He 2019

Journal of Accounting and Economics xxx (xxxx) xxx
Contents lists available at ScienceDirect
Journal of Accounting and Economics

journal homepage: www.journals.elsevier.com/
journal-of-accounting-and-economics
Earnings acceleration and stock returns*

Shuoyuan He a, b, Ganapathi (Gans) Narayanamoorthy a, *
a
Tulane University, USA
b
San Francisco State University, USA
a r t i c l e i n f o a b s t r a c t
Article history: We document that earnings acceleration, defined as the quarter-over-quarter change in
Received 1 October 2018 earnings growth, has significant explanatory power for future excess returns. These excess
Received in revised form 4 April 2019 returns are robust to a wide range of previously documented anomalies and a battery of
Accepted 20 June 2019
risk controls. The future return predictability appears to be consistent with investors
Available online xxx
assuming a seasonal random walk model for quarterly earnings and missing predictable
implications of earnings acceleration for future earnings growth. Finally, the excess returns
JEL classification:
from the basic earnings acceleration strategy can be enhanced further by focusing on profit
G14
M41
firms, low earnings volatility firms and on specific patterns of earnings acceleration.
© 2019 Elsevier B.V. All rights reserved.
Keywords:
Earnings acceleration
Trading strategy
Active investing
Mispricing
Anomaly
1. Introduction
Earnings acceleration, or the change in earnings growth, has anecdotally been discussed as a viable trading strategy in the
popular press.1 Despite anecdotal references to its use in the investing world, and in contrast to the vast number of studies on
the capital market implications of earnings growth (Kothari, 2001; Lakonishok et al., 1994; Dechow and Sloan, 1997; Bernard
and Thomas, 1989; Akbas et al., 2017), earnings acceleration has received limited research attention. In this study, we examine
the implications of earnings acceleration for future stock returns.
We measure earnings acceleration as the change in earnings growth from one quarter to the next, where earnings growth
is the scaled change in earnings over the corresponding quarter a year ago. Using a sample of 377,907 observations spanning
*
We thank S.P. Kothari for multiple discussions and several insightful comments on an earlier draft of this paper. We thank Salman Arif, Ray Ball, Patrick
Beaudan, Gus DeFranco, Ted Fee, Oleg Gredil, Wayne Guay (editor), Haibo Jiang, David Lesmond, Danny Meidan, Nick Pan, John Page, Bohan Song, Theodore
Sougiannis, Barrett Wheeler, Ruizhong Zhang, Morad Zekhnini, two anonymous referees and seminar participants at Tulane University, Temple University,
San Francisco State University, American Accounting Association Annual Meeting, and the National Association of Active Investment Managers conference
for their comments.
* Corresponding author.
E-mail addresses: [email protected] (S. He), [email protected] (G. Narayanamoorthy).
1
For example, American Century Investments appears to actively focus on companies with accelerating earnings (see http://americancenturyblog.com/
wp-content/uploads/Chart-of-the-Week-Mar.-4-2012-Earnings-Acceleration.pdf). Investor's Business Daily, a historically influential newspaper targeted at
investors, promotes a stock picking formula that incorporates “earnings acceleration” (see O'Neil, 1999). Several additional examples are cited in Cao et al.
(2011).
https://doi.org/10.1016/j.jacceco.2019.101238
0165-4101/© 2019 Elsevier B.V. All rights reserved.
Please cite this article as: He, S., Narayanamoorthy, G., Earnings acceleration and stock returns, Journal of Accounting and
Economics, https://doi.org/10.1016/j.jacceco.2019.101238
2 S. He, G.(G. Narayanamoorthy / Journal of Accounting and Economics xxx (xxxx) xxx
8,824 different firms and 176 fiscal quarters from 1972 to 2015, we find that earnings acceleration is a significant predictor of
future stock returns.2 A trading strategy that involves going long in the top decile of quarterly earnings acceleration and short
in the bottom decile of earnings acceleration produces large market-adjusted returns, both in one-month and quarter-long
trading windows that start two days after an earnings announcement. We find market-adjusted returns of 1.8% (3.4%)
over the month-long (quarter-long) window, which translates to annualized returns in excess of 23% (14%). The significant
excess returns persist even when low priced stocks (less than $5) and/or low capitalization stocks (up to $0.5 billion) are
excluded from the trading strategy. While the primary trading strategy involves buying/selling stocks two days after the
earnings announcement, we still obtain significant excess returns when a conservative trading strategy involving calendar
month rebalancing is adopted.
We conduct a battery of tests to rule out two potential explanations for the excess returns: (a) the returns are a mani-
festation of an already known anomaly, and (b) the analysis omits a risk factor. First, to rule out the known anomaly
explanation, we demonstrate that the excess returns are robust to the inclusion of several known anomalies, namely post-
earnings announcement drift (PEAD, Bernard and Thomas, 1990), profit trend anomaly (Akbas et al., 2017), combination of
known mispricing factors (Stambaugh and Yuan, 2017), gross profit anomaly (Novy-Marx, 2013), accrual anomaly (Sloan,
1996), past earnings volatility (Cao and Narayanamoorthy, 2012), return momentum (Jegadeesh and Titman, 1993), total
asset growth anomaly (Cooper et al., 2008), as well as the size and book-to-market anomalies. Second, to preclude a missing
risk factor explanation, we document the robustness of our results to the use of Fama-French three-factor and Carhart four-
factor adjusted returns. More recently, Fama and French (2015) have proposed a five-factor model that augments the previous
three-factor model with two additional factors e investment and profitability. The results are robust to the use of Fama-
French five-factor adjusted returns as well. Additionally, the trading strategy spanning 176 quarters rarely produces losses
e for a risk explanation, the frequency of losses would be significantly higher.
After ruling out the known anomaly and risk explanations, we then explore the nature of the price relevant information in
earnings acceleration that is apparently missed by the market leading to the abnormal returns. We specifically examine
whether earnings acceleration has implications for subsequent earnings growth and whether these implications are missed
by the market. Recall that, in the PEAD context, the measure of earnings growth (which is basically earnings change from a
seasonal random walk model e labeled Standardized Unexpected Earnings or SUE) has been shown to predict future earnings
growth, especially one quarter ahead. We test and find that earnings acceleration has implications for earnings growth,
especially two and three quarters ahead. Such prediction is incremental to the predictions of past earnings growth, which
have been documented before in the context of the PEAD. We also find significant short-window three-day abnormal returns
surrounding earnings announcements of these two future quarters. Significant short-window announcement returns make a
risk-based explanation for the anomaly unlikely (Brown and Warner, 1980, 1985; Rangan and Sloan, 1998).
Since earnings acceleration is associated with both future returns and future earnings growth, we structure a formal
efficiency test to directly associate the future returns from earnings acceleration with the implications of earnings acceler-
ation for future earnings growth. Along the lines of Sloan (1996), we employ the Mishkin test and document that the
abnormal returns are associated with the relation between earnings acceleration and future earnings growth.3
Having shown that the excess returns are consistent with the market missing the implications of earnings acceleration for
future earnings growth, we then explore a possible information processing bias underlying the market missing these im-
plications.4 It is well known that most quarterly earnings announcements are accompanied by a comparison to earnings from
the corresponding fiscal quarter a year ago. The abnormal returns documented in this study are consistent with investors
anchoring on earnings change from four quarters ago (i.e., assuming earnings follows seasonal random walk) and ignoring
intermediate earnings outcomes. If the true earnings process follows a seasonal random walk (SRW), there would be zero
correlation between earnings acceleration and future earnings growth. However, the empirically observed non-zero corre-
lations between earnings acceleration and future earnings growth (and future returns) occur when we assume that the true
earnings process is non-seasonal like, for example, IMA (1,1). Thus, our finding of the market missing the implications of
earnings acceleration for future earnings growth is consistent with the market adopting a SRW model (thus assuming zero
correlations between earnings acceleration and future growth) when the true process is IMA (1,1).5
Furthermore, we analytically show that the correlations between earnings acceleration and three of the next four quarters'
earnings growths monotonically decrease in the IMA (1,1) parameter, and the correlation for the other quarter is nearly
constant. The IMA (1,1) parameter represents the extent of mean reversion in the earnings process. Thus, if the market is
missing the above correlations, the excess returns will be larger with lower mean reversion. Consistent with this prediction,
we document that earnings acceleration effect is stronger among profit firms and among firms with low earnings volatility,
both of which have been shown to be associated with lower mean reversion (Brooks and Buckmaster, 1976; Dichev and Tang,
2
While majority of our tests focus on earnings acceleration, we show that general acceleration strategies involving changes in growth of sales and
profitability demonstrate similar abnormal returns.
3
Some recent studies employing the Mishkin test framework include Chen and Shane (2014), Hui et al. (2016) and Ma and Markov (2016).
4
Fama (1998) argues that documenting abnormal returns alone is not enough and recommends the presentation of a valid alternative model doc-
umenting information processing biases that can explain the observed results. A notable feature of our study is that we provide such an alternative.
5
While the signs of the correlations observed between earnings acceleration and future earnings growth can be generated by a number of non-SRW
processes, the IMA (1,1) assumption appears to predict the magnitudes of the correlations well and hence appears to be a reasonable approximation
for the true earnings process in this specific context.
S. He, G.(G. Narayanamoorthy / Journal of Accounting and Economics xxx (xxxx) xxx 3
2009). Additionally, going long on high earnings acceleration firms represented by consecutive positive earnings growth
quarters (which signifies low mean reversion) and going short on low earnings acceleration firms represented by positive
earnings growth followed by negative earnings growth can improve the anomalous returns by nearly 45% (from 1.8% to 2.6%
over a month).
Recently, there are heightened p-hacking concerns with studies documenting anomalies. Harvey et al. (2016) recommend
a higher hurdle (t-statistic greater than 3.0) for any new variable purporting to explain the cross-section of returns. Earnings
acceleration comfortably beats this hurdle. Green et al. (2017) document that post-2003 returns from several well-
documented anomalies are insignificantly different from zero. The earnings acceleration anomaly continues to perform
well even in the post-2003 period. Several previously documented anomalies do not remain significant when equal-weighted
portfolios are replaced by value-weighted portfolios (Hou et al., 2017). Our results remain robust to the construction of value-
weighted portfolios.
To our knowledge, this is the first study to examine the implications of earnings acceleration for future returns. Cao et al.
(2011) find that annual earnings acceleration is associated with long window contemporaneous returns.6 We make two
significant departures from Cao et al. (2011) in this study. First, we focus on future returns rather than contemporaneous
returns since we are interested in whether the market is efficient in incorporating the effects of earnings acceleration. Second,
given the anomaly context, we focus on quarterly earnings growth rather than annual earnings growth since much of the
information to the market in an annual earnings number has been pre-empted by three quarterly earnings numbers and likely
is assimilated in the stock price well before the annual earnings information is released (see Lee, 1992, and Landsman and
Maydew, 2002 for evidence on the speed with which accounting information is assimilated into stock prices).7 Conse-
quently, earnings acceleration in this study is measured as the change in earnings growth from one quarter to the next, while
earnings growth is the change in earnings over the corresponding quarter a year ago.8
In sum, we document that abnormal returns can be earned by employing a trading strategy that goes long in high past
earnings acceleration stocks and short in low past earnings acceleration stocks. The excess returns are robust to risk ad-
justments and other known anomalies, and are particularly relevant given the recent general disillusionment with several
accounting based anomalies not being robust (Linnainmaa and Roberts, 2017; Hou et al., 2017). Additionally, we show that the
anomaly is consistent with a specific information processing bias - anchoring on a SRW model for quarterly earnings. While
prior literature has documented future return implications of earnings levels (Novy-Marx, 2013), earnings changes (Bernard
and Thomas, 1990; Akbas et al., 2017) and earnings volatility (Cao and Narayanamoorthy, 2012), this study contributes to the
literature by documenting the future return implications of earnings acceleration or the second derivative of earnings.
While we consider the abnormal returns associated with earnings acceleration a new anomaly, we acknowledge that our
results can be interpreted alternatively as premium from an as yet unknown risk factor. Based on a simple dividend discount
model, Fama and French (2006) demonstrate a positive relation between expected return and future profitability. Since
earnings acceleration is related to future profitability, from a rational pricing perspective, earnings acceleration can alter-
natively be labeled a risk factor (Penman and Zhu, 2014). Regardless of the labeling, the main take-way from this study is that
earnings acceleration is associated with future excess returns, and the association cannot be explained by known risk factors.
The rest of the study is organized as follows. In section 2, we present the data and basic excess return results from the
acceleration-based trading strategy. We also discuss how the anomaly is distinct from previously documented anomalies and
is robust to risk adjustments. In section 3, we present results on the implications of earnings acceleration for future earnings
growth and short-window abnormal return tests around future earnings announcements. We also conduct formal statistical
tests of the association between abnormal returns and the earnings acceleration e future earnings growth relation. In section
4, we analytically examine the relation between earnings acceleration and future earnings growth assuming the true earnings
process follows IMA (1,1). In section 5, we consider alternative earnings acceleration patterns and discuss additional
robustness tests. In section 6, we provide our concluding remarks.
2. Earnings acceleration and future returns
In this section, we begin with discussing our sample selection and variable construction. Next, we document abnormal
returns from portfolio tests of the acceleration-based trading strategy. We also document the robustness of these returns to risk
as well as to other known anomalies. Finally, we augment our portfolio test results by conducting regression tests of the anomaly.
6
Others have used an earnings acceleration measure in their empirical tests (Aboody et al., 2004; Copeland et al., 2004; Chen and Zhang, 2007).
However, they include the forecasted change in earnings growth in their return models, while we include the current (realized) change in earnings growth.
Additionally, while earnings acceleration is simply a control variable in these prior studies, it is the focus of this study.
7
A trading strategy involving annual earnings acceleration as defined in Cao et al. (2011) does not yield excess returns.
8 et al. (1969) that documents 4th difference in earnings to be the relevant growth measure in
There is a vast amount of literature beginning with Latane
the context of quarterly earnings.
2.1. Data and descriptive statistics
2.1.1. Sample selection

Data used in this study is obtained from CRSP-Compustat Merged (quarterly), CRSP (daily), and I/B/E/S (detail) databases.
The sample selection procedure includes all quarterly earnings announcements from CRSP-Compustat Merged database
between 1972 and 2015. We delete observations if a firm has (i) more than one earnings announcement on any date, (ii)
earnings announcement date within 30 days of a previous earnings announcement date or (iii) earnings announcement date
either prior to or more than 180 days after the corresponding fiscal period-end, as these observations are potentially subject
to data errors. We restrict the sample to NYSE, AMEX and NASDAQ firms that have stock return data in CRSP. We exclude
financial and utility firms with SIC codes from 6000 to 6999 and from 4900 to 4949, as financial and utility firms were highly
regulated during much of our sample period, which can result in unusual earnings-return relationships. We require that every
observation has non-missing data to calculate market capitalization (SIZE), book-to-market ratio (BM), one-month market-
adjusted return (VMAR) and the various earnings acceleration measures. Our final sample consists of 8,824 firms and 377,907
firm-quarter observations from 1972 to 2015. Table 1 summarizes the sample selection procedure.
2.1.2. Construction of earnings acceleration and other variables

The measurements for all the variables used in this study are summarized in Table 2. The primary variable of interest is our
earnings acceleration measure, which we define as the earnings growth in quarter t minus the earnings growth in quarter t-1.
Earnings growth in quarter t is calculated as the deflated change in earnings per share (EPS) from quarter t-4 to quarter t (that
is, seasonally differenced EPS).9 We consider three alternative deflators for our earnings growth measure: the absolute value
of EPS in quarter t-4, the stock price at the end of quarter t-1, and the standard deviation of EPS calculated from the most
recent eight quarters (including quarter t). In other words, our earnings growth measures, EGA, EGP and EGV, are scaled
measures of EPSt e EPSt-4. In addition, we also consider sales growth and profitability growth as alternative growth mea-
sures10. We define sales growth as seasonally differenced sales per share (SPS), deflated by sales per share four quarters ago,
and profitability growth as the seasonally differenced return-on-assets (ROA). Thus, our first definition of earnings acceler-
ation (EAA) is calculated as:
EPSi;t EPSi;t4 EPSi;t1 EPSi;t5

EAAi;t ¼ EGAi;t EGAi;t1 ¼
EPSi;t4 EPSi;t5
Our second definition of earnings acceleration (EAP) is calculated as:

EAPi;t ¼ EGPi;t EGPi;t1 ¼
Stock Pricei;t1 Stock Pricei;t2
Table 1
Sample Selection.
This table reports the sample selection procedures. Data are firm-quarter observations from 1972 to 2015.
All Compustat-CRSP merged database firm-quarters between 1972 and 2015 1,027,392 100%
Drop observations with missing earnings announcement date (231,338) 23%
Drop observations with more than one earnings announcement on the same date for the same firm (2,215) 0%
Drop observations whose current quarter earnings announcement date (1,483) 0%
is before or more than 180 days after current quarter fiscal period end date
Drop if current earnings announcement is less than 30 days away (3,976) 0%
from the previous earnings announcement
Keep NYSE, AMEX or NASDAQ firms (177,462) 17%
Drop firms with SIC codes between 6000 and 6999 or between 4900 and 4949 (141,008) 14%
Keep observations with nonmissing CRSP daily price at the earnings announcement date (8,218) 1%
Keep domestic common stocks (38,998) 4%
Drop observations with missing SIZE or BM (4,036) 0%
Drop observations with all five acceleration measures (i.e., EAA, EAP, EAV, SA and PA) missing (40,224) 4%
Drop observations with missing VMAR (527) 0%
Total 377,907 37%
9
Following Cao et al. (2011) and Chan et al. (2003), we define earnings acceleration on a per share basis to account for the effects of mergers and
acquisitions, as well as to strip out any predictability due to changes in the scale of the firm's operations.
10
Chan et al. (2003) suggest using sales growth as an additional growth measure due to problems with negative base-period earnings. We include
profitability growth because it is similar to the measure used by Cao et al. (2011).
Table 2
Variable Definitions.
This table summarizes variable definitions.
Variables Descriptions
VMAR Value-weighted market-adjusted buy-and-hold return during the one-month window, defined as the raw return (two days through 30 days
after quarter t earnings announcement date) adjusted for the same period CRSP value-weighted index return
VMARQ Value-weighted market-adjusted buy-and-hold return during the quarter-long window, defined as the raw return (two days after quarter t
earnings announcement date through one day after quarter tþ1 earnings announcement date) adjusted for the same period CRSP value-
weighted index return
EMAR Equal-weighted market-adjusted buy-and-hold return during the one-month window following earnings announcement date
SAR Size-adjusted buy-and-hold return during the one-month window following earnings announcement date
FF3 Fama-French three-factor-adjusted buy-and-hold return during the one-month window following earnings announcement date
FFM Fama-French three-factor and momentum-adjusted buy-and-hold return during the one-month window following earnings announcement
date
FF5 Fama-French five-factor-adjusted buy-and-hold return during the one-month window following earnings announcement date
EGA Earnings growth (deflated by absolute value of earnings), defined as quarter t's earnings per share (EPS) minus quarter t-4's EPS, scaled by
the absolute value of quarter t-4's EPS; where EPS is calculated as income before extraordinary items (IBQ), divided by shares outstanding
(CSHOQ). Shares are adjusted for stock splits and stock dividends.
EGP Earnings growth (deflated by price), defined as quarter t's EPS minus quarter t-4's EPS, scaled by the stock price at the end of quarter t-1
EGV Earnings growth (deflated by standard deviation of earnings), defined as quarter t's EPS minus quarter t-4's EPS, scaled by the standard
deviation of EPS in the most recent 8 quarters (including quarter t)
SG Sales growth, defined as quarter t's sales per share (SPS) minus quarter t-4's SPS, scaled by quarter t-4's SPS; where quarter t's SPS is
calculated as net sales (SALEQ), divided by shares outstanding (CSHOQ). Shares are adjusted for stock splits and stock dividends.
PG Profitability growth, defined as quarter t's return-on-assets (ROA) minus quarter t-4's ROA; where ROA is defined as operating income after
depreciation (OIADPQ) per share at quarter t, divided by total assets (ATQ) per share at quarter t-1. Shares are adjusted for stock splits and
stock dividends.
EAA Earnings acceleration (absolute value of earnings deflated), defined as quarter t's EGA minus quarter t-1's EGA
EAP Earnings acceleration (price deflated), defined as quarter t's EGP minus quarter t-1's EGP
EAV Earnings acceleration (standard deviation of earnings deflated), defined as quarter t's EGV minus quarter t-1's EGV
SA Sales acceleration, defined as quarter t's SG minus quarter t-1's SG
PA Profitability acceleration, defined as quarter t's PG minus quarter t-1's PG
MSCORE Mispricing factor, from Yu Yuan's website (http://www.saif.sjtu.edu.cn/facultylist/yyuan/)
TREND Trend in quarterly gross profitability, measured as b1 from estimating the following trend regression each quarter:
GPQ ¼ a0 þ b1*t þ b2*D1 þ b3*D2 þ b4*D3 þ 3; where GPQ is calculated as sales (SALEQ) minus cost of goods sold (COGSQ), divided by total
assets (ATQ)
SIZE Market capitalization, defined as market price at earnings announcement date multiply by the total number of shares outstanding
BM Book-to-market ratio, defined as the book value of equity at the end of quarter t divided by the market capitalization at earnings
announcement date
PASTRET Past return, defined as the value-weighted market-adjusted stock return during the [-180,-2] window before quarter t earnings
announcement date
GP Gross profitability, defined as quarter t's SALEQ minus COGSQ, divided by ATQ
ACC Accruals, defined as quarter t's (DACTQ - DCHEQ - DLCTQ þ DDLCQ þ DTXPQ)/Average ATQ, where ACTQ, CHEQ, LCTQ, DLCQ, TXPQ represent
current assets, cash and short-term investments, current liabilities, debt in current liabilities and income tax payable, respectively
VOL Earnings volatility, defined as standard deviation of EPS in the most recent 8 quarters (including quarter t)
AG1 Asset growth (definition 1), defined as quarter t's total assets per share minus quarter t-1's total assets per share, divided by quarter t-1's
total assets per share
AG2 Asset growth (definition 2), defined as quarter t's total assets per share minus quarter t-4's total assets per share, divided by quarter t-4's
total assets per share
VMAR3 Value-weighted market-adjusted buy-and-hold return during the earnings announcement window, defined as the raw return (one day
before through one day after quarter t earnings announcement date) adjusted for the same period CRSP value-weighted index return
Our third definition of earnings acceleration (EAV) is calculated as:

EAVi;t ¼ EGVi;t EGVi;t1 ¼
SDEPSi;t SDEPSi;t1
Our definition of sales acceleration (SA) is calculated as:
SPSi;t SPSi;t4 SPSi;t1 SPSi;t5

SAi;t ¼ SGi;t SGi;t1 ¼
SPSi;t4 SPSi;t5
Our definition of profitability acceleration (PA) is calculated as:

PAi;t ¼ PGi;t PGi;t1 ¼ ROAi;t ROAi;t4 ROAi;t1 ROAi;t5
While our basic portfolio tests employ all five acceleration measures, we use EAP as our primary measure of focus in
subsequent tests for two reasons. First, earnings is the most widely accepted optimal measure of firm performance. Second,
the use of price as deflator is consistent with the broad capital markets literature in accounting.
To mitigate the impact of outliers, we follow prior research (see, for example, Rangan and Sloan, 1998; Livnat and
Mendenhall, 2006) and transform our earnings acceleration measures into decile ranks. The decile cutoffs are based on
the distribution of the previous fiscal quarter's earnings accelerations. The decile ranks are initially numbered 0 through 9. We
then convert the numbers to scaled ranks by dividing by 9 and subtracting 0.5. The resulting scaled ranks vary from 0.5 to
þ0.5 with a mean of zero and a range of one. The range of one implies that the coefficient on earnings acceleration in a return
regression represents the abnormal return from a zero investment strategy of going long on the highest earnings acceleration
decile and short on the lowest earnings acceleration decile. This choice of range facilitates a comparison of the economic
magnitudes of our main results to prior research.
The primary abnormal return measures in our study are calculated over two windows: (a) a window beginning two days
after quarter t's earnings announcement date and ending on day 30, and (b) a window beginning two days after quarter t's
earnings announcement date and ending one day after quarter tþ1's earnings announcement date. We use value-weighted
market-adjusted return as our measure for abnormal returns, and calculate the return as the difference between a firm's buy-
and-hold raw return and the same period CRSP value-weighted index return.11
2.1.3. Descriptive statistics

Table 3 presents descriptive statistics for the sample. In Panel A, the mean and median of all five earnings acceleration
measures are either negative or zero, indicating that the average firm's growth rate decreased over the time-period used in
our study. In Panel B, we report Spearman and Pearson correlations among variables. Both one-month and quarter-long
abnormal returns are positively correlated with all five earnings acceleration measures. The Spearman (Pearson) correla-
tion between one-month value-weighted market-adjusted return (VMAR) and EAA, EAP, EAV, SA and PA are 0.037 (0.013),
0.043 (0.018), 0.044 (0.039), 0.034 (0.022) and 0.045 (0.029), respectively; which indicates a possible positive association
between earnings acceleration and future returns. We also note that the earnings acceleration variables are negatively
correlated with the profitability trend (TREND) variable, suggesting that our earnings acceleration measures represent a
phenomenon different from the profitability trend phenomenon documented in Akbas et al. (2017).12
2.2. Basic empirical results
In Table 4, we present the results for both the one-month and quarter-long market-adjusted returns sorted into accel-
eration deciles for the five measures of acceleration. In Panel A, we report the results for equal-weighted average portfolio
returns. The month-long VMAR for the bottom decile on EAP is 0.2% while the VMAR for the topmost decile is 1.6%. This
represents a hedge portfolio return of 1.8% over one month, which in annualized terms is an excess return exceeding 23%.
Over the quarter-long window, the corresponding hedge return is 3.4%. The hedge portfolio returns for the other four
measures of acceleration over both return windows are comparable. Additionally, moving from the bottom decile to the top
decile, the stock returns are monotonically increasing, showing that the anomaly gradually increases in acceleration decile
and is not concentrated in a particular decile. In Fig. 1, we show the evolution of the cumulative abnormal return over the
month-long (day 2 to day 30) window for the top and bottom decile of earnings acceleration. Decile one has a small negative
return in the immediate aftermath of the earnings announcement and then remains at roughly that level for the entire month.
Decile ten, on the other hand, increases virtually monotonically to reach 1.6% at day 30. Recent research has documented that
several anomalies vanish when equal-weighted portfolios are replaced by value-weighted portfolios (Hou et al., 2017). In
Panel B, we report the results for value-weighted average portfolio returns, and we find that the returns remain robust to this
portfolio construction. The value-weighted portfolio for EAP, our primary acceleration variable, yields a hedge return of 1.5%
over a month compared to 1.8% for the equal-weighted portfolio. The table also presents results when other acceleration
variables are used. The other four acceleration measures (EAA, EAV, SA and PA) all yield consistently positive abnormal
returns.
In Table 5, we present the results for the robustness of the anomaly to alternative risk adjustments. Columns one through
five present returns for equal-weighted portfolios and columns six through ten provide the results for value-weighted
portfolios.13 Recall that our base results already adjust for value-weighted market index returns. In column one, we pre-
sent stock returns adjusted for equal-weighted market index returns (EMAR). The excess returns again show a monotonically
11
If a stock is delisted subsequent to portfolio formation, we compute the remaining return using the CRSP delisting return if it is available. Thereafter we
reinvest any remaining proceeds in the market portfolio until the end of the holding period.
12
Notwithstanding the negative correlation, our earnings acceleration measure is also conceptually different from the trend variable in Akbas et al. (2017).
In Akbas et al. (2017), the trend in profitability is defined as the trend coefficient in an earnings level regression, meaning that the trend variable is an
average of past quarter over quarter growth rates. In contrast, our earnings acceleration measure represents the change in the past quarter over quarter
growth rates.
13
While we present the results for one earnings acceleration measure (EAP) in the month-long return window, the results for this measure in the quarter-
long window are similarly significant. Additionally, the results for the other four acceleration measures (EAA, EAV, SA and PA) in both return windows are
similar to the results for EAP.
Table 3
Descriptive Statistics.
This table provides descriptive statistics for our sample. See Table 2 for variable definitions. Panel A reports the summary statistics. Panel B reports correlations among variables. Spearman (Pearson) correlations are
presented above (below) the diagonal. Correlations that are significant at the 1% significance level are marked in bold.
Panel A: Summary statistics
N Mean Std. Dev. Q1 Median Q3

VMAR 377,907 0.006 0.148 0.075 0.005 0.070
VMARQ 368,653 0.014 0.237 0.115 0.006 0.112
S. He, G.(G. Narayanamoorthy / Journal of Accounting and Economics xxx (xxxx) xxx
EAA 377,618 0.035 6.822 0.530 0.014 0.475
EAP 377,620 0.001 0.296 0.007 0.000 0.006
EAV 352,331 0.006 1.458 0.800 0.009 0.794
SA 373,821 0.008 0.290 0.082 0.002 0.074
PA 333,298 0.000 0.031 0.009 0.000 0.009
MSCORE 303,952 48.382 13.116 38.830 47.600 57.190
EGP 355,573 0.010 0.261 0.005 0.001 0.006
TREND 319,224 0.000 0.008 0.003 0.000 0.002
SIZE (millions) 377,907 2060.959 7917.914 49.826 205.669 915.994
BM 377,907 0.706 0.652 0.309 0.543 0.913
PASTRET 377,906 0.043 0.466 0.217 0.023 0.199
GP 360,298 0.087 0.128 0.051 0.089 0.135
ACC 288,806 0.003 0.046 0.015 0.003 0.023
VOL 358,806 0.374 1.088 0.050 0.118 0.286
AG1 358,858 0.019 0.108 0.021 0.011 0.046
AG2 359,861 0.099 0.292 0.030 0.064 0.174
Panel B: Pearson (below), Spearman (above) correlations
VMAR VMARQ EAA EAP EAV SA PA MSCORE EGP TREND SIZE BM PASTRET GP ACC VOL AG1 AG2
VMAR 1 0.597 0.037 0.043 0.044 0.034 0.045 ¡0.030 0.029 0.012 0.048 0.017 0.017 0.031 ¡0.029 ¡0.024 0.015 0.006
VMARQ 0.617 1 0.044 0.050 0.053 0.049 0.055 ¡0.059 0.058 0.035 0.004 0.046 0.039 0.047 ¡0.033 ¡0.039 0.020 ¡0.008
EAA 0.013 0.015 1 0.697 0.732 0.280 0.466 0.050 0.398 ¡0.032 0.001 ¡0.010 0.029 0.036 0.059 0.000 0.064 ¡0.035
EAP 0.018 0.018 0.161 1 0.858 0.287 0.537 0.046 0.436 ¡0.038 0.012 ¡0.009 0.019 0.038 0.057 ¡0.017 0.078 ¡0.020
EAV 0.039 0.045 0.337 0.2692 1 0.318 0.563 0.058 0.414 ¡0.050 ¡0.007 0.003 0.016 0.026 0.044 0.008 0.060 ¡0.040
SA 0.022 0.032 0.094 0.0511 0.2453 1 0.443 ¡0.036 0.171 ¡0.065 0.002 ¡0.008 0.060 0.037 0.056 ¡0.009 0.141 0.046
PA 0.029 0.040 0.188 0.179 0.455 0.374 1 0.052 0.249 ¡0.074 ¡0.006 0.007 0.025 0.039 0.052 0.008 0.060 ¡0.055
MSCORE ¡0.020 ¡0.044 0.034 0.015 0.061 ¡0.035 0.045 1 ¡0.124 ¡0.204 ¡0.144 0.097 ¡0.274 ¡0.337 ¡0.006 0.110 ¡0.091 ¡0.009
EGP 0.024 0.026 0.087 0.475 0.141 0.040 0.104 ¡0.031 1 0.347 0.004 ¡0.096 0.257 0.121 0.098 0.032 0.131 0.041
TREND 0.020 0.036 ¡0.005 ¡0.017 ¡0.034 ¡0.058 ¡0.056 ¡0.152 0.086 1 ¡0.018 ¡0.085 0.210 0.172 0.045 0.001 0.028 ¡0.099
SIZE ¡0.007 ¡0.013 0.000 0.000 0.001 0.003 0.001 ¡0.135 ¡0.009 ¡0.013 1 ¡0.452 0.160 ¡0.020 ¡0.005 0.090 0.132 0.180
BM 0.040 0.068 ¡0.005 0.004 0.004 ¡0.006 0.004 0.088 ¡0.036 ¡0.038 ¡0.140 1 ¡0.249 ¡0.191 ¡0.049 0.107 ¡0.125 ¡0.189
PASTRET 0.012 0.023 0.012 ¡0.014 0.014 0.041 0.017 ¡0.204 0.063 0.160 0.017 ¡0.205 1 0.122 0.059 ¡0.079 0.159 0.119
GP 0.010 0.018 0.016 0.007 0.018 0.030 0.066 ¡0.199 0.003 0.193 0.015 ¡0.162 0.062 1 0.083 ¡0.211 0.173 0.153
ACC ¡0.025 ¡0.028 0.049 0.029 0.051 0.050 0.075 ¡0.008 0.036 0.047 ¡0.010 ¡0.048 0.052 0.063 1 ¡0.071 0.187 0.140
VOL ¡0.014 ¡0.017 0.001 ¡0.012 0.006 ¡0.010 0.005 0.089 0.179 0.001 0.003 0.022 ¡0.030 ¡0.070 ¡0.033 1 ¡0.192 ¡0.275
AG1 ¡0.007 ¡0.009 0.048 0.029 0.045 0.113 0.042 ¡0.038 0.003 0.024 0.024 ¡0.101 0.149 0.086 0.191 ¡0.100 1 0.518
AG2 ¡0.026 ¡0.042 ¡0.015 ¡0.015 ¡0.032 0.044 ¡0.024 0.067 ¡0.065 ¡0.071 0.034 ¡0.143 0.099 0.044 0.116 ¡0.132 0.491 1
7
Table 4
Portfolios Formed Based on Acceleration Measures.
This table reports the average market-adjusted returns for portfolios formed based on acceleration deciles. See Table 2 for variable definitions. t-statistics are
reported in parentheses, and are calculated based on the time-series of the portfolio market-adjusted stock returns.
Panel A: Equal-weighted portfolio returns
One-month abnormal returns (VMAR) Quarter-long abnormal returns (VMARQ)
EA deciles EAA EAP EAV SA PA EAA EAP EAV SA PA

Lowest 0.001 0.002 0.002 0.000 0.002 0.010 0.007 0.001 0.001 0.003
(0.404) (0.617) (1.196) (0.038) (0.56) (2.515) (1.152) (0.397) (0.225) (0.466)
2 0.000 0.000 0.002 0.002 0.001 0.004 0.001 0.002 0.005 0.002
(0.213) (0.199) (0.956) (1.203) (0.346) (0.995) (0.353) (0.594) (1.316) (0.573)
3 0.001 0.001 0.003 0.003 0.003 0.003 0.001 0.008 0.007 0.006
(0.951) (0.596) (1.867) (1.54) (1.438) (0.959) (0.195) (2.352) (2.275) (1.684)
4 0.003 0.003 0.003 0.005 0.005 0.006 0.006 0.010 0.012 0.008
(1.821) (1.902) (1.923) (3.063) (3.271) (1.826) (2.187) (3.437) (3.704) (2.978)
5 0.004 0.006 0.007 0.006 0.006 0.009 0.012 0.017 0.014 0.013
(2.967) (5.159) (4.008) (3.933) (3.778) (3.326) (5.689) (5.078) (5.224) (4.603)
6 0.009 0.011 0.010 0.008 0.010 0.019 0.023 0.022 0.017 0.017
(6.121) (8.682) (5.439) (5.018) (6.283) (6.638) (10.864) (6.841) (6.142) (6.222)
7 0.013 0.012 0.011 0.011 0.012 0.027 0.023 0.025 0.024 0.022
(8.213) (7.872) (6.445) (7.248) (7.288) (7.096) (8.586) (7.04) (8.165) (7.331)
8 0.013 0.013 0.014 0.013 0.016 0.027 0.026 0.030 0.028 0.030
(7.284) (7.959) (7.565) (7.594) (8.046) (7.596) (8.704) (8.186) (8.732) (8.329)
9 0.014 0.015 0.016 0.014 0.018 0.033 0.031 0.032 0.032 0.036
(7.223) (7.279) (8.535) (7.628) (8.31) (8.529) (7.81) (9.168) (8.824) (9.172)
Highest 0.015 0.016 0.018 0.013 0.020 0.037 0.042 0.038 0.035 0.046
(6.225) (5.131) (8.484) (5.868) (6.775) (8.233) (6.19) (9.306) (7.405) (6.604)
Highest - Lowest 0.014 0.018 0.020 0.013 0.017 0.026 0.034 0.039 0.034 0.042
(9.218) (10.254) (13.736) (8.975) (7.648) (11.078) (11.451) (15.866) (12.724) (10.314)
Panel B: Value-weighted portfolio returns
EA deciles EAA EAP EAV SA PA EAA EAP EAV SA PA
Lowest 0.002 0.007 0.005 0.001 0.003 0.002 0.014 0.009 0.005 0.015
(0.824) (2.672) (2.846) (0.481) (1.087) (0.515) (2.791) (3.584) (1.418) (3.727)
2 0.001 0.003 0.002 0.001 0.000 0.003 0.009 0.001 0.007 0.003
(0.443) (1.849) (0.947) (0.657) (0.109) (1.277) (2.651) (0.423) (2.212) (1.114)
3 0.002 0.003 0.000 0.000 0.001 0.007 0.007 0.004 0.001 0.002
(1.109) (1.573) (0.191) (0.004) (0.371) (2.862) (2.884) (1.496) (0.467) (0.731)
4 0.001 0.001 0.002 0.001 0.000 0.001 0.003 0.001 0.001 0.001
(0.747) (0.433) (1.192) (0.529) (0.321) (0.306) (1.359) (0.335) (0.738) (0.293)
5 0.002 0.005 0.001 0.003 0.000 0.002 0.003 0.002 0.004 0.000
(1.559) (3.838) (1.105) (1.868) (0.143) (1.2) (1.267) (0.66) (1.887) (0.152)
6 0.003 0.003 0.004 0.003 0.003 0.003 0.005 0.001 0.004 0.004
(2.379) (2.254) (3.073) (1.708) (2.117) (1.303) (2.528) (0.474) (1.695) (2.007)
7 0.007 0.003 0.004 0.004 0.003 0.007 0.004 0.004 0.005 0.002
(5.04) (1.987) (2.918) (3.066) (2.12) (3.044) (1.749) (1.789) (2.374) (0.672)
8 0.004 0.002 0.005 0.005 0.005 0.005 0.000 0.005 0.004 0.005
(2.416) (1.143) (2.888) (3.614) (2.451) (1.859) (0.085) (1.771) (1.544) (1.54)
9 0.002 0.005 0.005 0.005 0.009 0.004 0.008 0.008 0.007 0.011
(0.832) (2.326) (2.233) (2.682) (4.186) (1.499) (2.522) (2.84) (2.604) (3.021)
Highest 0.005 0.008 0.005 0.006 0.004 0.013 0.008 0.008 0.010 0.011
(2.122) (2.985) (2.803) (2.636) (1.48) (3.406) (1.921) (2.902) (3.061) (2.165)
Highest - Lowest 0.007 0.015 0.010 0.007 0.006 0.014 0.022 0.017 0.015 0.026
(2.582) (4.394) (3.993) (2.14) (1.737) (3.734) (3.876) (4.895) (3.51) (4.281)
increasing trend across the EAP deciles and the hedge portfolio return is again 1.8%. In column two, we present the results of
the EAP strategy using size-adjusted returns and again obtain a hedge portfolio return of 1.8%. In columns three and four, we
employ returns that are adjusted by the typical Fama French (FF) factors. Column three presents the results with the tradi-
tional three factor model and column four uses the Fama French three factor plus Momentum adjustment. Recently, Fama and
French (2015) have developed and tested a five-factor model that extends their original three-factor model with investment
and profitability factors. They argue that this augmented model explains a number of well-documented anomalies. In column
five, we test the robustness of the earnings acceleration strategy to this augmented risk model and show that the excess hedge
portfolio return remains significant over the month long window. Columns six through ten present qualitatively similar
results for value-weighted portfolio returns and provide confidence in the robustness of the results to various risk adjust-
ments. In all the remaining tests, we continue to employ the VMAR measure for excess returns.
Fig. 1. Earnings Acceleration Strategy over Different Horizons.

This figure depicts the difference in value-weighted market-adjusted returns (VMAR) between top and bottom earnings acceleration (EAP) deciles over different
time horizons (after earnings announcement). The x-axis represents the number of days after the earnings announcement date. The y-axis represents the VMAR
for the top and bottom EAP deciles as well as their difference averaged over 176 fiscal quarters from 1972 till 2015. See Table 2 for variable definitions.
Table 5
Portfolios Formed Based on Earnings Acceleration: Alternative Risk Adjustments.
This table reports the average one-month risk-adjusted returns for portfolios formed based on EAP deciles. See Table 2 for variable definitions. t-statistics are
reported in parentheses, and are calculated based on the time-series of the portfolio risk-adjusted stock returns.
Equal-weighted portfolio returns Value-weighted portfolio returns
EAP deciles EMAR SAR FF3 FFM FF5 EMAR SAR FF3 FFM FF5
Lowest 0.010 0.002 0.004 0.004 0.003 0.015 0.008 0.008 0.007 0.007
(5.199) (1.028) (2.273) (2.449) (1.786) (5.264) (3.072) (3.601) (3.761) (3.49)
2 0.009 0.001 0.003 0.003 0.002 0.012 0.004 0.003 0.004 0.003
(6.499) (0.414) (2.327) (2.776) (2.388) (4.631) (2.108) (2.113) (2.367) (2.112)
3 0.008 0.001 0.001 0.002 0.001 0.011 0.003 0.001 0.001 0.001
(6.021) (0.538) (1.735) (1.894) (1.412) (4.203) (1.603) (0.347) (0.699) (0.599)
4 0.006 0.002 0.000 0.000 0.001 0.009 0.001 0.000 0.001 0.001
(4.578) (2.173) (0.678) (0.419) (0.863) (3.425) (0.525) (0.239) (0.499) (0.74)
5 0.003 0.005 0.003 0.003 0.003 0.003 0.005 0.003 0.003 0.003
(1.804) (5.834) (4.227) (3.687) (4.112) (1.16) (3.808) (2.74) (2.216) (2.565)
6 0.001 0.010 0.007 0.007 0.007 0.006 0.003 0.003 0.002 0.003
(1.043) (10.16) (10.81) (10.66) (10.52) (2.339) (2.353) (3.145) (2.772) (2.541)
7 0.003 0.011 0.008 0.008 0.008 0.005 0.003 0.003 0.002 0.003
(2.139) (11.142) (10.398) (10.205) (10.71) (2.082) (1.906) (2.731) (2.161) (2.355)
8 0.004 0.013 0.008 0.008 0.008 0.006 0.002 0.001 0.001 0.000
(3.282) (11.807) (9.796) (9.898) (10.219) (2.344) (1.051) (1.059) (0.948) (0.245)
9 0.006 0.014 0.009 0.009 0.010 0.004 0.004 0.003 0.004 0.004
(3.731) (10.512) (7.787) (8.066) (8.452) (1.278) (2.088) (2.211) (2.409) (2.703)
Highest 0.007 0.016 0.011 0.011 0.012 0.000 0.007 0.005 0.005 0.006
(3.183) (7.084) (5.769) (6.107) (6.372) (0.101) (2.96) (3.012) (2.805) (3.472)
Highest - Lowest 0.018 0.018 0.015 0.015 0.015 0.015 0.015 0.013 0.012 0.013
(9.973) (9.946) (9.899) (9.642) (9.637) (4.338) (4.535) (4.982) (5.061) (5.108)
In Table 6, we examine the robustness of the earnings acceleration strategy to other well-documented anomalies.
Stambaugh and Yuan (2017) create a comprehensive mispricing measure e M-Score that incorporates several well-
documented anomalies. In Panel A, we examine the robustness of the earnings acceleration strategy hedge portfolio
excess returns to this M-Score measure. We follow Liu et al. (2017) in constructing portfolios independently sorted on both
the M-Score measure as well as our variable of interest, namely earnings acceleration. As we move from column one to
column five, we move from the lowest quintile to the highest quintile of earnings acceleration. Similarly, when we move from
row one to row five, we move from the lowest M-Score (which represents underpricing) to the highest M-score (which
represents overpricing). The last column depicts the returns from a hedge portfolio strategy of going long on highest quintile
of earnings acceleration and short on the lowest quintile of earnings acceleration. The trading strategy yields consistently
positive returns across all rows showing the robustness of the strategy to other well documented anomalies captured in M-
Table 6
Portfolios Formed Based on Earnings Acceleration and Other Anomaly Variables.
This table reports the average one-month market-adjusted returns for equal-weighted portfolios formed based on EAP deciles and other anomaly variables
(using independent sorting). See Table 2 for variable definitions. t-statistics are reported in parentheses, and are calculated based on the time-series of the
portfolio market-adjusted stock returns.
Panel A: Two-way sorting, controlling for Mscore effect
EAP
Mscore effect Lowest 2 3 4 Highest Highest -

Lowest
Underpriced 0.008 0.002 0.001 0.009 0.015 0.020 0.021
(6.221) (0.881) (0.895) (7.151) (10.055) (8.304) (7.438)
2 0.007 0.003 0.001 0.008 0.014 0.016 0.019
(5.29) (1.417) (0.904) (6.195) (8.872) (7.305) (7.981)
3 0.006 0.006 0.001 0.009 0.012 0.011 0.017
Mscore
(4.555) (3.07) (0.669) (6.361) (7.203) (4.945) (7.769)
4 0.006 0.004 0.002 0.009 0.010 0.011 0.015
(3.802) (1.445) (1.377) (5.4) (5.459) (4.701) (6.261)
Overpriced 0.001 0.007 0.000 0.003 0.006 0.005 0.012
(0.684) (2.366) (0.164) (1.538) (2.545) (1.998) (5.445)
Underpriced - Overpriced 0.006 0.005 0.002 0.006 0.009 0.014
(3.029) (1.613) (0.694) (2.223) (3.401) (4.667)
Panel B: Two-way sorting, controlling for PEAD effect
EAP
SUE effect Lowest 2 3 4 Highest Highest -
Lowest
EGP Lowest 0.000 0.002 0.002 0.002 0.004 0.009 0.011
(0.174) (0.839) (0.846) (0.521) (1.367) (2.392) (4.182)
2 0.003 0.003 0.000 0.007 0.006 0.010 0.012
(2.269) (1.015) (0.02) (4.193) (3.109) (3.235) (3.749)
3 0.008 0.002 0.000 0.010 0.011 0.016 0.019
(6.603) (0.681) (0.273) (8.249) (6.513) (4.168) (3.797)
4 0.010 0.000 0.004 0.007 0.016 0.013 0.013
(6.128) (0.155) (1.886) (3.483) (9.342) (4.923) (4.039)
Highest 0.015 0.004 0.003 0.010 0.015 0.020 0.016
(5.849) (1.11) (0.98) (2.131) (5.626) (7.109) (5.833)
Highest - Lowest 0.014 0.006 0.001 0.012 0.011 0.011
(8.77) (1.873) (0.381) (1.86) (3.503) (3.988)
Panel C: Two-way sorting, controlling for Profit Trend effect
EAP
TREND effect Lowest 2 3 4 Highest Highest -
Lowest
TREND Lowest 0.006 0.001 0.004 0.011 0.010 0.011 0.011
(3.162) (0.354) (1.907) (5.827) (4.858) (3.541) (4.014)
2 0.008 0.002 0.002 0.007 0.013 0.016 0.014
(5.295) (0.765) (1.073) (4.664) (6.788) (6.445) (5.325)
3 0.008 0.002 0.001 0.009 0.013 0.017 0.016
(5.349) (0.549) (0.554) (6.105) (7.644) (5.886) (5.739)
4 0.009 0.001 0.002 0.008 0.015 0.023 0.022
(5.54) (0.459) (1.228) (5.182) (7.716) (6.454) (6.841)
Highest 0.015 0.005 0.008 0.012 0.021 0.026 0.020
(5.32) (1.744) (3.569) (4.982) (7.269) (6.369) (7.033)
Highest - Lowest 0.009 0.006 0.004 0.003 0.011 0.015
(5.298) (2.242) (2.216) (1.291) (4.13) (5.475)
Score.14 More importantly, the lowest excess return is still economically large at 1.2% over the month long window. The
returns to the M-score strategy, depicted along the last row, are typically lower than the magnitude of the returns for the
earnings acceleration strategy. Additionally, they do not remain consistently significant across all the earnings acceleration
quintiles.
In Panel B, we follow the same methodology as Panel A, but examine the joint returns from the earnings acceleration and
post-earnings announcement drift (PEAD) strategies. The key variable used in traditional PEAD studies is Standardized Un-
expected Earnings (SUE). The “Unexpected” earnings are the surprise from a seasonal random walk model for earnings. Thus,
they are identical to the seasonal growth in earnings, which is the growth measure we use to compute earnings acceleration.
As such, we use the variable EGP to represent the same SUE variable that has been employed in the PEAD literature. Again, we
14
The same pattern is observed when we use deciles instead of quintiles.
obtain significant excess returns to a strategy of going long in the highest EAP quintile and short in the lowest EAP quintile
across all rows (as shown in the last column) and the lowest hedge return across the rows is 1.1%. In contrast, the PEAD
strategy does not yield significant results uniformly across all EAP quintiles. Akbas et al. (2017) have recently documented a
profitability trend anomaly that can potentially overlap with the earnings acceleration anomaly presented in this study. A
direct comparison of the two anomalies is presented in Panel C. Once again, the earnings acceleration strategy tends to
produce much higher excess returns than the profitability trend anomaly. More importantly, the acceleration anomaly is
present across all profitability trend partitions, while the profitability trend anomaly appears to be significantly smaller in
several of the earnings acceleration partitions.15
In addition to the hedge portfolio return tests, a typically more conservative test of the anomaly entails a regression
analysis across all deciles. Besides including data from all deciles, another advantage of the regression approach is the ability
to control for several risk factors and anomalies simultaneously. In Table 7, we present regression analysis of excess returns
from the earnings acceleration strategy after including all the controls. For columns one through three, the dependent var-
iable is the month-long excess return and for columns four through six, it is the quarter-long excess return. The main
regression model estimated is:
VMARðQ Þi;t ¼ a0 þ a1 EGPi;t þ a2 EAPi;t þ a3 SIZEi;t þ a4 TRENDi;t þ a5 BMi;t þ a6 PASTRETi;t þ a7 GPi;t þ a8 ACCi;t
(1)
þ a9 VOLi;t þ a10 AGi;t þ εi;t
where EAP is earnings acceleration, defined as quarter t earnings growth minus quarter t-1 earnings growth, and earnings
growth is the seasonal change in EPS scaled by the stock price at the end of quarter t-1 (see Table 2).
The standard errors employed are from Fama-Macbeth regressions (with Newey-West correction with six lags) and hence
are controlled for cross-sectional and serial correlation in the panel data. Model one includes controls for the PEAD strategy
and size and model two includes additional controls for the profitability trend, book-to-market, past returns, profitability,
accruals, earnings volatility and asset growth. In the discussion, we focus on the results from model two. The regression
coefficient on EAP, our earnings acceleration variable, is 0.016. Recall that the EAP decile rank variable has been scaled to have
Table 7
Earnings Acceleration and Stock Returns: Regression Analysis.
This table reports the regression results testing the relation between earnings acceleration and stock returns. See Table 2 for variable definitions. Standard
errors are from a Fama-MacBeth estimation with Newey-West correction for up to six lags. t-statistics are reported in parentheses. ***, **, * indicate
significantly different from zero at the 1%, 5%, 10% level, respectively.
Variables (1) (2) (3) (4) (5) (6)

Constant 0.007*** 0.008*** 0.008*** 0.016*** 0.017*** 0.016***
(4.109) (4.104) (4.289) (4.742) (4.823) (4.470)
EGP 0.009*** 0.012*** 0.012*** 0.042*** 0.038*** 0.040***
4.956 (5.025) (5.102) (12.878) (8.041) (7.952)
EAP 0.016*** 0.016*** 0.015*** 0.023*** 0.023*** 0.021***
(10.628) (9.221) (9.429) (8.792) (8.428) (8.251)
SIZE 0.009*** 0.004 0.003 0.041*** 0.032*** 0.028***
(3.295) (1.212) (0.951) (6.797) (4.344) (3.944)
TREND 0.005** 0.004* 0.013*** 0.012***
(2.217) (1.930) (3.809) (4.484)
BM 0.014*** 0.014*** 0.025*** 0.024***
(4.303) (4.350) (3.521) (3.431)
PASTRET 0.010*** 0.011*** 0.005 0.006
(3.328) (3.457) (0.932) (1.097)
GP 0.011*** 0.012*** 0.017*** 0.018***
(3.632) (3.778) (3.296) (3.696)
ACC 0.013*** 0.012*** 0.022*** 0.019***
(9.909) (9.416) (9.502) (8.127)
VOL 0.011*** 0.012*** 0.028*** 0.032***
(6.461) (6.993) (8.598) (10.535)
AG1 0.001 0.003
(0.588) (0.995)
AG2 0.001 0.014***
(0.219) (3.817)
Observations 355,492 244,864 244,864 347,802 239,353 239,353
R-squared 0.010 0.053 0.053 0.019 0.067 0.068
Number of groups 176 162 162 176 162 162
15
We get similar results using Fama-French three- and five-factor adjusted returns instead of VMAR for all three panels A, B and C.
a range of one and a mean of zero. Thus, the coefficient of 0.016 can be interpreted as the return (1.6%) from a hedge portfolio
that entails going long in the highest decile of earnings acceleration and short on the lowest decile. In column five, the
corresponding incremental return is 2.3%. These returns are comparable in magnitude to the PEAD anomaly, which had
returns of 1.2% over the first month and 3.8% over a quarter. Book-to-market and gross profitability have incremental returns
of 1.4% (2.5%) and 1.1% (1.7%) over the next month (quarter), respectively. In contrast, the profit trend anomaly only yields
incremental returns of 0.5% over the month and 1.3% over the quarter.16
3. Earnings acceleration and future earnings growth
So far, we have demonstrated that earnings acceleration can predict future stock returns and that these returns are robust
to adjustments for risk and other known anomalies. We now explore the nature of the information contained in earnings
acceleration. We examine whether earnings acceleration has incremental predictive ability for future earnings growth and
whether the future abnormal return from the earnings acceleration strategy documented in section 2 is associated with this
predictive ability.17 If so, the abnormal return we document likely manifests because investors do not consider fully the
implications of earnings acceleration for future earnings growth.
3.1. Implications of earnings acceleration for future earnings growth
We estimate a regression of future earnings growth on past earnings acceleration. Since past earnings growth has been
shown to predict future earnings growth (in the PEAD context), we also control for past earnings growth to document that the
implications of earnings acceleration are incremental.
EGPtþk ¼ a þ bEAPt þ gEGPt þ εtþk (2)
Here k takes on the values 1, 2 and 3 meaning that EGPtþk represents the seasonal earnings growth one, two and three
quarters in the future.
Table 8 reports the regression results testing the relation between earnings acceleration and the earnings growth for each
of the three subsequent quarters. Columns one, four and seven represent the basic relation between earnings acceleration and
future earnings growth one, two and three quarters, respectively, in the future. While the coefficient for one quarter ahead
growth is negative, the coefficients for the two subsequent quarters (i.e., 0.046 and 0.237) are significantly positive.18 The
other columns examine the effect of earnings acceleration for future earnings growth after including various controls for a
wide range of potential explanatory variables for earnings growth. These control variables are defined in Table 2. With
controls, the coefficients on EAP for every subsequent quarter are positive, though they are consistently stronger for earnings
growth two and three quarters into the future than for one quarter into the future. In columns five and eight, the coefficients
are significantly positive at 0.056 and 0.248, respectively, suggesting that earnings acceleration is a significant predictor of
future two- and three-quarters-ahead earnings growth. Economically, moving from the bottom decile to the top decile of
scaled earnings acceleration leads to a nearly 25% incremental change in the decile of earnings growth three quarters hence.
For comparison, the EGP coefficient, that is relevant in the PEAD context, is 32% for one quarter ahead earnings growth
(column two), and is actually negative (4.5% in column eight) for three quarters ahead earnings growth.
3.2. Short-window returns around future earnings announcement dates
While our primary results employ one-month and quarter-long abnormal returns and are robust to controlling for a litany
of risk factors, a further intuitive test involves shorter-window returns, which are typically less susceptible to risk consid-
erations (Bernard and Thomas, 1990; Sloan, 1996). Specifically, since we wish to assess whether the earnings acceleration
anomaly is attributable to the market missing, at least partially, the implications of earnings acceleration for earnings growth
two and three quarters in the future, we examine whether abnormal returns occur in short windows around earnings an-
nouncements two and three quarters ahead.
Table 9 reports the regression results testing the relation between earnings acceleration and the three-day abnormal
return surrounding each of the three subsequent earnings announcements. The coefficient on EAP is positive and significant
in all columns (ranging in magnitude from 0.2% to 0.8%). The short-window excess returns are largest around the third
subsequent quarter's earnings announcement (ranging from 0.7% to 0.8% in columns seven through nine). These magnitudes
are comparable to or larger than historically reported three-day returns in the PEAD context. Our finding that earnings ac-
celeration is positively associated with three-day returns around all three subsequent earnings announcements strongly
indicates that investors do not appear to incorporate fully the implications of earnings acceleration for subsequent earnings in
16
These results remain robust to estimating Fama-MacBeth regressions based on market-value-weighted least squares method as in Green et al. (2017).
17
Such a test is analogous to the PEAD context, where current earnings growth from a seasonal random walk model had implications for future earnings
growth and these implications have been shown to be associated with PEAD (see Bernard and Thomas, 1990; Rangan and Sloan, 1998, among others).
18
When EGP is excluded, the coefficient on EAP is significantly positive and economically large for one quarter ahead as well.
Table 8
Earnings Acceleration and Future Earnings Growth.
This table reports the regression results testing the relation between earnings acceleration and future earnings growth. See Table 2 for variable definitions.
Standard errors are from a Fama-MacBeth estimation with Newey-West correction for up to six lags. t-statistics are reported in parentheses. ***, **, * indicate
significantly different from zero at the 1%, 5%, 10% level, respectively.
One-quarter-ahead earnings growth Two-quarters-ahead earnings growth Three-quarters-ahead earnings growth
EGPtþ1 EGPtþ2 EGPtþ3
Variables (1) (2) (3) (4) (5) (6) (7) (8) (9)
Constant 0.001 0.000 0.002 0.002 0.002 0.001 0.001 0.002 0.002
(0.670) (0.134) (0.803) (0.832) (0.791) (0.660) (0.349) (0.480) (0.549)
EGP 0.416*** 0.323*** 0.331*** 0.220*** 0.166*** 0.176*** 0.006 0.045*** 0.035***
(56.963) (35.022) (37.623) (34.501) (24.196) (26.078) (0.786) (7.011) (5.727)
EAP 0.018*** 0.011*** 0.005 0.046*** 0.056*** 0.046*** 0.237*** 0.248*** 0.238***
(4.642) (2.813) (1.277) (10.331) (14.611) (12.373) (41.149) (39.663) (39.761)
SIZE 0.024*** 0.060*** 0.045*** 0.047*** 0.074*** 0.056*** 0.070*** 0.088*** 0.073***
(8.788) (18.104) (14.236) (13.194) (18.407) (14.589) (18.082) (21.543) (18.452)
TREND 0.030*** 0.016*** 0.007 0.021*** 0.013* 0.003
(8.583) (4.336) (1.145) (5.276) 1.682 (0.723)
BM 0.046*** 0.052*** 0.052*** 0.063*** 0.055*** 0.066***
(10.363) (11.891) (10.384) (12.571) (10.578) (12.720)
PASTRET 0.124*** 0.131*** 0.083*** 0.088*** 0.043*** 0.049***
(23.265) (21.600) (17.250) (17.881) (6.353) (6.199)
GP 0.011*** 0.017*** 0.005 0.011** 0.007 0.002
(2.692) (4.540) (0.791) (2.047) (1.096) (0.330)
ACC 0.009*** 0.000 0.019*** 0.010*** 0.021*** 0.014***
(3.535) (0.091) (6.254) (3.452) (6.900) (4.887)
VOL 0.072*** 0.054*** 0.077*** 0.055*** 0.081*** 0.061***
(10.556) (9.398) (10.156) (7.747) (11.610) (9.950)
AG1 0.006 0.021*** 0.031***
(1.195) (6.276) (6.651)
AG2 0.077*** 0.118*** 0.117***
(18.510) (21.962) (16.671)
Observations 335,264 231,678 231,678 321,755 222,671 222,671 318,129 219,348 219,348
R-squared 0.171 0.210 0.216 0.068 0.112 0.122 0.068 0.115 0.124
Number of groups 175 161 161 174 160 160 173 159 159
a timely fashion. Although a significant portion of the mispricing is corrected in the one-month following an earnings
announcement, some of the correction only takes place when future quarterly earnings are announced (especially two and
three future fiscal quarters).
3.3. The association between earnings acceleration's implications for future earnings growth and for future stock returns
Having documented that earnings acceleration has implications for both earnings growth up to three quarters in the future
and for three-day returns surrounding each of the three subsequent earnings announcements, we then conduct a formal
statistical analysis of market efficiency, i.e., the Mishkin test. In earnings-based anomaly literature, several prior studies
conduct the Mishkin test to examine whether the signs and magnitudes of abnormal stock returns reflect the market's
understanding of the earnings process.19 A primary advantage of the Mishkin test is that the test of market efficiency remains
valid regardless of what other correlated (omitted) variables help to predict earnings (Lewellen, 2010). In our Mishkin test, we
simultaneously estimate two equations (one for future abnormal returns [VMAR] and one for future earnings growth
[EGPtþk]). EGPt is also included as an independent variable in both the equations to ensure that the effect documented is
incremental to the previously known effect of earnings growth on future earnings growth and future stock return. Simul-
taneous estimation establishes whether the relation between future growth (EGPtþk) and past acceleration (EAPt) implicit in
future abnormal returns ðVMARÞ is the same as the directly observable relation between future earnings growth and past
earnings acceleration.
If the market correctly understands the implications of the earnings process depicted in Eq. (2), any price response
subsequent to the earnings announcement should not be related to past earnings acceleration and only be related to the
earnings growth surprise εt þk. Specifically, assuming the price response to εt þk over the month subsequent to the current
earnings announcement is linear, we have
19
See, for example, Sloan (1996), Dechow and Sloan (1997), Rangan and Sloan (1998), Collins and Hribar (2000), Narayanamoorthy (2006), Cao and
Narayanamoorthy (2012), Chen and Shane (2014), Hui et al. (2016) and Ma and Markov (2016).
Table 9
Earnings Acceleration and Future Three-Day Abnormal Returns around Earnings Announcements.
This table reports the regression results testing the relation between earnings acceleration and future three-day abnormal returns around earnings an-
nouncements. See Table 2 for variable definitions. Standard errors are from a Fama-MacBeth estimation with Newey-West correction for up to six lags. t-
statistics are reported in parentheses. ***, **, * indicate significantly different from zero at the 1%, 5%, 10% level, respectively.
Three-day abnormal return (tþ1) Three-day abnormal return (tþ2) Three-day abnormal return (tþ3)
VMAR3tþ1 VMAR3tþ2 VMAR3tþ3
Variables (1) (2) (3) (4) (5) (6) (7) (8) (9)
Constant 0.004*** 0.004*** 0.003*** 0.004*** 0.003*** 0.003*** 0.004*** 0.003*** 0.003***
(6.804) (5.637) (5.198) (6.866) (4.875) (5.098) (6.817) (3.039) (2.921)
EGP 0.006*** 0.003 0.003* 0.000 0.000 0.000 0.006*** 0.005*** 0.005***
(3.495) (1.595) (1.704) (0.259) (0.060) (0.318) (7.626) (4.740) (4.438)
EAP 0.002*** 0.002*** 0.002*** 0.005*** 0.004*** 0.004*** 0.008*** 0.007*** 0.007***
(3.716) (3.286) (3.235) (8.506) (5.864) (5.454) (11.102) (8.416) (8.120)
SIZE 0.007*** 0.004*** 0.004*** 0.007*** 0.005*** 0.004*** 0.007*** 0.004*** 0.003**
(5.318) (3.291) (2.949) (5.695) (3.220) (2.865) (5.666) (2.767) (2.369)
TREND 0.000 0.001** 0.000 0.000 0.001 0.001
(0.116) (2.283) (0.339) (0.312) (0.668) (1.168)
BM 0.008*** 0.008*** 0.006*** 0.006*** 0.007*** 0.007***
(9.067) (8.352) (6.814) (6.100) (7.580) (6.879)
PASTRET 0.002** 0.002*** 0.001 0.001 0.002* 0.001
(2.548) (2.770) (0.791) (1.025) (1.736) (1.506)
GP 0.006*** 0.007*** 0.007*** 0.007*** 0.007*** 0.007***
(4.683) (5.145) (5.041) (5.289) (4.179) (4.430)
ACC 0.001 0.001 0.004*** 0.004*** 0.002*** 0.002***
(1.601) (0.751) (6.985) (6.143) (3.827) (3.494)
VOL 0.006*** 0.006*** 0.006*** 0.007*** 0.007*** 0.008***
(6.928) (7.918) (8.116) (9.174) (7.953) (8.356)
AG1 0.002*** 0.000 0.001
(2.661) (0.401) (0.892)
AG2 0.004** 0.004*** 0.004***
(2.080) (5.099) (4.384)
Observations 335,314 231,700 231,700 321,806 222,693 222,693 318,171 219,372 219,372
R-squared 0.007 0.041 0.040 0.005 0.039 0.040 0.005 0.040 0.041
Number of groups 175 161 161 174 160 160 173 159 159
VMARtþ1 ¼ k0 þ k1 εtþk þ mtþ1 (3)
If the market is efficient in incorporating the effect of earnings acceleration, it will only respond to the actual innovation
(or noise) after controlling for acceleration. In other words, if the market correctly infers the relation in Eq. (2), εt þk in Eq. (3)
will be the same as εtþk in Eq. (2). Substituting εtþk with EGPtþk a bEAPt gEGPt from Eq. (2), we then get
VMARtþ1 ¼ k0 þ k1 ðEGPtþk a bEAPt gEGPt Þ þ mtþ1
which, on rearranging, yields
VMARtþ1 ¼ ðk0 k1 a Þ þ k1 EGPtþk k1 b EAPt k1 g EGPt þ mtþ1 (4)
In Eq. (2), b is the actual coefficient on our variable of interest while in Eq. (4), b* is the coefficient inferred from the
market's expectation of EGPtþk .20 The restriction b ¼ b * yields a likelihood ratio statistic that has a chi-square distribution
with one degree of freedom. If b ¼ 0; then we conclude that the market appears to completely ignore the implications of
earnings acceleration for future earnings growth. If b ¼ b, then the market appears to understand the implications of
earnings acceleration perfectly. If 0 < b < b, then the market partially understands the earnings acceleration. If b < 0; then the
market not only underestimates the implications of earnings acceleration, but also prices stocks as if they expect current
earnings acceleration to have negative implications for future earnings growth.
We present the results of the Mishkin Test in Table 10. Panel A presents the results from jointly estimating the one-quarter-
ahead earnings forecasting equation and the pricing equation. While the coefficient (b) on the actual correlation between

earnings acceleration and one-quarter-ahead earnings growth is 0.031, the coefficient (b ) on market's assessment (based
on one month excess returns) of the effect is 0.564. Panel B presents the results from jointly estimating the two-quarters-
20
Standard OLS does not readily provide direct estimates and associated standard errors for b*. The Mishkin framework, in contrast, extracts the un-
derlying parameter estimates by using non-linear least squares estimation and is asymptotically equivalent to OLS. Thus, the Mishkin approach allows us to
directly estimate and test hypotheses relating to the implications of earnings acceleration for future stock returns (Dechow et al., 2011).
Table 10
Test of Stock Market Efficiency for the Earnings Acceleration Effect.
This table reports the regression results from nonlinear generalized least squares estimation of the following two equations.
EGPtþk ¼ a þ bEAPt þ gEGPt þ εtþk
VMARtþ1 ¼ ðk0 k1 a Þ þ k1 EGPtþk k1 b EAPt k1 g EGPt þ mtþ1
See Table 2 for variable definitions. The likelihood ratio statistic is distributed asymptotically as c2 with 1 degree of freedom.
Panel A: One-quarter-ahead
Parameter Coef. z-statistics p-value

g 0.403 225.710 0.000
b 0.031 17.400 0.000
g* 0.215 6.480 0.000
b* 0.564 15.840 0.000
Test of market efficiency: g ¼ g* b ¼ b*
Likelihood ratio statistics: 31.78 223.290
Marginal significant level: 0.000 0.000
Panel B: Two-quarters-ahead
g 0.207 107.150 0.000
b 0.039 19.990 0.000
g* 0.022 0.660 0.509
b* 0.500 14.160 0.000
Panel C: Three-quarters-ahead
g 0.023 11.970 0.000
b 0.238 121.660 0.000
g* 0.222 6.430 0.000
b* 0.331 8.890 0.000
ahead earnings forecasting equation and the pricing equation. While b is 0.039, indicating a positive actual association be-
tween earnings acceleration and two-quarters-ahead earnings growth, b is 0.5, suggesting market's negative assessment of
the effect of earnings acceleration on future earnings growth. Panel C presents the results from jointly estimating the three-
quarters-ahead earnings forecasting equation and the pricing equation. Estimated b of 0.238 and b of 0.331 suggest that
while there is a significant positive association between earnings acceleration and three-quarters-ahead earnings, the
market's assessment of the association between the two is incrementally negative. In all three panels, the likelihood ratio test

reject that b ¼ b , indicating that market underestimates the implication of earnings acceleration for future one-, two- and
three-quarters-ahead earnings. In particular, we provide unique evidence that the market appears to not only underestimate
the magnitude of the effect of earnings acceleration on two- and three-quarters-ahead earnings growth, but assumes it is
incrementally negative.21
4. Why do investors miss the implications of earnings acceleration?
Our results in Section 3 confirm that the post-earnings acceleration returns are consistent with investors not under-
standing fully some of the implications of earnings acceleration for future earnings growth. In this section, we explore
whether the missing of these implications is consistent with investors assuming a particular time series process for quarterly
earnings when the underlying model is different.
Earnings has long been modeled in accounting research as arising either from a random walk plus noise process (Beaver
et al., 1980) or as comprising of “permanent” and “transitory” components (Ali and Zarowin, 1992). Econometrically, both of
these characterizations are captured by an IMA (1,1) model (Kothari, 2001). While this model assumption has primarily been
applied to annual earnings, Narayanamoorthy (2006) employs it for quarterly earnings to represent a non-seasonal random
21
b*k being negative for k ¼ 1, 2 and 3 is similar to the results obtained in prior studies employing the Mishkin test (see, for example, Rangan and Sloan
[1998]). If investors completely ignored the implications of earnings acceleration for future earnings growth, then b*k would be exactly zero. Thus, the
negative values indicate that investors not only ignore the earnings acceleration implications, but also price stocks as if they expect earnings acceleration to
have negative implications for future earnings.
walk process.22 Here, we exploit the same IMA (1,1) model assumption and examine its implications for the relation between
earnings acceleration and future earnings growth.
Let earnings follow a simple IMA (1,1) process with parameter q s 0. q is a measure of the extent of mean reversion in
earnings changes. The higher the q, the higher the extent of mean reversion. If Xt is the earnings at time t, then the IMA (1,1)
process characterizing Xt is:
Xt ¼ Xt1 þ εt qεt1
If we seasonally difference this series,
D4 Xt ¼ Xt Xt4 ¼ ðXt Xt1 Þ þ ðXt1 Xt2 Þ þ ðXt2 Xt3 Þ þ ðXt3 Xt4 Þ

¼ εt qεt1 þ εt1 qεt2 þ εt2 qεt3 þ εt3 qεt4
D4 Xt1 ¼ εt1 qεt2 þ εt2 qεt3 þ εt3 qεt4 þ εt4 qεt5
Then, earnings acceleration is:
EAt ¼ D4 Xt D4 Xt1 ¼ εt qεt1 εt4 þ qεt5
We are interested in the predictability of EAt for D4Xtþ1, D4Xtþ2, D4Xtþ3 and D4Xtþ4. Thus,
D4 Xtþ1 ¼ εtþ1 þ ð1 qÞεt þ ð1 qÞεt1 þ ð1 qÞεt2 qεt3
ð1 qÞ2
dtþ1 ¼ CorrðEAt ; D4 Xtþ1 Þ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 2
2 1þq 1 þ 3ð1 qÞ2 þ q
D4 Xtþ2 ¼ εtþ2 þ ð1 qÞεtþ1 þ ð1 qÞεt þ ð1 qÞεt1 qεt2
ð1 qÞ2
ffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 2
2 1þq 1 þ 3ð1 qÞ2 þ q
D4 Xtþ3 ¼ εtþ3 þ ð1 qÞεtþ2 þ ð1 qÞεtþ1 þ ð1 qÞεt qεt1
q þ ðq 1Þ2
2 2
2 1þq 1 þ 3ð1 qÞ2 þ q
D4 Xtþ4 ¼ εtþ4 þ ð1 qÞεtþ3 þ ð1 qÞεtþ2 þ ð1 qÞεtþ1 qεt

q
2
2 1þq 1 þ 3ð1 qÞ2 þ q2
Using a representative q value of 0.5, the correlation values are 0.112, 0.112, 0.335 and 0.224, respectively.23 In other
words, the model shows that earnings acceleration has small positive implications for earnings growth one and two quarters
ahead and a much larger positive effect three quarters ahead. The implications for four quarters-ahead earnings growth are
negative. Earnings acceleration's correlation with earnings growth one, two and three quarterseahead is driven by the
presence of two common error terms: εt and εt1 . While both the error terms reinforce each other in their implication for
earnings growth three quarters ahead, they act in opposite directions in determining the correlation with earnings growth
one and two quarters ahead.
22
Several prior studies model quarterly earnings with seasonal differencing (e.g., Brown and Rozeff, 1979; Griffin, 1977; Watts, 1975; Foster, 1977). We do
not employ these models here for two reasons. First, these models are shown to have limited out of sample prediction performance. Second, while quarterly
earnings are shown to have both seasonal and non-seasonal characteristics, we specifically explore the theory that the market misprices the latter. Hence
we need a model without seasonal differencing. We note that several other non-seasonal models like, for example, AR (1) [see Brown and Han (2000)]
would yield results similar to IMA (1,1). In other words, it is not necessary for the earnings process to be IMA (1,1) for our theory of the market seasonally
differencing a non-seasonal process to hold. That said, IMA (1,1) represents a wide range of economic data including earnings well (Brown, 1993) and is
more readily tractable analytically.
23
In comparison to the calculated values of the model predictions depicted above, the actual observed correlations appear to be reasonably close at 0.150,
0.132, 0.225 and 0.115, respectively. Note that the coefficient on EAP in Table 8 will be different from these correlations since the multivariate regression
tests in Table 8 include controls for EGP while the model here does not.
In contrast to IMA (1,1), if the true process were a seasonal random walk (SRW) model, earnings acceleration will be εt
εt1 and the future earnings growths would be εtþ1 ; εtþ2 ; εtþ3 and εtþ4 ; respectively: In such a case, the correlations between
earnings acceleration and future earnings growth would be zero. Thus, if the market uses a naïve SRW model to form earnings
expectations, it assumes zero correlation between earnings acceleration and future earnings growth. Consequently, it will
miss the implications (correlations) of current earnings acceleration for future earnings growth that are a result of true
earnings being a process like IMA (1,1) (which appears to have empirical validity e see footnote 23).
Representing the true earnings as IMA (1,1) with mean reversion parameter, q, allows for cross-sectional tests of the
earnings acceleration effect with variation in mean reversion. In Appendix, we depict the sensitivities of the correlations
between earnings acceleration and future earnings growth to q. We find that three of the four correlations monotonically
decrease in q and the fourth correlation is almost unchanged in q: If the market were missing the correlations between
earnings acceleration and future earnings growth due to a fixation on the SRW model, we will observe greater returns when
the correlations are algebraically larger, that is, when the mean reversion parameter is lower.
To empirically test this prediction, we examine the relation between earnings acceleration and future abnormal returns
conditional on two measures that have been previously shown to be associated with mean reversion in earnings: losses and
earnings volatility. Specifically, Brooks and Buckmaster (1976) document higher mean reversion for loss firms. Dichev and
Tang (2009) find that higher volatility firms mean-revert more. Thus, under the hypothesis that market fixates on the
SRW model, firms with losses and more volatile earnings should demonstrate lower future abnormal returns from the
earnings acceleration strategy. In Table 11, Panel A, we compare returns from the acceleration-based strategy for loss versus
profit firms. Firms with EPS in quarter t smaller than zero are classified as loss firms and the rest are classified as profit firms.
The top quintile of earnings acceleration for loss firms produces an abnormal return of 0.9% while the same for profit firms
produces a return of 1.7%. The top minus bottom quintile abnormal return is 0.8% for loss firms while it is 1.8% for profit firms.
We test for the significance of difference in hedge returns using an approach similar to Paternoster et al. (1998).24 The dif-
ference in hedge returns between loss firms and profit firms is 1% and statistically significant.
In Panel B, we stratify companies according to past earnings volatility. Firms with above median earnings volatility in each
quarter are classified as high volatility firms and the rest are classified as low volatility firms. The top minus the bottom
Table 11
Losses and Earnings Volatility on the Effect of Earnings Acceleration.
This table reports the portfolio results testing the relation between earnings acceleration and stock returns, conditional on losses and earnings volatility. See
Table 2 for variable definitions. t-statistics are calculated based on the time-series of the portfolio market-adjusted stock returns.
Panel A: Loss and the effect of earnings acceleration
Loss firms Profit firms
EAP quintiles VMAR t-statistics VMAR t-statistics

Lowest 0.001 0.263 0.001 0.351
2 0.006 2.128 0.001 0.714
3 0.002 0.466 0.008 7.535
4 0.008 1.970 0.013 8.898
Highest 0.009 2.429 0.017 8.403
Highest - Lowest 0.008 3.537 0.018 13.683
Test for significance of difference in hedge returns between profit firms and loss firms:
Difference in hedge returns 0.010
z-statistics 3.766
Panel B: Earnings volatility and the effect of earnings acceleration
High earnings volatility firms Low earnings volatility firms
EAP quintiles VMAR t-statistics VMAR t-statistics
Lowest 0.000 0.154 0.001 0.422
2 0.002 1.097 0.002 1.553
3 0.003 2.074 0.009 8.377
4 0.008 4.976 0.016 9.796
Highest 0.014 5.726 0.021 6.849
Highest - Lowest 0.014 10.950 0.020 7.403
Test for significance of difference in hedge returns between low earnings volatility and high earning volatility firms:
z-statistics 2.044
24
Specifically, we evaluate the significance of the difference in returns between two groups (G1 and G2) using z-statistic, computed as: z ¼
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ,
meanG1meanG2 where meanG1 and meanG2 refer to mean returns in hedge return groups G1 and G2; SEG1 and SEG2 refer to the standard errors of returns
SEG12 þSEG22
in hedge return groups G1 and G2.
quintile abnormal return is 2% for low volatility firms and 1.4% for high volatility firms. The earnings acceleration effect being
stronger for profit firms and for low volatility firms (which are firms with lower mean reversion in earnings) is consistent
with our prediction, and it provides added corroboration for the hypothesis that the market misses the correlations between
earnings acceleration and future earnings growth due to a fixation on the SRW model for earnings.
5. Additional tests
In this section, we address three additional topics: (1) additional ways to enhance returns from the acceleration-based
strategy, (2) robustness of the results to the adoption of alternative definitions/deflators, and (3) preliminary tests
regarding the implementability of the strategy.
5.1. Enhancing the earnings acceleration strategy returns by considering alternative earnings acceleration patterns
In Section 4 above, we show that the basic earnings acceleration strategy can be enhanced by focusing on profit firms or
low earnings volatility firms. Here, we explore additional ways to enhance the strategy. Specifically, following Cao et al.
(2011), we partition EAP into six patterns as follows:
Pattern 1: Both current and previous quarter's earnings growth are positive, and current quarter's earnings growth is
higher than previous quarter's.
Pattern 2: Current quarter's earnings growth is positive, while previous quarter's earnings growth is negative.
Pattern 3: Both current and previous quarter's earnings growth are negative, and current quarter's earnings growth is
higher than previous quarter's.
Pattern 4: Both current and previous quarter's earnings growth are positive, and current quarter's earnings growth is
smaller than previous quarter's.
Pattern 5: Current quarter's earnings growth is negative, while previous quarter's earnings growth is positive.
Pattern 6: Both current and previous quarter's earnings growth are negative, and current quarter's earnings growth is
smaller than previous quarter's.
We examine the relation between earnings acceleration and stock returns conditional on each of the earnings acceleration
patterns. As discussed in Section 4, our theory suggests that the earnings acceleration effect is stronger when mean reversion
is lower. Basu (1997) documents that earnings decreases mean revert more than earnings increases. Thus, among the six
earnings acceleration patterns, the one pattern which should have the lowest mean reversion, and consequently, the highest
future returns, is Pattern 1 (since it is an increase-in-increase pattern). Every other pattern has at least one kind of decrease,
and hence will have some degree of mean reversion.25 Consistent with this prediction, we find that pattern 1 has the largest
EAP coefficient among all the patterns (untabulated). In Table 12, Panel A, we show that employing only pattern 1 obser-
vations has EAP coefficient of 5.6% (in column three) relative to 1.6% when using the full sample (column one).
While pattern 1 has the strongest earnings acceleration effect due to low mean reversion in earnings, it is not possible to
build a hedge portfolio strategy using pattern 1 alone. By construction, pattern 1 observations have positive earnings ac-
celerations and hence negative earnings acceleration deciles will not be represented in a hedge portfolio. As such, we need to
combine pattern 1 with patterns 4, 5 or 6 (which, by construction, populate the majority of the negative earnings acceleration
deciles) to form a hedge portfolio. Patterns 4, 5 and 6 all have higher mean reversion than pattern 1. However, there is no
theoretical guide for which of patterns e 4, 5 or 6 e have relatively lower mean reversion. We find that pattern 5, when
combined with pattern 1, produces the largest hedge portfolio returns (Table 12, Panel A, columns five and six). This com-
bination, despite producing higher hedge portfolio returns than the full sample, still yields lower returns in regression tests
relative to using pattern 1 alone (3.3% in column five versus 5.6% in column three), which is consistent with our theory. In
Table 12, Panel B, we report the hedge portfolio returns for portfolios double sorted (independently) based on EAP deciles and
whether the stock belongs to either pattern 1 or pattern 5. A trading strategy that focuses only on pattern 1 or pattern 5
generates one-month hedge return of 2.6%. In contrast, our base strategy only yielded a hedge return of 1.8% (Table 4). Thus,
focusing on specific patterns of earnings acceleration can enhance the excess returns by nearly 45%.
5.2. Alternative acceleration definitions/deflators
While we show that the relation between earnings acceleration and stock returns is present under each of the five def-
initions of acceleration (i.e., EAA, EAP, EAV, SA and PA) in section 2, we only present results for EAP when investigating the
relation between earnings acceleration and future earnings growth in section 3. Un-tabulated results show that the relation
between earnings acceleration and future earnings growth is present and remarkably robust across each of the other four
acceleration measures.26
25
Patterns 2, 3, 5, 6 all have at least one period of negative earnings growth. For pattern 4, although earnings growth in both periods is positive, the
change in earnings growth rate is negative (or in other words, there is a decrease in earnings growth rate).
26
We find that EAV consistently produces even stronger results than EAP.
Table 12
Patterns of Earnings Acceleration
This table reports the regression and portfolio results testing the relation between earnings acceleration and stock returns, conditional on different earnings
acceleration patterns. See Table 2 for variable definitions. Standard errors are from a Fama-MacBeth estimation with Newey-West correction for up to six
lags. t-statistics are reported in parentheses. ***, **, * indicate significantly different from zero at the 1%, 5%, 10% level, respectively.
Panel A: Full sample, pattern 1 only, and pattern 1 or 5
One-month abnormal returns (VMAR)
Full sample Pattern 1 firms only Pattern 1 or 5 firms only
Variables (1) (2) (3) (4) (5) (6)

Constant 0.008*** 0.008*** 0.006** 0.006** 0.010*** 0.010***
(4.104) (4.289) (2.177) (2.125) (4.965) (4.983)
EGP 0.012*** 0.012*** 0.013 0.014 0.001 0.002
(5.025) (5.102) (1.124) (1.151) (0.108) (0.247)
EAP 0.016*** 0.015*** 0.056*** 0.058*** 0.033*** 0.034***
(9.221) (9.429) (4.428) (4.752) (4.616) (4.802)
SIZE 0.004 0.003 0.018*** 0.017*** 0.010*** 0.009***
(1.212) (0.951) (3.895) (3.770) (2.883) (2.820)
TREND 0.005** 0.004* 0.007** 0.009** 0.004 0.004
(2.217) (1.930) (1.977) (2.364) (1.591) (1.650)
BM 0.014*** 0.014*** 0.020*** 0.021*** 0.012*** 0.013***
(4.303) (4.350) (3.870) (4.041) (3.042) (3.089)
PASTRET 0.010*** 0.011*** 0.002 0.002 0.005 0.005
(3.328) (3.457) (0.552) (0.460) (1.540) (1.551)
GP 0.011*** 0.012*** 0.014*** 0.015*** 0.010*** 0.010***
(3.632) (3.778) (3.637) (3.843) (2.765) (2.885)
ACC 0.013*** 0.012*** 0.011*** 0.010*** 0.013*** 0.012***
(9.909) (9.416) (3.986) (3.682) (6.392) (6.097)
VOL 0.011*** 0.012*** 0.026*** 0.026*** 0.014*** 0.014***
(6.461) (6.993) (6.560) (6.554) (5.764) (5.928)
AG1 0.001 0.008** 0.002
(0.588) (2.482) (0.831)
AG2 0.001 0.007** 0.000
(0.219) (2.155) (0.120)
Observations 244,864 244,864 55,724 55,724 94,491 94,491
R-squared 0.053 0.053 0.079 0.079 0.059 0.060
Number of groups 162 162 160 160 160 160
Panel B: Non-pattern 1 or 5 versus Pattern 1 or 5
Non-pattern 1 or 5 Pattern 1 or 5
EAP deciles VMAR t-statistics VMAR t-statistics

Lowest 0.000 0.036 0.005 1.645
2 0.004 1.575 0.005 2.456
3 0.002 1.338 0.002 1.290
4 0.004 2.411 0.001 0.344
5 0.006 4.557 0.007 3.987
6 0.009 5.101 0.011 7.998
7 0.007 4.157 0.016 8.748
8 0.009 5.184 0.018 8.532
9 0.013 6.078 0.022 7.809
Highest 0.015 4.621 0.022 4.831
Highest - Lowest 0.014 6.290 0.026 7.093
Test for significance of difference in hedge returns between pattern 1 or 5 and non-pattern 1 or 5 firms:
z-statistics 2.777
We also examine how different deflators for EGP and EAP affect our results. Specifically, we examine a total of 16 scenarios,
as a combination of using each of the following four variables as deflators for EGP and for EAP: last quarter's stock price, four-
quarters-prior stock price, last quarter's total asset, and four-quarters-prior total assets. The returns results remain qualita-
tively unchanged under these different EGP and EAP deflators.
Lastly, we estimate a Taylor-series expansion of the earnings process using the past 8 quarters' earnings, and define
earnings acceleration as the coefficient on the square of the time variable. This definition is analogous to acceleration in
physics, and it frees us from the need to use a deflator in defining earnings acceleration. The acceleration variable defined in
this manner also positively predicts future abnormal returns (untabulated).
5.3. Implementability of the earnings acceleration strategy
We have already discussed in section 2 the robustness of the strategy to the use of value-weighted portfolios instead of
equal-weighted portfolios. We further examine here the implementability of the earnings acceleration strategy along three
dimensions e stability of the excess returns over time, the exclusion of low price/low market capitalization stocks and the use
of calendar month rebalancing. In Fig. 2, we depict the one-month hedge returns to the earnings acceleration strategy for each
of the 176 quarters in our sample. The hedge return is positive in 140 out of the 176 fiscal quarters (80%), which suggests that
the relation between earnings acceleration and subsequent stock returns is quite stable over time. This also alleviates con-
cerns that the excess returns are a result of unidentified risk factors. More importantly, Fig. 2 shows that the trading strategy is
equally successful in recent years compared to earlier periods. From 2004 to 2015, the strategy yields positive excess returns
in 41 out of 48 fiscal quarters (85%). This finding is relevant in view of the recent finding by Green et al. (2017) that a majority
of the previously well documented anomalies do not generate returns significant from zero in the post-2003 period.
Notwithstanding that a detailed cost-benefit analysis of the trading strategy is beyond the scope of this study, our second set
of results indicate significant positive excess returns of between 0.9% and 1.5% per month even when low priced stocks (less than
$5) and/or various partitions of small capitalization stocks (up to $0.5 billion) are excluded from the trading strategy.
Fig. 2. Stability of Earnings Acceleration Strategy over Time.

This figure depicts the one-month return by fiscal quarter to a hedge portfolio taking a long position in the stock of firms in the highest decile of EAP and an equal
sized short position in the stock of firms in the lowest decile of EAP. The x-axis represents fiscal quarters. The y-axis represents the one-month hedge portfolio
returns.
Table 13
Alphas and Factor Loadings on Portfolios Sorted on Earnings Acceleration.
This table reports calendar-month average returns to portfolios sorted on earnings acceleration, and results of time series regressions of these portfolios'
returns on the Fama and French five factors [the market factor (MKT), the size factor small-minus-large (SMB), the value factor high-minus-low (HML), the
profitability factor robust-minus-weak (RMW), and the investment factor conservative-minus-aggressive (CMA)]. t-statistics are reported in parentheses.
***, **, * indicate significantly different from zero at the 1%, 5%, 10% level, respectively.
EAP deciles Average raw return Alpha MKT SMB HML RMW CMA
Lowest 0.015*** 0.004* 1.132*** 1.180*** 0.188 0.613*** 0.009
(4.283) (1.916) (22.600) (13.370) (1.380) (6.082) (0.040)
2 0.013*** 0.002 1.086*** 0.971*** 0.236*** 0.173* 0.117
(4.506) (1.335) (32.250) (15.136) (2.709) (1.964) (0.961)
3 0.012*** 0.001 1.069*** 0.872*** 0.147** 0.029 0.163
(4.49) (0.777) (38.848) (15.812) (1.967) (0.338) (1.427)
4 0.013*** 0.002** 1.056*** 0.759*** 0.059 0.106* 0.063
(5.048) (2.164) (58.170) (22.714) (1.323) (1.827) (0.906)
5 0.014*** 0.004*** 1.040*** 0.613*** 0.060 0.168*** 0.052
(5.969) (6.527) (51.321) (19.940) (1.421) (3.866) (0.922)
6 0.017*** 0.006*** 1.054*** 0.569*** 0.043 0.197*** 0.005
(6.976) (9.445) (60.761) (20.356) (1.021) (4.453) (0.093)
7 0.017*** 0.006*** 1.060*** 0.692*** 0.013 0.091 0.029
(6.825) (8.468) (48.754) (18.080) (0.221) (1.424) (0.509)
8 0.018*** 0.007*** 1.096*** 0.792*** 0.044 0.010 0.090
(6.871) (8.752) (37.940) (14.381) (0.627) (0.107) (0.950)
9 0.020*** 0.008*** 1.097*** 0.936*** 0.158** 0.139* 0.067
(7.035) (8.253) (38.265) (16.880) (2.091) (1.756) (0.664)
Highest 0.024*** 0.012*** 1.175*** 1.152*** 0.217 0.451*** 0.200
(6.963) (6.156) (20.851) (11.461) (1.572) (3.147) (0.959)
Highest - Lowest 0.009*** 0.008*** 0.044 0.028 0.029 0.162** 0.191**
(8.784) (7.116) (1.451) (0.550) (0.481) (2.219) (2.018)
The trading strategy outlined earlier involved buying and selling stocks two days after an earnings announcement. Such a
strategy can potentially lead to significant portfolio rebalancing costs. In our third set of tests, we adopt a conservative cal-
endar month-based rebalancing strategy. At the beginning of each calendar month, we sort stocks of companies that
announced earnings in the previous three months into earnings acceleration deciles. Table 13 presents the results of the
calendar month rebalancing strategy using equal-weighted portfolio returns.27 A hedge portfolio going long in the top
earnings acceleration decile and short in the bottom decile still yields about 0.9% excess returns over the month-long window.
The table also presents factor loadings on the five Fama-French factors.
6. Conclusion
We document that earnings acceleration is an important variable that active investors can focus on in their stock picking
efforts to earn significant excess returns. We find economically significant excess returns to an earnings acceleration-based
strategy over a quarter following an earnings announcement (with a significant portion accruing over the first month). The
returns are robust to a battery of controls for risk and are distinct from previously documented anomalies. In portfolio tests,
the incremental excess returns at 1.8% over a month translate to an annualized returns of over 23%. The returns are also
remarkably stable over a long period of time (we report results for 176 past quarters).
Our results indicate that the abnormal returns are consistent with investors not incorporating fully the implications of
current earnings acceleration for future earnings growth, especially two and three quarters in the future. Notably, current
earnings acceleration appears to be associated with significant positive returns in these quarters' earnings announcement
windows. Our direct tests of market efficiency also confirm that the positive returns from the earnings acceleration strategy
and the positive implications of earnings acceleration for future earnings growth are strongly associated. Last but not least, we
show that the results are consistent with investors employing a SRW model for quarterly earnings when the true underlying
process is not SRW. The test for whether investors do adopt such a model is only possible, however, in an experimental
setting.
Appendix
Correlations between Earnings Acceleration and Future Earnings Growth: Sensitivity to Mean Reversion
Recall that expressions for dj, the correlations between earnings acceleration and future earnings growths, were derived
under an IMA (1,1) model assumption in Section 4. Sensitivities of the correlations, dj, to the IMA (1,1) parameter q, the extent
of mean reversion, are computed analytically below.
4 3
ddtþ1 5q 6q þ 6q 5
¼ 3 3
dq 4 q þ 1 2 2q2 3q þ 2 2
2
4 3
¼ 3 2 3
dq 2
4 q þ 1 2 2q 3q þ 2 2
4 3
ddtþ3 q 3q þ 3q 1
¼ 3 2 3
dq 2
4 q þ 1 2 2q 3q þ 2 2
4 3
¼ 3 2 3
dq 2
4 q þ 1 2 2q 3q þ 2 2
In the range of 0 < q < 1, the derivatives of dtþ1, dtþ2 and dtþ4 are always negative. While the derivative of dtþ3 is not always
negative, it is very small, which makes the correlation between earnings acceleration and future earnings growth nearly flat
lying within a narrow range of 0.33e0.35.
References
Aboody, D., Barth, M.E., Kasznik, R., 2004. SFAS No. 123 stock-based compensation expense and equity market values. Account. Rev. 79 (2), 251e275.
Akbas, F., Jiang, C., Koch, P., 2017. The Trend in Firm Profitability and the Cross Section of Stock Returns. The Accounting Review.
Ali, A., Zarowin, P., 1992. Permanent versus transitory components of annual earnings and estimation error in earnings response coefficients. J. Account.
Econ. 15, 249e264.
27
The results are nearly unchanged if we consider earnings announcements only in the previous month instead of the previous three months or if we use
value-weighted portfolio returns.
Basu, S., 1997. The conservatism principle and the asymetric timeliness of earnings. J. Account. Econ. 24, 3e37.
Beaver, W., Lambert, R., Morse, D., 1980. The information content of security prices. J. Account. Econ. 2, 3e28.
Bernard, V.L., Thomas, J.K., 1989. Post-earnings-announcement drift: delayed price response or risk premium? J. Account. Res. 27 (1989), 1e36.
Bernard, V.L., Thomas, J.K., 1990. Evidence that stock prices do not fully reflect the implications of current earnings for future earnings. J. Account. Econ. 13,
305e340.
Brooks, L., Buckmaster, D., 1976. Further evidence of the time series properties of accounting income. J. Financ. XXXI (5).
Brown, L.D., 1993. Earnings forecasting research: its implications for capital markets research. Int. J. Forecast. 9, 295e320.
Brown, L.D., Han, J.C.Y., 2000. Do stock prices fully reflect the implications of current earnings for future earnings for AR1 firms ? J. Account. Res. 38 (1),
149e164.
Brown, L.D., Rozeff, M.S., 1979. Univariate time-series models of quarterly accounting earnings per share: a proposed model. J. Account. Res. 17 (1), 179e189.
Brown, S.J., Warner, J.B., 1980. Measuring security price performance. J. Financ. Econ. 8 (3), 205e258.
Brown, S.J., Warner, J.B., 1985. Using daily stock returns: the case of event studies. J. Financ. Econ. 14 (1), 3e31.
Cao, S.S., Narayanamoorthy, G.S., 2012. Earnings volatility, post-earnings announcement drift, and trading frictions. J. Account. Res. 50 (1), 41e74.
Cao, Y., Myers, L.A., Sougiannis, T., 2011. Does earnings acceleration convey information? Rev. Account. Stud. 16 (4), 812e842.
Chan, L.K.C., Karceski, J., Lakonishok, J., 2003. The level and persistence of growth rates. J. Financ. 58 (2), 643e684.
Chen, J.Z., Shane, P.B., 2014. Changes in cash: persistence and pricing implications. J. Account. Res. 52 (3), 599e634.
Chen, P., Zhang, G., 2007. How do accounting variables explain stock price movements? Theory and evidence. J. Account. Econ. 43 (2e3), 219e244.
Collins, D.W., Hribar, P., 2000. Earnings-based and accrual-based market anomalies: one effect or two? J. Account. Econ. 29 (1), 101e123.
Cooper, M.J., Gulen, H., Schill, M.J., 2008. Asset growth and the cross-section of stock returns. J. Financ. 63 (4), 1609e1651.
Copeland, T., Aaron, D., Alberto, M., 2004. The role of expectations in explaining the cross-section of stock returns. Rev. Account. Stud. (9), 149e188.
Dechow, P.M., Khimich, N., Sloan, R.G., 2011. The Accrual Anomaly. The Handbook Of Equity Market Anomalies: Translating Market Inefficiencies into
Effective Investment Strategies. Wiley Finance (Chapter 2).
Dechow, P.M., Sloan, R.G., 1997. Returns to contrarian investment strategies: tests of naive expectations hypotheses. J. Financ. Econ. 43, 3e27.
Dichev, I.D., Tang, V.W., 2009. Earnings volatility and earnings predictability. J. Account. Econ. 47 (1e2), 160e181.
Fama, E.F., 1998. Market efficiency, long-term returns, and behavioral finance. J. Financ. Econ. 49 (3), 283e306.
Fama, E.F., French, K.R., 2006. Profitability, investment and average returns. J. Financ. Econ. 82 (3), 491e518.
Fama, E.F., French, K.R., 2015. A five-factor asset pricing model. J. Financ. Econ. 116 (1), 1e22.
Foster, G., 1977. Quarterly accounting data: time-series properties and predictive-ability results. Account. Rev. 52 (1), 1e21.
Green, J., Hand, J.R.M., Zhang, X.F., 2017. The characteristics that provide independent information about average U.S. monthly stock returns. Rev. Financ.
Stud. 919e962.
Griffin, P., 1977. The time-series behavior of quarterly earnings: preliminary evidence. J. Account. Res. 15 (1), 71e83.
Harvey, C.R., Liu, Y., Zhu, H., 2016. ...and the cross-section of expected returns. Rev. Financ. Stud. 29 (1), 5e68.
Hou, K., Xue, C., Zhang, L., 2017. Replicating Anomalies. NBER Working Paper.
Hui, K.W., Nelson, K.K., Yeung, P.E., 2016. On the persistence and pricing of industry-wide and firm-specific earnings, cash flows, and accruals. J. Account.
Econ. 61 (1), 185e202.
Jegadeesh, N., Titman, S., 1993. Returns to buying winners and selling losers: implications for stock market efficiency. J. Financ. 48 (1), 65.
Kothari, S.P., 2001. Capital markets research in accounting. J. Account. Econ. 31 (1e3), 105e231.
Lakonishok, J., Shleifer, A., Vishny, R., 1994. Contrarian investment, extrapolation, and risk. J. Financ. XLIX (5).
Landsman, W.R., Maydew, E.L., 2002. Has the information content of quarterly earnings announcements declined in the past three decades? J. Account. Res.
40 (3), 797e808.
Latane , H.A., Tuttle, D.L., Jones, C.P., 1969. E/P ratios v. changes in earnings in forecasting future price changes. Financ. Anal. J. 25 (1), 117e123.
Lee, C.M.C., 1992. Earnings news and small traders: an intraday analysis. J. Account. Econ. 15 (2e3), 265e302.
Lewellen, J., 2010. Accounting anomalies and fundamental analysis: an alternative view. J. Account. Econ. 50 (2e3), 455e466.
Linnainmaa, J., Roberts, M., 2017. The History of the Cross Section of Stock Returns. Working Paper (August).
Livnat, J., Mendenhall, R.R., 2006. Comparing the post-earnings announcement drift for surprises calculated from analyst and time series forecasts. J.
Account. Res. 44 (1), 177e205.
Liu, J., Stambaugh, R.F., Yuan, Y., 2018. Absolving beta of volatility's effects. J. Financ. Econ. 128 (1), 1e15.
Ma, G., Markov, S., 2016. The market's assessment of the probability of meeting or beating the consensus. Contemp. Account. Res. 34 (1), 314e342.
Narayanamoorthy, G., 2006. Conservatism and cross-sectional variation in the post-earnings announcement drift. J. Account. Res. 44 (4), 763e789.
Novy-Marx, R., 2013. The other side of value: the gross profitability premium. J. Financ. Econ. 108 (1), 1e28.
O'Neil, W.J., 1999. 24 Essential Lessons for Investment Success: Learn the Most Important Investment Techniques from the Founder of Investor's Business
Daily. McGraw-Hill Education.
Penman, S., Zhu, J., 2014. Accounting anomalies, risk, and return. Account. Rev. 89 (5), 1835e1866.
Paternoster, R., Brame, R., Mazerolle, P., Piquero, A., 1998. Using the correct statistical test for the equality of regression coefficients. Criminology 36 (4),
859e866.
Rangan, S., Sloan, R.G., 1998. Implications of the integral approach to quarterly reporting for the post-earnings-announcement drift. Account. Rev. 73 (3),
353e371.
Sloan, R.G., 1996. Do stock prices fully reflect information in accruals and cash flows about future earnings? Account. Rev. 71 (3), 289e315.
Stambaugh, R.F., Yuan, Y., 2017. Mispricing Factors. Rev. Financ. Stud 30 (4), 1270e1315.
Watts, R., 1975. The Time Series Behavior of Quarterly Earnings. Working Paper.

He 2019

Uploaded by

Copyright:

Available Formats

He 2019

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

He 2019

Uploaded by

Copyright:

Available Formats

Journal of Accounting and Economics xxx (xxxx) xxx

Contents lists available at ScienceDirect

Journal of Accounting and Economics

Earnings acceleration and stock returns*

2. Earnings acceleration and future returns

2.1. Data and descriptive statistics

2.1.1. Sample selection

2.1.2. Construction of earnings acceleration and other variables

EPSi;t EPSi;t4 EPSi;t1 EPSi;t5

Our second deﬁnition of earnings acceleration (EAP) is calculated as:

EPSi;t EPSi;t4 EPSi;t1 EPSi;t5

Our third deﬁnition of earnings acceleration (EAV) is calculated as:

EPSi;t EPSi;t4 EPSi;t1 EPSi;t5

Our deﬁnition of sales acceleration (SA) is calculated as:

SPSi;t SPSi;t4 SPSi;t1 SPSi;t5

Our deﬁnition of proﬁtability acceleration (PA) is calculated as:

2.1.3. Descriptive statistics

2.2. Basic empirical results

Panel A: Summary statistics

N Mean Std. Dev. Q1 Median Q3

Panel B: Pearson (below), Spearman (above) correlations

Panel A: Equal-weighted portfolio returns

One-month abnormal returns (VMAR) Quarter-long abnormal returns (VMARQ)

EA deciles EAA EAP EAV SA PA EAA EAP EAV SA PA

Fig. 1. Earnings Acceleration Strategy over Different Horizons.

Equal-weighted portfolio returns Value-weighted portfolio returns

Panel A: Two-way sorting, controlling for Mscore effect

Mscore effect Lowest 2 3 4 Highest Highest -

One-month abnormal returns (VMAR) Quarter-long abnormal returns (VMARQ)

Variables (1) (2) (3) (4) (5) (6)

3. Earnings acceleration and future earnings growth

3.1. Implications of earnings acceleration for future earnings growth

EGPtþk ¼ a þ bEAPt þ gEGPt þ εtþk (2)

3.2. Short-window returns around future earnings announcement dates

One-quarter-ahead earnings growth Two-quarters-ahead earnings growth Three-quarters-ahead earnings growth

EGPtþ1 EGPtþ2 EGPtþ3

VMAR3tþ1 VMAR3tþ2 VMAR3tþ3

VMARtþ1 ¼ k0 þ k1 εtþk þ mtþ1 (3)

VMARtþ1 ¼ k0 þ k1 ðEGPtþk a bEAPt gEGPt Þ þ mtþ1

which, on rearranging, yields

VMARtþ1 ¼ ðk0 k1 a Þ þ k1 EGPtþk k1 b EAPt k1 g EGPt þ mtþ1 (4)

Parameter Coef. z-statistics p-value

4. Why do investors miss the implications of earnings acceleration?

If we seasonally difference this series,

D4 Xt ¼ Xt Xt4 ¼ ðXt Xt1 Þ þ ðXt1 Xt2 Þ þ ðXt2 Xt3 Þ þ ðXt3 Xt4 Þ

D4 Xt1 ¼ εt1 qεt2 þ εt2 qεt3 þ εt3 qεt4 þ εt4 qεt5

Then, earnings acceleration is:

EAt ¼ D4 Xt D4 Xt1 ¼ εt qεt1 εt4 þ qεt5

D4 Xtþ1 ¼ εtþ1 þ ð1 qÞεt þ ð1 qÞεt1 þ ð1 qÞεt2 qεt3

D4 Xtþ2 ¼ εtþ2 þ ð1 qÞεtþ1 þ ð1 qÞεt þ ð1 qÞεt1 qεt2

D4 Xtþ3 ¼ εtþ3 þ ð1 qÞεtþ2 þ ð1 qÞεtþ1 þ ð1 qÞεt qεt1

D4 Xtþ4 ¼ εtþ4 þ ð1 qÞεtþ3 þ ð1 qÞεtþ2 þ ð1 qÞεtþ1 qεt

Panel A: Loss and the effect of earnings acceleration

Loss firms Profit firms

EAP quintiles VMAR t-statistics VMAR t-statistics

5.2. Alternative acceleration deﬁnitions/deﬂators

Panel A: Full sample, pattern 1 only, and pattern 1 or 5

One-month abnormal returns (VMAR)

Full sample Pattern 1 ﬁrms only Pattern 1 or 5 ﬁrms only