01*** 10.86 <.01***
Health Care 0.25 0.61 0.22 0.64 14.56 <.01*** 14.13 <.01*** 34.42 <.01***
Financials 3.76 0.05* 13.20 <.01*** 5.64 0.01** 14.57 <.01*** 3.48 0.06*
Info Tech 1.34 0.24 1.79 0.18 0.48 0.49 2.40 0.12 0.71 0.39
Telecom Svcs. 2.18 0.15 6.40 0.02** 1.91 0.18 0.13 0.72 0.86 0.36
Utilities 0.15 0.70 1.18 0.28 0.14 0.70 0.00 0.95 0.02 0.87
This panel shows the results of F-tests that examine various properties of and relationships between coefficients within a regression model.
98
Table 12. Investigating the Effect of BE/ME
Panel A1. Several of the Estimated Coefficients of an OLS Regression
GICS Sector FVA1 FVA2 FVA3 BE/ME N Adj R2
Energy -0.18 1.16 *** 0.69 ** -1.90 *** 124 0.81
-0.52 (0.60) 4.82 (<.01) 2.15 (0.03) -4.82 (<.01)
Materials 0.15 5.97 *** 2.22 -0.85 ** 59 0.86
0.20 (0.84) 3.72 (<.01) 1.55 (0.12) -2.35 (0.02)
Industrials 1.07 *** 0.67 *** 0.66 -3.65 *** 191 0.99
5.03 (<.01) 3.23 (<.01) 1.61 (0.10) -7.86 (<.01)
Cnsmr Discret. 1.08 ** 2.10 *** 0.18 -0.93 *** 133 0.64
2.36 (0.01) 6.73 (<.01) 0.41 (0.68) -6.20 (<.01)
Cnsmr Staples 5.44 *** 57.45 *** -270.82 *** -1.43 35 0.90
4.55 (<.01) 7.73 (<.01) -7.28 (<.01) -1.03 (0.31)
Health Care 1.63 *** 1.78 *** 2.99 *** -3.08 *** 198 0.84
8.21 (<.01) 9.94 (<.01) 3.47 (<.01) -7.39 (<.01)
Financials 0.67 *** 0.71 *** 0.49 *** -1.57 *** 387 0.69
12.49 (<.01) 20.92 (<.01) 5.58 (<.01) -10.85 (<.01)
Info Tech 1.32 *** 1.56 *** 1.25 *** -2.10 *** 238 0.79
11.21 (<.01) 9.55 (<.01) 5.49 (<.01) -8.62 (<.01)
Telecom Svcs. 3.27 * 4.00 4.48 -1.08 23 0.67
1.78 (0.09) 1.62 (0.12) 0.82 (0.42) -1.03 (0.32)
Utilities 1.22 ** 0.54 0.01 -10.97 *** 51 0.78
2.58 (0.01) 0.98 (0.33) 0.01 (0.99) -4.95 (<.01)
Total 1439
The remaining coefficients are presented in the next half-panel.
99
Table 12 (cont’d)
Panel A2. Remaining Estimated Coefficients
GICS Sector Intercept NFVA NFVL FVL12 FVL3 NI
Energy 3.04 *** 1.02 *** -1.08 *** -1.09 *** -0.59 0.53 ***
2.65 (<.01) 13.68 (<.01) -9.46 (<.01) -8.35 (<.01) -0.59 (0.55) 4.66 (<.01)
Materials 1.31 1.20 *** -1.17 *** -0.80 ** -4.57 2.01 ***
0.97 (0.33) 9.69 (<.01) -7.59 (<.01) -2.53 (0.01) -0.39 (0.70) 6.11 (<.01)
Industrials 6.75 *** 0.77 *** -0.75 *** -1.26 *** -2.08 0.65 ***
7.59 (<.01) 16.83 (<.01) -12.82 (<.01) -3.98 (<.01) -1.58 (0.11) 3.35 (<.01)
Cnsmr Discret. 5.89 *** 0.57 *** -0.57 *** -1.01 *** -1.88 0.29 ***
7.30 (<.01) 7.96 (<.01) -6.58 (<.01) -3.15 (<.01) -0.44 (0.65) 3.75 (<.01)
Cnsmr Staples 1.95 0.92 *** -0.68 *** -1.68 ** -127.14 *** 2.66 ***
0.75 (0.45) 4.46 (<.01) -2.94 (<.01) -2.24 (0.03) -3.22 (<.01) 2.93 (<.01)
Health Care 3.45 *** 1.35 *** -1.53 *** 1.88 *** -1.38 *** -0.25
6.66 (<.01) 18.68 (<.01) -16.10 (<.01) 5.73 (<.01) -3.35 (<.01) -1.36 (0.17)
Financials 4.95 *** 0.63 *** -0.66 *** -1.20 *** -0.89 *** 0.45 ***
10.18 (<.01) 18.25 (<.01) -18.14 (<.01) -5.71 (<.01) -5.11 (<.01) 4.87 (<.01)
Info Tech 4.30 *** 0.59 *** -0.46 *** -0.00 1.23 0.89 ***
9.92 (<.01) 11.17 (<.01) -6.50 (<.01) -0.01 (0.99) 0.46 (0.64) 7.12 (<.01)
Telecom Svcs. 3.96 0.03 0.23 0.73 -1.32 2.39
1.31 (0.21) 0.08 (0.93) 0.55 (0.59) 0.47 (0.64) -0.37 (0.72) 1.30 (0.21)
Utilities 14.08 *** 0.53 *** -0.40 * -1.69 *** -2.06 1.76 **
5.34 (<.01) 3.02 (<.01) -1.78 (0.08) -3.08 (<.01) -1.54 (0.13) 2.06 (0.04)
Total
Half-panels A1 and A2 together show the coefficients of a modified Ohlson model estimated by
OLS regressions on the trimmed sample. The model is that of Panel A of Table 9 with the
BE/ME ratio as an additional regressor:
100
Table 12 Panel A2 (cont’d)
PRCit = α0 + α1NFVAit + α2FVA1it + α3FVA2it + α4FVA3it
+ α5NVFLit + α6FVL12it + α7FVL3it + α8 BE/MEit + β1NIit + εit
Additional columns, for the coefficient of BE/ME and its significance, would make a single panel
too wide to fit on a single page, and therefore the estimated coefficients are split into two half-
panels.
101
Table 12 (cont’d)
Panel B. Composition of the Low and High 45th
-percentile of BE/ME
In order to further examine the effect of BE/ME on the value relevance of the FVH, the sample
was split into two 45th percentiles. This panel shows the number of firms in each GICS sector
and the minimum and maximum values of the BE/ME in each 45th percentile.
BE/ME Low BE/ME High
BE/ME BE/ME
GICS Sector N Min Max N Min Max
Energy 56 0.13 0.97 56 1.14 9.90
Materials 29 0.06 0.59 27 0.73 13.11
Industrials 88 0.05 0.75 86 0.93 7.06
Cnsmr Discret. 65 0.09 0.91 60 1.07 28.63
Cnsmr Staples 16 0.03 0.53 16 0.67 4.54
Health Care 101 0.00 0.55 90 0.63 4.46
Financials 267 0.29 1.12 175 1.33 16.18
Info Tech 125 0.11 0.78 108 0.96 7.14
Telecom Svcs. 13 0.06 0.59 11 0.71 7.05
Utilities 23 0.08 0.67 23 0.83 2.66
783 0.00 1.12 652 0.63 28.63
102
Table 12 (cont’d)
Panel C. Estimated Coefficients from OLS Regression with BE/ME Low (Growth)
GICS Sector Int NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 NI N
Adj
R2
Energy 2.35 * 1.21 *** 0.39 1.16 *** 0.79 ** -1.25 *** -1.32 *** -0.68 0.52 *** 56 0.89
1.95 (0.05) 13.39 (<.01) 0.24 (0.81) 4.67 (<.01) 2.28 (0.02) -8.91 (<.01) -9.70 (<.01) -0.43 (0.67) 2.73 (<.01)
Materials 3.11 1.29 *** 0.40 0.02 2.32 -1.28 *** 3.93 -52.98 1.30 ** 29 0.88
1.59 (0.12) 7.96 (<.01) 0.48 (0.63) 0.00 (0.99) 1.26 (0.22) -6.42 (<.01) 1.13 (0.27) -0.47 (0.64) 2.37 (0.02)
Industrials 5.59 *** 1.01 *** 1.02 *** 0.95 ** 0.79 * -1.01 *** -2.33 *** 2.00 0.94 *** 88 0.74
5.17 (<.01) 10.05 (<.01) 3.51 (<.01) 2.41 (0.01) 1.88 (0.06) -8.26 (<.01) -4.23 (<.01) 0.72 (0.47) 2.95 (<.01)
Cnsmr Discret. 7.18 *** 0.64 *** 0.82 2.07 *** 0.56 -0.65 *** 1.54 -4.25 0.39 *** 65 0.62
5.17 (<.01) 4.24 (<.01) 1.19 (0.24) 5.49 (<.01) 0.24 (0.80) -3.53 (<.01) 0.79 (0.43) -0.10 (0.92) 2.74 (<.01)
Cnsmr Staples 1.90 0.17 6.55 *** 73.93 *** 0.00 . 0.14 4.49 0.00 . 1.58 16 0.93
0.61 (0.55) 0.38 (0.71) 4.60 (<.01) 6.82 (<.01) . . 0.29 (0.77) 0.73 (0.48) . . 1.34 (0.21)
Health Care 2.93 *** 1.44 *** 1.51 *** 1.31 *** 5.21 *** -1.60 *** 1.38 *** -1.68 *** -0.53 101 0.80
3.92 (<.01) 9.64 (<.01) 6.93 (<.01) 7.16 (<.01) 5.15 (<.01) -9.61 (<.01) 2.86 (<.01) -3.12 (<.01) -1.41 (0.16)
Financials 2.58 *** 0.77 *** 0.77 *** 0.82 *** 0.50 *** -0.79 *** -1.12 *** -0.67 *** 0.56 *** 267 0.91
4.51 (<.01) 36.03 (<.01) 28.24 (<.01) 40.71 (<.01) 5.53 (<.01) -37.54 (<.01) -9.59 (<.01) -4.67 (<.01) 3.57 (<.01)
Info Tech 2.84 *** 0.76 *** 2.02 *** 1.77 *** 1.59 *** -0.72 *** -0.27 0.69 0.58 ** 125 0.78
4.89 (<.01) 8.70 (<.01) 9.54 (<.01) 6.73 (<.01) 4.69 (<.01) -6.62 (<.01) -1.44 (0.15) 0.21 (0.83) 2.07 (0.04)
103
Table 12 Panel C (cont’d)
GICS Sector Int NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 NI N
Adj
R2
Telecom Svcs. 7.90 0.54 0.56 -197.24 13.16 -0.73 -1.15 4.73 3.62 13 0.16
1.18 (0.30) 0.57 (0.60) 0.45 (0.67) -0.77 (0.48) 0.96 (0.39) -0.75 (0.49) -0.46 (0.66) 0.67 (0.53) 1.29 (0.26)
Utilities 9.42 ** 0.26 0.30 0.36 1.99 -0.02 -0.32 -3.17 1.69 23 0.68
2.47 (0.02) 0.62 (0.54) 0.19 (0.85) 0.24 (0.81) 0.43 (0.67) -0.04 (0.97) -0.26 (0.79) -1.37 (0.19) 0.76 (0.46)
Total 783
This panel shows the estimated coefficients by GICS Sector of an OLS regression using the modified Ohlson model on the 35th-percentile with
BE/ME low, i.e. the growth firms. The model and variables are those given in Panel A of Table 9.
104
Table 12 (cont’d)
Panel D. Results of F-tests of Coefficients from OLS Regression with BE/ME Low
GICS Sector NFVA=1 FVA1=1 FVA2=1 FVA3=1 NFVL= –1 FVL12= –1
Energy 5.58 0.02** 0.14 0.71 0.42 0.52 0.36 0.55 3.19 0.08* 5.50 0.02**
Materials 3.21 0.08* 0.52 0.47 0.05 0.82 0.51 0.48 2.00 0.17 2.02 0.17
Industrials 0.01 0.93 0.00 0.94 0.02 0.88 0.26 0.61 0.01 0.93 5.82 0.01**
Cnsmr Discret. 5.55 0.02** 0.06 0.79 8.09 <.01*** 0.04 0.84 3.49 0.06* 1.69 0.19
Cnsmr Staples 3.69 0.08* 15.19 <.01*** 45.27 <.01*** . . 5.81 0.03** 0.80 0.39
Health Care 8.54 <.01*** 5.42 0.02** 2.86 0.09* 17.32 <.01*** 13.05 <.01*** 24.40 <.01***
Financials 115.93 <.01*** 69.40 <.01*** 84.24 <.01*** 29.46 <.01*** 100.17 <.01*** 1.09 0.29
Info Tech 7.53 <.01*** 23.17 <.01*** 8.58 <.01*** 3.04 0.08* 6.80 0.01** 15.57 <.01***
Telecom Svcs. 0.23 0.65 0.13 0.73 0.61 0.47 0.79 0.42 0.08 0.79 0.00 0.95
Utilities 3.12 0.09* 0.20 0.66 0.19 0.66 0.05 0.83 3.64 0.07* 0.30 0.59
105
Table 12 Panel D. (cont’d)
GICS Sector FVL3= –1 FVA1=FVA2 FVA1=FVA3 FVA2=FVA3 FVL12=FVL3
Energy 0.04 0.84 0.23 0.63 0.07 0.79 1.03 0.31 0.16 0.68
Materials 0.21 0.64 0.01 0.93 1.04 0.31 0.30 0.59 0.25 0.62
Industrials 1.15 0.28 0.02 0.89 0.21 0.64 0.09 0.76 2.32 0.13
Cnsmr Discret. 0.01 0.93 2.07 0.15 0.01 0.91 0.43 0.51 0.02 0.89
Cnsmr Staples . . 41.41 <.01*** . . . . . .
Health Care 1.60 0.20 0.50 0.48 13.11 <.01*** 16.27 <.01*** 30.17 <.01***
Financials 5.38 0.02** 2.68 0.10 7.99 <.01*** 11.70 <.01*** 5.99 0.01**
Info Tech 0.26 0.61 0.61 0.43 1.32 0.25 0.16 0.69 0.08 0.77
Telecom Svcs. 0.66 0.46 0.60 0.48 0.84 0.41 0.69 0.45 0.70 0.44
Utilities 0.88 0.36 0.00 0.98 0.13 0.72 0.11 0.75 1.26 0.28
This panel shows the results of F-tests that examine various properties of and relationships between coefficients within a regression model.
106
Table 12 (cont’d)
Panel E. Estimated Coefficients from OLS Regression with BE/ME High (Value)
GICS Sector Int NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 NI N Adj R2
Energy -1.92 ** 0.77 *** 0.57 ** 1.89 *** 0.80 ** -0.81 *** -1.07 *** -0.40 0.33 *** 56 0.89
-2.36 (0.02) 10.62 (<.01) 2.48 (0.01) 5.36 (<.01) 2.11 (0.04) -7.68 (<.01) -5.49 (<.01) -0.52 (0.60) 3.82 (<.01)
Materials -0.50 0.75 *** 1.28 0.55 -3.07 -0.78 *** -0.14 6.43 0.56 * 27 0.89
-0.72 (0.48) 6.78 (<.01) 1.54 (0.14) 0.14 (0.89) -1.07 (0.30) -6.18 (<.01) -0.66 (0.51) 1.30 (0.20) 2.03 (0.05)
Industrials -2.39 *** 0.81 *** 0.26 1.46 *** 0.30 -0.82 *** -0.99 *** -0.90 0.02 86 0.99
-2.78 (<.01) 12.03 (<.01) 0.79 (0.43) 5.52 (<.01) 0.45 (0.65) -9.22 (<.01) -3.23 (<.01) -0.62 (0.53) 0.11 (0.91)
Cnsmr Discret. -0.89 0.54 *** 0.74 -1.44 0.35 -0.51 *** -0.76 ** -2.03 0.04 60 0.64
-1.13 (0.26) 8.71 (<.01) 1.29 (0.20) -0.88 (0.38) 1.12 (0.26) -7.01 (<.01) -2.07 (0.04) -0.68 (0.49) 0.40 (0.68)
Cnsmr Staples 0.45 1.07 ** 0.56 26.77 -137.03 -1.28 * 7.33 -62.66 2.81 16 0.92
0.14 (0.89) 3.14 (0.01) 0.31 (0.76) 0.96 (0.37) -0.61 (0.55) -2.14 (0.06) 0.43 (0.67) -1.10 (0.30) 1.38 (0.21)
Health Care -0.42 1.33 *** 0.31 1.49 *** 0.62 -1.45 *** 0.97 -17.13 * -0.04 90 0.88
-0.73 (0.46) 14.33 (<.01) 0.88 (0.38) 7.56 (<.01) 0.52 (0.60) -10.13 (<.01) 0.73 (0.46) -1.98 (0.05) -0.17 (0.86)
Financials 1.05 ** 0.28 *** 0.28 *** 0.36 *** 0.34 *** -0.28 *** -0.55 *** -0.36 *** 0.35 *** 175 0.55
2.58 (0.01) 7.87 (<.01) 4.84 (<.01) 9.27 (<.01) 5.66 (<.01) -7.55 (<.01) -3.38 (<.01) -3.45 (<.01) 6.29 (<.01)
107
Table 12 Panel E. (cont’d)
GICS Sector Int NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 NI N
Adj
R2
Info Tech -0.25 0.74 *** 0.99 *** 1.10 *** 1.14 *** -0.72 *** -0.24 -0.96 0.33 ** 108 0.79
-0.76 (0.44) 12.11 (<.01) 8.85 (<.01) 6.38 (<.01) 5.06 (<.01) -6.80 (<.01) -0.28 (0.77) -0.19 (0.84) 2.58 (0.01)
Telecom Svcs. 0.50 0.69 4.28 6.75 ** 12.19 * -1.12 46.74 -9.49 2.32 11 0.97
0.30 (0.79) 1.06 (0.40) 2.13 (0.16) 4.49 (0.04) 3.94 (0.05) -1.03 (0.41) 1.76 (0.22) -2.48 (0.13) 2.41 (0.13)
Utilities -0.06 1.11 *** 1.42 ** 0.13 -0.74 -1.23 *** -1.50 -2.21 1.20 23 0.79
-0.02 (0.98) 6.26 (<.01) 2.26 (0.03) 0.16 (0.87) -0.31 (0.76) -4.93 (<.01) -1.55 (0.14) -1.03 (0.32) 1.75 (0.10)
Total 652
This panel shows the estimated coefficients by GICS Sector of an OLS regression using the modified Ohlson model on the 35th-percentile with
BE/ME high, i.e. the value firms. The model and variables are those given in Panel A of Table 9.
108
Table 12 (cont’d)
Panel F. Results of F-tests of Coefficients from OLS Regression with BE/ME High
GICS Sector NFVA=1 FVA1=1 FVA2=1 FVA3=1 NFVL= –1 FVL12= –1
Energy 10.06 <.01*** 3.59 0.06* 6.40 0.01** 0.27 0.60 3.26 0.07* 0.15 0.70
Materials 4.90 0.04** 0.11 0.74 0.01 0.91 2.00 0.17 2.96 0.10 15.26 <.01***
Industrials 8.09 <.01*** 5.18 0.02** 3.00 0.08* 1.04 0.31 3.99 0.04** 0.00 0.98
Cnsmr Discret. 57.20 <.01*** 0.20 0.65 2.20 0.14 4.29 0.04** 47.08 <.01*** 0.42 0.51
Cnsmr Staples 0.05 0.83 0.06 0.81 0.85 0.38 0.38 0.55 0.23 0.64 0.24 0.63
Health Care 12.81 <.01*** 3.87 0.05* 6.21 0.01** 0.10 0.75 9.97 <.01*** 2.22 0.14
Financials 419.58 <.01*** 160.93 <.01*** 270.78 <.01*** 124.50 <.01*** 362.48 <.01*** 7.58 <.01***
Info Tech 17.73 <.01*** 0.01 0.91 0.31 0.57 0.37 0.54 7.24 <.01*** 0.80 0.37
Telecom Svcs. 0.23 0.67 2.67 0.24 14.62 0.06* 13.07 0.06* 0.01 0.92 3.23 0.21
Utilities 0.40 0.53 0.45 0.51 1.19 0.29 0.53 0.47 0.83 0.37 0.27 0.61
109
Table 12 Panel F. (cont’d)
GICS Sector FVL3= –1 FVA1=FVA2 FVA1=FVA3 FVA2=FVA3 FVL12=FVL3
Energy 0.63 0.43 10.76 <.01*** 0.31 0.57 4.40 0.04** 0.75 0.39
Materials 2.26 0.14 0.03 0.85 1.67 0.21 0.60 0.44 1.76 0.20
Industrials 0.00 0.94 5.56 0.02** 0.00 0.95 2.60 0.11 0.00 0.94
Cnsmr Discret. 0.12 0.72 1.69 0.19 0.40 0.53 1.15 0.28 0.18 0.67
Cnsmr Staples 1.17 0.31 0.89 0.37 0.38 0.55 0.43 0.53 1.83 0.21
Health Care 3.47 0.06* 9.92 <.01*** 0.05 0.81 0.52 0.47 3.88 0.05*
Financials 37.94 <.01*** 3.95 0.04** 0.77 0.38 0.18 0.67 1.19 0.27
Info Tech 0.00 0.99 0.24 0.62 0.37 0.54 0.02 0.89 0.02 0.88
Telecom Svcs. 4.93 0.15 0.64 0.50 3.43 0.20 6.68 0.12 3.85 0.18
Utilities 0.32 0.58 5.22 0.03** 0.58 0.45 0.10 0.76 0.06 0.80
This panel shows the results of F-tests that examine various properties of and relationships between coefficients within a regression model.
110
Table 12 (cont’d)
Panel G. Significant Pairs of Coefficients of Fair Value Assets from Growth and Value
45th
Percentiles
Industry FVA1
Growth, Value
FVA2
Growth, Value
FVA3
Growth, Value
Expected
Financials 0.77***, 0.28*** 0.82***, 0.36*** 0.50***, 0.34***
Info Tech 2.02***, 0.99*** 1.77***, 1.10*** 1.59***, 1.14***
Not Expected
Energy
1.16***, 1.89*** 0.79**, 0.80**
Industrials
0.95**, 1.46***
Health Care
1.31***, 1.49***
This table summarizes findings from OLS regressions presented in Panels C and E. The first
coefficient in each pair is the growth coefficient from Panel C, while the second in each pair
is the corresponding value coefficient from Panel E. The upper part of the table shows the
coefficients in industry sectors whose pattern of coefficients follows the hypothesized pattern,
while the lower part shows the coefficients of industry sectors whose pattern of coefficients
does not follow the hypothesized pattern. The number of stars after each coefficient
represents its statistical significance, following the convention described in Panel A of Table
9.
111
Table 13. Investigation of the Effect of Industry Concentration (HHI)
Panel A. Descriptive Statistics of the Herfindahl-Hirschman Index (HHI)
GICS Sector N MIN Q1 MEDIAN Q3 MAX MEAN STD
Energy 20 286.10 462.90 639.70 639.70 802.20 575.68 180.99
Materials 41 23.50 238.30 578.50 976.60 2,661.60 753.79 678.42
Industrials 85 23.50 233.10 497.90 988.70 2,527.70 654.50 527.02
Cnsmr Discret. 44 19.00 177.95 650.50 1,198.45 2,815.50 781.76 692.72
Cnsmr Staples 25 221.30 529.90 806.50 895.70 1,524.20 766.40 311.56
Health Care 156 187.20 476.30 529.90 634.70 2,703.60 569.51 287.36
Info Tech 139 23.50 583.60 818.70 1,417.10 2,662.40 994.31 558.50
Summary 510 19.00 460.70 634.70 929.70 2,815.50 742.47 516.82
This panel shows descriptive statistics of the Herfindahl-Hirschman Index (HHI) by GICS Sector. The Census Bureau publishes the HHI
for selected industries. Q1 is the first quartile and Q3 is the third quartile. STD is the standard deviation.
112
Table 13 (cont’d)
Panel B. Composition of the Low and High 45th
-percentiles of HHI
This panel shows the number of firms in each 45th-percentile, and the minimum and
maximum value of the HHI within each 45th-percentile.
HHI Low HHI High
HHI HHI
GICS Sector N Min Max N Min Max
Energy 5 286.10 286.10 15 639.70 802.20
Materials 17 23.50 497.90 19 656.70 2,661.60
Industrials 38 23.50 442.50 39 583.60 2,527.70
Cnsmr Discret. 19 19.00 558.80 20 703.40 2,815.50
Cnsmr Staples 11 221.30 763.10 12 810.20 1,524.20
Health Care 46 187.20 476.30 110 529.90 2,703.60
Info Tech 41 23.50 624.50 67 1,398.50 2,662.40
177 19.00 763.10 282 529.90 2,815.50
113
Table 13 (cont’d)
Panel C. Estimated Coefficients from OLS Regression on the 45th
-percentile with HHI Low
GICS Sector Int NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 NI N
Adj
R2
Energy 14.89 . -1.01 . 11.23 . 414.64 . 0.00 . 1.66 . 0.00 . 0.00 . 0.00 . 5
. . . . . . . . . . . . . . . . . .
Materials 1.14 0.82 ** 5.40 -8.45 1.70 -0.46 -3.28 -31.11 1.84 * 17 0.73
0.25 (0.81) 2.42 (0.04) 0.78 (0.45) -0.87 (0.40) 0.32 (0.76) -1.30 (0.22) -0.61 (0.55) -0.30 (0.77) 2.26 (0.05)
Industrials 4.02 * 0.44 ** 1.91 -2.88 -2.92 -0.36 -0.29 -3.18 2.37 *** 38 0.64
1.79 (0.08) 2.36 (0.02) 0.82 (0.41) -0.67 (0.50) -0.23 (0.81) -1.08 (0.28) -0.68 (0.50) -1.17 (0.24) 4.03 (<.01)
Cnsmr Discret. 1.88 0.17 2.43 ** 17.13 -12.01 -0.11 3.98 0.00 . 0.26 19 0.78
1.32 (0.21) 0.72 (0.48) 2.33 (0.04) 1.77 (0.10) -1.40 (0.18) -0.33 (0.74) 0.40 (0.69) . . 0.87 (0.40)
Cnsmr Staples 0.22 2.53 ** 2.93 -0.16 0.00 . -3.76 ** 44.32 * 0.00 . -3.49 11 0.81
0.08 (0.93) 3.99 (0.01) 0.97 (0.38) -0.01 (0.98) . . -3.18 (0.03) 2.14 (0.09) . . -1.68 (0.16)
Health Care 0.99 1.21 *** 2.90 *** 0.84 * 6.43 *** -1.34 *** 1.96 ** 170.54 1.15 ** 46 0.88
1.08 (0.28) 7.71 (<.01) 5.83 (<.01) 1.72 (0.09) 3.60 (<.01) -6.06 (<.01) 2.16 (0.03) 0.34 (0.73) 2.41 (0.02)
Info Tech 1.62 0.73 *** 1.06 ** 1.07 *** 0.99 -0.61 * -0.12 4.07 0.75 *** 41 0.86
1.61 (0.11) 5.15 (<.01) 2.56 (0.01) 3.15 (<.01) 1.39 (0.17) -1.96 (0.05) -0.59 (0.55) 0.05 (0.95) 3.29 (<.01)
177
114
Table 13 Panel C (cont’d)
This panel shows the estimated coefficients by GICS Sector of an OLS regression using the modified Ohlson model on the 45th-percentile
where the Herfindahl-Hirschman Index (HHI) is low, i.e. the firms in unconcentrated industries. The model and variables are those given in
Panel A of Table 9.
Panel D
This panel shows the results of F-tests that examine various properties of and relationships between coefficients within a regression model.
115
Table 13 (cont’d)
Panel D. Results of F-Tests of Coefficients from the Previous Panel
GICS Sector NFVA=1 FVA1=1 FVA2=1 FVA3=1 NFVL= –1 FVL12= –1
Energy . . . . . . . . . . . .
Materials 0.29 0.60 0.40 0.54 0.96 0.35 0.02 0.90 2.33 0.16 0.18 0.68
Industrials 8.66 <.01*** 0.15 0.69 0.82 0.37 0.10 0.75 3.64 0.06* 2.71 0.11
Cnsmr Discret. 12.84 <.01*** 1.88 0.19 2.78 0.12 2.31 0.15 7.02 0.02** 0.25 0.62
Cnsmr Staples 5.84 0.07* 0.41 0.55 0.01 0.92 . . 5.44 0.08* 4.79 0.09*
Health Care 1.77 0.19 14.58 <.01*** 0.11 0.74 9.25 <.01*** 2.37 0.13 10.66 <.01***
Info Tech 3.79 0.06* 0.02 0.88 0.04 0.84 0.00 0.98 1.61 0.21 17.33 <.01***
GICS Sector FVL3= –1 FVA1=FVA2 FVA1=FVA3 FVA2=FVA3 FVL12=FVL3
Energy . . . . . . . . . .
Materials 0.09 0.77 1.99 0.19 0.14 0.72 0.76 0.40 0.07 0.79
Industrials 0.65 0.42 0.55 0.46 0.14 0.70 0.00 0.99 1.12 0.29
Cnsmr Discret. . . 2.18 0.16 2.99 0.11 4.41 0.05* . .
Cnsmr Staples . . 0.06 0.82 . . . . . .
Health Care 0.11 0.73 10.59 <.01*** 3.63 0.06* 7.86 <.01*** 0.11 0.74
Info Tech 0.00 0.94 0.00 0.99 0.01 0.93 0.01 0.92 0.00 0.95
116
Table 13 (cont’d)
Panel E. Estimated Coefficients from OLS Regression with 45th
-percentile with HHI High
GICS Sector Int NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 NI N
Adj
R2
Energy 1.99 0.94 ** -5.37 -2.93 33.21 -1.05 * 9.42 * -46.83 1.63 * 15 0.74
0.40 (0.70) 3.29 (0.01) -1.55 (0.17) -0.69 (0.51) 0.54 (0.61) -2.29 (0.06) 1.95 (0.09) -0.54 (0.60) 2.24 (0.06)
Materials -0.17 0.71 1.84 8.14 *** 5.07 ** -0.78 -0.07 1.02 2.20 *** 19 0.91
-0.07 (0.94) 1.63 (0.13) 1.61 (0.13) 4.32 (<.01) 2.45 (0.03) -1.56 (0.15) -0.06 (0.95) 0.09 (0.93) 3.52 (<.01)
Industrials -2.53 1.10 *** -0.68 5.47 -6.51 ** -0.99 *** -2.06 -384.94 0.95 ** 39 0.76
-1.10 (0.28) 5.60 (<.01) -0.68 (0.50) 1.52 (0.13) -2.59 (0.01) -3.56 (<.01) -1.00 (0.32) -0.59 (0.56) 2.33 (0.02)
Cnsmr Discret. -1.11 1.24 *** 1.74 * 5.87 -2.21 -1.12 *** -1.67 ** -0.49 1.60 ** 20 0.57
-0.30 (0.76) 3.96 (<.01) 2.05 (0.06) 0.40 (0.69) -1.04 (0.32) -3.79 (<.01) -2.95 (0.01) -0.07 (0.94) 2.62 (0.02)
Cnsmr Staples 2.40 0.75 * 0.12 58.35 -318.87 -0.44 -0.93 -75.99 0.83 12 0.97
1.25 (0.29) 2.70 (0.07) 0.10 (0.93) 1.87 (0.15) -1.39 (0.25) -1.46 (0.23) -0.77 (0.49) -1.64 (0.20) 0.35 (0.75)
Health Care 0.29 1.01 *** 1.79 *** 1.99 *** 0.93 -0.75 *** 2.78 *** -6.20 -0.34 110 0.87
0.59 (0.55) 7.11 (<.01) 6.78 (<.01) 7.89 (<.01) 1.10 (0.27) -3.61 (<.01) 6.09 (<.01) -0.59 (0.55) -1.20 (0.23)
117
Table 13 Panel E (cont’d)
GICS Sector Int NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 NI N
Adj
R2
Info Tech 1.59 ** 0.51 *** 1.78 *** 1.74 *** 1.09 ** -0.50 ** 9.85 6.42 1.62 *** 67 0.69
2.13 (0.03) 3.05 (<.01) 5.80 (<.01) 5.02 (<.01) 2.48 (0.01) -2.02 (0.04) 1.62 (0.10) 0.36 (0.72) 4.05 (<.01)
Total 282
This panel shows the estimated coefficients by GICS Sector of an OLS regression using the modified Ohlson model on the 45th-percentile
where the value of the Herfindahl-Hirschman Index (HHI) is high, i.e. the firms in concentrated industries. The model and variables are those
given in Panel A of Table 9.
118
Table 13 (cont’d)
Panel F. Results of F-Tests on Coefficients of the Previous Panel.
GICS Sector NFVA=1 FVA1=1 FVA2=1 FVA3=1 NFVL=_1 FVL12=_1
Energy 0.04 0.84 3.38 0.11 0.87 0.38 0.27 0.62 0.01 0.91 4.65 0.07*
Materials 0.42 0.52 0.55 0.47 14.34 <.01*** 3.87 0.07* 0.20 0.66 0.61 0.45
Industrials 0.26 0.61 2.81 0.10 1.55 0.22 8.93 <.01*** 0.00 0.96 0.26 0.61
Cnsmr Discret. 0.60 0.45 0.76 0.40 0.11 0.74 2.28 0.15 0.17 0.68 1.40 0.26
Cnsmr Staples 0.85 0.42 0.49 0.53 3.39 0.16 1.95 0.25 3.33 0.16 0.00 0.95
Health Care 0.00 0.97 8.91 <.01*** 15.45 <.01*** 0.01 0.93 1.42 0.23 68.60 <.01***
Info Tech 8.35 <.01*** 6.46 0.01** 4.57 0.03** 0.04 0.84 4.08 0.04** 3.20 0.07*
GICS Sector FVL3=–1 FVA1=FVA2 FVA1=FVA3 FVA2=FVA3 FVL12=FVL3
Energy 0.28 0.61 0.23 0.64 0.39 0.55 0.33 0.58 0.42 0.54
Materials 0.03 0.86 7.61 0.02** 3.18 0.10 1.31 0.27 0.01 0.92
Industrials 0.35 0.56 2.35 0.13 4.57 0.04** 11.26 <.01*** 0.34 0.56
Cnsmr Discret. 0.01 0.94 0.08 0.78 3.12 0.10 0.27 0.61 0.03 0.86
Cnsmr Staples 2.61 0.20 3.61 0.15 1.93 0.25 2.10 0.24 2.70 0.19
Health Care 0.24 0.62 0.32 0.57 0.81 0.37 1.46 0.22 0.72 0.39
Info Tech 0.17 0.68 0.01 0.94 1.96 0.16 1.27 0.26 0.03 0.85
This panel shows the results of F-tests that examine various properties of and relationships between coefficients within a regression model.
119
Table 13 (cont’d)
Panel G. Significant Pairs of Coefficients from OLS Regressions on HHI Low and High
45th
Percentiles
GICS Sector FVA1
Competitive, Conc.
FVA2
Competitive, Conc.
FVA3
Competitive, Conc.
Expected
Info. Tech. 1.06**, 1.78*** 1.07***, 1.74***
Not Expected
Cons. Disc. 2.43**, 1.74*
Mixed
Health Care 2.90***, 1.79*** [N] 0.84*, 1.99*** [E]
This table summarizes findings from OLS regressions presented in Panels C and E. The first
coefficient in each pair is the coefficient from Panel C the subsample with HHI low, while the
second in each pair is the corresponding coefficient from Panel E, the subsample with HHI
high. The upper part of the table shows the coefficients in the industry sector whose pattern of
coefficients followed the hypothesized pattern, while the middle part shows the coefficients of
the industry sector whose pattern of coefficients does not follow the hypothesized pattern. The
lower part shows the coefficients of the industry sector with mixed results. One pair of
significant coefficients exhibits (does not exhibit) the hypothesized pattern, and is followed by
an [E] ([N]) to indicate this. The number of stars after each coefficient represents its statistical
significance, following the convention described in Panel A of Table 9.
120
Table 13 (cont’d)
Panel H. Results of IRLS Regression on 45th
-percentile with HHI Low
GICS Sector Icept NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 NI N
Adj
R2
Materials 1.07 0.81 *** 5.82 *** -12.35 *** 1.41 -0.41 *** -2.47 *** -30.38 * 1.54 *** 17 0.30
1.88 (0.17) 198.42 (<.01) 24.57 (<.01) 57.11 (<.01) 2.41 (0.12) 47.08 (<.01) 7.38 (<.01) 3.03 (0.08) 125.32 (<.01)
Industrials 1.03 0.77 *** 1.80 -0.68 1.45 -0.79 *** -0.70 ** -2.40 1.51 *** 38 0.58
0.46 (0.49) 36.61 (<.01) 1.29 (0.25) 0.05 (0.81) 0.03 (0.86) 12.08 (<.01) 5.74 (0.01) 1.71 (0.19) 14.41 (<.01)
Cnsmr Discret. 0.79 ** 0.34 *** 2.64 *** 15.00 *** 0.53 -0.21 ** -9.26 *** 0.00 . 0.29 *** 19 0.43
4.35 (0.03) 31.08 (<.01) 90.69 (<.01) 34.13 (<.01) 0.05 (0.81) 5.38 (0.02) 12.33 (<.01) . . 13.00 (<.01)
Health Care 0.86 1.17 *** 2.95 *** 1.47 *** 6.24 *** -1.28 *** 1.69 * 200.52 1.30 *** 46
0.78 (0.37) 49.12 (<.01) 31.17 (<.01) 8.10 (<.01) 10.81 (<.01) 29.68 (<.01) 3.08 (0.07) 0.14 (0.71) 6.65 (<.01) 0.60
Info Tech 1.74 * 0.67 *** 0.86 ** 1.18 *** 1.07 -0.52 * -0.08 8.80 0.78 *** 41
3.03 (0.08) 23.12 (<.01) 4.31 (0.03) 12.31 (<.01) 2.29 (0.13) 2.88 (0.08) 0.13 (0.71) 0.01 (0.91) 11.96 (<.01) 0.63
Total 161
This panel shows the estimated coefficients by GICS Sector of an IRLS regression using the modified Ohlson model on the 45th-percentile
where the Herfindahl-Hirschman Index (HHI) is low, i.e. the firms in unconcentrated industries. The model and variables are those given in
Panel A of Table 9.
121
Table 13 (cont’d)
Panel I. Results of IRLS Regression on 45th
-percentile with HHI High
GICS Sector Icept NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 NI N
Adj
R2
Energy -1.03 1.06 *** -5.39 *** -2.51 -13.65 -1.14 *** 8.76 *** 18.99 1.52 *** 15 0.42
0.29 (0.59) 92.40 (<.01) 16.26 (<.01) 2.38 (0.12) 0.33 (0.56) 41.09 (<.01) 22.06 (<.01) 0.32 (0.57) 29.45 (<.01)
Materials 0.41 1.04 *** 0.90 8.24 *** 3.79 *** -1.15 *** -0.62 1.14 2.33 *** 19 0.21
0.07 (0.78) 13.58 (<.01) 1.51 (0.21) 46.57 (<.01) 8.19 (<.01) 13.08 (<.01) 0.67 (0.41) 0.02 (0.88) 33.87 (<.01)
Industrials -2.07 1.10 *** -0.63 5.01 -6.76 ** -0.99 *** -2.01 -329.06 0.95 ** 39 0.55
0.73 (0.39) 28.07 (<.01) 0.36 (0.54) 1.75 (0.18) 6.52 (0.01) 11.42 (<.01) 0.86 (0.35) 0.23 (0.63) 4.83 (0.02)
Cnsmr Discret. -5.80 1.50 *** 1.85 ** 5.63 -4.46 ** -1.16 *** -1.89 *** -6.39 2.37 *** 20 0.46
2.60 (0.10) 24.03 (<.01) 4.99 (0.02) 0.15 (0.69) 4.61 (0.03) 16.14 (<.01) 11.75 (<.01) 0.91 (0.33) 16.00 (<.01)
Health Care 0.62 * 0.57 *** 1.35 *** 1.55 *** 1.77 *** -0.11 3.89 *** 2.45 0.29 110
2.85 (0.09) 28.88 (<.01) 47.64 (<.01) 67.96 (<.01) 7.91 (<.01) 0.47 (0.49) 131.47 (<.01) 0.10 (0.75) 1.89 (0.16) 0.33
Info Tech 1.08 * 0.54 *** 1.25 *** 2.01 *** 1.02 *** -0.50 ** 6.70 5.96 1.36 *** 67
2.75 (0.09) 13.66 (<.01) 21.74 (<.01) 43.62 (<.01) 7.02 (<.01) 5.26 (0.02) 1.60 (0.20) 0.14 (0.70) 15.18 (<.01) 0.55
Total 270
This panel shows the estimated coefficients by GICS Sector of an IRLS regression using the modified Ohlson model on the 45th-percentile
where the Herfindahl-Hirschman Index (HHI) is high, i.e. the firms in concentrated industries. The model and variables are those given in
Panel A of Table 9.
122
Table 13 (cont’d)
Panel J. Significant Pairs of Coefficients from IRLS Regressions on HHI Low and High
45th
Percentiles
GICS Sector FVA1
Competitive, Conc.
FVA2
Competitive, Conc.
FVA3
Competitive, Conc.
Expected
Info. Tech. 0.86**, 1.25*** 1.18***, 2.01***
Not Expected
Cons. Disc. 2.64***, 1.85**
Mixed
Health Care 2.95***, 1.35*** [N] 1.47***, 1.55*** [E] 6.24***, 1.77*** [N]
This table summarizes findings from IRLS regressions presented in Panels H and I. The first
coefficient in each pair is the coefficient from Panel H the subsample with HHI low, while the
second in each pair is the corresponding coefficient from Panel I, the subsample with HHI
high. The upper part of the table shows the coefficients in industry sector whose pattern of
coefficients followed the hypothesized pattern, while the middle part shows the coefficients of
industry sector whose pattern of coefficients did not follow the hypothesized pattern. The
lower part shows the coefficients of the industry sector with mixed results. The pairs (pair) of
significant coefficients that exhibit (does not exhibit) the hypothesized pattern are (is)
followed by an [E] ([N]) to indicate this. The number of stars after each coefficient represents
its statistical significance, following the convention described in Panel A of Table 9.
123
Table 14. Three-year Average Herfindahl of Net Sales
Panel A. Descriptive Statistics
Three-year Average H
GICS Sector N MIN Q1 MEDIAN Q3 MAX MEAN STD
Energy 125 0.04 0.05 0.05 0.05 0.05 0.05 0.01
Materials 62 0.05 0.05 0.05 0.05 0.24 0.06 0.03
Industrials 194 0.03 0.03 0.07 0.11 0.26 0.08 0.05
Cnsmr Discret. 139 0.03 0.03 0.04 0.11 0.20 0.07 0.05
Cnsmr Staples 36 0.06 0.06 0.08 0.22 0.41 0.12 0.09
Health Care 210 0.05 0.06 0.07 0.15 0.15 0.09 0.05
Financials 480 0.03 0.03 0.03 0.04 0.19 0.06 0.05
Info Tech 256 0.05 0.05 0.11 0.12 0.33 0.10 0.04
Telecom Svcs. 26 0.06 0.06 0.06 0.10 0.10 0.08 0.02
Utilities 52 0.04 0.04 0.05 0.08 0.25 0.08 0.07
Summaries 1580 0.03 0.03 0.05 0.11 0.41 0.08 0.05
This panel shows descriptive statistics of the average of three-year Herfindahl of net sales by
GICS Sector. The details of the computation of the average of three-year Herfindahl of net sales
are provided in Appendix B. Q1 is the first quartile and Q3 is the third quartile. STD is the
standard deviation.
124
Table 14 (cont’d)
Panel B. The Low and High 35th
-percentiles (the mezzanine 30-percentile is not shown)
Havg Low Havg High
Havg Havg
GICS Sector N Min Max N Min Max
Energy 125 0.04 0.05 106 0.05 0.05
Materials 38 0.05 0.05 24 0.05 0.24
Industrials 71 0.03 0.06 82 0.08 0.26
Cnsmr Discret. 63 0.03 0.03 53 0.08 0.20
Cnsmr Staples 16 0.06 0.06 20 0.08 0.41
Health Care 76 0.05 0.06 75 0.15 0.15
Financials 219 0.03 0.03 187 0.04 0.19
Info Tech 100 0.05 0.08 117 0.11 0.33
Telecom Svcs. 17 0.06 0.06 26 0.06 0.10
Utilities 22 0.04 0.04 24 0.07 0.25
Total 747 0.03 0.08 714 0.04 0.41
This table shows the minimum and maximum values of the three year average Herfindahl
of net sales within the high and low 35th-percentiles by GICS sector. Thirty-fifth-
percentiles produces insufficient separation for the following three industries: Energy,
Materials, and Telecommunication Services. This can be seen by observing that the
maximum value of the low 35th-percentile equals the minimum value of the high 35th-
percentile. Furthermore, Consumer Staples, with 16 (20) observations in the low (high)
35th-percentile has insufficient observations for regression studies. Therefore, these four
industries will be dropped in the regression studies that follow.
125
Table 14 (cont’d)
Panel C. Number of Firms in the Low 35th
-percentile, Between 30th
-percentile, and
High 35th
-percentile of Three-year Average Herfindahl of Net Sales by Industry
Sector
GICS Sector Sample Low 35th
Between High 35th
Industrials 194 71 41 82
Cnsmr Discret. 139 63 23 53
Health Care 210 76 59 75
Financials 480 219 74 187
Info Tech 256 100 39 117
Utilities 52 22 6 24
Total 1331 551 242 538
This panel shows the number of firms in the low and high 35th percentiles of the three-year
average Herfindahl by GICS Sector. Due to lack of separation, as shown in the previous
panel, three sectors have been removed from this portion of the study: Energy, Materials,
and Telecommunication Services. The column entitled “Between” is the number of firms
in the 30th-percenthile that lies between the low and high 35th-percentiles. The value in the
column entitled “Sample” is the subtotal of the three columns to the right: “Low 35th”-
percentile, between, and “High 35th”-percentile.
126
Table 14. (cont’d)
Panel D. Estimates of Coefficients by IRLS of the Subsample with Average Herfindahl of Net Sales Low (Low 35th
-percentile)
GICS Sector Icept NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 NI N
Adj
R2
Industrials 3.02 * 0.55 *** 0.54 0.24 0.05 -0.44 *** -0.64 0.77 1.37 *** 71 0.34
2.95 (0.08) 15.49 (<.01) 1.55 (0.21) 0.24 (0.62) 0.01 (0.92) 6.64 (<.01) 1.59 (0.20) 0.06 (0.81) 14.6 (<.01)
Cnsmr Discret. 1.93 0.56 *** 1.70 7.75 *** 1.51 -0.52 *** -0.84 67.99 * 0.39 ** 63 0.28
1.83 (0.17) 17.40 (<.01) 1.13 (0.28) 7.97 (<.01) 0.31 (0.57) 11.94 (<.01) 0.87 (0.35) 3.03 (0.08) 5.93 (0.01)
Health Care 2.12 ** 0.94 *** 1.84 *** 0.64 * 0.35 -0.98 *** 4.35 *** 102.5 1.65 *** 76
6.44 (0.01) 54.28 (<.01) 19.03 (<.01) 2.75 (0.09) 0.09 (0.75) 29.76 (<.01) 47.73 (<.01) 1.01 (0.31) 15.7 (<.01) 0.48
Financials 3.88 *** 0.37 *** 0.36 *** 0.50 *** 0.59 *** -0.40 *** -0.82 *** -3.58 1.23 *** 219
30.51 (<.01) 28.85 (<.01) 12.24 (<.01) 45.84 (<.01) 9.47 (<.01) 27.89 (<.01) 8.43 (<.01) 1.98 (0.15) 93.8 (<.01) 0.40
Info Tech 1.87 *** 0.61 *** 1.57 *** 1.71 *** 1.13 *** -0.61 *** 0.06 1.01 0.73 *** 100
8.06 (<.01) 34.59 (<.01) 31.65 (<.01) 36.32 (<.01) 9.07 (<.01) 13.23 (<.01) 0.10 (0.75) 0.00 (0.95) 13.4 (<.01) 0.56
Utilities 4.65 -0.26 -1.73 0.85 1.29 0.56 0.65 -2.00 5.07 22
0.34 (0.55) 0.24 (0.62) 0.42 (0.51) 0.11 (0.74) 0.05 (0.81) 0.73 (0.39) 0.16 (0.69) 0.67 (0.41) 2.61 (0.10) 0.38
Total 731
127
Table 14 Panel D (cont’d)
This panel shows the estimated coefficients by GICS Sector of an IRLS regression using the modified Ohlson model on the 35th-percentile
where the average Herfindahl of net sales is low, i.e. the firms in unconcentrated industries. The model and variables are those given in Panel
A of Table 9.
128
Table 14 (cont’d)
Panel E. Estimates of Coefficients by IRLS of the Subsample with Average Herfindahl of Net Sales High (High 35th
-percentile)
GICS Sector Icept NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 NI N
Adj
R2
Industrials 0.71 0.90 *** 1.96 *** -0.23 0.35 -0.89 *** -1.88 *** -4.47 ** 0.53 82 0.39
0.28 (0.59) 45.52 (<.01) 14.42 (<.01) 0.19 (0.66) 0.01 (0.93) 25.71 (<.01) 8.46 (<.01) 3.99 (0.04) 1.67 (0.19)
Cnsmr Discret. 2.86 ** 0.83 *** 1.71 ** 1.85 *** 0.18 -0.98 *** 3.58 -21.53 0.39 *** 53 0.53
5.69 (0.01) 27.47 (<.01) 5.86 (0.01) 24.24 (<.01) 0.13 (0.72) 17.06 (<.01) 2.52 (0.11) 0.16 (0.68) 7.73 (<.01)
Health Care 0.47 * 0.94 *** 0.77 *** 1.37 *** 0.53 0.50 *** 3.41 *** 3.24 0.39 *** 75 0.45
3.55 (0.05) 36.35 (<.01) 26.30 (<.01) 85.37 (<.01) 1.12 (0.29) 9.75 (<.01) 71.93 (<.01) 0.54 (0.46) 7.67 (<.01)
Financials 1.38 ** 0.79 *** 0.80 *** 0.85 *** 0.57 *** -0.82 *** -0.97 ** -0.80 *** 0.38 *** 187 0.51
5.43 (0.01) 1,253.80 (<.01) 717.42 (<.01) 1,906.74 (<.01) 50.11 (<.01) 1,334.51 (<.01) 5.00 (0.02) 27.29 (<.01) 9.88 (<.01)
Info Tech 0.66 0.65 *** 1.25 *** 1.64 *** 1.19 *** -0.41 *** -0.60 4.95 * 1.25 *** 117 0.48
1.78 (0.18) 85.50 (<.01) 75.36 (<.01) 45.02 (<.01) 14.47 (<.01) 18.80 (<.01) 0.32 (0.57) 2.87 (0.09) 37.86 (<.01)
Utilities 5.21 * 1.12 *** 3.30 *** 1.94 * -25.49 * -1.24 *** -4.38 *** 9.78 2.40 ** 24 0.60
3.81 (0.05) 19.75 (<.01) 13.25 (<.01) 2.92 (0.08) 3.45 (0.06) 12.81 (<.01) 10.66 (<.01) 0.43 (0.50) 5.66 (0.01)
714
129
Table 14 Panel E (cont’d)
This panel shows the estimated coefficients by GICS Sector of an OLS regression using the modified Ohlson model on the 35th-percentile
where the average Herfindahl of net sales is high, i.e. the firms in concentrated industries. The model and variables are those given in Panel A
of Table 9
130
Table 14 (cont’d)
Panel F. Pairs of Significant Coefficients of Fair Value of Assets for Low and High
Herfindahl
GICS Sector FVA1
Competitive, Conc.
FVA2
Competitive, Conc.
FVA3
Competitive, Conc.
Not Expected
Cons. Disc. 7.75***, 1.85***
Mixed
Health Care 1.84***, 0.77*** [N] 0.64*, 1.37*** [E]
Financials 0.36***, 0.80*** [E] 0.50***, 0.85*** [E] 0.59***, 0.57*** [N]
Info. Tech. 1.57***, 1.25*** [N] 1.71***, 1.64*** [N] 1.13***, 1.19*** [E]
This table summarizes findings from IRLS regressions presented in Panels D and E. The first
coefficient in each pair is the coefficient from Panel D the subsample where the average
Herfindahl of net sales is low, while the second in each pair is the corresponding coefficient from
Panel E, the subsample where the average Herfindahl of net sales is high. The upper part of the
table shows the coefficients in industry sectors whose pattern of coefficients follows the
hypothesized pattern, while the middle part shows the coefficients of industry sectors whose
pattern of coefficients does not follow the hypothesized pattern. The lower part shows the
coefficients of the industry sector with mixed results. One pair of significant coefficients exhibits
(does not exhibit) the hypothesized pattern, and is followed by an [E] ([N]) to indicate this. The
number of stars after each coefficient represents its statistical significance, following the
convention described in Panel A of Table 9.
131
Table 14 (cont’d)
Panel G. Whole Sample with Herfindahl Indicator
Intercept NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 H_Ind N Adj R2
Coefficient 2.76*** 0.79*** 0.82*** 0.85*** 0.64*** –0.83*** –1.11*** –0.64*** 0.036 1,580 0.39
Chi-Square 157.86 13135 2099 105 105 10647 408 33.2 0.01
p-value (<0.01) (<0.01) (<0.01) (<0.01) (<0.01) (<0.01) (<0.01) (<0.01) 0.09
This table shows the estimated coefficients of an IRLS regression of the entire sample with an indicator variable, H_Ind, which takes the
value 0 for unconcentrated industries and 1 for concentrated industries. The model is a modified Ohlson model with the indicator variable
included as an additional term:
PRCit = α0 + α1NFVAit + α2FVA1it + α3FVA2it + α4FVA3it
+ α5NVFLit + α6FVL12it + α7FVL3it + α8 H_Indit + β1NIit + εit (3)
Other variables are those given in Panel A of Table 9.
132
Table 14 (cont’d)
Panel H1. First half of Coefficients of IRLS regression with an Indicator Variable for
Herfindahl above Median by GICS Sector
GICS Sector FVA1 FVA2 FVA3 H_Ind N Adj R2
Energy -0.32 1.19 *** 0.79 ** 3.78 ** 125 0.50
0.67 (0.41) 19.95 (<.01) 4.94 (0.02) 4.97 (0.02)
Materials 0.21 6.90 *** 2.47 -0.30 62 0.58
0.07 (0.78) 15.38 (<.01) 2.40 (0.12) 0.01 (0.93)
Industrials 1.05 *** 0.63 *** 0.36 -2.15 ** 194 0.38
18.15 (<.01) 6.82 (<.01) 0.58 (0.44) 3.94 (0.04)
Cnsmr Discret. 1.16 ** 2.25 *** 0.20 0.14 139 0.35
5.02 (0.02) 42.03 (<.01) 0.16 (0.68) 0.02 (0.89)
Cnsmr Staples 7.93 *** 65.59 *** -329.57 *** -6.17 *** 36 0.64
89.73 (<.01) 150.08 (<.01) 159.13 (<.01) 6.92 (<.01)
Health Care 1.53 *** 1.37 *** 4.95 *** 0.33 210 0.44
76.11 (<.01) 150.73 (<.01) 49.81 (<.01) 0.22 (0.63)
Financials 0.74 *** 0.82 *** 0.57 *** 1.15 * 480 0.41
866.00 (<.01) 2,246.50 (<.01) 66.44 (<.01) 3.51 (0.06)
Info Tech 1.36 *** 1.82 *** 1.16 *** -0.38 256 0.50
106.65 (<.01) 107.38 (<.01) 20.51 (<.01) 0.66 (0.41)
Telecom Svcs. 1.04 ** 0.26 -6.29 ** 4.46 *** 26 0.56
5.08 (0.02) 0.05 (0.82) 4.04 (0.04) 7.12 (<.01)
Utilities 1.32 ** 0.34 -0.42 -0.38 52 0.42
3.96 (0.04) 0.17 (0.67) 0.02 (0.88) 0.02 (0.88)
1580
The remaining coefficients of the regression are shown in the next half-panel.
133
Table 14 (cont’d)
Panel H2. Remaining half of Coefficients of IRLS regression with an Indicator Variable
for Herfindahl above Median by GICS Sector
GICS Sector Icept NFVA NFVL FVL12 FVL3 NI
Energy -3.47 ** 1.10 *** -1.18 *** -1.11 *** -1.19 0.51 ***
4.00 (0.04) 174.48 (<.01) 84.74 (<.01) 57.13 (<.01) 1.12 (0.29) 16.77 (<.01)
Materials -0.21 1.35 *** -1.38 *** -0.95 *** -2.94 2.10 ***
0.02 (0.87) 78.30 (<.01) 58.86 (<.01) 7.52 (<.01) 0.05 (0.82) 26.76 (<.01)
Industrials 3.44 *** 0.70 *** -0.65 *** -1.18 *** -1.72 1.18 ***
12.16 (<.01) 179.36 (<.01) 96.24 (<.01) 10.36 (<.01) 1.26 (0.26) 30.76 (<.01)
Cnsmr Discret. 3.59 *** 0.57 *** -0.55 *** -1.08 *** -1.69 0.44 ***
14.90 (<.01) 49.59 (<.01) 31.79 (<.01) 8.95 (<.01) 0.12 (0.72) 25.25 (<.01)
Cnsmr Staples 1.61 1.19 *** -0.97 *** -2.20 *** -
121.08
*** 1.85 ***
0.85 (0.35) 69.53 (<.01) 37.38 (<.01) 16.04 (<.01) 17.81 (<.01) 7.40 (<.01)
Health Care 0.76 1.34 *** -1.54 *** 1.96 *** -1.28 *** -0.08
1.62 (0.20) 341.52 (<.01) 289.08 (<.01) 33.78 (<.01) 9.44 (<.01) 0.20 (0.65)
Financials 0.96 ** 0.74 *** -0.77 *** -1.06 *** -0.78 *** 0.71 ***
4.17 (0.04) 1,602.19 (<.01) 1,672.55 (<.01) 99.37 (<.01) 48.64 (<.01) 59.08 (<.01)
Info Tech 1.76 *** 0.59 *** -0.42 *** -0.03 3.16 1.15 ***
13.42 (<.01) 107.31 (<.01) 28.59 (<.01) 0.03 (0.85) 1.09 (0.29) 71.57 (<.01)
Telecom Svcs. 0.00 0.80 *** -0.44 * -0.18 -3.71 ** 1.48 **
0.00 (0.99) 15.48 (<.01) 3.31 (0.06) 0.04 (0.84) 5.16 (0.02) 4.63 (0.03)
Utilities 6.77 ** 0.32 -0.16 -2.16 *** -1.38 2.59 **
4.18 (0.04) 1.68 (0.19) 0.27 (0.60) 8.26 (<.01) 0.50 (0.48) 4.76 (0.02)
These two half-panels together show the estimated coefficients by GICS Sector of an IRLS
regression on the entire sample using the modified Ohlson model with an indicator variable for
high or low average Herfindahl of net sales. The model is shown in Panel G.
134
Table 15. Entering, Incumbent, and Exiting Firms
Panel A. The 2008 Cohort from 2001–2009
Data Year - Fiscal
GICS Sector 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Energy 67 75 83 95 103 113 120 125 125 125 125
Materials 45 48 49 53 57 59 59 60 61 61 61
Industrials 146 161 165 173 180 187 194 195 195 195 195
Cnsmr Discret. 107 119 125 129 134 136 138 138 138 138 137
Cnsmr Staples 28 30 31 33 33 34 35 35 36 36 36
Health Care 138 153 170 187 198 203 207 208 210 210 210
Financials 318 345 375 399 423 450 471 476 481 481 481
Info Tech 186 203 211 234 243 249 254 256 256 256 255
Telecom Svcs. 13 17 18 21 22 22 24 24 26 26 26
Utilities 47 47 47 48 50 51 51 52 52 52 52
Total 1095 1198 1274 1372 1443 1504 1553 1569 1580 1580 1578
Panel B. Change in the Cohort between Each Fiscal Year and 2008
Data Year - Fiscal
GICS Sector 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Energy 58 50 42 30 22 12 5 0 0 0 0
Materials 16 13 12 8 4 2 2 1 0 0 0
Industrials 49 34 30 22 15 8 1 0 0 0 0
Cnsmr Discret. 31 19 13 9 4 2 0 0 0 0 1
Cnsmr Staples 8 6 5 3 3 2 1 1 0 0 0
Health Care 72 57 40 23 12 7 3 2 0 0 0
Financials 163 136 106 82 58 31 10 5 0 0 0
Info Tech 70 53 45 22 13 7 2 0 0 0 1
Telecom Svcs. 13 9 8 5 4 4 2 2 0 0 0
Utilities 5 5 5 4 2 1 1 0 0 0 0
Total Change 485 382 306 208 137 76 27 11 0 0 2
135
Table 15 (cont’d)
Panel C. Results of OLS Regression on Subsample of Entering Firms
GICS
Sector Icept NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 NI N
Adj
R2
Energy 0.13 0.86 *** -0.51 0.84 *** 0.72 ** -0.82 *** -0.82 *** -1.18 0.48 ** 58 0.40
0.01 (0.92) 29.21 (<.01) 0.24 (0.62) 8.79 (<.01) 5.70 (0.01) 15.28 (<.01) 18.71 (<.01) 1.08 (0.29) 3.98 (0.04)
Industrials 2.74 ** 0.25 ** 0.90 * -1.31 -0.30 -0.23 -0.80 * 0.10 0.43 49 0.17
6.45 (0.01) 5.84 (0.01) 3.41 (0.06) 1.09 (0.29) 0.14 (0.70) 2.37 (0.12) 2.88 (0.08) 0.01 (0.93) 2.28 (0.13)
Cnsmr
Discret.
1.20 0.76 *** 2.41 ** 5.39 *** -0.34 -0.73 ** 6.65 ** 24.79 0.58 *** 31 0.38
0.61 (0.43) 6.76 (<.01) 4.61 (0.03) 8.31 (<.01) 0.63 (0.42) 4.43 (0.03) 5.31 (0.02) 0.52 (0.47) 7.75 (<.01)
Health
Care
1.28 ** 0.57 *** 0.68 *** 0.74 *** 0.37 -0.62 *** 6.13 1.46 -0.19 72 0.23
5.96 (0.01) 15.83 (<.01) 6.97 (<.01) 17.50 (<.01) 0.15 (0.70) 9.74 (<.01) 0.29 (0.58) 0.03 (0.85) 1.07 (0.30)
Financials 2.83 *** 0.45 *** 0.67 *** 0.54 *** 0.42 *** -0.48 *** -0.31 -0.59 *** 0.42 *** 163 0.31
21.63 (<.01) 96.63 (<.01) 55.49 (<.01) 152.80 (<.01) 30.24 (<.01) 101.90 (<.01) 0.78 (0.37) 26.39 (<.01) 12.31 (<.01)
Info Tech 1.49 * 0.43 *** 2.10 *** 1.49 *** 2.52 -0.24 1.10 2.76 1.03 *** 70 0.43
3.63 (0.05) 7.51 (<.01) 26.70 (<.01) 16.64 (<.01) 0.98 (0.32) 0.78 (0.37) 0.30 (0.58) 0.66 (0.41) 9.38 (<.01)
Total 443
This panel shows the estimated coefficients by GICS Sector of an OLS regression using the modified Ohlson model on the entering firms.
The model and variables are those given in Panel A of Table 9.
136
Table 15 (cont’d)
Panel D. Results of OLS Regression on Incumbents (Including Unidentifiable Exiting Firms)
GICS
Sector Icept NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 NI N
Adj
R2
Energy -1.57 1.08 *** -0.51 4.02 *** -0.41 -1.14 *** -1.68 *** 0.86 0.64 *** 67 0.67
0.75 (0.38) 88.89 (<.01) 1.15 (0.28) 12.37 (<.01) 0.11 (0.74) 41.84 (<.01) 23.26 (<.01) 0.21 (0.64) 16.23 (<.01)
Materials -0.51 1.52 *** -0.09 7.14 *** 3.18 -1.60 *** -1.17 *** -22.6 2.25 *** 46 0.66
0.12 (0.72) 118.39 (<.01) 0.01 (0.90) 21.02 (<.01) 2.41 (0.12) 80.72 (<.01) 13.53 (<.01) 0.20 (0.65) 27.64 (<.01)
Industrials 3.16 *** 0.67 *** 0.97 *** 0.48 * 0.39 -0.59 *** -0.92 ** -1.84 1.41 *** 145 0.43
10.41 (<.01) 143.53 (<.01) 12.97 (<.01) 3.42 (0.06) 0.57 (0.45) 66.41 (<.01) 4.54 (0.03) 0.85 (0.35) 36.14 (<.01)
Cnsmr
Discret.
3.16 *** 0.57 *** 1.29 ** 2.22 *** 1.63 -0.56 *** -1.15 *** -1.78 0.41 *** 108 0.38
9.34 (<.01) 44.27 (<.01) 4.92 (0.02) 35.98 (<.01) 1.07 (0.30) 29.02 (<.01) 8.32 (<.01) 0.12 (0.72) 12.42 (<.01)
Cnsmr
Staples
-0.36 1.15 *** 6.70 *** 58.89 *** -300 *** -0.91 *** -2.09 ** -105 ** 2.40 * 28 0.64
0.02 (0.88) 28.93 (<.01) 21.22 (<.01) 40.88 (<.01) 47.04 (<.01) 14.81 (<.01) 6.49 (0.01) 5.72 (0.01) 2.96 (0.08)
Health Care 1.30 ** 1.36 *** 1.43 *** 1.42 *** 5.26 *** -1.58 *** 1.74 *** -1.49 *** 0.12 138 0.58
5.12 (0.02) 299 (<.01) 51.53 (<.01) 125.74 (<.01) 42.40 (<.01) 264 (<.01) 24.98 (<.01) 12.05 (<.01) 0.20 (0.65)
Financials 1.94 *** 0.73 *** 0.74 *** 0.82 *** 0.56 *** -0.77 *** -1.10 *** -0.96 *** 0.83 *** 317 0.44
10.87 (<.01) 1,117 (<.01) 646 (<.01) 1,619 (<.01) 25.03 (<.01) 1,175 (<.01) 82.72 (<.01) 28.23 (<.01) 45.69 (<.01)
Info Tech 1.12 ** 0.64 *** 1.19 *** 2.06 *** 1.08 *** -0.42 *** -0.17 -2.31 1.10 *** 186 0.50
6.38 (0.01) 106 (<.01) 71.62 (<.01) 99.78 (<.01) 17.79 (<.01) 26.53 (<.01) 1.34 (0.24) 0.07 (0.79) 55.36 (<.01)
137
Table 15 Panel D (cont’d)
GICS Sector Icept NFVA FVA1 FVA2 FVA3 NFVL FVL12 FVL3 NI N
Adj
R2
Telecom
Svcs.
1.59 *** -0.28 *** 3.39 *** 8.37 *** 9.53 *** 0.31 *** 1.40 *** 10.00 *** 6.03 *** 13 0.29
623 (<.01) 981 (<.01) 1,894 (<.01) 45565 (<.01) 10595 (<.01) 418 (<.01) 2,365 (<.01) 17145 (<.01) 91687 (<.01)
Utilities 7.85 *** -0.11 1.28 ** -1.03 -1.18 0.29 -1.74 ** -0.56 5.60 *** 47 0.48
7.34 (<.01) 0.15 (0.69) 4.42 (0.03) 1.18 (0.27) 0.19 (0.65) 0.70 (0.40) 5.83 (0.01) 0.11 (0.74) 11.05 (<.01)
Total 1095
This panel shows the estimated coefficients by GICS Sector of an OLS regression using the Modified Ohlson model on the incumbent firms.
Although this group is called “incumbent firms,” it also includes unidentifiable exiting firms. The model and variables are those given in
Panel A of Table 9.
138
Table 15 (cont’d)
Panel E. Significant Pairs of Fair Value of Assets of Entering and Incumbent Firms
GICS Sector FVA1
Entering, Incumb.
FVA2
Entering, Incumb.
FVA3
Entering,
Incumb.
Expected Pattern
Energy 0.84***, 4.02***
Industrials 0.90*, 0.97***
Health Care 0.68***, 1.43*** 0.74***, 1.42***
Financials 0.67***, 0.74*** 0.54***, 0.82*** 0.42***, 0.56***
Not Expected Pattern
Cons. Discret. 2.41**, 1.29** 5.39***, 2.22***
Mixed
Info. Tech. 2.10***, 1.19*** [N] 1.49***, 2.06*** [E]
This table summarizes findings from OLS regressions presented in Panels C and D. The
first coefficient in each pair is the coefficient from Panel C the subsample of entering firms,
while the second in each pair is the corresponding coefficient from Panel D, the subsample
of incumbents and unidentifiable exiting firms. The upper part of the table shows the
coefficients in industry sectors whose pattern of coefficients follows the hypothesized
pattern, while the middle part shows the coefficients of the industry sector whose pattern of
coefficients did not follow the hypothesized pattern. The lower part shows the coefficients
of the industry sector with mixed results. One pair of significant coefficients exhibits (does
not exhibit) the hypothesized pattern, and is followed by an [E] ([N]) to indicate this. The
number of stars after each coefficient represents its statistical significance, following the
convention described in Panel A of Table 9.
139
Table 15 (cont’d)
Panel F. Results of One-sided t-Test of log(AT) by Status as Entering or Incumbent
Firms
Group Method
N
Mean 95% CL Mean Std Dev
95% CL Std
Dev
Entr = 0 1,095 6.97 6.85 7.08 1.94 1.86 2.03
Entr = 1 485 6.33 6.18 6.48 1.65 1.55 1.76
Diff (Entr=0 less
Entr=1)
Satterthwaite 0.64 0.48 Infty
This panel shows the results of a one-sided t-Test for the logarithm of total assets, log(AT),
by the firm’s status as an entering or incumbent.
Method Variances DF t-value p-value
Satterthwaite Unequal 1078.9 6.68 <.01
140
Table 15 (cont’d)
Panel G1. Certain Coefficients of OLS Regressions with an Indicator for Status, Size, and Their Interaction
Model FVA1 FVA2 FVA3 Entr
log(A)
Entr ×
logA
Adj R2
Status Indicator 0.84 *** 0.83 *** 0.67 *** -2.07 *** 0.93
t-value (p-value) 36.45 (<.01) 56.00 (<.01) 8.48 (<.01) -5.15 (<.01)
With Size 0.85 *** 0.82 *** 0.61 *** 0.69 *** 0.93
t-value (p-value) 37.05 (<.01) 56.12 (<.01) 7.77 (<.01) 6.37 (<.01)
With Status & Size 0.85 *** 0.83 *** 0.63 *** -1.77 *** 0.63 *** 0.93
t-value (p-value) 37.09 (<.01) 56.50 (<.01) 8.06 (<.01) -4.41 (<.01) 5.78 (<.01)
With Interaction 0.84 *** 0.83 *** 0.63 *** 3.68 ** 0.83 *** -0.84 *** 0.93
t-value (p-value) 37.02 (<.01) 56.77 (<.01) 8.08 (<.01) 2.40 (0.02) 6.85 (<.01) -3.68 (<.01)
N 1,580
The remaining coefficients of the regression are shown in the next half-panel.
141
Table 15 (cont’d)
Panel G2. Other Coefficients of OLS Regressions with an Indicator for Status, Size, and Their Interaction
Model Icept NFVA NFVL FVL12 FVL3 NI
Status Indicator 4.39 *** 0.79 *** -0.83 *** -0.83 *** -0.79 *** 0.81 ***
t-value (p-value) 16.61 (<.01) 88.94 (<.01) -80.40 (<.01) -11.81 (<.01) -5.57 (<.01) 14.12 (<.01)
With Size -0.60 0.78 *** -0.83 *** -0.88 *** -0.84 *** 0.82 ***
t-value (p-value) -0.85 (0.40) 88.94 (<.01) -80.88 (<.01) -12.47 (<.01) -5.95 (<.01) 14.32 (<.01)
With Status & Size 0.37 0.78 *** -0.83 *** -0.88 *** -0.83 *** 0.82 ***
t-value (p-value) 0.50 (0.61) 89.46 (<.01) -81.42 (<.01) -12.60 (<.01) -5.90 (0.72) 14.32 (<.01)
With Interaction -1.07 0.78 *** -0.83 *** -0.89 *** -0.80 *** 0.82 ***
t-value (p-value) -1.27 (0.20) 89.86 (<.01) -81.81 (<.01) -12.73 (<.01) -5.70 (<.01) 14.36 (<.01)
These two half-panels together show the estimated coefficients by GICS Sector of an OLS regression using the modified Ohlson model
on the entering firms:
PRCit = α0 + α1NFVAit + α2FVA1it + α3FVA2it + α4FVA3it
+ α5NVFLit + α6FVL12it + α7FVL3it
+ α8 Entrit + α9 log(Ait) + α10 Entrit ×log(Ait) + β1NIit + εit
Entri is an indicator variable that takes the value 1 for entering firms and zero for established. The variable log(Ait) is the natural
logarithm of the firm’s total assets. Other variables are those given in Panel A of Table 9.
142
Table 16. Investigation of the Effect of Liquidity, Using the Quick Ratio (QR) and
Operating Cash Flow Ratio (CR)
Panel A. Presence and Absence of Measures of Liquidity (QR and CR)
GICS Sector Trimmed QR Missing QR Present CR Missing CR Present
Energy 125 49 76 0 125
Materials 62 25 37 2 60
Industrials 194 103 91 12 182
Cnsmr Discret. 139 79 60 7 132
Cnsmr Staples 36 21 15 0 36
Health Care 210 136 74 3 207
Financials 480 471 9 455 25
Info Tech 256 130 126 1 255
Telecom Svcs. 26 13 13 0 26
Utilities 52 34 18 0 52
Total 1580 1061 519 480 1100
This panel shows the number of firms in each GICS Sector, the number of firms for which
the Quick Ratio (QR) is missing and present, and number of firms for which the Operating
Cash Flow Ratio (CR) is missing or present. The QR and CR were computed as follows:
LCT
XPPINVTACTQR
and
LCT
OANCFCR
where ACT = Current Assets – Total, INVT = Inventories – Total, XPP = Prepaid Expenses,
LCT = Liabilities Current – Total, and OANCF = Operating Activities – Net Cash Flow.
143
Table 16 (cont’d)
Panel B. Descriptive Statistics of the Quick Ratio (QR) by GICS Sector
GICS Sector N MIN Q1 MEDIAN Q3 MAX MEAN STD
Energy 76 0.13 0.68 1.08 1.46 7.85 1.35 1.26
Materials 37 0.21 0.87 1.10 1.91 48.66 2.71 7.82
Industrials 91 0.39 1.01 1.29 1.85 11.74 1.67 1.40
Cnsmr Discret. 60 0.21 0.65 1.06 1.57 11.50 1.45 1.60
Cnsmr Staples 15 0.56 0.64 0.87 1.34 22.27 2.37 5.52
Health Care 74 0.58 1.43 2.37 4.69 13.82 3.35 2.67
Financials 9 0.69 1.00 2.52 3.79 12.27 3.67 3.63
Info Tech 126 0.36 1.41 2.19 4.00 25.08 3.28 3.28
Telecom Svcs. 13 0.08 0.59 1.06 1.18 12.04 1.80 3.12
Utilities 18 0.11 0.58 0.73 1.00 1.42 0.78 0.33
Total 519 0.08 0.91 1.39 2.52 48.66 2.33 3.28
This panel shows descriptive statistics of Quick Ratio (QR) by GICS Sector. Q1 is the first
quartile and Q3 is the third quartile. STD is the standard deviation.
144
Table 16 (cont’d)
Panel C1. First Half of the Estimated Coefficients from an IRLS Regression with QR
Alone
GICS Sector FVA1 FVA2 FVA3 QR N Adj R2
Energy -0.83 1.47 * 0.94 ** 0.14 76 0.49
0.59 (0.44) 3.80 (0.05) 6.62 (0.01) 0.05 (0.81)
Materials 0.05 5.44 2.02 0.01 37 0.63
0.00 (0.95) 2.49 (0.11) 0.87 (0.35) 0.01 (0.92)
Industrials 0.95 ** 0.35 0.47 -0.37 91 0.35
5.38 (0.02) 0.96 (0.32) 0.58 (0.44) 0.32 (0.57)
Cnsmr Discret. 2.76 2.18 *** 2.57 -0.61 60 0.32
2.69 (0.10) 20.36 (<.01) 0.80 (0.37) 0.78 (0.37)
Health Care 0.83 *** 0.75 *** 5.61 *** -0.30 * 74 0.49
9.15 (<.01) 24.17 (<.01) 10.59 (<.01) 2.90 (0.08)
Info Tech 1.33 *** 2.20 *** 1.08 *** -0.19 * 126 0.51
32.06 (<.01) 79.66 (<.01) 9.32 (<.01) 3.34 (0.06)
Total 464
The remaining estimated coefficients are in the next half-panel.
145
Table 16 (cont’d)
Panel C2. Remaining Estimated Coefficients from an IRLS Regression with QR Alone
GICS
Sector Icept NFVA NFVL FVL12 FVL3 NI
Energy -0.37 1.12 *** -1.21 *** -1.09 *** -0.81 0.41 ***
0.05 (0.82) 114.73 (<.01) 59.82 (<.01) 47.11 (<.01) 0.26 (0.61) 7.70 (<.01)
Materials 0.16 1.42 *** -1.45 *** -0.05 -4.32 1.99 ***
0.01 (0.92) 82.48 (<.01) 52.73 (<.01) 0.00 (0.98) 0.13 (0.71) 19.60 (<.01)
Industrials 2.32 0.66 *** -0.54 *** -1.96 ** -0.95 1.63 ***
1.39 (0.23) 68.83 (<.01) 22.51 (<.01) 4.79 (0.02) 0.08 (0.77) 18.91 (<.01)
Cnsmr
Discret.
3.42 ** 0.49 *** -0.47 *** 1.15 -2.74 0.23
3.99 (0.04) 17.07 (<.01) 11.56 (<.01) 0.38 (0.53) 0.22 (0.63) 2.57 (0.10)
Health Care 3.13 *** 0.84 *** -0.81 *** 4.02 *** -0.26 0.34
11.41 (<.01) 75.28 (<.01) 34.23 (<.01) 90.77 (<.01) 0.39 (0.53) 1.75 (0.18)
Info Tech 2.88 *** 0.46 *** -0.31 *** 0.13 29.51 1.33 ***
19.32 (<.01) 33.51 (<.01) 7.71 (<.01) 0.61 (0.43) 0.29 (0.59) 53.10 (<.01)
These two half-panels together show the estimated coefficients by GICS Sector of an IRLS
regression on the entire sample using the modified Ohlson model with an additional
explanatory variable, the Quick Ratio (QR).
PRCit = α0 + α1NFVAit + α2FVA1it + α3FVA2it + α4FVA3it
+ α5NVFLit + α6FVL12it + α7FVL3it + α8 QRit + β1NIit + εit
The other variables are those given in Panel A of Table 9.
146
Table 16 (cont’d)
Panel D1. Several Estimated Coefficients from an IRLS Regression with FVA3 x QR
GICS
Sector FVA1 FVA2 FVA3 QR FVA3xQR N
Adj
R2
Energy -0.81 1.43 * 1.88 * 0.43 -0.59 76 0.49
0.57 (0.45) 3.58 (0.05) 3.43 (0.06) 0.42 (0.51) 1.00 (0.31)
Materials 0.01 5.36 1.43 0.01 0.36 37 0.61
0.00 (0.98) 2.10 (0.14) 0.08 (0.78) 0.00 (0.95) 0.02 (0.90)
Industrials 1.09 *** 0.29 -21.11 ** -
0.67
17.50 ** 91 0.39
7.64 (<.01) 0.72 (0.39) 5.53 (0.01) 1.12 (0.29) 5.78 (0.01)
Cnsmr
Discret.
2.67 2.14 *** 3.51 -
0.30
-1.09 60 0.31
2.38 (0.12) 16.96 (<.01) 0.50 (0.48) 0.05 (0.83) 0.05 (0.82)
Health Care 1.10 *** 0.83 *** 11.34 *** -
0.15
-2.67 ** 74 0.49
18.23 (<.01) 30.34 (<.01) 14.02 (<.01) 0.71 (0.39) 5.68 (0.01)
Info Tech 1.31 *** 2.21 *** 0.96 -
0.20
* 0.09 126 0.51
29.70 (<.01) 78.67 (<.01) 2.30 (0.12) 3.08 (0.07) 0.05 (0.81)
Total 464
The remaining estimated coefficients are in the next half-panel.
147
Table 16 (cont’d)
Panel D2. Remaining Estimated Coefficients from an IRLS Regression with FVA3 x
QR
GICS
Sector Icept NFVA NFVL FVL12 FVL3 NI
Energy -0.63 1.13 *** -1.23 *** -1.10 *** -0.85 0.43 ***
0.15 (0.70) 118.17 (<.01) 62.42 (<.01) 48.42 (<.01) 0.28 (0.59) 8.48 (<.01)
Materials 0.12 1.42 *** -1.45 *** 0.01 -3.79 2.03 ***
0.00 (0.94) 73.60 (<.01) 47.01 (<.01) 0.00 (0.99) 0.08 (0.77) 15.11 (<.01)
Industrials 2.66 0.67 *** -0.55 *** -2.06 ** -0.71 1.56 ***
2.01 (0.15) 78.40 (<.01) 25.76 (<.01) 5.86 (0.01) 0.05 (0.82) 18.98 (<.01)
Cnsmr
Discret.
3.10 0.49 *** -0.47 *** 1.16 -2.70 0.23
2.19 (0.13) 16.42 (<.01) 11.05 (<.01) 0.38 (0.53) 0.21 (0.64) 2.46 (0.11)
Health Care 2.39 *** 0.96 *** -0.99 *** 4.52 *** -0.47 0.20
6.74 (<.01) 90.36 (<.01) 45.54 (<.01) 105.53 (<.01) 1.42 (0.23) 0.62 (0.43)
Info Tech 2.90 *** 0.45 *** -0.30 ** 0.14 29.74 1.33 ***
18.72 (<.01) 28.17 (<.01) 6.07 (0.01) 0.67 (0.41) 0.29 (0.59) 52.61 (<.01)
These two half-panels together show the estimated coefficients by GICS Sector of an IRLS
regression on the entire sample using the modified Ohlson model with an additional
explanatory variable, the Quick Ratio (QR). The model is given by
PRCit = α0 + α1NFVAit + α2FVA1it + α3FVA2it + α4FVA3it + α5NVFLit
+ α6FVL12it + α7FVL3it + α8 QRit + α9QRit ×FVA3it + β1NIit + εit
The other variables are those given in Panel A of Table 9.
148
Table 16. (cont’d)
Panel E. Descriptive Statistics of the Operating Cash Flow Ratio (CR) by GICS Sector
GICS Sector N MIN Q1 MEDIAN Q3 MAX MEAN STD
Energy 125 -2.39 0.48 1.09 1.77 14.57 1.40 1.80
Materials 60 -11.16 -0.01 0.34 0.65 4.06 0.00 2.12
Industrials 182 -6.07 0.14 0.39 0.66 8.86 0.51 1.08
Cnsmr Discret. 132 -1.21 0.19 0.48 0.89 3.31 0.58 0.67
Cnsmr Staples 36 -1.33 0.25 0.51 0.81 1.36 0.48 0.53
Health Care 207 -13.47 -2.49 -0.05 0.58 4.39 -1.08 2.64
Financials 25 -1.31 -0.25 0.29 1.52 18.80 1.28 3.84
Info Tech 255 -10.50 0.05 0.41 0.81 2.18 0.30 1.18
Telecom Svcs. 26 -5.42 0.39 0.91 1.34 3.22 0.63 1.65
Utilities 52 -6.92 0.30 0.42 0.79 1.61 0.38 1.09
Total 1100 -13.47 0.06 0.40 0.86 18.80 0.26 1.86
This panel shows descriptive statistics of Operating Cash Flow Ratio (CR) by GICS Sector.
Q1 is the first quartile and Q3 is the third quartile. STD is the standard deviation.
149
Table 16. (cont’d)
Panel F1. First Half of the Estimated Coefficients from an IRLS Regression with CR
Alone
GICS Sector FVA1 FVA2 FVA3 CR N Adj R2
Energy -0.28 1.11 *** 0.75 ** 0.48 125 0.47
0.50 (0.47) 16.69 (<.01) 4.32 (0.03) 2.09 (0.14)
Materials -0.02 6.89 *** 2.03 -0.07 60 0.57
0.00 (0.98) 12.33 (<.01) 1.62 (0.20) 0.03 (0.85)
Industrials 0.92 *** 0.31 0.20 -0.65 182 0.43
14.34 (<.01) 1.76 (0.18) 0.19 (0.66) 1.74 (0.18)
Cnsmr Discret. 0.78 2.22 *** 1.27 2.83 *** 132 0.39
2.40 (0.12) 44.70 (<.01) 1.00 (0.31) 13.45 (<.01)
Cnsmr Staples 5.66 *** 56.69 *** -269.80 *** 5.92 ** 36 0.73
24.19 (<.01) 60.97 (<.01) 57.71 (<.01) 5.91 (0.01)
Health Care 1.81 *** 1.66 *** 3.28 *** 0.13 207 0.42
92.42 (<.01) 208.67 (<.01) 13.90 (<.01) 1.12 (0.28)
Info Tech 1.33 *** 1.82 *** 1.17 *** 0.26 255 0.50
101.49 (<.01) 106.65 (<.01) 20.99 (<.01) 1.81 (0.17)
Telecom Svcs. 0.57 6.29 *** 7.82 1.27 26 0.59
0.53 (0.46) 9.98 (<.01) 2.32 (0.12) 2.19 (0.13)
Utilities 1.32 ** 0.33 -0.28 -0.02 52 0.44
3.99 (0.04) 0.17 (0.68) 0.01 (0.92) 0.00 (0.98)
Total 1075
The remaining estimated coefficients are in the next half-panel.
150
Table 16 (cont’d)
Panel F2. Remaining Estimated Coefficients from an IRLS Regression with CR Alone
GICS
Sector Icept NFVA NFVL FVL12 FVL3 NI
Energy -0.73 0.98 *** -1.02 *** -0.95 *** -0.59 0.51 ***
0.41 (0.52) 134.96 (<.01) 61.87 (<.01) 41.70 (<.01) 0.27 (0.60) 16.20 (<.01)
Materials -0.02 1.28 *** -1.27 *** -1.45 -3.82 2.35 ***
0.00 (0.98) 82.73 (<.01) 51.80 (<.01) 2.31 (0.12) 0.09 (0.77) 39.02 (<.01)
Industrials 1.80 ** 0.71 *** -0.59 *** -1.14 *** -0.85 1.30 ***
4.72 (0.02) 199.37 (<.01) 82.72 (<.01) 10.06 (<.01) 0.27 (0.60) 38.24 (<.01)
Cnsmr
Discret.
1.91 * 0.58 *** -0.56 *** -0.81 ** -0.07 0.54 ***
3.65 (0.05) 55.15 (<.01) 35.13 (<.01) 5.41 (0.02) 0.00 (0.98) 37.33 (<.01)
Cnsmr
Staples
-2.49 0.84 *** -0.57 *** -1.18 -
123.16
*** 2.61 ***
1.04 (0.30) 19.55 (<.01) 7.18 (<.01) 2.46 (0.11) 9.92 (<.01) 8.83 (<.01)
Health Care 0.99 * 1.27 *** -1.47 *** 2.06 *** -1.12 ** -0.05
3.79 (0.05) 297.47 (<.01) 240.11 (<.01) 41.42 (<.01) 6.61 (0.01) 0.07 (0.78)
Info Tech 1.57 *** 0.57 *** -0.40 *** 0.02 2.99 1.10 ***
17.45 (<.01) 97.16 (<.01) 25.79 (<.01) 0.02 (0.89) 0.98 (0.32) 66.22 (<.01)
Telecom
Svcs.
5.34 ** -0.01 0.00 -0.20 3.12 3.67 ***
5.06 (0.02) 0.00 (0.97) 0.00 (0.99) 0.02 (0.89) 1.34 (0.24) 8.68 (<.01)
Utilities 6.54 ** 0.32 -0.16 -2.14 *** -1.35 2.59 **
4.34 (0.03) 1.68 (0.19) 0.27 (0.60) 8.24 (<.01) 0.49 (0.48) 4.69 (0.03)
Total
These two half-panels together show the estimated coefficients by GICS Sector of an IRLS
regression on the entire sample using the modified Ohlson model with an additional
explanatory variable, the operating cash flow fatio (CR)
PRCit = α0 + α1NFVAit + α2FVA1it + α3FVA2it + α4FVA3it
+ α5NVFLit + α6FVL12it + α7FVL3it + α8 CRit + β1NIit + εit .
The other variables are those given in Panel A of Table 9.
151
Table 16 (cont’d)
Panel G1. Several Estimated Coefficients from an IRLS Regression with FVA3 x CR
GICS
Sector FVA1 FVA2 FVA3 CR FVA3xCR N
Adj
R2
Energy -0.26 1.10 *** 0.97 0.54 -0.14 125 0.46
0.43 (0.51) 16.15 (<.01) 1.59 (0.20) 2.27 (0.13) 0.10 (0.75)
Materials -0.26 6.90 *** 1.96 -0.07 -1.51 60 0.54
0.10 (0.74) 11.97 (<.01) 1.39 (0.23) 0.03 (0.87) 0.60 (0.43)
Industrials 0.95 *** 0.29 -0.04 -0.59 -1.27 182 0.42
14.16 (<.01) 1.49 (0.22) 0.01 (0.92) 1.37 (0.24) 0.86 (0.35)
Cnsmr
Discret.
0.82 2.21 *** 1.60 3.06 *** -0.99 132 0.39
2.62 (0.10) 43.87 (<.01) 1.23 (0.26) 12.75 (<.01) 0.25 (0.61)
Cnsmr
Staples
6.37 *** 58.04 *** -
320.59
*** 5.74 ** 80.60 36 0.65
27.94 (<.01) 58.40 (<.01) 34.55 (<.01) 5.03 (0.02) 0.78 (0.37)
Health Care 1.69 *** 1.65 *** 4.68 *** 0.07 0.49 ** 207 0.43
84.47 (<.01) 218.00 (<.01) 26.52 (<.01) 0.35 (0.55) 4.84 (0.02)
Info Tech 1.34 *** 1.81 *** 1.00 *** 0.23 0.56 255 0.50
103.12 (<.01) 105.91 (<.01) 9.94 (<.01) 1.26 (0.26) 1.03 (0.31)
Telecom
Svcs.
1.29 ** -2.02 -36.65 *** 0.38 20.97 *** 26 0.47
6.41 (0.01) 2.04 (0.15) 23.31 (<.01) 0.38 (0.54) 23.24 (<.01)
Utilities 1.47 ** 0.40 -9.09 -0.27 8.37 * 52 0.43
5.07 (0.02) 0.25 (0.61) 2.48 (0.11) 0.05 (0.82) 3.00 (0.08)
Total 1075
The remaining estimated coefficients are in the next half-panel.
152
Table 16 (cont’d)
Panel G2. Remaining Estimated Coefficients from an IRLS Regression with
FVA3 x CR
GICS
Sector Icept NFVA NFVL FVL12 FVL3 NI
Energy -0.70 0.97 *** -1.01 *** -0.94 *** -0.65 0.52 ***
0.37 (0.54) 129.90 (<.01) 59.63 (<.01) 40.05 (<.01) 0.32 (0.57) 16.59 (<.01)
Materials -0.11 1.32 *** -1.33 *** -1.48 -3.11 2.60 ***
0.01 (0.93) 80.77 (<.01) 52.11 (<.01) 2.34 (0.12) 0.05 (0.81) 41.83 (<.01)
Industrials 1.75 ** 0.71 *** -0.59 *** -1.15 *** -0.96 1.31 ***
4.47 (0.03) 197.61 (<.01) 81.61 (<.01) 10.15 (<.01) 0.34 (0.56) 39.06 (<.01)
Cnsmr
Discret.
1.78 * 0.58 *** -0.55 *** -0.83 ** -0.01 0.54 ***
2.97 (0.08) 54.70 (<.01) 34.72 (<.01) 5.64 (0.01) 0.00 (0.99) 37.70 (<.01)
Cnsmr
Staples
-3.33 0.92 *** -0.65 *** -1.25 -
104.06
** 2.30 **
1.70 (0.19) 21.94 (<.01) 8.56 (<.01) 2.55 (0.11) 5.51 (0.01) 6.23 (0.01)
Health Care 0.85 * 1.30 *** -1.49 *** 1.95 *** -1.14 *** -0.13
2.94 (0.08) 323.93 (<.01) 261.68 (<.01) 39.07 (<.01) 7.24 (<.01) 0.54 (0.46)
Info Tech 1.58 *** 0.56 *** -0.39 *** 0.03 2.97 1.09 ***
17.59 (<.01) 94.07 (<.01) 24.45 (<.01) 0.05 (0.82) 0.97 (0.32) 64.79 (<.01)
Telecom
Svcs.
-1.04 1.65 *** -1.29 *** -1.38 -4.90 *** -0.28
0.38 (0.53) 45.49 (<.01) 22.28 (<.01) 1.99 (0.15) 7.13 (<.01) 0.12 (0.72)
Utilities 6.16 ** 0.40 * -0.26 -2.20 *** -0.19 2.57 **
3.93 (0.04) 2.73 (0.09) 0.72 (0.39) 8.87 (<.01) 0.01 (0.92) 4.67 (0.03)
Total
These two half-panels together show the estimated coefficients by GICS Sector of an IRLS
regression on the entire sample using the modified Ohlson model of the previous half-panels
with an additional term investigating the interaction of the operating cash flow ratio (CR)
and the coefficient of the Level 3 fair value of assets, FVA3
PRCit = α0 + α1NFVAit + α2FVA1it + α3FVA2it + α4FVA3it + α5NVFLit
+ α6FVL12it + α7FVL3it + α8 CRit + α9 CRit ×FVA3it + β1NIit + εit .
The other variables are those given in Panel A of Table 9
153
Table 17. Investigating the Value Relevance of the FAS 157 Ratio
Panel A1. Certain Coefficients of IRLS Regression including the Fair Value of Assets
Ratio
GICS Sector FVA1 FVA2 FVA3 FVArat N Adj R2
Energy -0.45 1.17 *** 0.77 * 0.86 125 0.46
1.00 (0.31) 10.75 (<.01) 3.04 (0.08) 0.00 (0.94)
Materials 0.14 6.89 *** 2.31 1.49 60 0.56
0.03 (0.86) 15.21 (<.01) 1.96 (0.16) 0.05 (0.81)
Industrials 0.90 *** 0.55 ** 0.34 7.47 182 0.42
7.82 (<.01) 5.17 (0.02) 0.47 (0.49) 0.96 (0.32)
Cnsmr Discret. 1.40 ** 2.25 *** 0.29 -2.57 132 0.39
3.84 (0.04) 40.48 (<.01) 0.30 (0.58) 0.15 (0.70)
Cnsmr Staples 8.42 *** 64.30 *** -337.64 *** -7.47 36 0.72
33.83 (<.01) 103.14 (<.01) 100.18 (<.01) 0.30 (0.58)
Health Care 1.47 *** 1.38 *** 5.12 *** 1.24 207 0.42
68.11 (<.01) 158.67 (<.01) 53.97 (<.01) 1.21 (0.27)
Info Tech 1.43 *** 1.93 *** 1.22 *** -1.47 255 0.50
86.31 (<.01) 97.92 (<.01) 22.26 (<.01) 1.11 (0.29)
Telecom Svcs. 0.97 6.36 *** 7.12 -2.15 26 0.56
0.42 (0.51) 8.14 (<.01) 1.43 (0.23) 0.03 (0.85)
Utilities 1.17 0.20 -1.38 29.36 52 0.42
2.42 (0.12) 0.05 (0.81) 0.16 (0.68) 0.39 (0.53)
1075
The remaining coefficients of the regression are shown in the next half-panel.
154
Table 17 (cont’d)
Panel A2. Remaining Estimated Coefficients
GICS
Sector Icept NFVA NFVL FVL12 FVL3 NI
Energy -0.32 1.05 *** -1.11 *** -1.05 *** -0.78 0.55 ***
0.06 (0.80) 154.68 (<.01) 72.78 (<.01) 50.18 (<.01) 0.47 (0.49) 18.66 (<.01)
Materials -0.36 1.36 *** -1.40 *** -0.96 *** -2.53 2.14 ***
0.06 (0.80) 98.86 (<.01) 69.51 (<.01) 7.87 (<.01) 0.04 (0.84) 36.26 (<.01)
Industrials 1.63 0.72 *** -0.65 *** -1.20 *** -1.59 1.15 ***
2.42 (0.11) 161.21 (<.01) 91.61 (<.01) 10.42 (<.01) 1.00 (0.31) 28.28 (<.01)
Cnsmr
Discret.
3.82 *** 0.56 *** -0.55 *** -1.20 *** -1.54 0.49 ***
16.82 (<.01) 50.23 (<.01) 32.60 (<.01) 7.49 (<.01) 0.10 (0.74) 30.99 (<.01)
Cnsmr
Staples
-1.15 1.26 *** -1.02 *** -1.96 *** -
101.20
*** 1.19
0.37 (0.54) 58.23 (<.01) 30.62 (<.01) 9.32 (<.01) 8.16 (<.01) 2.40 (0.12)
Health Care 0.32 1.37 *** -1.57 *** 1.87 *** -1.34 *** -0.12
0.22 (0.63) 343.67 (<.01) 299.68 (<.01) 31.52 (<.01) 10.47 (<.01) 0.41 (0.52)
Info Tech 1.87 *** 0.57 *** -0.42 *** 0.00 3.32 1.15 ***
12.36 (<.01) 90.45 (<.01) 30.10 (<.01) 0.00 (0.99) 1.21 (0.27) 73.19 (<.01)
Telecom
Svcs.
5.31 * -0.10 0.16 0.17 1.66 4.25 ***
3.34 (0.06) 0.06 (0.80) 0.14 (0.71) 0.01 (0.91) 0.32 (0.56) 11.71 (<.01)
Utilities 4.48 0.40 -0.24 -2.43 *** -1.29 2.65 **
0.99 (0.31) 2.32 (0.12) 0.56 (0.45) 8.26 (<.01) 0.42 (0.51) 4.83 (0.02)
These two half-panels together show the estimated coefficients by GICS Sector of an IRLS
regression using the modified Ohlson model with the Fair Value of Assets Ratio, FVARat,
as an additional explanatory variable.
PRCit = α0 + α1NFVAit + α2FVA1it + α3FVA2it + α4FVA3it
+ α5NVFLit + α6FVL12it + α7FVL3it + α8FVAratit + β1NIit + εit .
The basic model and other variables are those given in Panel A of Table 9.
155
155
Table 18. Sensitivity Analysis
Panel A. Steps Taken to Develop the Sample
Step Description 2,008 2,009 Total
Change Firms Change Firms Change Firms
0 Extract from Compustat for fiscal year with:
11 <= Stock Exchange Code <= 14
6,631 6,631 6,477 6,477 13,108 13,108
1 Less: Not Major Market (NYSE, AMEX, or NASDAQ) (580) 6,051 (556) 5,921 (1,136) 11,972
2 Less: Not Final Data (keep UPD = 3 only) (787) 5,264 (809) 5,112 (1,596) 10,376
3 Less: Firms with fiscal year-end not December (1,284) 3,980 (1,252) 3,860 (2,536) 7,840
4 Less: Missing AT, LT, NI, or CSHPRI (87) 3,893 (40) 3,820 (127) 7,713
5 Less: Missing complete set of FVH observations (1,240) 2,653 (942) 2,878 (2,182) 5,531
6 Less: Firms missing a GICS code (18) 2,635 (2) 2,876 (20) 5,511
7 Merged with CRSP Monthly 3/31/2008 and 3/31/2009,
and Google Finance on 3/31/2011 with PRC > $1.00
(471) 2,164 (318) 2,558 (789) 4,722
8 Less: Firms where the computed sum of Level 1, Level 2,
and Level 3 Assets or Liabilities did not equal the
respective observation of the Total from Compustat.
(346) 1,818 (433) 2,125 (779) 3,943
9 Less: Outliers (|Studentized Residual| > 2) (24) 1,794 (104) 2,021 (128) 3,815
This table shows the steps taken to develop the sample. More details on Step 9, identification of outliers, are provided in Panel D below.
156
156
Table 18 (cont’d)
Panel B. Composition of the Sample by GICS Sector and Fiscal Year at Step 8
GICS Sector 2008 2009 Total
Code Description N %(Row) N %(Row)
10 Energy 151 47 173 53 324
15 Materials 82 43 110 57 192
20 Industrials 228 46 270 54 498
25 Consumer Discretionary 173 43 229 57 402
30 Consumer Staples 49 48 53 52 102
35 Health Care 262 45 314 55 576
40 Financials 507 48 560 52 1,067
45 Information Technology 284 47 316 53 600
50 Telecommunication Services 31 41 45 59 76
55 Utilities 51 48 55 52 106
Total 1,818 46 2,125 54 3,943
This table shows the composition of the sample by Global Industry Classification Standard
(GICS®) Sector by year. For each year, the number of firms and the row percentage are shown.
157
Table 18 (cont’d)
Panel C1. Ratio of Levels 1–3 of Assets Reported at Fair Value to Total Assets (Percent) by GICS Sector for 2008, 2009, and
Combined
GICS Sector
FVA1 FVA2 FVA3
2008 2009 Combined 2008 2009 Combined 2008 2009 Combined
Avg StDv Avg StDv Avg StDv Avg StDv Avg StDv Avg StDv Avg StDv Avg StDv Avg StDv
Energy 0.9 3.3 2.0 8.6 1.5 6.7 3.0 5.5 1.1 3.9 2.0 4.8 1.3 3.7 0.5 2.1 0.9 3.0
Materials 2.9 8.2 5.6 11.6 4.5 10.4 1.0 3.2 0.8 2.5 0.8 2.8 0.4 1.6 0.5 2.4 0.4 2.1
Industrials 3.5 7.5 4.2 9.5 3.8 8.7 1.0 3.5 1.1 4.2 1.0 3.9 0.6 2.7 0.8 4.8 0.7 4.0
Consmr Discret 5.0 11.4 5.9 13.0 5.5 12.3 2.0 7.3 2.0 5.8 2.0 6.5 1.1 3.6 0.9 3.7 1.0 3.7
Consmr Stapls 4.8 11.8 5.7 11.6 5.2 11.6 0.9 2.9 0.8 2.8 0.8 2.8 0.1 0.5 0.1 0.3 0.1 0.4
Health Care 19.0 25.2 21.7 28.1 20.5 26.8 12.3 21.3 10.1 19.2 11.1 20.2 1.6 5.1 1.1 4.1 1.3 4.6
Financials 2.8 7.6 3.8 10.8 3.3 9.4 16.7 19.8 17.9 20.9 17.3 20.4 2.8 12.6 2.9 11.9 2.9 12.2
Info Tech 15.4 18.8 15.5 19.7 15.5 19.3 7.2 13.5 8.1 14.6 7.7 14.1 2.0 5.5 1.8 5.7 1.9 5.6
Telecom 3.6 10.3 5.4 10.0 4.6 10.1 0.9 2.4 0.9 2.7 0.9 2.6 0.5 1.7 0.2 1.0 0.3 1.3
Utilities 2.5 5.1 2.7 4.4 2.6 4.8 2.1 3.5 1.7 3.2 1.9 3.3 0.4 0.9 0.3 0.8 0.4 0.8
This table shows the ratio of fair value of assets for each of the three levels of the fair value hierarchy (FVH) to the firm’s total assets expressed
as a percentage. The ratios are shown by Sector of the Global Industry Classification Standard (GICS®) for fiscal 2008, 2009, and both years
combined together.
158
Table 18 (cont’d)
Panel C2. Ratio of Liabilities Reported at Fair Value to Total Liabilities (Percent) by GICS Sector for 2008, 2009, and Combined
GICS Sec
FVL1 FVL2 FVL12 FVL3
2008 2009 Comb 2008 2009 Comb 2008 2009 Comb 2008 2009 Comb
Avg StDv Avg StDv Avg StDv Avg StDv Avg StDv Avg StDv Avg StDv Avg StDv Avg StDv Avg StDv Avg StDv Avg StDv
Energy 1.0 6.2 1.8 9.6 1.4 8.2 2.6 8.4 3.3 9.6 3.0 9.1 3.6 11.0 5.1 13.8 4.4 12.6 0.6 2.5 1.5 7.2 1.1 5.6
Materials 1.1 6.2 1.2 7.2 1.1 6.8 3.7 8.9 2.9 9.6 3.2 9.3 4.7 13.1 4.1 11.9 4.4 12.4 0.1 0.4 0.1 0.6 0.1 0.5
Industrials 0.6 3.5 0.2 1.3 0.4 2.5 2.5 6.9 2.3 8.4 2.4 7.7 3.1 7.9 2.4 8.4 2.7 8.2 0.7 5.6 1.0 6.4 0.9 6.0
Cnsr Disc 1.3 7.9 1.6 9.7 1.5 9.0 1.3 4.5 3.1 11.1 2.3 8.9 2.6 9.3 4.7 14.7 3.8 12.7 0.5 4.9 0.7 5.4 0.6 5.2
Cnsr Stpls 1.2 6.7 1.9 9.5 1.6 8.2 1.9 5.7 3.9 10.1 2.9 8.3 3.1 9.7 5.8 13.4 4.5 11.8 0.1 0.7 0.4 2.4 0.3 1.8
Hlth Care 0.6 5.7 0.6 5.7 0.6 5.7 1.5 7.7 3.1 12.5 2.4 10.6 2.2 9.5 3.7 13.7 3.0 12.0 0.7 4.7 3.1 12.1 2.0 9.5
Financials 0.2 2.8 0.5 5.0 0.4 4.1 1.4 8.1 0.9 6.5 1.1 7.3 1.6 8.6 1.4 8.1 1.5 8.4 1.1 7.7 1.0 7.3 1.0 7.5
Info Tech 0.3 2.6 0.8 6.1 0.6 4.8 1.1 5.2 1.5 7.5 1.3 6.5 1.4 5.7 2.4 10.5 1.9 8.6 0.3 4.2 0.9 5.2 0.6 4.8
Telecom 1.4 7.6 3.1 14.5 2.4 12.2 3.5 11.7 2.8 12.0 3.1 11.8 4.9 13.8 5.9 18.5 5.5 16.6 0.4 1.5 0.8 2.5 0.6 2.2
Utilities 1.1 4.0 0.3 1.4 0.7 3.0 2.4 3.9 1.4 2.5 1.9 3.3 3.5 6.7 1.8 3.2 2.6 5.2 0.5 1.2 2.3 12.8 1.4 9.0
This table shows the ratio of fair value of liabilities for each level of the FVH to the total liabilities of each firm expressed as a percentage.
FVLn, where n = 1, 2, or 3, is the ratio of the fair value of Level n liabilities to total liabilities expressed as a percentage. FVL12 is the sum of
Level 1 and Level 2 fair value of liabilities divided by total liabilities. The ratios are shown by Sector of the Global Industry Classification
Standard (GICS®) for fiscal 2008, 2009, and both years combined together.
159
Table 18 (cont’d)
Panel D. The Number of Outliers in Each GICS Sector (Step 9)
Absolute Value of Studentized Residual
<= 2 > 2 (Outliers) Total
GICS Sector 2008 2009 Comb 2008 2009 Comb
Energy 151 159 310 0 14 14 324
Materials 78 104 182 4 6 10 192
Industrials 226 247 473 2 23 25 498
Consumer Discretionary 171 220 391 2 9 11 402
Consumer Staples 46 51 97 3 2 5 102
Health Care 258 305 563 4 9 13 576
Financials 501 530 1,031 6 30 36 1,067
Information Techonlogy 281 314 595 3 2 5 600
Telecommunication Services 31 40 71 0 5 5 76
Utilities 51 51 102 0 4 4 106
Total 1,794 2,021 3,815 24 104 128 3,943
This table shows the distribution of outliers, identified as the observations with Studentized
Residuals greater than two, by GICS Sector.
160
Table 18 (cont’d)
Panel E. Subsamples for OLS Models for Tests of Hypotheses
GICS Sector Trimmed H1
(BE/ME)
H4
(QR)
H4
(CR)
H4’
(CR & BE>=0)
Energy 310 305 174 310 287
Materials 182 167 118 181 161
Industrials 473 462 235 449 402
Consumer Discretionary 391 374 184 376 322
Consumer Staples 97 88 44 97 89
Health Care 563 532 212 558 504
Financials 1031 740 28 64 59
Information Technology 595 537 267 585 564
Telecommunication Services 71 61 37 71 63
Utilities 102 98 29 100 96
Total 3815 3364 1328 2791 2547
This table shows the composition of the samples for the tests of the hypotheses. The column
marked “Trimmed” is the result of Step 9 of Table C1, and is also given as the Combined
(“Comb”) column where Studentized Residual <= 2 in Table C4. This sample is closest to
that of STY. The column with the heading “H1 (BE/ME)” shows the composition of the
sample where the ratio of Book Equity-to-Market Equity is not missing. The numerator,
book equity (BE), was computed from Computstat variables as BE = AT – LT – MIB, where
AT is the total assets, LT is the total liabilities, and MIB is the minority interest as stated on
the balance sheet. The denominator was the Compustat variable MkValT (Market Value,
Total). The absence of this variable led to the missing value for the BE/ME ratio. OLS
models on this sample test hypothesis H1. The column with the heading “H4 (CR)” shows
the composition of the sample where the operating cash flow ratio (CR) is not missing. The
CR is the ratio of the Compustat variables OANCF (operating cash flow ratio), as the
numerator and LCT (Liabilities, Current, Total) as the denominator. Missing values for
either of these observations result in missing values of their ratio. This sample was used to
test hypothesis H4. The final column, with heading “H4’ (CR & BE >=0)” shows the
composition of the CR sample where the book equity, BE as defined above, is not negative.
161
Table 18 (cont’d)
Panel F. OLS Regressions of the Modified Ohlson Model on the Trimmed Sample by GICS Sector
Variable Energy Materials Indstrls Consmr
Discret
Consmr
Staples
Health
Care Financials
Info
Tech
Telecom
Svcs Utilities
Icept 11.13*** 5.43*** 7.47*** 7.44*** 9.04*** 7.37*** 3.53*** 1.91*** 5.08*** 5.45***
fva1 0.98** 3.56*** 0.86*** 0.06 1.01 1.45*** 0.76*** 2.2*** 0.09 0.49
fva2 0.05 4.85*** 0.68*** 1.93*** 2.82* 1.26*** 0.73*** 3.96*** 1.78* 0.39
fva3 -0.23 -3.93 0.45 -0.51 -21.59 3.8*** 0.49*** 3.06*** 2.89 2.23
fvl12 -0.23 -0.82*** -
1.31*** -1.43*** -0.08 -0.05 -0.96*** -1.06*** 0.23
-
1.26***
fvl3 0.5 -3.52 -2.53 -1.19 -71.13* -1.29 -0.99*** 10.91*** 0.06 -1.85*
nfva 0.32*** 0.82*** 0.73*** 1.37*** 0.19 1.33*** 0.71*** 1.37*** 0.24* 0.35**
nfvl -0.2** -0.59*** -
0.68*** -1.42*** -0.04
-
1.51*** -0.72*** -1.64*** 0.02 -0.17
nipsh 0.94*** -0.11 1.06*** 0.95*** 7.04*** 2.64*** 0.97*** 3.23*** 3.54*** 2.83***
N 310 182 473 391 97 563 1031 595 71 102
AdjR2 0.37 0.54 0.95 0.86 0.70 0.63 0.79 0.72 0.70 0.70
162
Table 18 F. (cont’d)
This table shows the results of regressions of the modified Ohlson model by ordinary least squares (OLS) for each sector of the Global
Industry Classification System (GICS). The first column on the left lists the variables of the model and subsequent columns list the
estimated coefficients followed by a number of stars that indicates the statistical significance as follows: no star indicates p-value >
0.1, * indicates the p-value <= 0.1, ** indicates the p-value <= 0.05, and *** indicates the p-value <= 0.01. N indicates the number of
observations in the data subset, and AdjR2 provides the adjusted coefficient of determination, Adjusted R2. The model and variables
are those given in Panel A of Table 9.
163
Table 18 (cont’d)
Panel G. OLS Regressions of the Modified Ohlson Model with BE/ME by GICS Sector (Test of H1)
Variable Energy Materials Indstrls Consmr
Discret
Consmr
Staples
Health
Care Financls
Info
Tech
Telecm
Svcs Utilities
Icept 14.06*** 10.07*** 12.15*** 7.26*** 10.08*** 10.18*** 6.71*** 6.45*** 9.29*** 12.8***
fva1 1.2*** 3.74*** 0.72*** 0.3 1.66* 2.2*** 0.79*** 2.55*** 0.14 0.56
fva2 0.27 5.9** 0.42** 1.57*** 3.51** 1.96*** 0.68*** 3.97*** 1.99** 1**
fva3 0.13 -0.37 0.72** -0.44 -41.25 3.22*** 0.47*** 3.49*** 2.28 -0.93
fvl12 -0.33** -1.1*** -1.59*** -1.51*** -0.73 -0.82** -0.93*** -1.95*** -0.11 -1.3***
fvl3 0.7 3.94 -2.42* -1.29 -57.79 -2.19*** -0.8*** 4.15 0.5 -1.75**
beme -4.52*** -6.94*** -6.43*** -0.01 -3.54 -10.84*** -2.4*** -9.1*** -6.55** -9.77***
nfva 0.41*** 0.99*** 0.88*** 1.41*** 0.48*** 1.61*** 0.65*** 1.77*** 0.5** 0.5***
nfvl -0.34*** -0.84*** -0.85*** -1.46*** -0.35* -1.81*** -0.66*** -2.06*** -0.38 -0.33*
nipsh 0.93*** -0.18 0.51*** 1.01*** 6.2*** 2.05*** 0.56*** 2.11*** 2.69*** 1.75**
N 296 153 430 351 84 516 725 524 56 93
Adj R2 0.43 0.64 0.97 0.83 0.67 0.73 0.72 0.80 0.65 0.80
This table shows the results of regressions by OLS of the modified Ohlson model including a term for the book-to-market equity
(BE/ME) for each sector of the GICS. Other formatting conventions follow those of Panel F. The model and variables are those
given in Panel A of Table 9.
164
Table 18 (cont’d)
Panel H. IRLS Regression of the Modified Ohlson Model with QR (for H4a)
Variable Energy Materials Indstrls Consmr
Discret
Consmr
Staples
Health
Care Financials Info Tech
Telecom
Svcs Utilities
Icept. 6.19*** 3.97* 6.68*** 6.34*** 8.4** 8.31*** 14.77*** 4.74*** 5.51** 11.09*
fva1 -0.07 3.77*** 0.41 1.33*** 2.33* 1.01** -0.27 1.48*** 0.32 4.09**
fva2 -0.46 5.05*** 0.6** 2.97*** -0.97 1.04*** 0.36 2.74*** 2.1** -0.92
fva3 0.13 -6.53** 0.21 -2.19 12.28 1.17 0.4 1.57*** 25.46* -8.74
fvl12 -0.86*** -1.02*** -1.68*** -0.83*** 0.52 1 -0.74 -0.16 -1.99 1.54
fvl3 0.32 7.6 -3.1 -2.48* 25.1 -1.65* -0.32 47.38*** 0.84 1.57
qr -0.07 -0.08 0.18 -0.28 -0.58 0.02 -0.54 -0.35*** -0.26 -7.18
nfva 0.85*** 0.88*** 0.64*** 1.02*** 0.32* 0.97*** 0.04 0.62*** -0.04 0.11
nfvl -0.79*** -0.63*** -0.51*** -0.97*** -0.31 -1.24*** 0.33 -0.48*** 0.34 -0.01
nipsh 0.56*** -0.12 1.73*** 1.1*** 7.73*** 3.64*** 2.7** 2.01*** 3.89*** 5.48**
N 174 118 235 184 44 212 28 267 37 29
AdjR2 0.48 0.63 0.97 0.94 0.80 0.55 0.98 0.41 0.68 0.55
This table shows the results of regressions by Iteratively Reweighted Least Squares (IRLS) of the modified Ohlson model including a
term for the quick ratio (QR) for each sector of the GICS. Other formatting conventions follow those of Panel F. The model and
variables are those given in Panels C1 and C2 of Table 16.
165
Table 18 (cont’d)
Panel I. IRLS Regression of the Modified Ohlson Model with Interaction between QR and FVA3 (for H4b)
Variable Energy Materials Indstrls Consmr
Discret
Consmr
Staples
Health
Care Financials Info Tech
Telecom
Svcs Utilities
Icept. 6.33*** 4.23* 6.68*** 6.36*** 6.93* 8.57*** 14.02*** 4.51*** 5.98 9.76*
fva1 -0.04 3.86*** 0.4 1.34*** 1.96 0.93** -0.24 1.51*** 0.35 3.95***
fva2 -0.46 4.85*** 0.6** 2.97*** -9.98 1.03*** 0.44 2.74*** 2.17** -1.2
fva3 0.88 -18 0.38 -2.26 -111.54 -2.91 10.85 2.64*** 23.78 23.94
fvl12 -0.86*** -0.99*** -1.68*** -0.82*** 0.57 0.89 -0.82 -0.22 -1.89 2.02
fvl3 0.32 10.07 -3.11 -2.48 530.74 -1.64* -0.15 45.67*** 0.6 1.95
nfva 0.85*** 0.86*** 0.64*** 1.02*** 0.38** 0.95*** 0.12 0.66*** -0.03 0.02
qr -0.06 -0.2 0.18 -0.3 -0.48 -0.1 -0.53 -0.26** -0.59 -4.01
fva3 × qr -0.43 11.29 -0.16 0.04 91.94* 2.55 -0.57 -0.84 0.67 -36.89
nfvl -0.79*** -0.61*** -0.51*** -0.97*** -0.28 -1.21*** 0.21 -0.54*** 0.32 0.11
nipsh 0.58*** -0.12 1.74*** 1.1*** 8.16*** 3.7*** 2.62** 2.02*** 3.88*** 5.09**
N 174 118 235 184 44 212 28 267 37 29
AdjR2 0.48 0.63 0.97 0.94 0.78 0.56 0.98 0.41 0.68 0.67
166
Table 18 (cont’d)
Panel I
This table shows the results of regressions by IRLS of the modified Ohlson model including a term for the QR and its interaction with
the Level 3 fair value assets (FVA3) for each sector of the GICS. Other formatting conventions follow those of Panel F. The model
and variables are those given in Panels D1 and D2 of Table 16.
Panel J
This table shows the results of regressions by IRLS of the modified Ohlson model including a term for the operating cash flow ratio
(CR) for each sector of the GICS. Other formatting conventions follow those of Panel F. The model and variables are those given
Panels F1 and F2 of Table 16.
Panel K
This table shows the results of regressions by IRLS of the modified Ohlson model including a term for the CR and its interaction with
the FVA3 for each sector of the GICS. Other formatting conventions follow those of Panel F. The model and variables are those
given Panels G1 and G2 of Table 16.
167
Table 18 (cont’d)
Panel J. IRLS Regressions on Modified Ohlson Model that includes the Operating Cash Flow Ratio, CR (H4a)
Variable Energy Materials Indstrls Consmr
Discret
Consmr
Staples
Health
Care Financials
Info
Tech
Telecom
Svcs Utilities
Icept. 7.31*** 4.17*** 6.02*** 3.36*** 5.47** 6.29*** 8.85*** 3.35*** 2.64 6.24***
fva1 0.69** 3.01*** 0.8*** 1.51*** 1.2 1.45*** 0.36 1.55*** 0.22 0.32
fva2 0.33 4.52*** 0.51*** 3.06*** 2.7* 1.24*** 0.32 2.5*** 1.69** 0.99*
fva3 -0.13 -3.73 0.4 -0.42 -22.78 1.78 -0.4 1.57*** 1.07 1.78
fvl12 -0.47*** -1*** -1.15*** -0.89*** 0.43 0.06 -0.57 -0.18 0.14 -1.5***
fvl3 0.2 -5.53 -1.12 -0.94 -65.68* -1.35* 0.54 9.8*** 1.01 -1.79*
cr 0.89*** 0.48 0.3 3.39*** 4.02* 0.16 0.31 0.68*** 2.26** -0.27
nfva 0.51*** 0.8*** 0.71*** 0.99*** 0.19 1.23*** 0.12 0.71*** 0.3** 0.3*
nfvl -0.43*** -0.54*** -0.62*** -0.94*** -0.02 -
1.35*** 0.26
-
0.54*** -0.08 -0.09
nipsh 0.7*** 0.04 1.15*** 0.85*** 6.87*** 2.15*** 0.76** 1.99*** 3.02*** 2.22***
N 310 181 449 376 97 558 64 585 71 100
AdjR2 0.52 0.52 0.95 0.87 0.74 0.55 0.92 0.33 0.64 0.72
168
Table 18 (cont’d)
Panel K. IRLS Regressions on Modified Ohlson Model that includes the Interaction between CR and FVA3 (H4b)
Variable Energy Materials Indstrls Consmr
Discret
Consmr
Staples
Hlth
Care Financials
Info
Tech
Telecom
Svcs Utilities
Icept. 7.35*** 3.54** 5.94*** 3.03*** 5.43** 6.35*** 8.84*** 3.35*** 2.72 6.13***
fva1 0.71** 3.31*** 0.8*** 1.57*** 1.21 1.45*** 0.35 1.56*** 0.22 0.29
fva2 0.33 4.59*** 0.47*** 3.04*** 2.83 1.25*** 0.32 2.49*** 1.64* 0.99*
fva3 0.63 0.25 -0.04 1.1 -25.08 1.56 -0.38 1.52*** -1.04 1.6
fvl12 -0.47*** -1.03*** -1.14*** -0.87*** 0.47 0.06 -0.57 -0.18 0.18 -1.44***
fvl3 0 -6.86 -1 -0.89 -65.22 -1.35* 0.53 9.8*** 0.98 -1.76*
cr 0.98*** 0.57 0.16 3.95*** 3.94* 0.19 0.32 0.66*** 1.95 -0.3
cr × fva3 -0.49 -7.5 1.23 -3.21 6.56 -0.3 -0.01 0.29 1.65 0.62
nfva 0.5*** 0.82*** 0.72*** 0.98*** 0.19 1.23*** 0.12 0.71*** 0.29** 0.31*
nfvl -0.42*** -0.55*** -0.61*** -0.93*** -0.02 -1.35*** 0.25 -0.54*** -0.06 -0.09
nipsh 0.71*** 0.05 1.15*** 0.86*** 6.86*** 2.18*** 0.76** 1.98*** 3.07*** 2.18***
N 310 181 449 376 97 558 64 585 71 100
AdjR2 0.51 0.56 0.95 0.87 0.74 0.55 0.92 0.33 0.64 0.72
169
Table 18 (cont’d)
Panel L. IRLS Regressions on sample with BE >= 0 using Modified Ohlson Model that includes the CR (H4’a)
Variable Energy Materials Indstrls Consmr
Discret
Consmr
Staples
Hlth
Care Financials
Info
Tech
Telecom
Svcs Utilities
Icept. 6.94*** 3.23** 5.66*** 3.31*** 6.46*** 5.19*** 8.82*** 3.16*** 2.42 6.46***
fva1 0.59 3.24*** 0.79*** 0.97*** 0.84 1.63*** 0.36 1.62*** 0.22 0.6
fva2 0.44* 6.01*** 0.27 3.21*** 2.52* 1.48*** 0.24 2.5*** 2.06** 1.07**
fva3 -0.15 -0.36 0.52 -1.07 -18.72 1.91* -0.41 1.66*** 0.19 -1.68
fvl12 -0.55*** -1.13*** -1.57*** -0.94*** 0.47 0.48 -0.5 -0.27* 0.08 -1.81***
fvl3 0.5 -9.8 -0.88 -2.46* -65.81* -1.6** 0.85 7.7*** 0.73 -1.37
cr 0.86** 0.4 0.17 3.03*** 3.9* 0.11 0.29 0.67*** 2.68** -0.3
nfva 0.58*** 0.88*** 0.78*** 1.09*** 0.1 1.38*** 0.08 0.75*** 0.37*** 0.3*
nfvl -0.53*** -0.62*** -0.68*** -1.09*** 0.1 -1.56*** 0.32 -0.6*** -0.17 -0.09
nipsh 0.74*** 0.05 1.12*** 0.78*** 7.32*** 1.92*** 0.71** 1.98*** 3.17*** 2.25***
N 287 161 402 322 89 504 59 564 63 96
AdjR2 0.56 0.59 0.96 0.82 0.78 0.58 0.93 0.33 0.66 0.72
170
Table 18 (cont’d)
Panel L
This table shows the results of regressions by IRLS of the modified Ohlson model including a term for the CR for each sector of the
GICS. Firms with BE/ME < 0 have been removed from the dataset for this sample, as described in the right-most column with
heading H4’ in Panel E. Aside from the difference in dataset, the analysis of this table is the same as that shown in Panel J. Other
formatting conventions follow those of Panel F.
Panel M
This table shows the results of regressions by IRLS of the modified Ohlson model including a term for the CR and its interaction with
FVA3 for each sector of the GICS. Firms with BE/ME < 0 have been removed from the dataset for this sample, as described in the
right-most column with heading H4’ in Panel E. Aside from the difference in dataset, the analysis of this table is the same as that
shown in Panel K. Other formatting conventions follow those of Panel F. The model and variables are those given Panels F1 and F2
of Table 16.
171
Table 18 (cont’d)
Panel M. IRLS Regressions on sample with BE >= 0 using Modified Ohlson Model that includes the Interaction between CR
and FVA3 (H4’b)
Variable Energy Materials Indstrls Consmr
Discret
Consmr
Staples
Hlth
Care Financials
Info
Tech
Telecom
Svcs Utilities
Icept. 6.97*** 3.06* 5.59*** 3.09*** 6.46*** 5.23*** 8.82*** 3.16*** 2.48 6.19***
fva1 0.6* 3.35*** 0.79*** 0.98*** 0.84 1.64*** 0.36 1.62*** 0.23 0.41
fva2 0.44* 6.1*** 0.23 3.21*** 2.5 1.48*** 0.24 2.49*** 2.05** 1.18**
fva3 0.42 1.02 0.09 -0.3 -18.38 1.76* -0.4 1.61*** -1.36 -10.48**
fvl12 -0.54*** -1.14*** -1.55*** -0.94*** 0.47 0.48 -0.5 -0.27* 0.14 -1.56***
fvl3 0.35 -10.06 -0.77 -2.46* -65.85 -1.6** 0.84 7.69*** 0.62 -0.74
cr 0.93*** 0.48 0.03 3.39*** 3.9* 0.13 0.29 0.64*** 2.46* -0.47
cr × fva3 -0.37 -6.38 1.2 -1.71 -0.69 -0.23 -0.01 0.3 1.27 8.39**
nfva 0.58*** 0.88*** 0.78*** 1.09*** 0.1 1.38*** 0.08 0.75*** 0.35** 0.28*
nfvl -0.53*** -0.62*** -0.68*** -1.08*** 0.1 -1.56*** 0.32 -0.6*** -0.13 -0.04
nipsh 0.74*** 0.06 1.13*** 0.78*** 7.32*** 1.94*** 0.71** 1.98*** 3.23*** 2.23***
N 287 161 402 322 89 504 59 564 63 96
AdjR2 0.56 0.60 0.96 0.82 0.77 0.58 0.93 0.33 0.67 0.73
172
Table 18 (cont’d)
Panel N. IRLS Regression on a sample with BE/ME >=0 with CR and an Indicator Variable for the Year
Variable Energy Materials Indstrls Consmr
Discret
Consmr
Staples
Hlth
Care Financials
Info
Tech
Telecom
Svcs Utilities
Icept. 1.71 -0.18 1.56* -0.78 4.92* 3.39*** 6.17*** 1.19** 1.64 4.84**
fva1 0.37 3.08*** 0.67*** 0.98*** 0.85 1.59*** 0.32 1.66*** 0.16 0.35
fva2 0.51** 5.95*** 0.22 3.3*** 2.66* 1.49*** 0.02 2.43*** 2.01** 0.97*
fva3 0.29 -0.64 0.46 -1.04 -19.66 2.5** -0.31 1.69*** 0.53 -0.25
fvl12 -
0.53*** -1.11*** -1.31*** -0.98*** 0.12 0.36 -0.39 -0.31** 0.04 -1.49***
fvl3 0.03 -9.75 0.16 -3.14** -68.46* -1.57** 1.79 7.35*** 1.2 -1.71*
cr 1.1*** 0.45 0 3.82*** 3.41 0.12 0.3 0.81*** 2.62** -0.43
yr09 7.52*** 6*** 7.17*** 6.24*** 4.29* 3.15*** 4.55* 3.39*** 1.28 2.46
nfva 0.59*** 0.83*** 0.76*** 1.07*** 0.1 1.39*** 0.08 0.77*** 0.37** 0.28*
nfvl -0.5*** -0.55*** -0.64*** -1.05*** 0.1 -
1.57*** 0.34
-
0.62*** -0.18 -0.06
nipsh 0.86*** 0.06 1.35*** 0.67*** 7.04*** 1.8*** 0.52 1.84*** 3.15*** 2.47***
N 287 161 402 322 89 504 59 564 63 96
AdjR2 0.65 0.61 0.96 0.82 0.76 0.59 0.93 0.34 0.66 0.72
173
Table 18 (cont’d)
Panel N
This table shows the results of regressions by IRLS of the modified Ohlson model including a
term for the CR and an indicator variable for the year, yr09, for each sector of the GICS. The
indicator variable, yr09, takes the value 0 when the year is 2008 and 1 when the year is 2009.
Firms with BE/ME < 0 have been removed from the dataset for this sample, as described in the
right-most column with heading H4’ in Panel E. Aside from the difference in dataset, the
analysis of this table is the same as that shown in Panel J. Other formatting conventions follow
those of Panel F. The model is given by
PRCit = α0 + α1NFVAit + α2FVA1it + α3FVA2it + α4FVA3it + α5NVFLit+ α6FVL12it
+ α7FVL3it + α8 CRit + α9 Yr09it + α10 CRit ×FVA3it+ α11 CRit ×Yr09it
+ α12 FVA3it × Yrit + α13 CRit ×FVA3it ×Yr09it + β1NIit + εit
where other variables are given in Panel A of Table 9.
Panel O
This table shows the results of regressions by IRLS of the modified Ohlson model including the
CR, yr09, and their interaction for each sector of the GICS. Firms with BE/ME < 0 have been
removed from the dataset for this sample, as described in the right-most column with heading
H4’ in Panel E. Other formatting conventions follow those of Panel F. The model and
variables are those of Panel N.
Panel P
This table shows the results of regressions by IRLS of the modified Ohlson model including the
CR, yr09, and the interaction between CR and FVA3 for each sector of the GICS. Firms with
BE/ME < 0 have been removed from the dataset for this sample, as described in the right-most
column with heading H4’ in Panel E. Other formatting conventions follow those of Panel F.
The model and variables are those of Panel N.
Panel Q
This table shows the results of regressions by IRLS of the modified Ohlson model including the
CR, yr09, and the interaction between CR and FVA3 for each sector of the GICS. Firms with
BE/ME < 0 have been removed from the dataset for this sample, as described in the right-most
column with heading H4’ in Panel E. Other formatting conventions follow those of Panel F.
The model and variables are those of Panel N.
174
Table 18 (cont’d)
Panel O. IRLS Regression Modified Ohlson with Interaction between CR and Year Dummy
Variable Energy Materials Indstrls Consmr
Discret
Consmr
Staples
Hlth
Care Financials
Info
Tech
Telecom
Svcs28
Utilities
Icept. 2.55* -0.2 1.83** -0.86 4.1 3.25*** 5.96*** 1.34*** -1 4.78**
fva1 0.4 3.08*** 0.69*** 0.99*** 0.96 1.6*** 0.32 1.65*** 0.11 0.37
fva2 0.49** 5.95*** 0.25 3.3*** 2.78* 1.48*** 0.01 2.43*** 2.43** 0.97*
fva3 0.29 -0.6 0.45 -1.04 -22.43 2.41** -0.29 1.7*** 0.86 -0.27
fvl12 -0.52*** -1.13*** -1.25*** -0.98*** 0.2 0.35 -0.39 -0.31** 0.11 -1.51***
fvl3 0.05 -9.95 0.04 -3.11** -69.66* -1.56** 1.86 7.28*** 1.56 -1.7*
cr 0.7 0.32 -0.57 3.92*** 5.23 -0.11 0.36 0.47 5.08** -0.28
yr09 6.65*** 5.95*** 6.06*** 6.39*** 5.8 3.32*** 4.83* 3.19*** 4.62 2.64
cr × yr09 0.65 0.16 2.25** -0.27 -2.92 0.33 -0.14 0.65 -3.06 -0.38
nfva 0.57*** 0.84*** 0.75*** 1.07*** 0.1 1.39*** 0.08 0.77*** 0.36** 0.28*
nfvl -0.48*** -0.56*** -0.64*** -1.05*** 0.1 -1.56*** 0.34 -0.62*** -0.18 -0.06
nipsh 0.87*** 0.07 1.33*** 0.67*** 6.96*** 1.85*** 0.5 1.85*** 3.18*** 2.46***
N 287 161 402 322 89 504 59 564 63 96
AdjR2 0.63 0.61 0.96 0.82 0.75 0.59 0.93 0.34 0.67 0.72
28 The IRLS regression for Telecommunication Services did not converge.
175
Table 18 (cont’d)
Panel P. Modified Ohlson with Interaction between FVA3 and Year Dummy
Variable Energy Materials Indstrls Consmr
Discret
Consmr
Staples Hlth Care Financials Info Tech
Telecom
Svcs29
Utilities
Icept. 1.63 0.02 1.59* -0.98 5.25** 3.35*** 5.97*** 1.33*** 1.7 5.02**
fva1 0.38 3.1*** 0.67*** 0.98*** 0.7 1.59*** 0.31 1.67*** 0.17 0.34
fva2 0.51** 5.93*** 0.21 3.32*** 0.23 1.49*** 0.01 2.44*** 2.04** 0.98*
fva3 0.43 -2.89 0.16 -0.14 -45.22 2.69* -0.24 1.04*** 0.51 -0.93
fvl12 -0.53*** -1.11*** -1.31*** -0.97*** 0.06 0.37 -0.38 -0.33** 0.06 -1.48***
fvl3 0.41 -9.3 0.19 -3.1** -63.88 -1.56** 1.87 7.34*** 1.15 -1.74*
cr 1.12*** 0.45 -0.02 3.83*** 3.72 0.13 0.32 0.79*** 2.6** -0.43
yr09 7.79*** 5.67*** 7.08*** 6.44*** 2.51 3.2*** 4.85** 3.17*** 1.23 2.23
fva3 × yr09 -1.48 4.18 0.41 -1.25 70.53* -0.56 -0.38 0.88** 0.65 1.32
nfva 0.59*** 0.84*** 0.75*** 1.07*** 0.1 1.39*** 0.08 0.78*** 0.36** 0.27*
nfvl -0.51*** -0.56*** -0.64*** -1.05*** 0.11 -1.57*** 0.34 -0.64*** -0.16 -0.05
nipsh 0.86*** 0.06 1.35*** 0.67*** 7.43*** 1.79*** 0.5 1.87*** 3.17*** 2.48***
N 287 161 402 322 89 504 59 564 63 96
AdjR2 0.64 0.61 0.96 0.82 0.78 0.59 0.93 0.34 0.65 0.72
29 The IRLS regression for Telecommunication Services did not converge.
176
Table 18 (cont’d)
Panel Q. IRLS Regression including the Year Dummy with Interaction between FVA3 and CR
Variable Energy Materials Indstrls Consmr
Discret
Consmr
Staples
Health
Care Financials
Info
Tech
Telecom
Svcs30
Utilities
Icept. 1.68 -0.35 1.58* -1.06 4.93* 3.43*** 5.97*** 1.2** 1.61 4.75**
fva1 0.39 3.24*** 0.66*** 1*** 0.85 1.59*** 0.3 1.66*** 0.16 0.21
fva2 0.5** 6.03*** 0.2 3.31*** 2.73 1.49*** 0 2.43*** 2.02* 1.1**
fva3 1.08 0.84 0.16 -0.05 -20.97 2.3** -0.21 1.6*** 0.71 -8.95*
fvl12 -0.52*** -1.13*** -1.31*** -0.97*** 0.12 0.36 -0.39 -0.31** 0.03 -1.3***
fvl3 -0.2 -10.28 0.18 -3.17** -68.38* -1.56** 1.82 7.33*** 1.22 -1.06
cr 1.21*** 0.53 -0.13 4.26*** 3.39 0.15 0.33 0.73*** 2.64* -0.59
yr09 7.58*** 5.91*** 7.1*** 6.26*** 4.3* 3.17*** 4.83** 3.41*** 1.31 2.24
nfva 0.58*** 0.85*** 0.76*** 1.07*** 0.1 1.39*** 0.08 0.77*** 0.38** 0.25*
nfvl -0.49*** -0.57*** -0.64*** -1.05*** 0.1 -1.57*** 0.34 -
0.62*** -0.19 -0.01
nipsh 0.87*** 0.08 1.34*** 0.67*** 7.04*** 1.82*** 0.5 1.84*** 3.14*** 2.39***
fva3 × cr -0.5 -5.45 0.78 -2.06 2.94 -0.28 -0.06 0.58 -0.15 7.82**
N 287 161 402 322 89 504 59 564 63 96
AdjR2 0.64 0.63 0.96 0.83 0.76 0.59 0.93 0.34 0.65 0.73
30 The IRLS regression for Telecommunication Services did not converge.
177
Table 18 (cont’d)
R. Modified Ohlson with Triple Interaction
Variable Energy Materials Indstrls Consmr
Discret
Consmr
Staples Hlth Care Financials Info Tech
Telecom
Svcs Utilities
Icept. 2.15 -0.03 1.77** -1.84 4.67 3.19*** 5.92*** 1.49*** 1 5.04**
fva1 0.39 3.28*** 0.7*** 1.01*** 0.77 1.59*** 0.31 1.67*** 0.05 0.36
fva2 0.5** 5.9*** 0.21 3.25*** 0.38 1.45*** -0.05 2.42*** 2.72*** 1.02*
fva3 1.02 -1.28 0.15 1.72 -107.21 3.06** -0.2 0.84* -10.13 -9.63*
fvl12 -0.52*** -1.15*** -1.24*** -0.99*** 0.22 0.36 -0.42 -0.33** 0.08 -1.41***
fvl3 0.39 -10.22 -0.08 -3.1** -65.74* -1.54** 1.42 7.25*** 1.9 -1.28
cr 0.88* 0.57 -0.62 5.16*** 4.33 -0.25 0.36 0.4 2.82 -0.45
yr09 7.14*** 5.39*** 6.11*** 7.37*** 3.36 3.42*** 4.66* 3.01*** 2.18 1.93
nfva 0.58*** 0.87*** 0.75*** 1.08*** 0.09 1.38*** 0.11 0.78*** 0.32** 0.28*
nfvl -0.5*** -0.6*** -0.63*** -1.05*** 0.12 -1.56*** 0.3 -0.64*** -0.11 -0.05
nipsh 0.88*** 0.06 1.33*** 0.67*** 7.55*** 1.93*** 0.48 1.87*** 3.24*** 2.43***
fva3 × cr -0.39 -9.95 0.22 -3.99 128.59 0.4 -0.07 0.55 9.35 8.47**
fva3 × yr09 -2.34 2.14 -0.64 -4.12 129.68* 0.36 4.29 0.8 28.84 9.11
cr × yr09 0.5 -0.13 1.99* -1.63 -1.54 0.51 -0.09 0.48 -0.46 -0.24
fva3×cr×yr09 0.52 11.62 1.64 6.07 -123.36 -2.39*** -0.57 1.28 -14.96* -8.8
N 287 161 402 322 89 504 59 564 63 96
AdjR2 0.63 0.62 0.96 0.82 0.79 0.59 0.92 0.34 0.68 0.73
178
Table 18 (cont’d)
Panel R (cont’d)
This table shows the results of regressions by IRLS of the modified Ohlson model including the CR,
yr09, their pair-wise interaction with FVA3, and the triple interaction between CR, yr09, and FVA3
for each sector of the GICS. Firms with BE/ME < 0 have been removed from the dataset for this
sample, as described in the right-most column with heading H4’ in Panel E. Other formatting
conventions follow those of Panel F. The model and variables are those given in Panel N.
179
References
Bain, J. S. 1954. Economies of Scale, Concentration, and the Condition of Entry in Twenty
Manufacturing Industries. The American Economic Review 44 (1):15-39.
Ball, R., and P. Brown. 1968. An Empirical Evaluation of Accounting Income Numbers.
Journal of Accounting Research 6 (2):159-172.
Banko, J. C., C. M. Conover, and G. R. Jensen. 2006. The Relationship between the Value
Effect and Industry Affiliation. Journal of Business 79 (5):2595-2616.
Banz, Rolf W., 1981. The relationship between return and market value of common stocks,
Journal of Financial Economics. 9:1, 3-18.
Barth, M. E., W. H. Beaver, and W. R. Landsman. 2001. The relevance of the value
relevance literature for financial accounting standard setting: another view. Journal
of Accounting and Economics 31 (1-3):77-104.
Barth, M. E., and G. Clinch. 2009. Scale Effects in Capital Markets-Based Accounting
Research. Journal of Business Finance & Accounting 36 (3 & 4):253-288.
Bernard, V. L. 1989. Capital Markets Research in Accounting During the 1980s: A Critical
Review. In The State of Accounting Research as We Enter the 1990s. Champaign,
IL: The University of Illinois at Urbana-Champaign, 72-120.
Bhojraj, S., C. M. C. Lee, and D. K. Oler. 2003. What's My Line? A Comparison of Industry
Classification Schemes for Capital Market Research. Journal of Accounting
Research 41 (5):745-774.
Biddle, G. C., and G. S. Seow. 1991. The Estimation and Determinants of Associations
Between Returns and Earnings: Evidence from Cross-industry Comparisons.
Journal of Accounting, Auditing & Finance 6 (2):183-232.
Brown, P., and R. Ball. 1967. Some Preliminary Findings on the Association between the
Earnings of a Firm, Its Industry, and the Economy. Journal of Accounting Research
5 (Empirical Research in Accounting):55-77.
Cao, Jie. SAS Macro (Winsorization)
http://ihome.cuhk.edu.hk/~b121456/tools/Winsorize_Macro.txt
(accessed July 15, 2011).
Clarke, R. N. 1989. SICs as Delineators of Economic Markets. The Journal of Business 62
(1):17-31.
Daniel, S., B. Jung, H. Pourjalali, and E. Wen. 2010. Firm Characteristics Influencing
Responses Towards Adoption of the Fair Value Accounting Option: A Survey of
Chief Financial Officers of U.S. Firms: SSRN, 1579326.
Dichev, I. D. 2008. On the Balance Sheet-Based Model of Financial Reporting. Accounting
Horizons 22 (4):453-470.
El-Gazzar, S., P. Finn, and C. Tang. 2009. The Value Relevance of Earnings and
Nonearnings Information in Regulated and Deregulated Markets: The Case of the
Airline Industry. International Advances in Economic Research 15 (1):88-101.
Fama, E. F. 1970. Efficient Capital Markets: A Review of Theory and Empirical Work. The
Journal of Finance 25 (2):383-417.
———. 1991. Efficient Capital Markets: II. The Journal of Finance 46 (5):1575-1617.
Fama, E. F., and K. R. French. 1992. The Cross-Section of Expected Stock Returns. The
Journal of Finance 47 (2):427-465.
———. 1997. Industry costs of equity. Journal of Financial Economics 43 (2):153-193.
———. 2008. Dissecting Anomalies. Journal of Finance 63 (4):1653-1678.
180
Fama, E. F., and J. D. MacBeth. 1973. Risk, Return, and Equilibrium: Empirical Tests. The
Journal of Political Economy 81 (3):607-636.
Financial Accounting Standards Board. 2004. FASB Issues Proposal on Fair Value
Measurements. Norwalk, CT: News Release, June 23, 2004.
———. 2006. Fair Value Measurements. Norwalk, CT: Financial Accounting Standards
Board, FAS 157.
———. 2008. Effective Date of FASB Statement No. 157, edited by F. S. Position.
Norwalk, CT: Financial Accounting Standards Board, FSP FAS 157-2.
Fries, S., M. Miller, and W. Perraudin. 1997. Debt in Industry Equilibrium. The Review of
Financial Studies 10 (1):39-67.
Goh, B. W., J. Ng, and K. Ow Yong. 2009. Market Pricing of Banks' Fair Value Assets
Reported under SFAS 157 During the 2008 Economic Crisis: SSRN, 1335848.
Guenther, D. A., and A. J. Rosman. 1994. Differences between COMPUSTAT and CRSP
SIC codes and related effects on research. Journal of Accounting and Economics 18
(1):115-128.
Hou, K., and D. T. Robinson. 2006. Industry Concentration and Average Stock Returns. The
Journal of Finance 61 (4):1927-1956.
Huber, P. J. 1981. Robust Statistics. 1st ed. New York: John Wiley & Sons.
Jagannathan, R., and Z. Wang. 1996. The Conditional CAPM and the Cross-Section of
Expected Returns. The Journal of Finance 51 (1):3-53.
JP Morgan Chase, 2008. Annual Report.
Kile, C. O., and M. E. Phillips. 2009. Using Industry Classification Codes to Sample High-
Technology Firms: Analysis and Recommendations. Journal of Accounting,
Auditing & Finance 24 (1):35-58.
King, B. F. 1966. Market and Industry Factors in Stock Price Behavior. The Journal of
Business 39 (1):139-190.
Kolev, K. 2009. Do Investors Perceive Marking-to-Model as Marking-to-Myth? Early
Evidence from FAS 157 Disclosure: SSRN, 1336368.
Kothari, S. P. 2001. Capital Markets Research in Accounting. Journal of Accounting and
Economics 31 (1-3):105-231.
Lang, M., and R. Lundholm. 1996. The Relation Between Security Returns, Firm Earnings,
and Industry Earnings. Contemporary Accounting Research 13 (2):607-629.
Laux, C., and C. Leuz. 2009. The crisis of fair-value accounting: Making sense of the recent
debate. Accounting, Organizations and Society 34 (6-7):826-834.
Lev, B., and J. A. Ohlson. 1982. Market-Based Empirical Research in Accounting: A
Review, Interpretation, and Extension. Journal of Accounting Research 20
(Supplement):249-322.
Livingston, M. 1977. Industry Movements of Common Stocks. The Journal of Finance 32
(3):861-874.
MacKay, P., and G. M. Phillips. 2005. How Does Industry Affect Firm Financial Structure?
Review of Financial Studies 18 (4):1433-1466.
Maksimovic, V., and J. Zechner. 1991. Debt, Agency Costs, and Industry Equilibrium. The
Journal of Finance 46 (5):1619-1643.
MSN Money, Interactive Chart for INX (S&P 500 Index),
http://investing.money.msn.com/investments/charts?symbol=$US:INX#symbol=$I
NX&event=&BB=off&CCI=off&EMA=off&FSO=off&MACD=off&MFI=off&PS
AR=off&RSI=off&SMA=off&SSO=off&Volume=off&period=Custom&linetype=
Line&scale=Auto&dates=1/1/2007,1/1/2012&comparelist=$indu,$compx,$inx,
(accessed January 10, 2012).
181
Moskowitz, T. J., and M. Grinblatt. 1999. Do Industries Explain Momentum? The Journal of
Finance 54 (4):1249-1290.
Nwaeze, E. T. 1998. Regulation and the Valuation Relevance of Book Value and Earnings:
Evidence from the United States. Contemporary Accounting Research 15 (4):547-
573.
Ohlson, J. A. 1995. Earnings, book value and dividends in security valuation. Contemporary
Accounting Research 11 (2):661–687.
Penman, S. H. 2007. Financial Reporting Quality: Is Fair Value a Plus or a Minus?
Accounting and Business Research (International Accounting Policy Forum):33-44.
Riedl, E. J., and G. Serafeim. 2011. Information Risk and Fair Values: An Examination of
Equity Betas. Journal of Accounting Research 49 (4):1083-1122.
SAS Institute. 2008 SAS/STAT® 9.2 User’s Guide.
Saunders, N. C. 1999. The North American Industry Classification System: Change on the
Horizon. Occupational Outlook Quarterly 43 (3):34-37.
Schumpeter, J. 1912. The Theory of Economic Development. Cambridge, MA: Harvard
University Press.
Song, C. J., W. B. Thomas, and H. Yi. 2010. Value Relevance of FAS No. 157 Fair Value
Hierarchy Information and the Impact of Corporate Governance Mechanisms.
Accounting Review 85 (4):1375-1401.
Spiegel, M. 2011. The Academic Analysis of the 2008 Financial Crisis: Round 1. Review of
Financial Studies, 24(6): 1773 -1781.
Teets, W. 1992. The Association between Stock Market Responses to Earnings
Announcements and Regulation of Electric Utilities. Journal of Accounting
Research 30 (2):274-285.
US Department of Justice, “Herfindahl-Hirshman Index”:
http://www.justice.gov/atr/public/testimony/hhi.htm, (accessed July 15, 2011).
Weiner, C. 2005. The Impact of Inudstry Classification Schemes on Financial Research.
Berlin, Germany: Humbolt University School of Business and Economics
Discussion Paper 2005-062, 48.
Whited, T. M. 1992. Debt, Liquidity Constraints, and Corporate Investment: Evidence from
Panel Data. The Journal of Finance 47 (4):1425-1460.
Wikipedia. 2010. Studentized Residual: http://en.wikipedia.org/wiki/Studentized_residual.
Williams, J.T. 1995. Financial and Industrial Structure with Agency. The Review of
Financial Studies 8 (2): 431-474.
182
Appendix A
This appendix compares the Standard Deviation to the Median Absolute Deviation, MAD.
The well-known standard deviation is the square-root of the variance, where the well-known
variance is the second central moment. The MAD is the median of the absolute deviations
from the median, and is defined as follows:
where xi is an observation and
A large difference between the standard deviation and MAD suggests the presence of
outliers for a particular variable (SAS 2008).
183
Table A1. Standard Deviation and Standardized Median Absolute Deviation (MAD)
Untreated Winsorized Trimmed
GIC Sectors Variable StdDev MAD StdDev MAD StdDev MAD
Energy nfva 31.13 26.24 30.36 26.24 26.12 24.23
fva1 1.67 0.00 0.86 0.00 1.72 0.00
fva2 2.56 0.36 1.50 0.36 2.57 0.29
fva3 1.66 0.00 1.04 0.00 1.75 0.00
nfvl 19.70 15.89 19.07 15.89 16.75 14.93
fvl12 5.20 0.07 2.95 0.07 5.51 0.06
fvl3 0.53 0.00 0.48 0.00 0.55 0.00
nidsh 5.14 2.78 4.65 2.78 5.01 2.88
PRC 15.02 12.56 14.95 12.56 13.48 11.33
Materials nfva 24.59 23.50 24.59 23.50 21.98 24.07
fva1 1.05 0.00 1.05 0.00 1.10 0.00
fva2 0.59 0.02 0.59 0.02 0.49 0.01
fva3 0.53 0.00 0.53 0.00 0.57 0.00
nfvl 18.10 15.46 18.10 15.46 16.57 15.19
fvl12 3.71 0.12 3.71 0.12 3.13 0.08
fvl3 0.06 0.00 0.06 0.00 0.06 0.00
nidsh 4.31 1.83 4.31 1.83 2.85 1.80
PRC 22.78 16.55 22.78 16.55 14.68 10.31
Industrials nfva 112.72 22.28 25.99 22.28 114.97 22.68
fva1 7.31 0.03 2.22 0.03 7.46 0.01
fva2 13.26 0.00 1.22 0.00 13.53 0.00
fva3 1.15 0.00 0.53 0.00 1.17 0.00
nfvl 51.09 14.34 22.03 14.34 52.08 14.19
fvl12 1.93 0.08 1.36 0.08 1.97 0.07
fvl3 0.39 0.00 0.32 0.00 0.39 0.00
nidsh 8.77 1.78 2.89 1.78 8.92 1.66
PRC 71.21 10.36 12.78 10.36 72.42 10.13
184
Table A1. (cont’d)
Untreated Winsorized Trimmed
GIC Sectors Variable StdDev MAD StdDev MAD StdDev MAD
Cnsmr Discret. nfva 42.40 18.64 34.21 18.64 29.63 19.32
fva1 3.71 0.02 2.82 0.02 2.15 0.00
fva2 2.20 0.00 2.11 0.00 1.59 0.00
fva3 1.00 0.00 0.42 0.00 1.05 0.00
nfvl 26.72 13.41 25.31 13.41 24.47 13.99
fvl12 4.18 0.04 4.16 0.04 3.16 0.05
fvl3 0.11 0.00 0.02 0.00 0.11 0.00
nidsh 6.21 2.01 5.51 2.01 6.12 1.95
PRC 37.43 9.37 18.81 9.37 9.05 7.84
Cnsmr Staples nfva 30.72 21.22 30.72 21.22 31.75 18.77
fva1 1.73 0.02 1.73 0.02 1.79 0.03
fva2 5.20 0.00 5.20 0.00 5.41 0.00
fva3 0.79 0.00 0.79 0.00 0.82 0.00
nfvl 24.50 13.34 24.50 13.34 25.44 13.53
fvl12 2.47 0.11 2.47 0.11 2.57 0.12
fvl3 0.68 0.00 0.68 0.00 0.71 0.00
nidsh 2.56 1.57 2.56 1.57 2.65 1.55
PRC 24.07 17.27 24.07 17.27 24.37 15.45
Health Care nfva 19.40 5.10 18.02 5.10 18.78 3.51
fva1 1.84 0.44 1.67 0.44 1.86 0.47
fva2 4.57 0.01 4.03 0.01 4.80 0.00
fva3 0.57 0.00 0.40 0.00 0.58 0.00
nfvl 15.60 2.27 12.30 2.27 16.03 1.79
fvl12 1.01 0.00 0.90 0.00 1.00 0.00
fvl3 0.79 0.00 0.31 0.00 0.84 0.00
nidsh 2.07 1.43 1.90 1.43 1.92 1.24
PRC 16.09 8.00 15.15 8.00 9.96 5.75
185
Table A1. (cont’d)
Untreated Winsorized Trimmed
GIC Sectors Variable StdDev MAD StdDev MAD StdDev MAD
Financials nfva 109.55 76.12 89.94 76.12 81.12 74.49
fva1 31.23 0.07 16.54 0.07 19.10 0.05
fva2 52.74 16.25 42.36 16.25 41.52 17.67
fva3 6.07 0.00 4.83 0.00 5.19 0.00
nfvl 137.64 93.80 108.90 93.80 102.65 92.39
fvl12 20.33 0.00 2.69 0.00 2.43 0.00
fvl3 2.58 0.00 2.20 0.00 2.55 0.00
nidsh 4.26 1.22 2.90 1.22 2.97 1.16
PRC 23.99 7.43 12.55 7.43 15.24 6.53
Info Tech nfva 10.99 5.06 10.19 5.06 9.19 4.65
fva1 1.93 0.80 1.81 0.80 1.94 0.77
fva2 1.80 0.02 1.34 0.02 1.34 0.02
fva3 0.99 0.00 0.51 0.00 1.04 0.00
nfvl 8.68 2.17 7.25 2.17 6.70 1.96
fvl12 2.21 0.00 0.47 0.00 2.04 0.00
fvl3 0.07 0.00 0.02 0.00 0.07 0.00
nidsh 1.86 0.97 1.79 0.97 1.77 0.94
PRC 17.70 5.92 15.30 5.92 6.88 4.71
Telecom Svcs. nfva 18.43 16.13 18.43 16.13 19.09 17.62
fva1 1.80 0.17 1.80 0.17 1.85 0.28
fva2 1.13 0.00 1.13 0.00 1.17 0.00
fva3 0.33 0.00 0.33 0.00 0.35 0.00
nfvl 13.31 12.19 13.31 12.19 13.69 11.85
fvl12 0.84 0.02 0.84 0.02 0.87 0.02
fvl3 0.58 0.00 0.58 0.00 0.60 0.00
nidsh 1.58 1.13 1.58 1.13 1.63 1.13
PRC 10.46 11.90 10.46 11.90 10.02 10.83
186
Table A1. (cont’d)
Untreated Winsorized Trimmed
GIC Sectors Variable StdDev MAD StdDev MAD StdDev MAD
Utilities nfva 30.00 25.76 30.00 25.76 30.07 25.24
fva1 3.86 0.59 3.86 0.59 3.89 0.60
fva2 3.64 0.54 3.64 0.54 3.67 0.62
fva3 0.51 0.09 0.51 0.09 0.51 0.10
nfvl 24.72 21.80 24.72 21.80 24.88 19.48
fvl12 4.12 0.60 4.12 0.60 4.12 0.60
fvl3 0.67 0.00 0.67 0.00 0.67 0.00
nidsh 1.86 1.24 1.86 1.24 1.88 1.24
PRC 12.81 12.10 12.81 12.10 12.63 11.98
The Untreated column compares the standard deviation and MAD of the original sample
before Winsorization and the Winsorized column contains the same statistics, but after
Winsorization. If outliers are not present, the standard deviation and MAD should be similar.
The definitions of the variables are provided in Panel A of Table 9.
187
Appendix B. Computation of the Herfindahl of Sales
This appendix describes the computation of the Herfindahl based on net sales using data
from Compustat. Following Hou and Robinson (2006), the Herfindahl for this dissertation
was based on net sales for three years. The SALE variable31 for 2006 through 2008 was
extracted from Compustat (Fundamentals Annual with Monthly Updates). For each year,
the Herfindahl for industry j, Hj, is defined as:
where sij is the share of firm i in industry j, and the summation is for all J firms in industry j.
In order to perform the computation, the following definition of the share:
was used, where SUMj,
is the sum of the net sales for all firms in the industry. Inserting the expression for sij into
the definition of Hj gives
For each industry j, the SUMj is a constant, and can therefore be factored out of the terms of
the sum:
31 Compustat’s SALE variable represents gross sales (the amount of actual billings to customers for
regular sales completed during the period) reduced by cash discounts, trade discounts, and returned
sales and allowances for which credit is given to customers, for each operating segment.
188
Finally, the three-year average
where Hj,2006 is the Herfindahl of net sales for industry j in 2006, etc, and the overbar on the
Hj indicates the computation of an average was computed and used.
Therefore, for each firm, the square was computed as SALEi ×SALEi. Next, for each industry
in each year, the numerator of the Hj was computed as the sum of the SALEi×SALEi, the
denominator as the square of industry sum, SUMj (the sum of the SALEi), and finally Hj
itself as the quotient of the numerator and denominator. Finally, the average of the three Hj
of each year was computed to obtain . Note that when data from Compustat was
extracted, in order to be able to compute the Herfindahl on a wide range of firms, data on
both Active and Inactive firms was included and in particular firms were not eliminated
based on the month on which their fiscal year ends.