Accepted Manuscript

Does IFRS make analysts more efficient in using fundamental information in-
cluded in financial statements?

Nandana P.W. Pathiranage, Christine A. Jubb

PII: S1815-5669(18)30148-6
DOI: https://doi.org/10.1016/j.jcae.2018.10.004
Reference: JCAE 141

To appear in: Journal of Contemporary Accounting & Economics

Received Date: 10 April 2017

Revised Date: 12 September 2018
Accepted Date: 2 October 2018

Please cite this article as: Pathiranage, N.P.W., Jubb, C.A., Does IFRS make analysts more efficient in using
fundamental information included in financial statements?, Journal of Contemporary Accounting & Economics
(2018), doi: https://doi.org/10.1016/j.jcae.2018.10.004

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
 Does IFRS make analysts more efficient in using
fundamental information included in financial statements?

Nandana P. W. Pathiranagea

and

Christine A. Jubbb

Acknowledgements: The authors thank Professor Asheq Rahman, Auckland University of Technology, who was
the discussant at the JCAE Mid-year Conference 2017, University of South Australia, Adelaide, for his
comments.
a
Swinburne Business School, Swinburne University of Technology, John Street, Hawthorn, Victoria, Australia,
3122, and University of Kelaniya, Sri Lanka.
b
Swinburne Business School, Swinburne University of Technology, John Street, Hawthorn, Victoria, Australia,
3122,
Address for Correspondence: Nandana Wasantha Pathiranage, Swinburne University of Technology, John
Street, Hawthorn, Vic, Australia 3122, email: nwasanthapathiranage@swin.edu.au

Does IFRS make analysts more efficient in using

fundamental information included in financial statements?

ABSTRACT

This paper investigates the efficiency with which analysts use fundamental signals when
forecasting one-year-ahead change in earnings per share (EPS) in Australian and European
contexts and the impact of International Financial Reporting Standards (IFRS) on this
efficiency. Results reveal that adoption of IFRS seems to increase analysts’ awareness of
fundamental signals useful for predicting future changes in EPS. However, overall, analysts
remain only as efficient as they were pre-IFRS in using these fundamental signals. While
their efficiency in using the earnings signal decreased, it increased for non-earnings signals
in the post- compared to pre-IFRS period. Furthermore, analysts substantially underutilise
the earnings signal in common compared to code law countries. These findings are likely to
be of interest to analysts and market participants when making forecasts and investment
decisions, and to standard setters and regulators in evaluating the impact of accounting
standards.

Key words: analysts’ efficiency, analysts’ forecasts, fundamental signals, IFRS

JEL: M41
 Does IFRS make analysts more efficient in using
fundamental information included in financial statements?

1.0 Introduction

Analysts play an important intermediary role in enhancing the informativeness of financial

statements (Sun 2011) and facilitating information flows in capital markets (Ramnath et al.
2008, Bradshaw 2011). They are an important user group in terms of financial statement
fundamental signals, using this information in making their earnings forecasts and valuing
shares (Lev & Thiagarajan 1993). If analysts are efficient in selecting and using appropriate
information, their outputs will be more accurate, minimising information asymmetry. Here
‘efficiency’ refers to analysts’ ability to use information content regarding future earnings
embedded in fundamental information.

Despite analysts being major users of accounting information, evidence exists of their less
than optimal efficiency in using fundamental accounting signals (e.g., De Bondt and Thaler
1990, Easterwood and Nutt 1999, Bradshaw et al. 2001, Abarbanell and Bushee 1997,
Swanson et al. 2003, Wahab et al. 2015). This study examines whether the adoption of
International Financial Reporting Standards (IFRS), which supposedly increases accounting
information quality (e.g., Barth et al. 2008), improves analysts’ efficiency.

Adoption of IFRS brought about changes in terms of recognition, measurement and

presentation of financial statement elements and, more generally, affected the
transparency and comparability of financial statements. The impact of IFRS on analyst
efficiency provides another perspective through which to view the change in the accounting
environment induced by IFRS, a perspective which to date has received little research
attention.

Adoption of IFRS has been shown to increase the quality of the accounting information
environment (e.g., Barth et al. 2008, Christensen et al. 2015, Chalmers et al. 2011, Cotter et
al. 2012, Beuselinck et al. 2010, De George et al. 2016, Charitou et al. 2018), thereby being
likely to affect analysts’ efficiency in using fundamental accounting signals in their earnings
forecasts. However, IFRS also introduced greater complexity, reflected in higher audit fees

1
(Kim et al. 2012) and associated in particular with complex fair value reporting (Goncharov
et al. 2014), potentially rendering the analysts’ task more difficult. A better understanding of
the impact of IFRS on analysts’ efficiency is important for analysts, investors, financial
statement preparers and regulators.

Filling the gap in research, this study provides evidence of IFRS impact on analysts’ efficiency
in using fundamental signals and in a non-US context. Further, a much larger pooled sample1
compared with relevant prior studies (Abarbanell and Bushee 1997, Swanson et al. 2003,
Wahab et al. 2015) provides compelling and robust evidence in respect of analysts’
efficiency.

The context of this study is 11 European countries plus Australia, countries that adopted
IFRS at a similar time, controlling to the extent possible for confounding events. Further,
given that the quality of accounting information also depends on investor protection (Houqe
et al. 2012) and the legal origin of the rule of law (e.g., Alford et al. 1993), the sample
countries chosen have similarly high levels of investor protection (WEF 2012). Whether the
country’s legal system is based on code or common law was controlled, which also accounts
for the difference between generally accepted accounting principles (GAAP) and IFRS (GAAP
difference). While existing research indicates that countries with higher rather than lower
differences between GAAP and IFRS experience greater increase in earnings quality upon
IFRS adoption (e.g., Bae et al. 2008, Ding et al. 2007), this paper provides compelling
evidence of pre- and post-IFRS analyst efficiency in using fundamental signals.

In their seminal study, Lev and Thiagarajan (1993) identified 12 non-earnings fundamental
signals referencing analysts’ written pronouncements. Abarbanell and Bushee (1997) used
eight of these to study analysts’ efficiency in using these fundamental signals. The current
study used earnings plus eight non-earnings fundamental signals from Abarbanell and
Bushee (1997) and four additional non-earnings fundamental signals useful for analyst
forecasts (Dempsey et al. 1997) to examine one-year-ahead change in earnings per share
(EPS) forecasts and the IFRS impact on this efficiency.

1
Abarbanell and Bushee (1997) used 2,609 observations (1983–1990) from a US context; Swanson et al.
(2003) used 354 observations from Mexico; and Wahab et al. (2015) used 219 quarterly observations
(2010-2011) from the US.

2
The findings reveal that analysts seem to be aware of the information content about future
earnings embedded in most of the fundamental signals examined; however, they appear
inefficient in using some of this information. As such, they appear to underreact or
overreact to certain fundamental signals when forecasting one-year-ahead change in EPS.
Furthermore, analysts seem to understand the importance of the included fundamental
signals after rather than before the adoption of IFRS. Evidence of inefficiency remains in
using these signals despite evidence of increased accounting information quality after IFRS
adoption (e.g., Barth et al. 2008). It must be noted that this apparent inefficiency may be
wholly or partly attributable to measurement errors associated with ‘noisy data’ and/or
analysts’ incentives, particularly the incentive for optimism (Ramnath et al. 2008). Findings
also reveal that IFRS impact on analysts’ efficiency is different between earnings and non-
earnings signals when forecasting one-year-ahead change in EPS and analysts’ efficiency in
using the earnings signal varies between code and common law countries.

The remainder of this paper proceeds as follows. The next section reviews the literature and
develops the hypotheses. Subsequently, section 3 explains the method and models used for
testing the hypotheses. Section 4 reports the results and section 5 concludes.

2.0 Literature review and hypotheses development

Financial statements provide useful fundamental information for market participants to

predict future earnings and abnormal returns (e.g., Ou and Penman 1989, Frankel and Lee
1998, Lev and Thiagarajan 1993, Abarbanell and Bushee 1997, Al-Debie and Walker 1999,
Ohlson and Penman 1992, El-Gazzar et al. 2009). Identification of what accounting
information is useful has a long history. Dempsey et al. (1997) identified a set of financial
information used frequently by analysts and other stakeholders based on a survey of 420
financial statement users. Ou and Penman (1989) identified 68 financial statement items,
later reduced to 16 based on statistical significance, useful in predicting annual EPS change,
referring to these items as ‘fundamental signals’.2 Lev and Thiagarajan (1993) identified 12

2
Fundamental signals represent changes in balance sheet or income statement items in relation to each
other.

3
non-earnings3 items used by analysts, referring to pronouncements by financial analysts
published in sources, such as the Wall Street Journal, Barron’s (1984–1990) and Value Line
publications, and major security firms’ (e.g., brokers’) commentaries.

Using one earnings and eight of the 12 non-earnings fundamental signals identified by Lev
and Thiagarajan (1993), Abarbanell and Bushee (1997) studied analysts’ efficiency in using
these fundamental signals in a US context. They showed that analysts fail to impound all the
information about future earnings included in these signals and, generally, analysts
underreact to some fundamental signals when making their forecast revisions. Swanson et
al. (2003) replicated Abarbanell and Bushee’s research in a Mexican context and found
similar evidence.

More recently, Wahab et al. (2015) examined analysts’ and whisperers’4 efficiency in using
fundamental signals when forecasting change in one-quarter-ahead EPS in a US context,
again finding that not all information included in fundamental signals is incorporated when
forecasting future EPS. However, whisperers’ efficiency in using fundamental signals was
higher than that of analysts, and their forecasts included unique information incremental to
analysts’ forecasts. More granularly, analysts underreacted to change in earnings and in
gross margin, whilst they overreacted to change in selling and administrative expenses and
in cash flows from operations.

Outside the fundamental signals literature, inefficiency has been found in analysts’ use of
other accounting and economic information. For instance, De Bondt and Thaler (1990)
found that analysts overreact to past earnings changes, resulting in overly optimistic
forecasts. Abarbanell (1991) reported that analysts’ forecasts underreact to information in
prior share price changes. Abarbanell and Bernard (1992) documented that analysts
underestimate the serial correlation in quarterly earnings, but to a lesser extent than
investors do through share prices. Easterwood and Nutt (1999) found that analysts
underreact to negative information and overreact to positive information, both creating
overly optimistic reactions and Bradshaw et al. (2001) showed that analysts underreact to
predictable earnings patterns following extreme accruals.

3
These included inventory, accounts receivable, capital expenditure, R&D expenditure, gross margin, sales
and administrative expenses, provision for doubtful receivables, effective tax rate, order backlog, labour
force, last-in-first-out (LIFO) earnings and audit qualification.
4
Whisperers are ‘an alternative anonymous source of EPS forecasts’ (Wahab et al. 2015, p. 2).

4
However, it is important to note that in reviews of the analyst literature, Ramnath et al.
(2008) and Bradshaw (2011) commented on the potential for analysts’ incentives to provide
an alternative explanation for their apparent inefficient processing of information. Since
these incentives are likely to be present both pre- and post-IFRS, examining IFRS impact on
apparently inefficient information processing by analysts is likely to shed light on the impact
of IFRS. A change in accounting associated with higher quality of financial reporting
theoretically (intuitively) should increase analysts’ efficiency in using fundamental
information.

Analysts’ earnings forecasts provide useful and relevant information to investors who make
investment decisions. The literature indicates that fundamental values based on consensus
forecasts are highly correlated with contemporaneous stock prices (Ali et al. 2003). If
analysts are efficient in using the information in fundamental signals, their forecasts may be
more accurate and therefore value relevant. Researchers have given limited attention to
investigating analysts’ efficiency in using fundamental signals post-2000 and outside the US.
The international accounting environment has undergone significant changes in the 21st
century and so analysts (especially those outside the US) may be unaware of their efficiency
in using these fundamental signals.

The limited literature available reveals that analysts do not fully incorporate the information
included in fundamental signals when making their earnings forecasts and are therefore
inefficient. As such, the following hypotheses have been developed:

H1a: Analysts are inefficient in choosing the fundamental signals that are useful in
predicting change in one-year-ahead earnings per share from a given suite of signals.

H1b: Analysts are inefficient in using an earnings fundamental signal for forecasting one-
year-ahead change in earnings per share.

H1c: Analysts are inefficient in using non-earnings fundamental signals for forecasting
one-year-ahead change in earnings per share.

Several researchers have found that IFRS adoption had a significant impact on many
financial statement fundamentals in terms of recognition, measurement, classification and
presentation (e.g., Blanchette et al. 2011, Goodwin et al. 2008, Stent et al. 2010). IFRS

5
impact on individual fundamental signals varies and may affect financial reporting quality
positively or negatively. However, in this study, we focus on the overall impact of IFRS
adoption on analysts’ efficiency in using one earnings and several non-earnings signals for
their earnings forecasts.

Many studies have investigated the impact of IFRS adoption on analysts’ information
environment and their forecast accuracy (e.g., Chalmers et al. 2012, Cheong et al. 2010,
Cheong and Al Masum 2010, Cotter et al. 2012, Preiato et al. 2013, Charitou et al. 2018),
finding a positive relationship between the two. For example, Chalmers et al. (2012) found
that adoption of IFRS involves more precise recognition of intangible assets than
predecessor standards, which conveys useful information to financial analysts in making
their earnings forecasts. Cheong et al. (2010) documented that intangibles capitalised under
IFRS aided analysts in forecasting future earnings, and Charitou et al. (2018) reported that
analysts’ recommendation revisions are more informative after the mandatory European
IFRS adoption. However, audit-fee-based evidence (Kim et al. 2012; Goncharov et al. 2014)
of increased post-IFRS complexity may indicate greater difficulty in efficient processing of
information. Examining analysts’ efficiency in using fundamental accounting signals in pre-
and post-IFRS periods can shed light on this potential trade-off. We developed the following
hypotheses based on an expected positive impact from IFRS adoption on analysts’ forecast
accuracy.

H2a: Analysts are more efficient after adoption of IFRS in choosing the given fundamental
signals that are useful for predicting one-year-ahead change in earnings per share
from the given suite of signals.

H2b: Analysts’ efficiency in using the earnings signal for forecasting one-year-ahead
change in earnings per share improved after adoption of IFRS.

H2c: Analysts’ efficiency in using non-earnings signals for forecasting one-year-ahead

change in earnings per share improved after adoption of IFRS.

3.0 Research method

We used one earnings and 12 non-earnings signals to examine analysts’ efficiency. Eight of
these non-earnings signals were selected from Lev and Thiagarajan (1993); namely,

6
inventories (INV), accounts receivable (AR), capital expenditure (CAPX), selling and
administrative expenses (SA), effective tax rate (ETR), gross margin (GM), labour force (LF)
and audit qualification (AQ). The other four non-earnings signals, financial leverage (LEV),
cash flows from operations (CF), goodwill (GW) and discretionary accruals (DACCR), were
selected from the literature that reports them to be useful to analysts (Dempsey et al.
1997). They were constructed in a manner consistent with Lev and Thiagarajan (1993). By
construction, we expect all non-earnings signals to have a negative relationship with one-
year-ahead change in EPS (CEPS1) and forecasted change in one-year-ahead change in EPS
(FCEPS1).

Table 1 explains the construction of these non-earnings independent variables, but first we
use inventory as an illustration. We measured inventory (INV) as annual percentage change
in inventory less annual percentage change in sales.

INVt=

and
=

where average (Avg) is calculated as the average of the relevant numbers for the last two
years for all variables.

It is important to note that with two exceptions, we calculated non-earnings measures as

percentage changes. The exceptions are an indicator variable for audit qualification and
discretionary accruals (absolute value difference between average for prior two years and
current year).

Table 1 about here

Abarbanell and Bushee (1997) examined analysts’ efficiency in using given fundamental
signals when making forecast revisions, and Wahab et al. (2015) examined analysts’
efficiency when forecasting one-quarter-ahead change in EPS. In this study, we investigated
analysts’ efficiency in relation to one-year-ahead change in EPS and the following regression
models were used, with the outcomes compared.

(1)

7
 (2)

where

CEPS1 = One-year-ahead change in earnings per share (EPS) ( 1 = [ +1 −

]÷ −1)
FCEPS1 = Mean analysts’ forecasts of one-year-ahead EPS made one month after
earnings announcement date in year t – actual EPS in year t, deflated by
stock price at end of t-1
CHGEPS = Change in current EPS (change in EPS between year t-1 and t, deflated by
stock price at end of t-1)
Signals = Percentage change (Lev and Thiagarajan 1993) in inventories (INV), accounts
receivable (AR), capital expenditure (CAPX), selling and administrative
expenses (SA), effective tax rate (ETR), gross margin (GM), labour force (LF),
financial leverage (LEV), cash flows from operations (CF), goodwill (GW) and
audit qualification (AQ) (‘clean’ or not, where ‘not clean’ = 1), and
discretionary accruals (DACCR) (absolute difference between average for
prior two years and current year)
Yr = Dichotomous indicator for 2001–2012, where individual year = 1 (2005, year
of IFRS adoption, excluded)
In = Dichotomous industry indicator (basic materials, consumer goods, consumer
services, health care, industrials, oil and gas, technology,
telecommunications and utilities), where industry = 1
Cty = Dichotomous country indicator (Spain, Finland, Czech Republic, Belgium,
Portugal, Poland, Italy, France, Denmark, Sweden [code law] and UK and
Australia [common law]), where individual country = 1
Codelaw = Dichotomous legal origin indicator, where code law countries = 1
IFRS = Dichotomous IFRS indicator where post-IFRS (2006–2012) = 1 and pre-IFRS
(2001–2004) = 0
Ɛ = Error term

Model 1 provides information about the fundamental signals useful in predicting one-year-
ahead change in EPS, while Model 2 provides information about the signals used by analysts

8
to forecast change in one-year-ahead EPS.5 Therefore, if the results reveal that certain
fundamental signals are significantly associated with FCEPS1 (Model 2) in the same way as
with CEPS1 (Model 1), then it can be inferred that analysts are efficient in selecting such
fundamental signals appropriately for their earnings forecasts. Hence, it can be concluded
that analysts are aware of the fundamental signals that are useful for forecasting one-year-
ahead change in EPS.

In order to examine the efficiency with which analysts use fundamental signals, analysts’
forecasts for one-year-ahead change in earnings per share (FCEPS1) and all fundamental
signals were regressed with one-year-ahead change in earnings per share (CEPS1). This
resulted in Model 3, where all variables are as defined and measured previously.

Codelaw+ (3)

If analysts are fully efficient in using these fundamental signals, FCEPS1 should embed all
information about CEPS1 included in the fundamental signals. If this is the case, there will be
no incremental R-squared contribution from including CHGEPS and the non-earnings signals,
and nor will there be a significant association between CEPS1 and the fundamental signals in
estimating Model 3.

If any fundamental signal(s) remain(s) significant when estimating Model 3 and an

incremental R-squared contribution over FCEPS1 occurs, then analysts’ inefficiency in using
the fundamental signal(s) can be inferred. When measuring analysts’ efficiency, we always
compared results from the estimation of Model 3 with those from Model 1, that is, Model 1
became the benchmark for analysts’ efficiency. Based on this benchmark, we measured a
percentage for analysts’ inefficiency of use of the earnings signal and non-earnings signals as
a whole. We calculated this analysts’ inefficiency in using the earnings signal as the
incremental R-squared from including CHGEPS over FCEPS1 in Model 3,6 divided by the R-
squared value from estimating earnings alone in Model 1. In the next step, we calculated

5
If a fundamental signal is significantly associated with one-year-ahead change in EPS (CEPS1), that signal is
useful in predicting CEPS1. If a fundamental signal is significantly associated with FCEPS1, analysts use that
signal to predict one-year-ahead change in EPS.
6
First FCEPS1 is regressed with CEPS1, then CHGEPS is included and regressed again with CEPS1, then the
incremental R-Squared from including CHGEPS is calculated.

9
analysts’ inefficiency in using non-earnings signals as the incremental R-squared from
including non-earnings signals over FCEPS1 and CHGEPS in Model 3, divided by the
incremental R-squared value from including non-earnings signals over earnings in Model 1.

In order to isolate the impact of inefficient use of individual fundamental signals, the
fundamental signals were next regressed with analysts’ forecast error (FE), resulting in
Model 4.

(4)

where FE = and all other variables are as defined previously.

Abarbanell and Bushee (1997, Footnote 18) have explained that if a fundamental signal is
significantly associated with forecast error in the same direction as its association with
CEPS1 (Model 1), ‘analysts fail to adjust their forecasts sufficiently high when a signal
conveys good news and sufficiently low when a signal conveys bad news’ (Abarbanell and
Bushee 1997, p. 17), that is, analysts underreact to that fundamental signal. When the
coefficient for the signal is of the opposite sign, analysts overreact to that signal.

In order to examine the impact of IFRS on analysts’ efficiency in selecting and using the
fundamental signals for making one-year-ahead change in EPS forecasts, the sample was
partitioned into pre-IFRS (2001–2004) and post-IFRS (2006–2012) periods and Models 1, 2, 3
and 4 were estimated separately for both sub-samples. To assess IFRS impact on analysts’
efficiency in selecting appropriate fundamental signals for forecasting one-year-ahead
change in EPS, significant variables from estimation of Models 1 and 2 were compared for
pre- and post-IFRS periods. We examined IFRS impact on analysts’ efficiency in using
fundamental signals by estimating Model 3 for both pre- and post-IFRS periods and
comparing these with Model 1 output for both periods. To isolate analysts’ inefficient use of
fundamental signals, Model 4 was estimated for both pre- and post-IFRS periods and
compared with outputs from Models 1 and 2.

The selection of sample countries (10 code and two common law countries) was based on
GAAP difference (Bae et. al. 2008) and the level of investor protection. All the selected code
law countries have a higher GAAP difference compared with the common law countries,

10
while all the selected countries have a high, but similar, level of investor protection (WEF
2012). Data was collected for 2001 to 2012 for all listed companies excluding financial
companies for which all the required data was available from Thomson Reuters Eikon
database or Bloomberg.

The initial sample size was 8,217 firm-year observations. We removed observations that
produced studentised residuals greater than three or where the Cook’s distance statistic
was greater than one to address multivariate outliers (Abarbanell and Bushee 1997; Lev and
Thiagarajan 1993), eliminating 246 firm-year observations. We also removed observations
for the IFRS transition year (2005), eliminating 605 firm-year observations. The final pooled
sample was 7,366 firm-year observations with analyst following (1,780 [24 per cent] pre-
IFRS and 5,586 [76 per cent] post-IFRS).

4.0 Empirical analysis

Ou (1990) reported that observations involving fundamentals include many outliers.
Following Barth et al.’s (2008) and Barth et al.’s (2012) method, all continuous variables
were winsorised at 2 per cent. Descriptive statistics presented in Table 2 are after
winsorisation. Regression results were White (1980) adjusted.

TABLE 2 about here

As Table 2 documents, all variables have non-zero means. It is notable that standard
deviations for some variables are high as their measures represent percentage change from
one year to the next. When calculating descriptive statistics for forecast error, absolute
forecast error was used. As can be seen from Table 2, FCEPS1, CEPS1, INV, LF, AQ, CAPX, CF,
GW and DACCR are significantly different between pre- and post-IFRS periods. We report
Pearson’s correlations in Table 3. None of the correlations is of a level that creates
multicollinearity concerns, with no correlation being higher than 0.433.

TABLE 3 about here

4.1 Analysts’ efficiency in choosing fundamental signals

Table 4 (Panel A) reports the results from estimating CEPS1 (Model 1) and FCEPS1 (Model 2)
using Abarbanell and Bushee’s (1997) replicated variables separately and all fundamental

11
signals. Results for both analyses reveal that the earnings signal (CHGEPS) is significant and
positively associated with both CEPS1 (Model 1) and FCEPS1 (Model 2), that is, analysts
appear to be aware of the information content in CHGEPS with regard to future earnings
and seem to understand the reversing nature of current year accruals and mean reversion in
annual earnings changes (Abarbanell and Bushee 1997, Swanson et al. 2003). As such,
analysts are efficient in choosing the earnings signal for their forecasts of change in one-
year-ahead EPS.

With respect to the non-earnings signals, the eight fundamental signals adopted from
Abarbanell and Bushee (1997) (for both the full sample in Panel A and the sub-sample
excluding the Global Financial Crisis (GFC) of 2007–2009 in Panel B) are significant in
estimating CEPS1 (Model 1). LF, ETR, GM, and CAPX are associated also with estimating
FCEPS1 (Model 2) in the same direction and, therefore, analysts are efficient in selecting
these signals. SA and AQ are significantly associated with estimating FCEPS1 (Model 2) but
not with estimating CEPS1 (Model 1), so analysts are not efficient in selecting these two
signals.

TABLE 4 about here

As seen from Table 4, the four added non-earnings fundamental signals (LEV, CF, GW and
CDACCR) are significant in estimating CEPS1 (Model 1). However, of these four, only three
(LEV,7 CF and GW) are significantly associated with estimating FCEPS1 (Model 2) in the same
direction. This result indicates that analysts are efficient in selecting LEV, CF and GW, but
inefficient in selecting CDACCR when forecasting one-year-ahead change in EPS. Overall,
eight of the nine fundamental signals significant in estimating CEPS1 are associated also
with estimation of FCEPS1 and in the same direction (CHGEPS, LF, ETR, GM, CAPX, LEV, CF
and GW), leading to the overall qualitative conclusion that analysts are efficient in their
choice of signals, therefore rejecting H1a.

As seen from Table 4 (Panels A and B), R-squared values for estimation of FCEPS1 (Model 2)
are 0.251 (Panel A), and 0.224 (Panel B) respectively. Moreover, the incremental R-squared
of 1.3 per cent from including the non-earnings signals over earnings for Model 2 is
significant using a partial F-test. This indicates that analysts use the information content in

7
LEV is significant for the sample that excludes the GFC period.

12
these non-earnings signals when making their earnings forecasts (Abarbanell and Bushee
1997).

4.2 Analysts’ efficiency in using fundamental signals

Comparing the results from estimating CEPS1 (Model 1) and CEPS1 after including FCEPS1 as
an independent variable (Model 3) for the full sample and the sub-sample excluding the
GFC, as reported in Table 5 Panels A and B respectively, the inclusion of FCEPS1 in
estimating CEPS1 (Model 3) does not change the significance of many fundamental signals.
Therefore, analysts seem to be inefficient in using the fundamental signals. Comparison of
the significance of coefficients between outputs from estimation of CEPS1 in Model 1 and
Model 3 reveals considerable decrease in the significance level for ETR and GW, while the
significance level for LF decreases from 5 to 10 per cent for Model 3 compared to Model 1.
However, GM’s significance level increases for Model 3 compared to Model 1. This could be
why GM is negatively significant when estimating CEPS1 (Model 1) whilst positively
significant when estimating FCEPS1 (Model 2). It would seem that analysts do not
understand fully the signal from GM when predicting FCEPS1. Apart from these variations,
the levels of significance for other variables (CHGEPS, LEV, CF and CDACCR) do not change
after inclusion of FCEPS1 in estimating Model 3.

TABLE 5 about here

The incremental R-squared contribution from fundamental signals in estimating CEPS1 after
including FCEPS1 (Model 3) reported in Table 5 for the full sample (Panel A) and that
excluding the GFC (Panel B) is 0.046 (0.040) for the earnings signal and 0.017 (0.017) for the
non-earnings signals over earnings. Partial F-tests indicate that the incremental R-squared
contributions from these signals are significant. These results point to the conclusion that
analysts do not fully incorporate the information content relating to one-year-ahead change
in EPS included in fundamental signals when making their earnings forecasts. Therefore,
analysts appear to be not fully efficient in using these fundamental signals. The inefficiency
percentage8 in comparison to estimating CEPS1 (benchmark Model 1) in using the earnings

8
When calculating analysts’ inefficiency, the benchmark model is Model 1. Earnings alone R-squared for
Model 1 is 0.213 and the incremental R-squared from the earnings signal for Model 3 (representing
inefficiency in using the earnings signal) is 0.046 (0.3100.264). Therefore, the percentage inefficiency in
using the earnings signal is 0.046/0.213 = 22 per cent and percentage utilisation of the signal is 78 per cent

13
signal is 22 per cent for the full sample (base of Panel A) and 23 per cent excluding the GFC
(base of Panel B), and 65 per cent for the non-earnings signals for both sub-samples,
supporting H1b and H1c.

In order to isolate the impact of this inefficiency, the fundamental signals were then
regressed with analysts’ forecast error (FE) (Model 4) and compared with Model 1 (refer
Table 6). If a particular signal is significantly associated with the dependent variable for
Models 1 and 4 in the same (opposite) direction, then it is inferred that analysts underreact
(overreact) to that signal (Abarbanell and Bushee 1997).

As Table 6 depicts, CHGEPS is positively significant with one-year-ahead change in earnings

per share (CEPS1) and negatively associated with forecast error (FE) for both samples. This
indicates that analysts overreact to CHGEPS when making their earnings forecasts. As such,
analysts appear to be inefficient in using earnings signals when forecasting one-year-ahead
change in EPS, supporting H1b.

TABLE 6 about here

Of the non-earnings fundamental signals, the results reported in Panels A and B of Table 6
indicate that analysts overreact to CAPX and GW when making their forecasts, since in both
panels these signals are significantly associated with the estimation of CEPS1 and FE
(Models 1, 2 and 4 respectively). However, the sign changes for the association when
estimating FE (Model 4) compared with CEPS1. In both Panels, SA is not significant with
CEPS1 (Model 1) but is significantly associated with FE (Model 4) with different signs. As
such, analysts fully overreact to this signal (Wahab et al., 2015). Similarly, the estimations
reported in Table 6 further reveal that analysts underreact to GM, CF and CDACCR. In
addition, the results based on the sample that excludes the GFC period reported in Table 6
(Panel B) indicate that analysts use LF, ETR and LEV efficiently when making their forecasts
of change in one-year-ahead EPS (CEPS1, Model1).

Hence, we conclude that analysts overreact to fundamental signals CHGEPS, SA, CAPX and
GW, whilst they underreact to GM, CF and DACCR. Abarbanell and Bushee (1997) have also

(1 - 22 per cent). In the same way, the incremental R-squared contribution from the non-earnings signals
over earnings is 0.017 for Model 1 and 0.011 for Model 3 and the percentage inefficiency is 0.011/0.017 =
65 per cent, with the percentage utilisation 35 per cent.

14
reported that analysts underreact to some fundamental signals, such as INV, GM and LF,
when making their forecast revisions. The findings do not support analyst underreaction to
LF; rather LF appears to be used efficiently. The INV variable is not useful in estimating
CEPS1. However, some of this study’s findings support those of Wahab et al. (2015), which
had documented analysts’ overreaction to SA and underreaction to GM.

The results indicate that for most of the non-earnings signals useful in predicting one-year-
ahead EPS, analysts either overreact or underreact when forecasting and are therefore
inefficient in using these signals, supporting H1c.

As a robustness test, we conducted the same analyses after controlling for analyst following,
and the results support the above conclusion fully. In addition, we conducted the same
analyses for the GFC only period on the basis that forecasts are more likely to be carefully
determined during this period. The results show that CHGEPS, AO, CF, GW and CDACR are
useful for predicting CEPS1 and analysts select CHGEPS, AO, CF and GW efficiently. Of these
signals, CHGEPS and CF are not efficiently used. Analysts overreact to CHGEPS and
underreact to CF, as is the case for the other periods.

4.3 IFRS and analysts’ efficiency

In order to assess IFRS impact on analysts’ efficiency in choosing and using the fundamental
signals, Models 1, 2, 3, and 4 were estimated for pre- and post-IFRS periods and then the
results were compared. Further, analysts’ percentage efficiency is equal to the complement
of inefficiency (i.e., 1- inefficiency) is calculated for pre-and post-IFRS and compared.

As can be seen from Table 7, of the five fundamental signals (CHGEPS, CAPX, LEV, CF and
GW) significantly associated with the estimation of CEPS1 (Model 1), only three (CHGEPS,
CAPX and GW) are significantly associated with the estimation of FCEPS1 (Model 2) pre-IFRS
(Panel A). However, when post-IFRS (Panel B) is considered, seven of the ten variables
(CHGEPS, SA, ETR, AQ, CAPX, CF and GW) significant in predicting CEPS1 (Model 1) are also
significantly associated with FCEPS1 (Model 2). Therefore, we conclude that analysts are
more efficient in choosing fundamental signals useful for making their forecasts post-
compared to pre-IFRS period, supporting H2a. This result is consistent with findings by
Chalmers et al. (2012), Cheong et al. (2010), Cotter et al. (2012) and Preiato et al. (2013)
that analyst forecast accuracy increased following adoption of IFRS. This outcome could be

15
due to a quality increase in the information environment post-IFRS (Horton et al. 2013) or
increased quantity and quality of financial reporting disclosures, increased comparability,
greater transparency or improved readability of financial statements (e.g., Barth et al. 2008,
Cheung and Lau 2016, Daske and Gebhardt 2006, Daske et al. 2008, Glaum et al. 2013).

Regarding IFRS impact on analysts’ efficiency in using the signals, Table 7 shows a significant
incremental R-squared contribution from including fundamental signals over FCEPS1 in
estimating CEPS1 (Model 3), both pre-IFRS (base of Panel A, Model 3, 0.027 [from earnings]
+ 0.015 [from non-earnings] = 0.042) and post-IFRS (base of Panel B, Model 3, 0.054 + 0.009
= 0.063). This indicates inefficiency in using these fundamental signals both pre- and post-
IFRS. The percentage inefficiency in use of fundamental signals by analysts pre-IFRS is 24 per
cent and post-IFRS is 27 percent9. Therefore, the efficiency in utilising fundamental signals
(that is 1 – inefficiency percentage) for forecasting one-year-ahead change in EPS is similar
at 76 per cent (1 – 24%) pre-IFRS and 73 per cent (1− 27%) post-IFRS. This points to the
conclusion that overall, analysts’ efficient utilisation of these fundamental signals is similar
in pre- and post-IFRS periods.

TABLE 7 about here

In extended analysis, we examined analysts’ efficiency in using earnings and non-earnings

signals. Table 7 (base of Panels A and B, Model 3) indicates that analysts’ efficient utilisation
of the earnings signal decreases from 82 per cent (complement of 18 per cent) pre-IFRS to
76 per cent (complement of 24 per cent) post-IFRS. When comparing coefficients of CHGEPS
from estimation of CEPS1 (Model 1) between pre- (0.366) and post-IFRS (0.425) periods, the
predictive ability of earnings increased following IFRS adoption. However, analysts fail to
utilise this increased predictive ability post-IFRS when making their earnings forecasts,
thereby rejecting H2b.

In addition, analysts seem to be better in using non-earnings signals post- compared to pre-
IFRS, given the increase in percentage of efficient utilisation of non-earnings signals over
earnings from 29 per cent (complement of 71 per cent) pre-IFRS to 36 per cent

9
Inefficiency in utilising the fundamental signals pre-IFRS is 24 per cent and is calculated as incremental R-
squared from including fundamental signals in Model 3 of 0.042 (made up of 0.027 + 0.015) pre-IFRS
divided by R-squared for Model 1 (0.173). Similarly, inefficiency post-IFRS of 27 per cent is calculated as
incremental R-squared from including fundamental signals in Model 3 post-IFRS of 0.063 (made up of
0.054 +0.009) divided by 0.236, which is the R-Square of Model 1 estimated post-IFRS.

16
(complement of 64 per cent) post-IFRS (refer Table 7, Panels A and B, final row of
calculations in Model 3). This could correspond to a decrease in analysts’ efficiency in using
the earnings signal post-IFRS. Analysts use more information content from the non-earnings
signals when forecasting one-year-ahead change in EPS in the post- compared to pre-IFRS
period. Overall, these results support H2c.

Continuing with Table 7, to isolate the impact of analysts’ inefficient use of fundamental
signals pre- and post-IFRS, we compared results from estimating CEPS1 and FE (Models 1
and 4) pre- and post-IFRS. The results show that analysts overreact to earnings signals in
both periods. In terms of non-earnings signals, analysts underreact to SA and CF pre-IFRS;
however, they seem to use CAPX and GW efficiently pre-IFRS. In terms of LEV, it is
significantly associated with CEPS1 but not with FCEPS1 or FE pre-IFRS. This indicates that
analysts do not use LEV but use other variables that capture the effects of LEV (Wahab et al.
2015) pre-IFRS. As shown in Table 7, although analysts’ efficiency in selecting and using
non-earnings fundamental signals is higher post- compared to pre-IFRS, they appear not to
fully utilise the information content embedded in these fundamental signals. Accordingly,
analysts underreact to fundamental signals GM, CF and CDACCR but overreact to AQ, CAPX
and GW. However, analysts are fully efficient in using SA and ETR post-IFRS.

Prior studies have reported that goodwill is more value relevant (Aharony et al. 2010) and
more representative of firms’ underlying investment opportunities post-IFRS (Godfrey and
Koh 2009). Further, Cheong et al. (2010) reported that intangible asset value under IFRS is
more useful than under prior GAAP for analysts in making their earnings forecasts. However,
the current study’s findings show that although analysts use GW for their earnings forecasts,
they overreact to GW to a certain extent post-IFRS.

In robustness tests, excluding both the GFC (2007–2009) and the full IFRS transition (2004–
2006), observations result in consistent findings, except CDACCR is no longer more
significant post-IFRS for any model. Further, IFRS impact on analysts’ efficiency was
examined after controlling for analyst following, with the results fully supporting the original
conclusion.

4.4 Legal origin and analysts’ efficiency

17
As a further robustness test, we examined analysts’ efficiency in using appropriate
fundamental signals separately for code and common law countries. The analysis reveals a
lower efficiency among analysts in using the earnings signal in common compared with code
law countries. Analysts overreact to the earnings signal in both cases, with a higher
overreaction occurring in common law countries. For the non-earnings signals, analysts are
efficient in selecting appropriate signals for code law countries, but it is less so for common
law countries. However, comparatively, analysts’ efficiency in using non-earnings signals is
low in code compared to common law countries as analysts underreact and overreact to
more non-earnings fundamental signals in code law countries and few signals are useful in
common law countries. As such, analysts are comparatively efficient in using fewer signals.
This result is not consistent with that of Barniv et al. (2005), who documented that analysts’
forecast accuracy for common compared to code law countries is higher due to corporate
governance that is more effective, stronger legal protection laws and higher quality financial
reporting.

5.0 Conclusion

Financial analysts are important users of financial statement information and provide useful
information for market participants. The efficiency with which analysts use this information
affects the accuracy of their forecasts and hence information asymmetry. However, the
efficiency with which analysts use financial information has been rarely researched. We
investigated analysts’ efficiency in using one earnings and 12 non-earnings fundamental
signals when forecasting one-year-ahead change in EPS in Australian and European contexts
(and for code law and common law country sub samples), using data from 2001 to 2012. We
also examined IFRS impact on analysts’ efficiency, thereby making a valuable contribution to
the literature.

Overall, the pooled analyses point to the conclusion that analysts are aware of most of the
selected fundamental signals useful in forecasting change in one-year-ahead EPS
(Abarbanell and Bushee, 1997), and they use this information when making their earnings
forecasts. However, analysts do not fully incorporate the information content about future
earnings embedded in fundamental signals when forecasting EPS and are therefore

18
inefficient. Analysts seem to overreact to fundamental signals CHGEPS, SA, CAPX and GW,
whilst they underreact to GM, CF and DACCR. They seem to use LF, ETR and LEV efficiently.

The analysis of IFRS impact also indicates that analysts are comparatively efficient in
selecting fundamental signals useful in predicting one-year-ahead change in EPS after
adoption of IFRS. However, overall, analysts’ ability to incorporate future earnings
information included in the fundamental signals is similar pre- and post-IFRS. Moreover,
analysts are generally inefficient in using fundamental signals both pre- and post-IFRS.
Further, their efficiency in using the earnings signal decreased, while increasing for non-
earnings signals over earnings, post- compared to pre-IFRS. However, analysts continue to
underreact and overreact to certain non-earnings signals post-IFRS. Furthermore, analysts
substantially underutilise the earnings signal in common compared to code law countries.

Limitations include investigation of IFRS overall impact rather than that for individual IFRS.
Similarly, our focus has been on all analysts rather than individual analysts with their
differing abilities. Additionally, firms at different lifecycle stages may present different
forecasting challenges. Nevertheless, these findings are likely to be of interest to financial
analysts and other market participants, as well as to standard setters and regulators.

This study attempts to minimise selected country GAAP versus IFRS difference, but future
research could use predominantly debt versus equity country partitioning rather than legal
origin. Future research could also use real rather than accruals-based earnings management
and include other fundamental signals or other country comparisons.

19
References

Abarbanell, J. S., 1991. Do analysts' earnings forecasts incorporate information in prior stock price
changes?. Journal of Accounting and Economics, 14 (2), 147 - 165.
Abarbanell, J. S., Bushee, B. J., 1997. Fundamental Analysis, Future Earnings, and Stock Prices.
Journal of Accounting Research, 35 (1), 1-24.
Abarbanell, J. S., Bernad, V. L., 1992. Tests of Analysts' Overreaction/Underreaction to Earnings
Information as an Explanation for Anomalous Stock Price Behavior. The Journal of Finance,
47(3), 1181-1207
Aharony, J., Barniv, R., Falk, H., 2010. The Impact of Mandatory IFRS Adoption on Equity Valuation of
Accounting Numbers for Security Investors in the EU. European Accounting Review, 19(3),
535-578.
AL-Debie, M., Walker, M., 1999. Fundamental Information Analysis: An Extension and UK Evidence.
The British Accounting Review, 31(3), 261-280.
Alford, A., Jones, J., Leftwich, R. Zmijewski, M., 1993. The Relative Informativeness of Accounting
Disclosures in Different Countries. Journal of Accounting Research, 31, 183-223.
Ali, A., Lee-Seok, H., Trombley, M. A., 2003. Residual-Income-Based Valuation Predicts Future Stock
Returns: Evidence on Mispricing vs. Risk Explanations. Accounting Review, 78(2), 377-397.
Bae, K-H., Tan, H. & Welker, M., 2008. International GAAP Differences: The Impact on Foreign
Analysts. The Accounting Review, 83(3), 593-628.
Barniv, R., Myring, M. J., Thomas, W. B., 2005. The Association between the Legal and Financial
Reporting Environments and Forecast Performance of Individual Analysts. Contemporary
Accounting Research, 22(4), 727-758.
Barth, M. E., Landsman, W. R., Lang, M., Williams, C., 2012. Are IFRS-Based and US GAAP-Based
Accounting Amounts Comparable? Journal of Accounting and Economics, 54(1), 68-93.
Barth, M. E., Landsman, W. R., Lang, M. H., 2008. International Accounting Standards and Accounting
Quality. Journal of Accounting Research, 46(3), 467-498.
Beuselinck, C., Joos, P., Khurana, I., Meulen, S. V., 2010. Mandatory Adoption of IFRS and Analysts'
Forecasts Information Properties, Tilburg University Working Paper.
Blanchette, M., Racicot, F. É., Girard, J. Y., 2011. The Effects of IFRS on Financial Ratios: Early
Evidence in Canada. Certified General Accountants Association of Canada.
Bradshaw, M. T., (2011). Analysts’ Forecasts: What Do We Know after Decades of Work?
doi:10.2139/ssrn.1880339
Bradshaw, M. T., Richardson, S. A., Sloan, R. G., 2001. Do Analysts and Auditors Use Information in
Accruals. Journal of Accounting Research, 39(1), 45-74.
Bradshaw, M. T., Huang, A. G., Tan, H., 2014. Analyst Target Price Optimism Around the World.
Midwest Finance Association 2013 Annual Meeting, Chicago.
Chalmers, K., Clinch, G., Godfrey, J. M., 2011. Changes in Value Relevance of Accounting Information
Upon IFRS Adoption: Evidence from Australia. Australian Journal of Management, 36(2),
151-173.
Chalmers, K., Clinch, G., Godfrey, J. M., Wei, Z., 2012. Intangible Assets, IFRS and Analysts’ Earnings
Forecasts. Accounting & Finance, 52(3), 691-721.
Charitou, A., Karamanou, I., Kopita, A., 2018. Are Analyst Stock Recommendation Revisions More
Informative in the Post-IFRS Period? Journal of Business Finance & Accounting, 45(1-2), 115-
139.
Cheong, C. S., Al Masum, M., 2010. Financial Analysts’ Forecast Accuracy: Before and After the
Introduction of AIFRS. Australasian Accounting Business and Finance Journal, 4(3), 65-81.
Cheong, C. S., Kim, S., Zurbruegg, R., 2010. The Impact of IFRS on Financial Analysts’ Forecast
Accuracy in the Asia-Pacific Region: The Case of Australia, Hong Kong and New Zealand.
Pacific Accounting Review, 22(2), 124-146.

20
Cheung, E., Lau, J., 2016. Readability of Notes to the Financial Statements and the Adoption of IFRS.
Australian Accounting Review, 26(2), 162-176.
Christensen, H. B., Lee, E., Walker, M., Zeng, C., 2015. Incentives or Standards: What Determines
Accounting Quality Changes Around IFRS Adoption? European Accounting Review, 24(1), 31-
61.
Cotter, J., Tarca, A., Wee, M., 2012. IFRS Adoption and Analysts’ Earnings Forecasts: Australian
Evidence. Accounting & Finance, 52(2), 395-419.
Daske, H., Gebhardt, G., 2006. International Financial Reporting Standards and Experts’ Perceptions
of Disclosure Quality. Abacus, 42(3-4), 461- 498.
Daske, H., Hail, L., Leuz, C., Verdi, R., 2008. Mandatory IFRS Reporting Around the World: Early
Evidence on the Economic Consequences. Journal of Accounting Research, 46(5), 1085-1142.
De Bondt, W. F., Thaler, R. H., 1990. Do Security Analysts Overreact? The American Economic Review,
80(2), 52-57.
De George, E. T., LI, X., Shivakumar, L., 2016. A Review of the IFRS Adoption Literature. Review of
Accounting Studies, 21, 898-1004.
Dempsey, S. J., Gatti, J. F., Grinnell, D. J., Cats-baril, W. L., 1997. The Use of Strategic Performance
Variables as Leading Indicators in Financial Analyst’s Forecasts. Journal of Financial
Statement Analysis, 2, 61-79.
Ding, Y., Hope, O. K., Jeanjean, T., Stolowy, H., 2007. Differences between Domestic Accounting
Standards and IAS: Measurement, Determinants and Implications. Journal of Accounting and
Public Policy, 26(1), 1-38
Easterwood, J. C., Nutt, S. R., 1999. Inefficiency in Analyst’s Earnings Forecasts: Systematic
Misreaction or Systematic Optimism? Journal of Finance, 54(5), 1777-1797.
El-Gazzar, S. M., Finn, P. M., Tang, C., 2009. The Value Relevance of Earnings and Non-earnings
Information in Regulated and Deregulated Markets: The Case of the Airline Industry.
International Advances in Economic Research, 15(1), 88-101.
Frankel, R., Lee, C. M. C., 1998. Accounting Valuation, Market Expectation, and Cross-Sectional Stock
Returns. Journal of Accounting & Economics, 25(3), 283-319.
Fudenberg, D., Tirole, J., 1995. A Theory of Income and Dividend Smoothing Based on Incumbency
Rents. Journal of Polical Economics, 103(1), 75-93
Glaum, M., Baetge, J., Grothe, A., Oberdörster, T., 2013. Introduction of International Accounting
Standards, Disclosure Quality and Accuracy of Analysts’ Earnings Forecasts. European
Accounting Review, 22(1), 79-116.
Godfrey, J. M., Koh, P. S., 2009. Goodwill Impairment as a Reflection of Investment Opportunities.
Accounting & Finance, 49(1), 117-140.
Goncharov, I., Riedl, E., Sellhorn, T., 2014. Fair Value and Audit Fees, Review of Accounting Studies,
19(1), 210-241.
Goodwin, J., Ahmed, K., Heaney, R. 2008. The Effects of International Financial Reporting Standards
on the Accounts and Accounting Quality of Australian Firms: A Retrospective Study. Journal
of Contemporary Accounting & Economics, 4(2), 89-119.
Guay, W.R., Kothari, S., Watts, R.L., 1996. A Market-Based Evaluation of Discretionary Accrual
Models. Journal of Accounting Research, 34, 83-105.
Horton, J., Serafeim, G., Serafeim, I., 2013. Does Mandatory IFRS Adoption Improve the Information
Environment? Contemporary Accounting Research, 30(1), 388-423.
Houqe, M. N., Van Zijl, T., Dunstan, K., Karim, A., 2012. The Effect of IFRS Adoption and Investor
Protection on Earnings Quality Around the World. The International Journal of Accounting,
47(3), 333-355.
Kim, J., Liu, X., Zheng, L., 2012. The Impact of Mandatory IFRS Adoption on Audit Fees: Theory and
Evidence. Accounting Review, 87(6), 2061-2094.
Kothari, S.P., Leone, A.J., Wasley, C.E., 2005. Performance Matched Discretionary Accrual Measures.
Journal of Accounting and Economics, 39(1), 163-197.

21
Lev, B., Thiagarajan, S. R., 1993. Fundamental Information Analysis. Journal of Accounting Research,
31(2), 190-215.
Ohlson, J. A., Penman, S. H., 1992. Disaggregated Accounting Data as Explanatory Variables for
Returns. Journal of Accounting, Auditing & Finance, 7(4), 553-573.
Ou, J. A., 1990. The Information Content of Nonearnings Accounting Numbers as Earnings Predictors.
Journal of Accounting Research 28(1), 144-163.
Ou, J. A., Penman, S. H., 1989. Financial Statement Analysis and the Prediction of Stock Returns.
Journal of Accounting & Economics, 11, 295-329.
Pathiranage, N. P. W., 2017. The Impact of IFRS on the Predictive Ability and Value Relevance of
Fundamental Signals and Analysts’ Efficiency in Using These Signals. PhD Thesis. Swinburne
University of Technology.
Preiato, J. P., Brown, P. R., Tarca, A., 2013. Mandatory Adoption of IFRS and Analysts’ Forecasts: How
Much Does Enforcement Matter? https://ssrn.com/abstract=1499625.
Ramnath, S., Rock, S., Shane, P., 2008. The Financial Analyst Forecasting Literature: A Taxonomy with
Suggestions for Further Research. International Journal of Forecasting, 24(1), 34 -75.
Salon, R.G., 2006. Do Stock Prices Fully Reflect Information in Accruals and Cash Flows about Future
Earnings?. The Accounting Review, 71(3), 289-315.
Stent, W., Bradbury, M., Hooks, J. 2010. IFRS in New Zealand: Effects on Financial Statements and
Ratios. Pacific Accounting Review, 22(2), 92-107.
Sun, J., 2011. The Effect of Analyst Coverage on the Informativeness of Income Smoothing
International Journal of Accounting, 46(3), 333-349.
Swanson, E. P., Rees, L., Juarez-Valdes, L. F., 2003. The Contribution of Fundamental Analysis after a
Currency Devaluation. The Accounting Review, 78(3), 875-902.
Wahab, S., Teitel, K., Morzuch, B., 2015. How Analysts and Whisperers Use Fundamental Accounting
Signals to Make Quarterly EPS Forecasts. Journal of Accounting, Auditing & Finance, 32(3),
401-422.
White, H. 1980. A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for
Heteroskedasticity. Econometrica, 48(4), 817–838.
World Economic Forum (WEF), 2012. The Global Competitiveness Report. World Economic
Forum. http://www.weforum.org/reports.

22
 Table 1: Measurement of non-earnings fundamental signals

Variable Measurement Description

Inventory (INV) Δ% Inventory – Δ% sales, Increased inventory compared to increased sales
Δ (for all variables) represents, is considered a negative signal regarding future
(Inventory t – inventory avg )÷ inventory avg earnings.
and Inventory avg = (Inventoryt-1 + Inventoryt-2) ÷ 2
Accounts Δ% Accounts Receivable – Δ% sales* Disproportionate increase in accounts receivable
Receivable (AR) relative to sales is considered a negative signal
regarding future earnings.
Selling & Δ% Selling & Administrative Expenses – Δ% sales* Disproportionate increase in selling and
Administrative administration expenses a negative signal
Expenses (SA) regarding future earnings.
Labour Force (LF) Measures efficiency of labour; positive value
indicates decreased labour force as measured by
contribution to revenue.
Effective Tax Rate Unusual decrease in effective tax rate generally
(ETR) where *100 considered negative signal for earnings
persistence.
EBT = Pre-tax Income + Amortisation/
Impairment of Intangibles
Gross Margin (GM) Δ% Sales – Δ% Gross Margin* Disproportionate decrease in gross margin
(relative to sales) negative signal.
Audit Qualification Coded 1 for qualified opinion, 0 otherwise, emphasis Qualified audit opinion provides negative signal.
(AQ) of matter paragraph relating to going concern, 1 is
assigned, 0 otherwise

Capital Δ% Industry Capital Expenditure – Δ% Firm Capital If firm’s change in capital expenditure is less
Expenditure expenditure* than that for industry, negative signal regarding
(CAPX) firm’s future growth and future earnings
performance.
Financial Leverage Increase in leverage from t-1 to t is negative
(LEV) signal.

Increase in CFO in year t relative to t-1 expected

Cash Flows (CF) to have a positive impact on earnings in year
t+1. Sloan (1996) reports that a higher
proportion of earnings attributable to CFO
signifies a higher quality of income that will
more likely persist.
Goodwill (GW) GW provides useful information about future
earnings potential and has a positive
relationship with future earnings (Cheong et al.,
2010), so capture earnings peristence.
Discretionary CDACCR = The theoretical model of Fudenberg and Tirole
Accruals (CDACCR) DACCRt is measured using Kothari et al. (2005) (1995) suggests that managers consider
performance matched discretionary accruals expected future earnings when making
discretionary accounting choices and there is a
positive relationship between current year
discretionary accruals and stock returns (Guay et
al. 1996)

*See calculation of inventory for treatment of percentage change. ** All non-earnings fundamental
signals measured as annual percentage change except AQ and CDACCR.

23
 Table 2: Descriptive statistics
Variables Pooled sample N = 7,366 Pre-IFRS N = 1,780 Post-IFRS N = 5,586 t-tests of
(24 per cent) (76 per cent) difference
Minimum Maximum Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. t
FCEPS1 -0.098 0.463 0.034 0.098 0.039 0.094 0.032 0.099 2.544**
ABSFE 0.000 0.428 0.062 0.082 0.063 0.083 0.061 0.078 -0.844
CEPS1 -0.280 0.349 0.009 0.097 0.025 0.098 0.004 0.095 7.771***
CHGEPS -0.361 0.313 -0.001 0.105 -0.002 0.103 -0.001 0.105 -0.444
INV -46.514 81.861 2.648 22.265 0.686 23.373 3.273 21.865 -4.130***
AR -54.062 54.808 -0.818 20.107 -0.272 18.921 -0.992 20.469 1.370
SA -44.144 48.541 -0.106 13.917 -0.195 15.145 -0.078 13.503 -0.290
LF -66.169 35.210 -6.034 18.670 -8.651 19.673 -5.200 18.262 -6.556***
ETR -99.498 82.628 -2.031 26.613 -1.646 25.045 -2.154 27.095 0.729
GM -101.425 66.367 -3.035 29.117 -3.925 29.623 -2.751 28.951 -1.465
CAPX -256.590 102.179 -11.484 69.134 -14.233 68.849 -10.608 69.208 -1.932*
LEV -0.837 1.540 0.037 0.440 0.042 0.407 0.036 0.450 0.521
CF -19.048 14.530 -1.071 6.599 -1.754 7.183 -0.853 6.387 -4.734***
GW -258.950 64.227 -27.198 66.452 -32.032 76.212 -25.658 62.955 -3.198***
CDACCR -0.218 0.198 -0.005 0.079 -0.019 0.089 0.000 0.075 -8.219***
Chi-Sq
AQ 0.000 1.000 0.006 0.012 0.00376 15.385***
Refer to Table 1 for variable definitions. IFRS = indicator coded 1 for observations post-IFRS (20062012), 0 otherwise (2001-2004); FCEPS1 = (mean analysts’ forecast for one-year-ahead EPS
made one month after earnings announcement date in year t – actual EPS in year t divided by price at the end of t-1; FE = (actual EPS in year t+1 – forecast EPS for year t+1) divided by price at
end of t-1; ABSFE = absolute forecast error; CEPS1 = one-year-ahead change in earnings per share ( 1=[ +1 − ] ÷ −1); CHGEPS = change in current earnings per share (change in
EPS between year t-1 and t deflated by the stock price at the end of t-1). ***, ** and * indicate statistical significance at 1 per cent, 5per cent and 10per cent respectively.

24
 Table 3: Pearson’s correlations for independent variables (N = 7366)
CHGEPS INV AR SA LF ETR GM AQ CAPX LEV CF GW
INV .003
AR .073*** .044***
SA .098*** .017 .063***
*** *** *** ***
LF .194 .088 .243 .131
*** **
ETR -.084 -.001 -.002 -.018 -.024
*** *** *** **
GM .162 -.018 .040 -.018 .080 -.028
AQ .014 -.003 .005 .004 -.016 -.009 -.014
** *** ** *** **
CAPX .026 .032 .018 .027 .040 -.021 .023 .004
*** ** *** *** ***
LEV -.064 .006 -.003 -.026 -.039 .012 -.372 .004 -.061
*** *** *** *** *** ***
CF .167 -.006 .065 .082 .169 -.006 .070 .012 .061 -.012
GW .025** .037*** -.013 -.013 .001 -.006 .028** -.006 .167*** -.105*** .077***
CDACCR -.184*** -.025** -.018 -.103*** -.036*** .041*** -.138*** .010 .020 .073*** .433*** .036***
Refer to Table 1 for variable definitions. *** Correlation significant at 1 per cent. ** Correlation significant at 5 per cent.

25
 Table 4: Analysts’ efficiency in choosing useful fundamental signals (H1a)
Sample Size Panel A: Full Sample N = 7,366 Panel B: Sub-sample that excludes GFC (2007-2009) N = 5,176
Models Abarbanell and Bushee Model Full Model Abarbanell and Bushee Model Full Model
Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Model 1 Model 2
Dependent Variable CEPS1 FCEPS1 CEPS1 FCEPS1 CEPS1 FCEPS1 CEPS1 FCEPS1
R2 of full Model (a) 0.221 0.248 0.230 0.251 0.221 0.248 0.188 0.224
Coef. Coef. Coef. Coef. Coef. Coef. Coef. Coef.
α 0.02290** 0.04942*** 0.02269** 0.05111*** 0.02290** 0.04942*** 0.0019 0.01826*
CHGEPS 0.38997*** 0.43108*** 0.40945*** 0.43483*** 0.38997*** 0.43108*** 0.37109*** 0.39958***
INV 0.00001 -0.00001 0.00001 -0.00001 0.00001 -0.00001 -0.00002 -0.00001
AR -0.00007 -0.00005 -0.00006 -0.00005 -0.00007 -0.00005 -0.00006 -0.00004
SA 0.00007 0.00027*** 0.00016 0.00030*** 0.00007 0.00027*** 0.00014 0.00039***
LF -0.00022*** -0.00014** -0.00015** -0.00010 -0.00022*** -0.00014** -0.00021** -0.00020**
ETR -0.01290** -0.01406** -0.01284** -0.01404** -0.01290** -0.01406** -0.01411** -0.02098***
GM -0.00069*** 0.00025* -0.00034** 0.00036* -0.00069*** 0.00025* -0.00049** 0.00068***
AQ -0.01298 -0.02247** -0.01213 -0.02149** -0.01298 -0.02247** -0.00816 -0.01600*
CAPX 0.00007*** 0.00011*** 0.00007*** 0.00010*** 0.00007*** 0.00011*** 0.00008*** 0.00011***
LEV 0.00799*** 0.00365 0.01125*** 0.00698*
CF -0.00153*** -0.00049** -0.00122*** -0.00035
GW 0.00529*** 0.00976*** 0.00341** 0.00927***
CDACCR -0.06279*** -0.00242 -0.04742** 0.01430
CHGEPS alone Model 0.213 0.238 0.213 0.238 0.171 0.206 0.171 0.206
R-Sq (b)
Incremental R-Sq 0.008 0.010 0.017 0.013 0.050 0.042 0.017 0.018
(a-b)
Controls Years, countries, industries, code law, IFRS
Refer to Table 1 for variable definitions. GFC = Global Financial Crisis. ***, ** and * indicate statistical significance at 1 per cent, 5 per cent and 10 per cent
respectively. Results are based on robust standard errors clustered by firm. All standard errors are less than 0.05. The highest VIF for any model is for CDACCR (1.71). All
year dummies are significant and positive for the full sample in full Model 1. IFRS is significant full sample Model 2.

26
 Table 5: Analysts’ efficiency in using fundamental signals (H1b & H1c)
Sample Size Panel A Full Sample Panel B Excluding GFC Period
(N = 7,366) (N = 5,176)
Models Model 1 Model 3 Model 1 Model 3
Dependent variable CEPS1 CEPS1 CEPS1 CEPS1
R-Sq of full model (a) 0.230 0.321 0.188 0.279
Coef. Coef. Coef. Coef.
α 0.02269** 0.00450 0.00190 -0.00458
CHGEPS 0.40945*** 0.25948*** 0.37109*** 0.23595***
INV 0.00001 0.00001 -0.00002 -0.00001
AR -0.00006 -0.00004 -0.00006 -0.00005
SA 0.00016 0.00005 0.00014 0.00001
LF -0.00015** -0.00012* -0.00021** -0.00014*
ETR -0.01284** -0.00805 -0.01411** -0.00706
GM -0.00034** -0.00047** -0.00049** -0.00072***
AQ -0.01213 -0.00473 -0.00816 -0.00276
CAPX 0.00007*** 0.00004*** 0.00008*** 0.00005***
LEV 0.00799*** 0.00661** 0.01125*** 0.00879***
CF -0.00153*** -0.00136*** -0.00122*** -0.00110***
GW 0.00529*** 0.00184 0.00341** 0.00018
CDACCR -0.06279*** -0.06191*** -0.04742** -0.05225**
FCEPS1 0.34492*** 0.33799***
CHGEPS/ FCEPS1 CHGEPS FCEPS1 CHGEPS FCEPS1
alone model R-Sq (b) 0.213 0.264 0.171 0.228
FCEPS1 and CHGEPS in 0.310 0.268
the Model R-Sq
Incremental R-Sq from 0.213 0.046 22%# 0.171 0.040 23%
the earnings signal
Incremental R-Sq from 0.017 0.011 65% 0.017 0.011 65%
non-earnings signals
Controls Years, countries, industries, code law, IFRS
Refer to Table 1 for variable definitions. GFC = Global Financial Crisis. ***, ** and * indicate statistical
significance at 1 per cent, 5per cent and 10 per cent respectively. Results based on robust standard errors
clustered by firm. All standard errors are less than 0.05.
#
The percentage adjacent to incremental R-squared from the earnings signal represents analysts’ inefficiency in
using the earnings signal calculated as 0.046/0.213 = 22% and 0.040/0.171 = 23%. The percentage adjacent to the
incremental R-squared from the non-earnings signal represents analysts’ inefficiency in using non-earnings signals
calculated as 0.011/0.017 = 65%. All year dummies are significant for the full sample in full Model 3. IFRS is not
significant.

27
 Table 6: Impact on analysts’ efficient use of fundamental signals (H1b & H1c)
Sample Size Panel A: Full Sample Panel B: Excluding GFC Period
N = 7,366 N = 5,176
Models Model 1 Model 4 Model 1 Model 4
Dependent CEPS1 FE CEPS1 FE
variable
R-Sq of full 0.230 0.055 0.188 0.067
model (a)
Coef. Coef. Coef. Coef.
α 0.02269** -0.03171* 0.00190 -0.01689
CHGEPS 0.40945*** -0.08202*** 0.37109*** -0.09043***
INV 0.00001 0.00001 -0.00002 0.00001
AR -0.00006 -0.00001 -0.00006 -0.00001
SA 0.00016* -0.00013 0.00014 -0.00025**
LF -0.00015** 0.00001 -0.00021** 0.00002
ETR -0.01284** -0.00381 -0.01411** 0.00051
GM -0.00034** -0.00075*** -0.00049** -0.00128***
AQ -0.01213 0.01493 -0.00816 0.01432
CAPX 0.00007*** -0.00004** 0.00008*** -0.00004*
LEV 0.00799*** 0.00250 0.01125*** 0.00039
CF -0.00153*** -0.00086*** -0.00122*** -0.00061**
GW 0.00529*** -0.00429** 0.00341** -0.00568***
CDACCR -0.06279*** -0.05084** -0.04742* -0.04865**
Controls Years, countries, industries, code law, IFRS
Refer to Table 1 for variable definitions. GFC = Global Financial Crisis. ***, ** and * indicate statistical
significance at 1 per cent, 5 per cent and 10 per cent respectively. Results are based on robust standard errors
clustered by firm. All standard errors are less than 0.05. All year dummies and IFRS are significant for full sample
in full Model 4.

28
 Table 7: IFRS impact on analysts’ efficiency in using fundamental signals (H2a, H2b & H2c)
Sample Panel A Pre-IFRS (n = 1,780) Panel B Post-IFRS (n = 5,586)
Models Model 1 Model 2 Model 3 Model 4 Model 1 Model 2 Model 3 Model 4
Dependent variable CEPS1 FCEPS1 CEPS1 FE CEPS1 FCEPS1 CEPS1 FE
Full model R-Sq 0.173 0.250 0.254 0.038 0.236 0.251 0.333 0.055
Coef. Coef. Coef. Coef. Coef. Coef. Coef. Coef.
α -0.00529 0.07470** -0.03107 -0.06129 0.00643 0.02238 -0.00135 -0.02396
CHGEPS 0.36608*** 0.42632*** 0.21897*** -0.07673** 0.42523*** 0.43676** 0.27344*** -0.08070***
INV -0.00002 0.00010 -0.00002 -0.00002 0.00001 -0.00001 0.00001 0.00002
AR -0.00012 -0.00010 -0.00008 0.00001 -0.00005 -0.00005 -0.00003 -0.00001
SA -0.00001 0.00030** -0.00011 -0.00031* 0.00025** 0.00031*** 0.00014 -0.00003
LF -0.00012 -0.00020 -0.00005 0.00015 -0.00017** -0.00004 -0.00015** -0.00009
ETR -0.01669 -0.02131 -0.00934 -0.00003 -0.01227* -0.01194* -0.00812 -0.00518
GM -0.00035 0.00074 -0.00061 -0.00142** -0.00034* 0.00033 -0.00046** -0.00065***
AQ -0.00607 0.01002 -0.00953 -0.01723 -0.02078** -0.04145*** -0.00637 0.02861***
CAPX 0.00012*** 0.00011*** 0.00008** 0.00001 0.00006*** 0.00010*** 0.00003 -0.00005**
LEV 0.02145*** 0.00637 0.01925** 0.00937 0.00428 0.00373 0.00299 -0.00016
CF -0.00134*** 0.00002 -0.00135*** -0.00087** -0.00159*** -0.00067*** -0.00136*** -0.00087***
GW 0.00767*** 0.00893*** 0.00459 -0.00133 0.00406** 0.00931*** 0.00032 -0.00548**
CDACCR -0.04647 0.01439 -0.05143 -0.03397 -0.06841*** -0.00838 -0.06550*** -0.05429**
FCEPS1 0.34507*** 0.34753***
CHGEPS/ FCEPS1 alone CHGEPS CHGEPS FCEPS1 CHGEPS CHGEPS FCEPS1
model R-Sq (b) 0.152 0.235 0.212 0.222 0.239 0.270
FCEPS1 and CHGEPS in 0.239 0.324
the model R-Sq
#
Incremental R-Sq from 0.152 0.027 18% 0.222 0.054 24%
the earnings signal
Incremental R-Sq from 0.021 0.015 71% 0.014 0.009 64%
non-earnings signals
Controls Years, countries, industries, code law
Refer to Table 1 for variable definitions. ***, ** and * indicate statistical significance at 1 per cent, 5 per cent and 10 per cent respectively. Results are based on robust
#
standard errors clustered by firm. All standard errors are less than 0.05. Pre-IFRS 2001-2004 and post-IFRS 2006-2012. The percentage adjacent to the incremental R-squared
from the earnings signal represents analysts’ inefficiency in using the earnings signal calculated as 0.027/0.152 = 18% and 0.054/0.222 = 24%. The percentage adjacent to the
incremental R-squared from the non-earnings signal represents analysts’ inefficiency in using non-earnings signals calculated as 0.015/0.021 = 71% and 0.009/0.014 = 64%.

29