Majoring in Minors: Contextual ﬁnancial statement
analysis, XBRL, and small-cap stocks
August 18, 2006
∗ Corresponding author. Hampton University. Hampton, VA 23666. Phone: 757-727-5166 Email:
A simple screen based on publicly-available ﬁnancial statements consistently selects
a subset of the Russell 2000 which outperforms the index every year from 2001-2005
and generates an annualized 4-factor alpha of 7.58% over the same period. Nearly all of
the outperformance is a selection effect, regardless of how one chooses to segment the
market, e.g. trading volume, price-to-book, beta, industry sector or analyst coverage. The
strategy alpha provides a rough estimate of the upper-bound on the total reduction in the
cost of capital that might result from the widespread adoption of XBRL by small public
In the traditional efﬁcient markets paradigm, market prices reﬂect available information. But
information gathering is a costly activity. Research analysts don’t come cheap and the work-
product of an analyst is difﬁcult for the average person to replicate. Grossman and Stiglitz
(1980) demonstrates that market prices can not be fully revealing when information gathering
costs money. Wang (1993) shows how uninformed traders require a risk premium to trade in
markets dominated by informed traders. Easley and O’Hara (2004) goes further, suggesting
ﬁrms can reduce their cost of capital by choosing wisely features like accounting treatments,
analyst coverage, and market microstructure.
One innovative approach a company might take to improve its surrounding information
structure is by adopting XBRL. eXtensible Business Reporting Language (XBRL) is an adap-
tation of eXtensible Markup Language (XML) for business, accounting and ﬁnancial data.
It is a set of tags which identify ﬁnancial data so that it can be processed uniformly by any
software program developed in accordance with the standard. XBRL is an open language,
which (in theory) means that software developers are free to specialize in code that facilitates
data analysis as opposed to data acquisition and manipulation. Recent experience suggests
open standards such as XBRL shift the supply curve to the right, yielding more output at
lower prices. Reductions in the cost of fundamental analysis (the work-product of the analyst)
should result in increased analyst coverage and, according to Easley and O’Hara (2004), a
lower cost of capital.
That widespread adoption of XBRL would lead to higher analyst coverage for many com-
panies that don’t now receive it is a beneﬁt assumed by its proponents. Yet, at the time of
this writing, the adoption of the standard in industry is quite low, limited to FDIC insured
commercial banks and 24 companies who have volunteered to use the standard in an SEC
pilot program.1 Ironically, many of the companies participating in the SEC pilot are large
companies with lots of analyst coverage already.
An obvious question is: why would something so valuable go unadopted? And so in this
paper, we provide some rough “back of the envelope” calculations of the potential reduction
in the cost of capital for small companies that adopt the XBRL standard for ﬁnancial reporting
and thus beneﬁt from an explosion of amateur and professional analyst coverage sparked by
the widespread availability of XBRL-enabled analytical tools.
Drawing upon the notion of “contextual ﬁnancial statement analysis” from the accounting
literature, we adapt a screen developed by Piotroski (2000) and measure the returns it generates
from 2001-2005. Contextual ﬁnancial statement analysis, according to Beneish, Lee, and
Tarpley (2001), refers to the limited use of ﬁnancial statements to gain greater understanding
of the likely performance of a subset of ﬁrms with common characteristics. We study the
stocks in the Russell 2000. At the end of 2000, up to 32% of the companies in this index had
no analyst coverage, according to IBES. The companies are the smallest 2000 of the largest
3000 publicly traded companies in the United States. The largest companies in the Russell
2000 have market capitalizations of less than $2 billion. Thus, the screen relies heavily on
data from the ﬁnancial statements of uncovered stocks–just the type of screen that (in theory)
would be more easily implemented in a world of widespread XBRL adoption.
Since there is a growing literature exploring the investment value of contextual ﬁnancial
statement analysis, e.g. Mohanram (2005), the returns to the screen are of independent interest.
1 The SEC Interactive Data Initiative, see http://www.sec.gov/spotlight/xbrl.htm
We extend the work of Piotroski (2000) in two ways. First, the screen generates an annualized
4-factor alpha of 7.58%. This is comparable to the 7.5% improvement he documents when
applying the screen to a universe of high book-to-market stocks. However, our result suggests
that his screen can work in a much broader context, i.e. the Russell 2000. Second, the alpha
is generated in a later time period (2001-2005 versus 1975-1995) suggesting that some of the
inefﬁciency he unearthed still remains on the ground.
Though a risk-based explanation of the returns seems unlikely (the selected stocks tend to
have better ﬁnancials), there could be important economic barriers to implementation, e.g. low
liquidity. These barriers would slow expected improvements in a company’s cost of capital
by making it difﬁcult for investors to trade on XBRL-powered ideas. The screen overweights
stocks in the lowest trading volume category. Yet, performance attribution analysis by trading
volume categories reveals that all of the outperformance results from selection effects, not
allocation effects (which are a drag on performance). Moreover, more than 40% of the total
selection effect comes from stocks with above-average daily trading volume (around 150, 000
shares/day for stocks in the Russell 2000). Clearly, liquidity issues reduce the expected XBRL-
related reductions in the cost of capital, but a signiﬁcant amount remains, say 3.03% (0.40 ×
It is worth noting that a clean(er) test of XBRL beneﬁts involves comparing the cost of
capital of current XBRL adopters to similarly situated non-adopters. Or alternatively one
might analyze FDIC insured commercial banks only. Banks have to report quarterly ﬁnancials
to the FDIC in XBRL format and the requirement extends from the banks with lots of analyst
coverage to those without any. Therefore the information “playing ﬁeld”, with respect to
covered and uncovered banks, is level and an information-related alpha would be unlikely.
Unfortunately, we don’t have sufﬁcient data to carry out the analysis.
An experiment for which we do have sufﬁcient data simulates adopters by assuming that
the underlying ﬁnancial data is available 60 days after quarter-end (as opposed to 90). Please
note that we are not assuming in our tests that 10-Qs become available to the public 60 or 90
days after quarter-end. In fact, the SEC requires 10-Qs be ﬁled 45 days after quarter-end.2
What we are simulating in our screen is the time it takes for a 10-Q posted on EDGAR (most
likely as a PDF ﬁle) to ﬁnd its way into machine readable form on a commercially available
With the 30-day reduction, the return to the screen increases by 1.75%. And so we con-
clude that widespread adoption of XBRL by small public companies could reduce their cost of
capital by up to [1.75%, 3.03%]. Our results suggest that the beneﬁts of XBRL adoption are
large enough for small public companies to justify the cost of adoption. In addition, the returns
to contextual ﬁnancial statement analysis seem large enough to generate demand from the an-
alyst community that companies adopt the standard. In other words, XBRL may convince
analysts that it makes sense to “major in minors”.
The paper proceeds as follows. The next section describes the XBRL standard and explains
the promise and challenge of its adoption. Section IV presents the results of the screen we
describe in Section III. Section V presents the results of two tests which taken together help
quantify the information premium in these stocks, particularly those without analyst coverage.
Section VI concludes.
II. XBRL - Promise and Challenge
It is often stated that the internet is changing the world in ways similar to the agricultural and
industrial revolutions. The adoption of open standards for representing and manipulating in-
formation has lead to increased collaboration which has increased the productivity of workers
and enabled opportunities for trade that didn’t exist before.
Standards are key. Because in a world of many different languages, cultures, customs
etc., it is impossible to communicate without a common understanding of some type. XBRL
2 See Bryant-Kutcher, Peng, and Zvinakis (2005)
is a special purpose XML for accounting, business and ﬁnancial data.3 It provides a set of
common tags for accounting, business and ﬁnancial data that everyone worldwide recognizes.
The beneﬁts of worldwide recognition are obvious: everyone can look at XBRL-tagged data
and agree on what it is.
Of course, no additional agreement on the interpretation of the data need result. None
of the current debates about income and expense recognition will cease with the adoption of
XBRL. What XBRL does is enable raw ﬁnancial data to be directly addressed and classiﬁed
in whatever way seems most rational to the entity reporting it and the person analyzing it.
It is this granularity that is its real promise, though the development of XBRL tags certainly
hastens additional standardization in worldwide accounting principles.
Once tagged, XBRL data can be read and processed by any software that is built to under-
stand the standard. This allows the software development community to devote its entire skill
set towards the production of analytical value-add, since they are freed (through the adoption
of the standards) from all questions about how ﬁnancial data should be represented and stored.
In a similar way, the standards allow the analyst community to devote its entire skill base to
fundamental analysis of the data companies report, instead of expending resources to cap-
ture and translate the data for use in its analytical tools. The result, it is hoped, is a signiﬁcant
rightward shift in the marginal cost curve for suppliers of fundamental analysis, allowing them
to expand coverage of stocks beyond those now covered and to add additional nuance to the
coverage of those stocks now covered.
Analyst coverage has never been a money maker for the institutions that provide it. Before
the deregulation of commissions, the coverage function was funded by commission revenue.
Before the dot com bust, it was funded by investment banking business (expected and actual)
from covered companies. In the wake of new regulations and laws since the bust like Sarbanes-
Oxley, the analyst function has searched for a new business proposition.
3 See White (2006)
Initial indications are that it has not found one. The USA Today reported in 2003 that the
number of covered companies fell since SOX.4 Cuts in analyst staff litter the MDNAs of in-
vestment banks and other institutions that traditionally have been big providers of coverage. A
recent discussion at the SEC about the coverage afforded small companies openly considered
that small companies should pay for coverage, clearly not an optimal solution from an ethical
standpoint, even with the obvious well-intentioned safeguards factored in.5
Into this situation comes XBRL, which carries the promise of reducing costs for traditional
coverage providers and perhaps allowing current providers of “quasi-coverage” to develop an
attractive business case for providing actual coverage. Even individual investor armed with
screening software based on XBRL standards may be able to provide the analysis they need
on their own.
That’s the promise. But of course there are obstacles in the road from here to there. First,
there are the costs to the companies of adopting the new standard. Though the costs of pro-
ducing an XBRL tagged report are small,6 it is not free. And, in percentage terms, adoption is
most expensive for companies that stand to beneﬁt the most–small public uncovered compa-
nies. It might be that adoption of the standard will require SEC edict, yet it’s not clear that the
required cooperation from reporting companies–necessary in building a meaningful taxonomy
of XBRL tags–will come once “the stick” is used to speed adoption.
What would seem to be needed is a market based rationale for adopting the standards.
That’s what we’re trying to provide. It is this: uncovered stocks are meaningfully underpriced
(even after accounting for their liquidity disadvantages). This underpricing would reverse
somewhat if their data could be more easily captured and analyzed. XBRL will enable this to
happen. Therefore, a small uncovered company can reduce its cost of capital by adopting the
standard. The beneﬁts are at least equal to the cost, under the most pessimistic of assumptions.
4 USA Today recently reported 4,499 companies had active analyst coverage in 2003, down from 6,072 four
5 See Wander and Thyen (2006)
6 At the SEC’s Interactive Data Roundtable held June 12, 2006, one panelist estimated $300 and 80 man-hours.
So, in the next section, we describe an investment approach which would be much easier
to implement and demonstrably more proﬁtable with XBRL standards (though it does not
require them for implementation or proﬁtability).
III. Contextual Financial Statement Analysis (an adaptation)
To get a sense of the gains from XBRL tagged data, we identify an investment approach that
is heavily dependent on company-issued ﬁnancial reports. In an efﬁcient market, the value
of these (and any other publicly available information for that matter) is zero. However, in
the world of Easley and O’Hara (2004) such information, widely disseminated and quickly
connected to powerful new analytical tools that also are widely available, might have value.
The investment approach we implement is called contextual ﬁnancial statement analysis.
The idea is ﬁnancial statements have value for in the analysis of uncovered companies because
they are a source of information for reports that would have been issued if someone had the
time/inclination to do the analysis. The statements have value, not in and of themselves, but
because the analysis one would do with them would not already exist in the marketplace.
The output of this missing analysis is presumably known only to the insiders of the ﬁrm.
Granted it would be “right there” if someone took the time to look. But since no one takes the
time to look, only the insiders end up knowing.
This is the sense in which we apply the insights of Easley and O’Hara (2004). In this world,
the company’s stock carries a risk premium because of the imbalance of private versus public
information. The company can’t shed the risk premium by talking more, because it won’t be
believed if its pronouncements are good and it has no incentive to make bad pronouncements
just to improve long run credibility because its existence in the long run is not assured.
So it just sits there, undervalued.
In our way of thinking, contextual ﬁnancial statement analysis is simply the automated
production of analyst reports. And the returns to the strategy are thus the captured deadweight
loss from a disadvantaged information structure.
The list of papers documenting returns to context ﬁnancial statement analysis is signiﬁcant
and growing. Our work builds on two of them: Piotroski (2000) and Mohanram (2005).
In Piotroski (2000), investors employ a simple screen based on the ﬁnancial statements of
high book-to-market ﬁrms. A composite of binary indicators of improvements in proﬁtability,
leverage, and operating efﬁciency selects ﬁnancially strong high BM ﬁrms that outperform the
full universe by 7.5% and shifts the return distribution to the right. Financial statements add
value in the analysis of high BM ﬁrms–it is argued–because the analyst community neglects
the ﬁrms, the ﬁrms’ voluntary disclosures are not believed, and/or the greater part of ﬁrm value
is derived from the performance of assets-in-place. All three conditions establish ﬁnancial
statements as the most accessible, reliable, and important source of information about these
In Mohanram (2005), investors employ an analogous screen on low book-to-market stocks.
A composite of binary indicators–this time tailored for growth ﬁrms–screens stocks, select-
ing ﬁnancially weak low BM ﬁrms that underperform the full universe by about the same
magnitude. Financial statements add value in the analysis of low BM ﬁrms because the ana-
lyst community overpraises the ﬁrms, the ﬁrms’ voluntary disclosures are not reviewed with
a critical eye, and/or the performance of assets-in-place provides a “ﬂoor” for growth ﬁrm
valuations. Again, all three conditions establish ﬁnancial statements as a valuable source of
We adapt the Piotroski (2000) screen so that it more closely approximates the work of a
“reasonably diligent” analyst. Though we use many of the same indicators as Piotroski (2000)
(and, like him, do not attempt to uncover an optimal set of indicators), we improve the screen
by seeking evidence of sustained improvement in the indicators over each quarter of the past
year. This adaptation models our assumption that most analysts who use one-year-old ﬁnancial
statements would also incorporate information from statements issued throughout that year to
avoid being misled by ”one shot wonders”. It also aids in the detection of companies who are
showing “organizational momentum”, that is continuous improvement in performance. Note
that the demands for company-issued ﬁnancial reports is more intense in our strategy than in
We apply the adapted Piotroski (2000) screen by calculating quarterly ﬁnancial ranks using
a rolling four-quarter method. The framework analyzes nine ﬁnancial ratios and assigns a “1”
if the ﬁnancial signal is positive and a “0” if the ﬁnancial signal is negative. The signals are
summed across the ﬁnancial ratios to rank stocks from ”0” where no signals are positive to
”9” where all nine are positive (See Table 1).
A value-weighted portfolio is constructed from those stocks that achieve a ranking of 6
or more for four consecutive quarters. The portfolios are held for a full year and rebalanced
annually. In the initial tables, we present results from a strategy that identiﬁes screened stocks
at December-end using the four quarters that preceded it ending with the quarter that ends
September 30. In a later section, we present results from the strategy under the assumption that
the screened stocks are identiﬁed at November-end. This shift simulates quicker availability
of quarter-end September 30 data due to widespread XBRL adoption.
One additional change is we adapt the screen of Piotroski (2000) for use on the stocks that
comprise the Russell 2000 index. These stocks are the 1,001 to 3,000 largest stocks in the
U.S. general equity universe, but they are predominantly small cap stocks (less than $2 billion
in market capitalization).
Piotroski’s paper focused on high BM ﬁrms (many of them companies with small market
capitalizations), reasoning that in this context, ﬁnancial statements were particularly valuable
for the reasons stated above. There are many reasons why the Russell 2000 is a promising
group of candidate stocks for our analysis. First, not all companies in the Russell 2000 are
high BM, making it possible to test the widespread applicability of the screen along this di-
mension, as well as along other dimensions like trading volume. Also, the Russell 2000 often
serves as an index for passive strategies, ensuring enough liquidity in the component stocks for
trading purposes. Finally, the vast majority of the stocks in the index have little or no analyst
coverage–the primary market condition that XBRL adoption will ameliorate.
The following quarterly ﬁnancials are pulled from COMPUSTAT for each company in the
index and run through commercially available screening and backtesting software developed
• Long-term Debt
• Cash Flow from Sale of Common Stock
• Net Sales
• Total Assets
• Income Taxes
• Interest Expense
• Cash Flow from Operations
• Current Assets
• Current Liabilities
• Net Income
Factset runs the screen and outputs attribution analysis automatically. It includes sound
approaches to the traditional data issues that plague empirical studies in ﬁnance, like survivor-
ship bias and look-ahead bias. It is output from this program which we report in the subsequent
Table 1. Financial Ratio Framework
Financial Ratio Deﬁnition Signal Value
ROA Current year’s return on assets Greater than zero? 1 if Yes,
0 if No
∆ROA Current year’s ROA less prior year’s ROA Improved ROA? 1 if Yes,
0 if No
CFO Total cash ﬂow from operations Greater than zero? 1 if Yes,
0 if No
ACCRUAL Current year’s net income less cash ﬂow from Greater than ROA? 1 if Yes,
operations, scaled by beginning of the year 0 if No
∆MARGIN Current year’s operating margin less prior year’s Improved 1 if Yes,
operating margin operating margin? 0 if No
∆LEV ER Current year’s ratio of total long-term debt to Improved long- 1 if Yes,
average total assets less prior year’s long-term term debt to 0 if No
total debt to average total assets asset ratio?
∆TURNOV ER Current year’s asset turnover ratio less prior Improved Asset 1 if Yes,
year’s asset turnover ratio turnover ratio? 0 if No
∆LIQUID Current year’s current ratio less prior year’s Improved current 1 if Yes,
current ratio ratio? 0 if No
EQOFFER Current year equity offering Cash from equity 1 if Yes,
greater than zero? 0 if No
Total Score 0 to 9
IV. Investment Performance & Attribution
Each year our screen selects a subset of the Russell 2000 and holds it for 12 months. The aver-
age characteristics of the selected stocks is presented in Table 2. There we see that the selected
stocks are larger, more proﬁtable, and have lower price multiples. Clearly, the characteristics
of the selected stocks are not typical for a ﬁrm in distress or loading on a “distress” factor.
Table 2. Selected Characteristics of Selected Stocks and the Russell 2000 Index
Characteristic Screen Index
Market Cap 539.29 444.40
Dividend Yield 1.08 1.29
Price/Cash Flow 16.15 27.49
Price/Book 3.53 4.05
Price/Sales 2.81 18.90
ROA 8.76 1.66
ROE 17.41 6.84
Operating Margin 17.53 13.69
Net Margin 10.51 5.61
LT Debt/Capital 28.29 30.14
The performance of the screen in each year (2001-2005) is presented in Table 3. There are
large raw return differences between the selected stocks and the index in every year studied.
On average, the screen outperforms the index by 14.42%. In fact, in none of the years does
the screen outperform the index by less than 4%. Over the entire study period, the alpha
with respect to a Fama-French 3-factor model plus the Carhart momentum factor is 0.61% per
month (t-stat: 2.48), or 7.58% on an annualized basis.
This performance is on the order of Piotroski (2000) who documented an average raw
return difference of 7.5% annually between the stocks that met his screen and his universe of
high book-to-market stocks.
Table 3. Investment Performance of Selected Stocks and the Russell 2000 Index
Year Screen Index
2001 24.43 2.46
2002 1.10 -20.45
2003 51.74 47.26
2004 32.15 18.14
2005 14.64 4.55
Averages 24.81 10.39
Taken together with the 20-year study in Piotroski (2000), the results point to a puzzling,
continuing value of contextual ﬁnancial statement analysis in the small cap market. Moreover,
the characteristics of the selected stocks point away from a risk-based explanation. The se-
lected stocks appear to have smaller amounts of ﬁnancial and operating risk, at least according
to the measures in Table 2.
The performance attribution analysis in Tables 4-8 further confounds a risk-based expla-
nation. In these tables, the performance of the screen relative to the index is decomposed into
an allocation effect and a selection effect in the standard way (see Reilly and Brown (2003)).
In each year, the effects are normalized by the return difference in that year so that an “aver-
age effect” over all years can be obtained. The attribution analysis is carried out along ﬁve
dimensions: trading volume, price-to-book ratio, beta, industry sector, and analyst coverage.
From the tables, we see that nearly all of the outperformance is due to selection effects.
not allocation effects. Though the screen overweights stocks with low trading volume and low
price-to-book, the contribution of those allocation effects are not more than 4% in either case.
The screen might pick a stock with low trading volume or low price-to-book, but apparently it
ﬁnds such stocks in a state of “neglect” not “distress”.
Similarly, the contribution of beta allocation effects is negative. Interestingly, the strategy
overweights low beta stocks, another difﬁculty for risk-based explanations.
The analysis by industry sector reveals that the screen managed to avoid the dot com bust.
However, the average difference between the index weights and the screen weights is generally
quite small. There does not appear to be a clear “industry story” here.
Table 4. Investment Performance Attribution Analysis: Trading Volume
Decile Screen Index Allocation Selection Total
Deﬁnition Weight Weight Effect Effect Effect
0.9-73.7 7.90 10.41 -0.07 0.04 -0.04
0.6-0.9 5.85 8.80 -0.02 0.09 0.07
0.4-0.6 4.84 9.33 -0.02 0.04 0.02
0.3-0.4 7.14 9.22 0.02 0.07 0.09
0.2-0.3 7.10 8.51 0.01 0.02 0.03
0.1-0.2 8.13 9.50 -0.02 0.27 0.25
0.1-0.1 9.43 9.72 0.00 0.04 0.05
0.1-0.1 11.38 10.08 -0.03 0.13 0.10
0.0-0.1 11.83 10.09 -0.00 0.23 0.23
0.0-0.0 22.49 9.20 -0.12 0.28 0.16
N/A 3.90 5.14 -0.00 0.06 0.05
Totals 100.00 100.00 -0.24 1.24 1.00
Deciles are formed based on the average daily trading volume (in millions of shares) of stocks in the Russell 2000 index,
as of December 31, 2000. Stocks are assigned to the N/A category if their trading volume is not reported.
Table 5. Investment Performance Attribution Analysis: Price-to-book ratio
Quintile Screen Index Allocation Selection Total
Deﬁnition Weight Weight Effect Effect Effect
3.9-1754.4 13.20 15.75 0.01 0.18 0.19
2.3-3.9 15.74 17.03 0.01 0.01 0.02
1.6-2.3 18.80 15.30 0.00 0.24 0.25
1.1-1.6 20.91 17.41 0.00 0.16 0.16
0.0-1.1 23.32 14.72 0.05 0.20 0.25
N/A 8.03 19.78 -0.03 0.15 0.12
Totals 100.00 100.00 0.04 0.96 1.00
Quintiles are formed based on price-to-book ratios of stocks in the Russell 2000 index, as of December 31, 2000.
Stocks are assigned to the N/A category if their book value at that time is negative.
Table 6. Investment Performance Attribution Analysis: Beta
Quintile Screen Index Allocation Selection Total
Deﬁnition Weight Weight Effect Effect Effect
1.9-14.0 6.04 12.81 0.00 0.08 0.08
1.2-1.9 13.93 17.63 -0.02 0.11 0.09
0.7-1.2 17.02 19.81 0.01 0.08 0.09
0.3-0.7 27.12 22.11 -0.02 0.25 0.23
-2.4-0.3 31.59 21.61 -0.02 0.47 0.45
N/A 4.31 6.03 -0.01 0.06 0.05
Totals 100.00 100.00 -0.05 1.05 1.00
Quintiles are formed based on equity betas with respect to a value-weighted index of all stocks on the NYSE.
Stocks are assigned to the N/A category if they have not traded publicly for at least 12 months prior to December 31, 2000.
Table 7. Investment Performance Attribution Analysis: Industry Sector
Decile Screen Index Allocation Selection Total
Deﬁnition Weight Weight Effect Effect Effect
Autos And Transportation 5.13 3.93 -0.00 -0.01 -0.01
Consumer Discretionary 20.35 18.16 0.02 0.28 0.30
Consumer Staples 5.40 2.38 -0.09 0.06 -0.03
Financial Services 18.21 22.91 -0.03 0.29 0.26
Health Care 11.75 12.51 0.02 0.15 0.17
Integrated Oils 0.36 0.08 0.01 0.01 0.01
Materials And Processing 7.35 9.16 -0.01 0.04 0.03
Other 0.79 0.63 -0.00 0.01 0.01
Other Energy 6.59 4.18 0.00 -0.01 -0.01
Producer Durables 7.51 8.28 -0.02 0.04 0.02
Technology 9.83 13.09 0.03 0.07 0.04
Utilities 5.94 4.68 0.02 0.05 0.06
Unassigned 0.79 0.02 0.13 0.01 0.14
Totals 100.00 100.00 0.01 0.99 1.00
Eleven industry sectors are deﬁned by Frank Russell Company. Stocks are assigned to the “Other” or “Unassigned” category if the Frank Russell Company
does not assign it to one of the eleven sectors.
Table 8. Investment Performance Attribution Analysis: Analyst Coverage
Category Screen Index Allocation Selection Total
Deﬁnition Weight Weight Effect Effect Effect
3+ 44.01 43.40 0.01 0.49 0.53
2 11.07 11.48 -0.01 0.13 0.12
1 14.97 13.15 0.01 0.05 0.06
0 18.74 13.56 -0.04 0.24 0.21
N/A 11.21 18.41 0.05 0.06 0.10
Totals 100.00 100.00 0.02 0.98 1.00
The number of analysts who have ﬁled earnings estimates on a particular company with IBES as of December 31, 2000.
The performance attribution analysis does not support a traditional risk-based explanation
of the screen’s proﬁtability. However, there is evidence for the notion of information structure
risk as described in Easley and O’Hara (2004). In Table 8, we again see the dominant role
of selection effects in explaining the outperformance (98%), but also clear is the signiﬁcant
contribution of selected stocks with no analyst coverage. Not every uncovered stock is a
“treasure awaiting discovery” and the screen seems able to make the distinction. Perhaps
surprising is the extremely strong contribution of selected stocks having 3 or more analysts.
Apparently, even some covered stocks are “diamonds in the rough” and the screen catches
these as well.
In the next section, we discuss two simple thought experiments that help quantify the
beneﬁts of XBRL adoption.
V. Information Structure Risk and the Cost of Capital
Easley and O’Hara (2004) offers the possibility of a company inﬂuencing its cost of capital
by inﬂuencing the information structure surrounding the trading in its stock. Though it seems
logical to suggest that a company could improve that structure by gaining more analyst cover-
age. In practice, the number of analysts covering the ﬁrm is not a decision variable for the ﬁrm.
It is more likely dependent on the supply decisions of the analysts themselves as they provide
coverage until marginal cost (the value of the time spent in alternative investment activities
if the next report is not written) equals the marginal revenue (commissions and investment
banking revenue resulting from the next report).7 However, a ﬁrm can control whether it tags
its data using XBRL.
7 At the SEC’s Interactive Data Roundtable held June 12, 2006, one analyst said that the typical analyst
covers between 20-40 stocks. Another suggested that of the 27 stocks he covered (along with 2 support staff), 12
were covered by everyone else in the industry and the remaining 15 were “pet projects”, based on his expected
The decision of the U.S. Federal Deposit Insurance Corporation (FDIC) to require that
member banks submit quarterly ”Call Reports” in XBRL format provides an opportunity for
a direct test of XBRL on the cost of capital of the ﬁrms. According to White (2006), the
FDIC decision has resulted in a “dramatic improvement in the accuracy of the data that its
8,300 member banks submit and the ability to analyze the information in the reports in days
rather than months.” From the perspective of investors, it has enabled the FDIC to generate a
standardized ﬁnancial picture of all member banks (regardless of analyst coverage) in a timely
way–exactly the promise of XBRL. If XBRL has a positive effect on banks’ cost of capital, it
would be surprising to see uncovered banks outperforming the covered banks for information-
related reasons (they all are adopters). Though such a study is beyond the scope of this paper,
the market experience of these ﬁrms is an interesting topic for further research.
As a second experiment, we run our screen with one slight modiﬁcation. Instead of se-
lecting stocks by the screen at the end of December (90 days after the 3rd quarter has ended),
we do so at the end of November.8 Note we are not suggesting that the simulation models
information timeliness in a market efﬁciency sense. The information in 10-Qs and 10-Ks is
available immediately on the day they are ﬁled (through Edgar or company website). What we
are modeling instead–in bringing the screen date closer to the quarter end–is the timely broad
dissemination of this information, i.e. the time it takes to get from the PDF ﬁle on Edgar or
the company website into a commercially available database like compustat. XBRL, at least
potentially, improves timeliness in this more nuanced sense.
The results are as follows: the assumption of quicker availability improves screen prof-
itability by 1.75%–from 14.42% to 16.17% according to Table 9. If the screen can be imple-
mented sooner after the 10Ks and 10Qs become public, affected companies can look forward
to reducing their cost of capital by this amount. In other words, their stock currently offers
investors at 1.75% bonus if they can carry out the Piotroski (2000) analysis 30 days quicker.
Until 2003, the 10-K annual report was due 3 months following the end of the ﬁscal year, but that due date
was shortened to 75 days as part of SOX. 10-Q quarterly reports are due 45 days following the end of the ﬁscal
Table 9. Performance of Selected Stocks & the Russell 2000 Index (Nov-Nov)
Year Screen Index
2001 24.27 4.83
2002 8.02 -10.59
2003 50.85 36.26
2004 33.82 17.09
2005 19.63 8.15
Averages 27.32 11.15
A simple screen based on publicly-available ﬁnancial statements consistently selects a subset
of the Russell 2000 which outperforms the index every year from 2001-2005 and generates an
annualized 4-factor alpha of 7.58% over the same period. Nearly all of the outperformance is
a selection effect, regardless of how one chooses to segment the market, e.g. trading volume,
price-to-book, beta, industry sector or analyst coverage. The strategy alpha provides a rough
estimate of the upper-bound on the total reduction in the cost of capital that might result from
the widespread adoption of XBRL by small public companies.
Beneish, M., C. Lee, and R. Tarpley, 2001, Contextual fundamental analysis through the prediction of
extreme returns, Review of Accounting Studies 6, 165–189.
Bryant-Kutcher, L., E. Peng, and K. Zvinakis, 2005, Timeliness and quality of 10-K ﬁlings: The impact
of the accelerated ﬁling deadline, Working paper, University of Oregon.
Easley, D., and M. O’Hara, 2004, Information and the cost of capital, The Journal of Finance 59,
Grossman, S., and J. Stiglitz, 1980, On the impossibility of informationally efﬁcient markets, American
Economic Review 70, 393–408.
Mohanram, P., 2005, Separating winners from losers among low book-to-market stocks using ﬁnancial
statement analysis, Review of Accounting Studies 10, 133–170.
Piotroski, J., 2000, Value investing: the use of historical ﬁnancial statement information to separate
winners from losers, Journal of Accounting Research 38, 1–41.
Reilly, F., and K. Brown, 2003, Investment Analysis and Portfolio Management. (Thomson South-
Western United States).
Wander, H., and J. Thyen, 2006, Final report of the advisory committee on smaller public companies
to the U.S. Securities and Exchange Commission.
Wang, J., 1993, A model of intertemporal asset prices under asymmetric information, Review of Eco-
nomic Studies 60, 249–282.
White, C., 2006, The accountant’s guide to XBRL. (Clinton E. White, Jr. www.skipwhite.com).