Justification - Inflation

DOI: 10.1111/j.1475-679X.2005.00181.
x
Journal of Accounting Research
Vol. 43 No. 4 September 2005
Printed in U.S.A.
Inflation Illusion and

Post-Earnings-Announcement Drift
TA R U N C H O R D I A ∗ A N D L A K S H M A N A N S H I VA K U M A R †
Received 23 December 2003; accepted 5 December 2004
ABSTRACT
This paper examines the cross-sectional implications of the inflation illusion

hypothesis for the post-earnings-announcement drift. The inflation illusion
hypothesis suggests that stock market investors fail to incorporate inflation
in forecasting future earnings growth rates, and this causes firms whose earn-
ings growths are positively (negatively) related to inflation to be undervalued
(overvalued). We argue and show that the sensitivity of earnings growth to in-
flation varies monotonically across stocks sorted on standardized unexpected
earnings (SUE) and, consistent with the inflation illusion hypothesis, show
that lagged inflation predicts future earnings growth, abnormal returns, and
earnings announcement returns of SUE-sorted stocks. Interestingly, control-
ling for the return predictive ability of inflation weakens the ability of lagged
SUE to predict future returns of SUE-sorted stocks.
1. Introduction
Ball and Brown [1968] were the first to note that stock returns continue
to drift in the direction of earnings surprises for several months after the
earnings are announced. Since then, several studies have confirmed the ro-
bustness of the post-earnings-announcement drift using more recent data
∗ Goizueta Business School, Emory University; †London Business School. We thank Eli Amir,
Ray Ball, Sudipta Basu, Michael Brennan, Greg Clinch, Paul Irvine, Ken Peasnell, Shivaram
Rajgopal, Abbie Smith, Mohan Venkatachalam, an anonymous referee, and seminar partici-
pants at Cranfield University, Case Western University, London Business School, Duke Uni-
versity, Vanderbilt University and London Business School Accounting Symposium for helpful
comments. The second author was supported by the Dean’s Fund for Research at the London
Business School. All errors are our own.
521
Copyright
C , University of Chicago on behalf of the Institute of Professional Accounting, 2005
522 T. CHORDIA AND L. SHIVAKUMAR
and using data from stock markets other than those in the United States,
where the phenomenon was first identified.1 These studies show that trad-
ing strategies that sort stocks on their standardized unexpected earnings
(SUE) yield significantly positive payoffs in the year following the earnings
announcement.
The post-earnings-announcement drift has been robust over three
decades, and it conflicts with the standard frictionless asset-pricing mod-
els. The drift is commonly interpreted as evidence that investors underreact
to earnings surprises (Bernard and Thomas [1990], Ball and Bartov [1996]).
Most empirical studies of the post-earnings-announcement drift have pro-
vided evidence consistent with market inefficiency and investor irrationality
(Bernard and Thomas [1990], Rangan and Sloan [1998], Brown and Han
[2000], Mendenhall [2004]). However, it is difficult to understand why stock
prices appear to not respond completely and immediately to information
as visible and freely available as publicly announced earnings. This paper
argues that part of the underreaction to earnings surprises is a result of
the “inflation illusion” hypothesis that was first proposed by Modigliani and
Cohn [1979] to explain the negative correlation between the aggregate
market yield and inflation.
Modigliani and Cohn [1979] argue that stock market investors fail to
incorporate the effect of inflation on nominal earnings growth rates when
valuing stocks. Thus, when inflation rises, investors do not adjust the future
earnings growth, even though they fully adjust the discount rates. This failure
to consider the impact of inflation on earnings growth causes market yields
to be depressed in periods of high inflation and to be excessive in periods
of deflation. A direct implication of this hypothesis is that, if the earnings
growth in response to inflation varies across stocks, inflation illusion would
induce mis-valuation in the cross section. Stocks with earnings growths that
are positively related to inflation would be undervalued, whereas those with
earnings growths negatively related to inflation would be overvalued.
This paper investigates the possibility that earnings growth in response to
inflation varies across stocks sorted on earnings growth as measured by SUE,
and that inflation illusion partly causes the post-earnings-announcement
drift. Since economy-wide variables are well known to play a significant role
in determining earnings changes of individual firms (e.g., Ball and Brown
[1967]), one can expect stocks sorted on earnings growth to have different
earnings sensitivities to the business cycle. Consider the effect of inflation
on earnings growth. Firms with high earnings sensitivities to inflation are
likely to have larger increases in earnings growth and, thus, be included
in high-SUE portfolios. Similar arguments indicate that firms with negative
or less positive earnings sensitivities to inflation are likely to be included
1 Foster, Olsen, and Shevlin [1984] and Bernard and Thomas [1990], among others, confirm
the robustness of the Ball and Brown [1968] findings using more recent data. Hew et al. [1996]
and Booth, Kallunki and Martikainen [1996] extend the post-earnings-announcement drift
evidence to non-U.S. data.
INFLATION ILLUSION 523
in low-SUE portfolios. Thus, sorting stocks on the basis of SUE could yield
portfolios that have different earnings sensitivities to inflation. The failure
by investors to consider inflation in predicting future earnings growth could
then explain part of the drift.
We document the following patterns in earnings and returns that are
consistent with the inflation illusion explanation for the post-earnings-
announcement drift: (1) the sensitivity of earnings growth to inflation does
vary monotonically across SUE-sorted stocks, (2) future earnings growths
and returns of stocks sorted on SUE are predicted by inflation, (3) infla-
tion also predicts the future earnings announcement returns of SUE-sorted
stocks, (4) inflation significantly reduces the ability of SUE to predict future
returns, indicating that the prior finding on the ability of SUE to predict
future returns is due partly to investors underreacting to the earnings in-
formation in inflation; and (5) risk-based asset-pricing models that include
the standard Fama and French [1993] model, as well as a factor related to
news about future inflation, do not capture the impact of SUE on the cross
section of stock returns.
We conclude that investor underestimation of the impact of inflation on
future earnings growth partly causes the post-earnings-announcement drift.
Such underestimation could be due either to investor naı̈veté or to rational
learning in a world of parameter uncertainty. Brav and Heaton [2002] show
that it is difficult to distinguish between these explanations because of their
mathematical and predictive similarities.
This paper’s findings take our understanding of the post-earnings-
announcement drift to a more basic, macroeconomic level. Whereas
Bernard and Thomas [1990], Ball and Bartov [1996], and Brown and Han
[2000] provide evidence that investors underestimate the magnitude of the
serial correlation in quarterly differenced earnings, our results suggest that
this underestimation arises partly from investors underestimating the mag-
nitude of the impact of inflation on future corporate earnings. Our results
also add to the recent evidence in the macroeconomics literature (Ball and
Croushore [2003]) that economic agents systematically misestimate the ef-
fects of monetary policies on future output.
The rest of this paper is organized as follows. The next section presents
the inflation illusion hypothesis. Section 3 discusses the data and presents
the main results. Section 4 considers alternative risk-based explanations for
the results, and section 5 concludes.
2. Inflation Illusion Hypothesis and Its Implications

for the Drift Anomaly
2.1 INFLATION ILLUSION HYPOTHESIS
Modigliani and Cohn [1979] suggest that, unlike bond market investors,
stock market investors are subject to inflation illusion. They fail to un-
derstand the effect of inflation on nominal earnings growth rates, and
consequently, capitalize real earnings growth at a nominal interest rate

rather than the economically correct real rate. To see this, consider the
Gordon growth model for the market portfolio,
D t+1 E t+1
Pt = = (1 − b) , (1)
r −g r −g
where:
r= long-term discount rate;
g= long-term growth rate of dividends (D t ) or earnings (E t );
b= plowback ratio; and
Pt = price.
The main influence on long-term nominal interest rates is the long-term
rate of inflation (Fama [1975]). Moreover, as stocks are claims to the real
economy, changes in the long-term expected inflation would move nominal
earnings growth rates one-for-one, offsetting the effect of inflation on nom-
inal rates and thus leaving the aggregate price–earnings or price–dividends
ratio unaffected. Modigliani and Cohn [1979] argue that the price–earnings
ratio moves with the nominal bond yields because stock market investors
irrationally fail to adjust the nominal earnings growth rate, whereas the
nominal discount rates, which are determined by bond market investors,
correctly incorporate the impact of inflation. Thus, stocks will be underval-
ued in periods of inflation, with the undervaluation varying with the level
of inflation.2
Modigliani and Cohn [1979] discuss inflation illusion in the context of
the aggregate market portfolio, but the phenomenon should be manifested
at the individual stock level as well. More specifically, consider the Gordon
growth model for an individual stock, i,
D it+1 E it+1
Pit = = (1 − b i ) , (2)
ri − gi ri − gi
where all variables are defined as before.
If, in response to inflation, the earnings growths of some firms are higher
than those of other firms, and if investors do not adjust for the effect of
inflation on earnings growth, then there should be cross-sectional differ-
ences in valuation. More specifically, if investors adjust the discount rate, r i ,
in response to increased inflation but fail to adjust g i , then there should be
cross-sectional differences in valuation. In the presence of inflation, a stock
whose earnings growth is positively related to inflation should be underval-
ued, and a stock whose earnings growth is negatively related to inflation
should be overvalued. Thus, inflation, which is highly persistent, would not
only predict the future earnings growths of firms, but also predict their
future abnormal returns, as the mispricing is corrected over time.
2 Campbell and Vuolteenaho [2003] provide empirical support for this hypothesis and show
that inflation illusion explains almost 80% of the time-series variation in mispricing of the S&P
500.
2.2IMPLICATIONS OF INFLATION ILLUSION HYPOTHESIS

FOR THE POST-EARNINGS-ANNOUNCEMENT DRIFT
We consider the possibility that there is cross-sectional variation in the
sensitivity to inflation of earnings of SUE-sorted stocks and, as a result, mis-
pricing due to inflation illusion varies across these stocks. One reason to
expect earnings sensitivity to vary across SUE-sorted stocks is that earnings
growth is, to a large extent, determined by business-cycle variables (Brown
and Ball [1967], Gonedes [1973], Magee [1974]). Firms with more positive
earnings sensitivities to inflation are likely to have larger increases in earn-
ings growth, and thus are likely to be included in high-SUE portfolios. Firms
with negative or less positive earnings sensitivities to inflation are likely to
be included in low-SUE portfolios.3
The validity of the above conjecture for SUE-sorted portfolios is eventu-
ally an empirical issue, and in section 3.2, we test whether sorting stocks on
SUE yields portfolios that have different earnings sensitivities to inflation.
The remainder of this paper examines whether investors’ failure to con-
sider inflation in predicting future earnings growth partly causes the post-
earnings-announcement drift.
3. Results
3.1 DATA AND DESCRIPTIVE STATISTICS
Our sample consists of all New York Stock Exchange–American Stock
Exchange (NYSE-AMEX) firms with data available in the monthly Center
for Research in Security Prices (CRSP) database and quarterly Compustat
files for the period December 1971 through December 2001. We focus only
on common stocks, and eliminate American Depositary Receipts, Real Estate
Investment Trusts, Americus Trust Components, units, and closed-end funds
from the sample.
Following most prior studies in the literature, our tests use the SUE to
capture the post-earnings-announcement drift. In each month, the SUE
for firm i is computed as the most recently announced earnings less the
earnings four quarters ago. This earnings change is standardized by its stan-
dard deviation estimated over the prior eight quarters.4 To avoid using stale
earnings, we require the most recent earnings to have been announced no
earlier than four months before the end of the month in which standardized
earnings are measured. We note that the main results of this paper are un-
affected if the analysis is restricted to stocks with earnings announcements
3 Similar arguments can be made for other macroeconomic variables as well, and thus sorting
stocks on SUE could yield portfolios that differ not only in their earnings sensitivities to inflation
but also in their sensitivities to other macroeconomic variables as well. Given the arguments of
Modigliani and Cohn [1979], we focus primarily on inflation.
4 We repeat the analyses after allowing for a drift in earnings, as in Bernard and Thomas
[1989]. The results remain qualitatively unchanged with this modification.

in the current month. In each month, sample firms are sorted into deciles
based on SUE, using the distribution of SUE from the prior three months to
determine the decile cutoffs. The decile portfolios are denoted P 1 through
P 10 , with P 1 (P 10 ) being the lowest (highest) SUE portfolio. In each month,
and within each portfolio, we average the SUE of firms constituting the
portfolio to obtain the portfolio’s SUE.
Table 1 presents the average SUE for each of the 10 SUE portfolios. Over
the sample period January 1972 through December 2001, the SUE for the
lowest unexpected earnings decile portfolio, P 1 , was −4.65, whereas the
SUE for portfolio P 10 was 3.24. Note also that the cross-sectional standard
deviations are high in the extreme SUE portfolios, that is, in portfolios P 1
and P 10 .5 Subperiod analysis shows that the average SUE in the 1970s is
higher than the average SUE in the 1980s and 1990s for all the portfolios.
The null hypothesis of equal unexpected earnings across the subperiods is
rejected for each of the decile portfolios.
3.2 EARNINGS EXPOSURE TO MACROECONOMIC VARIABLES
We now examine how the relationship between earnings growth, mea-
sured by SUE, and business conditions varies across the SUE portfolios. It
might appear that one could formally test for this cross-sectional variation
by regressing the SUE of individual firms on proxies for changes in busi-
ness conditions (BC t ), averaging the coefficients on BC t within each
SUE portfolio and then, finally, evaluating the average coefficients across
SUE portfolios. However, having a meaningful number of observations in
the individual regressions under this approach requires imposing parame-
ter stationarity assumptions for relatively long periods of time, owing to the
quarterly frequency of earnings observations. This assumption is unlikely
to be valid for several reasons. For instance, earnings exposures of firms to
business conditions change as firms continuously react to the changing en-
vironment by investing in new projects, mergers, acquisitions, divestitures,
restructurings, and plant closings (Ball, Kothari, and Watts [1993]). Fur-
ther, a firm’s earnings exposure to macroeconomic factors depends on the
nature of its contracts (e.g., nominal or real) with suppliers, customers, and
employees, which vary as contracts mature and new contracts are signed. In
addition, earnings exposure will also vary with changes in a firm’s produc-
tion (e.g., input mix and suppliers), marketing (e.g., pricing), and financial
strategies (e.g., holdings of cash and trading securities and hedge contracts)
through their effects on product prices, factor costs, and returns on financial
investments. Finally, as reported earnings are based on historical costs of in-
ventory and historical depreciation, earnings exposure to inflation will also
change with replacement of inventory and property, plant and equipment.
5 The large standard deviations in SUE for the P and P

1 10 portfolios are due to a few extreme
observations. To test the sensitivity of our results to outliers in the SUE, we repeated our analyses
after excluding, in each month, the extreme 1% of SUE on either side. Our conclusions remain
unaffected by this modification.
TABLE 1
Summary Statistics of Standardized Unexpected Earnings (SUE)
P1 P2 P3 P4 P5 P6 P7 P8 P9 P 10
Dec 1971–Dec 2001 Mean −4.65 −1.21 −0.60 −0.24 0.00 0.20 0.43 0.76 1.33 3.24
Std. dev. 6.12 0.33 0.28 0.21 0.17 0.18 0.24 0.30 0.39 0.62
Dec 1971–Dec 1981 Mean −3.02 −1.08 −0.50 −0.14 0.13 0.39 0.69 1.11 1.75 3.76
Std. dev. 0.71 0.39 0.32 0.24 0.19 0.18 0.20 0.23 0.32 0.62
Jan 1982–Dec 1991 Mean −4.46 −1.37 −0.75 −0.36 −0.10 0.09 0.28 0.55 1.05 2.73
Std. dev. 1.38 0.27 0.23 0.19 0.14 0.12 0.13 0.15 0.19 0.32
Jan 1992–Dec 2001 Mean −6.48 −1.18 −0.56 −0.21 −0.01 0.14 0.33 0.62 1.19 3.21
Std. dev. 10.22 0.25 0.19 0.13 0.08 0.07 0.10 0.13 0.20 0.37
F -test subperiods(p-value) 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
In each month, t, all NYSE-AMEX firms are sorted into deciles based on their standardized unexpected earnings (SUE it ), which is defined as SUE it = (E it − E it −4 )/σ it , where E it is
the most recently announced earnings for firm i, E it −4 represents the earnings four quarters ago, and σ it is the standard deviation of (E it − E it −4 ) over the prior eight quarters. The
SUE for each decile portfolio (or portfolio SUE) is then obtained by averaging the SUE i t of the individual firms constituting the portfolio. The table presents the time-series mean and
standard deviation for each portfolio SUE. The sample period comprises 360 months from January 1972 through December 2001. On average, each decile portfolio consists of 183 stocks.
INFLATION ILLUSION
527
The alternative is to assume parameter stationarity at the portfolio level,

as done by Ball, Kothari, and Watts [1993], and regress the portfolio SUE
on proxies for BC t .6 However, a problem with this approach is that the
regression of SUE on changes in business conditions would be misspecified
if estimated separately for portfolios sorted on SUE, as this entails estimating
regressions separately for observations sorted on the dependent variable.
To avoid this problem, we form portfolios by sorting on SUE in quarter q
and regress the quarter-ahead (q + 1) SUE for each portfolio on business
conditions, including growth in nominal gross domestic product (GDP),
growth in real GDP, industrial production growth, and change in Consumer
Price Index (i.e., inflation). As earnings as well as data on GDP are available
only on a quarterly frequency, these regressions are estimated with quarterly
data.
The results are presented in table 2. Earnings vary monotonically with the
business cycle, as measured by growth in the nominal GDP. For firms in the
lowest-SUE portfolio, the association between portfolio SUE and growth in
nominal GDP is a statistically significant 0.11. Moving across portfolios, we
observe that the coefficients on SUE increase almost monotonically across
the portfolios. For firms in the highest-SUE portfolio, the coefficient on port-
folio SUE is a statistically significant 0.29, which implies that a one standard
deviation increase in growth of nominal GDP (0.94) increases the portfolio
SUE by 0.27. Further, the null hypothesis that the coefficients across the 10
portfolios are the same is strongly rejected (p < 0.001).
In contrast, for coefficient estimates from regressions of growth in real
output, measured either as real GDP or as industrial production, the null
hypothesis that the coefficients are the same for all portfolios cannot be
rejected at a 5% level in either of these regressions, indicating that the port-
folios have little variation in their exposure to real output. The systematic
variation in coefficients in regression of nominal GDP growth, but not in
regression of real GDP growth, indicates that earnings of SUE portfolios are
likely to reflect information about inflation.
Regressions of inflation on portfolio SUE confirm the above conjecture
and yield conclusions that are similar to those discussed above for growth in
nominal GDP. In these regressions, the coefficient on portfolio SUE is signifi-
cantly negative only for the smallest SUE decile, and increases monotonically
from about −0.09 for the lowest-SUE portfolio to 0.39 for the highest-SUE
portfolio. These coefficients imply that a one standard deviation increase
in quarterly inflation of 0.88 will decrease the SUE of portfolio P 1 by 0.08,
whereas it will increase the SUE of portfolio P 10 by 0.34.
6 We find that individual stocks constantly change their SUE portfolios. For instance, a stock
that is included in the top quintile of SUE in month t has less than a 20% chance of continuing
to be in this portfolio by month t+12. This suggests that earnings exposures estimated at the
portfolio level would be attenuated if the earnings exposures of individual stocks were relatively
stable over time, thereby making it less likely to identify differences in earnings exposures across
portfolios.
TABLE 2
Regression of Standardized Unexpected Earnings on Changes in Business Conditions
Independent F -test
Variable P1 P2 P3 P4 P5 P6 P7 P8 P9 P 10 (p-value)
GGDP q,q Coef. 0.105 0.149 0.160 0.137 0.130 0.144 0.169 0.193 0.213 0.293 0.00
t-stat. 2.99 5.03 5.37 6.49 5.69 6.41 6.62 6.78 6.20 5.60
GRGDP q,q Coef. 0.171 0.170 0.160 0.130 0.128 0.118 0.109 0.087 0.062 0.001 0.06
t-stat. 4.40 4.96 4.53 5.06 4.65 4.19 3.31 2.28 1.37 0.02
IPG q,q Coef. 0.084 0.097 0.105 0.077 0.079 0.080 0.068 0.060 0.055 0.034 0.24
t-stat. 4.62 6.19 6.54 6.48 6.48 6.43 4.48 3.43 2.56 1.06
INF q,q Coef. −0.093 0.004 0.024 0.034 0.040 0.062 0.104 0.162 0.220 0.394 0.00
t-stat. −2.52 0.13 0.67 1.29 1.31 2.31 3.47 5.07 5.97 7.93
Each quarter, sample firms are sorted into deciles based on their standardized unexpected earnings in quarter q, which is defined as SUE i,q = (E iq − E iq −4 )/σ iq , where E iq is the
most recently announced earnings, E iq −4 represents the earnings from four quarters ago, and σiq is the standard deviation of (E iq – E ia −4 ) over the prior eight quarters. Portfolio
SUE are computed by averaging SUEiq +1 across all firms constituting a portfolio. The table presents the coefficient estimates and associated t-statistics from regressing the portfolio
SUE in quarter q+1 on macroeconomic conditions that include the three-month growth in GDP (GGDP), real GDP (GRGDP), industrial production (IPG), and the three-month
inf lation (INF). The regression is estimated separately for each SUE portfolio, using 119 quarterly observations. The column titled “F -test (p-value)” reports the p-values from the
F -test that the coefficients across all portfolios are the same. The sample covers the period from January 1972 through December 2001.
INFLATION ILLUSION
529
These results indicate that earnings sensitivity to inflation varies systemat-

ically across SUE-sorted portfolios, which raises the possibility that inflation
illusion partly causes the post-earnings-announcement drift. If, as argued
by Modigliani and Cohn [1979], investors fail to take into account inflation
while predicting future earnings growth, then stocks in the highest-SUE
portfolio would be undervalued whereas those in the lowest-SUE portfo-
lio would be overvalued. Thus, the results in table 2 combined with the
inflation illusion argument give rise to the following testable implications.
First, as inflation is highly persistent, current levels of inflation would pre-
dict cross-sectional variation in future earnings growth across SUE-sorted
portfolios. Second, if investors fail to consider this cross-sectional variation
in future earnings growth, then current levels of inflation would predict
cross-sectional variation in future abnormal returns to SUE-sorted portfo-
lios. We test these implications next.
In order to maximize the power of our tests, subsequent analyses focus
primarily on a zero-investment portfolio that is long on the highest-SUE
portfolio and short on the lowest-SUE portfolio.7 The findings in table 2
suggest that this zero-investment portfolio, which we refer to as PMN (for
positive minus negative SUE), maximizes the earnings exposure to inflation,
and has a significantly positive exposure to inflation.
3.3 INFLATION AND FUTURE EARNINGS GROWTH
To test whether inflation predicts future earnings growth, we regress the
difference in SUE between the highest- and the lowest-SUE portfolio, which
measures future earnings changes of the PMN portfolio, on lagged inflation.
The difference in SUE across the extreme-SUE portfolios (SUE pmn ) is mea-
sured over the four quarters subsequent to the SUE-portfolio formation
month (quarters +1 through +4). As inflation in a particular month is re-
ported only in the following month, we lag inflation by one month relative
to the PMN formation month (i.e., month t−1) to ensure that only infor-
mation available to investors as of the portfolio formation month is used
in the regressions. Inflation is measured for the quarter ending in month
t−2 (INF t −4,t −2 ) or for the year ending in month t−2 (INF t −11,t −2 ).8 As the
SUE correspond to a quarter, but are measured every month for the PMN
portfolio, there is overlap in the observations. To account for this overlap,
we report t-statistics based on Newey–West standard errors. In any case, our
results are qualitatively unaffected when we use nonoverlapping data at a
quarterly frequency to estimate these regressions. The regression results are
reported in table 3.
Panel A of table 3 shows that lagged inflation is related to SUE pmn for up
to four quarters ahead (i.e., quarters +1 through +4). We obtain this result
10 + P 9 + P 8 + P 7 +
7 Our results are robust to defining the zero-investment portfolio as P
P 6 − P 5 − P 4 − P 3 − P 2 − P 1.
8 The results are robust to using inflation measured in month t − 2.
TABLE 3
Regression of SUE of PMN Portfolio on Lagged Macroeconomic Variables and Lagged SUEpmn
SUE pmn ,q+1 SUE pmn ,q+2 SUE pmn ,q+3 SUE pmn ,q+4
I II III IV V VI VII VIII
Panel A: Excludes lagged SUE pmn
Intercept 1.73 1.59 0.82 0.74 0.30 0.22 −1.13 −1.12
(21.11) (14.92) (10.12) (7.55) (3.79) (2.23) (−12.32) (−10.00)
INF t −4,t −2 0.59 0.50 0.35 0.36
(8.71) (8.33) (5.92) (5.17)
INF t −11,t −2 0.18 0.14 0.10 0.09
(7.32) (7.22) (5.26) (4.01)
F -test 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
(p-value)
Adj. R 2 (%) 43.50 45.11 37.81 36.35 18.12 18.37 20.85 14.31
Panel B: Includes lagged SUE pmn
Intercept 0.34 0.34 0.00 −0.02 −0.17 −0.16 −1.75 −1.89
(2.26) (2.18) (−0.03) (−0.10) (−0.63) (−0.55) (−5.79) (−5.99)
INF t −4,t−2 0.27 0.26 0.20 0.19
(4.72) (4.42) (2.57) (2.59)
INF t −11,t−2 0.07 0.06 0.05 0.01
(3.00) (3.00) (2.09) (0.23)
SUE pmn ,q 0.00 0.00 −0.01 −0.01 0.00 0.00 −0.01 −0.01
(1.34) (0.88) (−1.72) (−1.80) (−0.98) (−1.08) (−0.92) (−0.99)
SUE pmn ,q−1 0.65 0.65 0.39 0.40 0.04 0.03 0.29 0.38
(7.43) (7.13) (4.10) (3.79) (0.34) (0.24) (2.10) (2.57)
SUE pmn ,q−2 0.09 0.06 −0.01 −0.03 0.34 0.32 0.01 0.04
(0.64) (0.43) (−0.09) (−0.23) (1.93) (1.74) (0.07) (0.25)
SUE pmn ,q−3 0.01 0.05 0.29 0.34 0.07 0.10 0.16 0.26
(0.09) (0.42) (2.47) (2.92) (0.50) (0.68) (0.98) (1.53)
F -test 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
(p-value)
Adj. R 2 (%) 60.86 58.86 46.68 44.05 21.49 20.52 26.9 22.9
This table presents results from the regression of SUE for the zero-investment portfolio, PMN, on lagged
inf lation. In each month t−1, all NYSE-AMEX firms are sorted into deciles based on their most recently
available standardized unexpected earnings (SUE i,q ) as described in table 1. The SUE of individual firms
are averaged across stocks in each SUE decile, and the difference in the average SUE across the extreme
deciles is the dependent variable for the regression. The dependent variable (SUE pmn,q +j , j = 1 to 4) is
measured for the four earnings announcements (i.e., earnings announcements in quarters +1 to +4)
subsequent to month t−1. Inf lation is measured for the quarter ending in month t−2 (INF t −4 ,t −2 ) or for
the year ending in month t−2 (INF t −11,t −2 ). Panel B includes lagged SUE pmn over the past four quarters
as dependent variables. The regressions use up to 360 monthly observations over the period January 1972
through December 2001. The t-statistics are based on Newey–West standard errors to correct for overlap in
SUE pmn .
irrespective of the period over which lagged inflation is measured. Further,

the adjusted R 2 s in these regressions are as high as 45%, which points to
the importance of inflation in determining future earnings growth. The
coefficient on quarterly inflation of 0.59 in regressions of the one-quarter-
ahead SUE implies that a one standard deviation increase in inflation (0.88)
increases the one-quarter-ahead SUE pmn by 0.52. This represents an increase
of over 20% on the average quarter-ahead SUE pmn of 2.45. In summary,
inflation has a differential impact on the future earnings of the highest-
and the lowest-SUE decile portfolios.
Given the evidence in previous studies of a positive serial correlation in

SUE, we repeat the above regression after controlling for up to four lags
of SUE pmn . From panel B, we observe that the inclusion of lagged SUE pmn
reduces the coefficient estimates on inflation as well as the intercepts in
the various regressions. However, inflation continues to be statistically sig-
nificant in almost all the regressions, with the only exception being the
regression of the four-quarter-ahead SUE on inflation measured over the
prior one year. With the exception of SUE pmn,q −1 , the lagged SUE tend to be
insignificant in almost all the regressions. This is, however, contrary to exist-
ing evidence, which, unlike the current portfolio-level regressions, is based
on analysis of firm-level data. So for better comparison with prior studies,
we repeat the regressions using firm-level data in section 3.5.9
3.4 INFLATION AND FUTURE RETURNS

This section initially presents summary statistics on the returns to SUE-
sorted portfolios and then tests whether the current inflation level predicts
future abnormal returns to the PMN portfolio.
3.4.1. Summary Statistics on Payoffs to SUE Portfolios. Table 4 documents the

average monthly returns from holding the SUE portfolios for a period of six
months following the portfolio formation month. The six-month holding
period was chosen because Watts [1978] and Bernard and Thomas [1989]
document that the post-earnings-announcement drift occurs most often
within six months of the earnings announcement date. We tested the ro-
bustness of our results to holding periods of 9 and 12 months and obtained
qualitatively similar results. To avoid test statistics based on overlapping re-
turns, we follow the Jegadeesh and Titman [1993] approach in measuring
these returns. This approach measures returns in calendar time, which is
useful given our objective of evaluating the relationship between returns
and macroeconomic conditions.
Over the entire sample period from January 1972 through December
2001, the monthly holding period returns increase monotonically from
0.78% for the lowest-SUE portfolio, P 1 , to 1.70% for the highest-SUE portfo-
lio, P 10 . The difference in returns between the highest- and the lowest-SUE
portfolios, P 10 − P 1 (denoted PMN), is a statistically and economically sig-
nificant 0.92% per month, with over 74% of the months having positive
payoffs to PMN. These results are consistent with those of Foster, Olsen,
and Shevlin [1984] and Bernard and Thomas [1989], although the magni-
tude of the payoffs reported in table 4 is lower than those in prior studies.
For instance, based on an event study for the period 1974–1986, Bernard
and Thomas [1989] report a significant payoff of 4.2% on a portfolio that is
9 In regressions that consider industrial production growth (IPG), the coefficient on IPG
tends to be insignificant for the quarter-ahead and two-quarter-ahead SUE, but is significantly
positive for the three- and four-quarter-ahead SUE.
TABLE 4
Monthly Returns on SUE Portfolios
P1 P 10 PMN =
(lowest) P2 P3 P4 P5 P6 P7 P8 P9 (highest) P 10 − P 1
Jan 1972–Dec 2001 Mean (%) 0.78 1.00 1.05 1.22 1.37 1.42 1.50 1.61 1.63 1.70 0.92
t-stat. 2.46 3.25 3.40 3.91 4.49 4.79 5.11 5.53 5.62 6.00 7.31
%>0 56.7 58.3 59.4 60.6 61.7 60.3 62.5 62.2 61.1 63.3 74.2
Jan 1972–Dec 1981 Mean (%) 0.71 1.06 1.12 1.28 1.47 1.43 1.62 1.71 1.79 1.81 1.10
t-stat. 1.05 1.57 1.67 1.89 2.20 2.21 2.53 2.76 2.92 3.11 4.08
%>0 51.7 53.3 53.3 55.0 56.7 54.2 57.5 57.5 57.5 57.5 79.2
Jan 1982–Dec 1991 Mean (%) 0.72 0.94 0.94 1.15 1.32 1.50 1.47 1.64 1.63 1.74 1.02
t-stat. 1.40 1.91 1.86 2.26 2.66 3.08 3.01 3.31 3.20 3.52 6.74
%>0 55.0 57.5 58.3 62.5 60.0 61.7 64.2 63.3 60.8 64.2 75.8
Jan 1992–Dec 2001 Mean (%) 0.91 1.02 1.10 1.22 1.32 1.34 1.39 1.47 1.47 1.54 0.63
t-stat. 2.09 2.46 2.73 3.03 3.40 3.56 3.93 4.02 4.12 4.12 2.96
%>0 63.3 64.2 66.7 64.2 68.3 65.0 65.8 65.8 65.0 68.3 67.5
F -test (p-value) 0.96 0.99 0.97 0.98 0.97 0.98 0.95 0.94 0.90 0.92 0.26
In each month, t, firms are sorted into deciles based on their standardized unexpected earnings (SUE it ) from the most recent earnings announcement. In each month the
decile classifications are computed using all announcements that were made in the prior three months. SUE it is defined as (E it − E it −4 )/σit , where E it is the most recently
announced earnings, E iq −4 represents the earnings from four quarters ago, and σit is the standard deviation of (E it − E it −4 ) over the prior eight quarters. The portfolios are
held for the following six-month period. The table reports the returns to these portfolios as well as the percentage of months with positive returns. The table also reports payoffs
from a strategy of being long on the highest-SUE portfolio (P 10 ) and short on the lowest-SUE portfolio (P 1 ). The p-value from the F -test for test of equality of payoffs across sub-
periods is presented in the last row. The sample consists of 360 months between January 1972 and December 2001. On average, each SUE portfolio consists of approximately 182 stocks.
INFLATION ILLUSION
533
long on P 10 and short on P 1 in the 60 event-days subsequent to an earnings

announcement.10
We also conduct a subperiod analysis for the periods January 1972 through
December 1981, January 1982 through December 1991, and January 1992
through December 2001. In each of the subperiods, the difference in
monthly holding period returns between the highest- and the lowest-SUE
portfolios is economically and statistically significant, and we are unable to
reject the null hypothesis that the P 10 − P 1 returns are the same across the
subperiods. In other words, the post-earnings-announcement drift is robust
over the entire sample as well as across each of the subperiods.
In order to assess whether the payoffs to PMN are explained by commonly
used factors or are subsumed by other anomalies, we regress PMN on the
market factor as well as on the Fama and French [1993] factors.11 To con-
serve space, we discuss these results without tabulating them. Regressing
the payoffs to PMN on the value-weighted market return alone results in an
insignificant coefficient on the market return and an adjusted R 2 of about
zero. This suggests that the PMN portfolio is a zero-beta portfolio relative to
the market. Regressing PMN on market, size, and book-to-market factors in-
creases the adjusted R 2 to 25%. The coefficients on size and book-to-market
factors are significantly negative. The significant relationships between PMN
and the size and book-to-market factors point to the need to control for these
factors in our analyses. However, the intercept is positive and statistically sig-
nificant in each case. The minimum intercept across all regressions is 0.95,
which suggests that the Fama and French [1993] factors do not subsume
the PMN portfolio returns.
3.4.2. Predictability of Future Returns to PMN . To test whether investors react

with a delay to the information in inflation as implied by the inflation illusion
hypothesis, we examine the relationship between lagged inflation and the
returns to PMN. Results from a univariate analysis are presented in panel
A of table 5. All sample months are sorted into low-, medium-, and high-
inflation groups based on inflation from two months prior. The two-month
lag is used to ensure that the inflation information is publicly announced
10 For better comparison with Bernard and Thomas [1989], we restrict our portfolios to
include only stocks with earnings announcements in the formation month, and examine payoffs
in the following three-month period. Considering only firms that made an announcement in
the formation month substantially reduces the average number of stocks in each portfolio, with
the extreme portfolios containing fewer than 20 stocks in about 15% of the sample months.
The average payoffs to PMN in this case are 1.4% per month for the full sample and 1.6% per
month for the 1974–1986 sample, which are comparable to the magnitude of the drift reported
in prior studies.
11 The Fama and French [1993] factors are: (1) return on the value-weighted market portfo-
lio, MKT; (2) returns on a zero-investment portfolio related to size (SMB, for small minus big);
and (3) returns on a zero-investment portfolio related to book-to-market ratio (HML, for high
minus low). The returns on the Fama and French [1993] factors are obtained from Professor
Kenneth French’s Web site (mba.tuck.dartmouth.edu/pages/faculty/ken.french/).
before the sample month in which the PMN payoff is measured. Panel A
reports the raw returns and the Fama–French-adjusted payoffs to PMN for
each of these inflation groups. Fama–French-adjusted returns are the alphas
from the regression of the time series of the PMN returns on the Fama and
French [1993] factors plus the residuals averaged across months in each
inflation group.
From panel A of table 5, it is clear that the raw returns as well as the Fama–
French-adjusted returns to PMN increase monotonically across the inflation
groups. For instance, the raw return to PMN is only 0.48% per month fol-
lowing low-inflation months, whereas it is 1.31% per month following high-
inflation months. The corresponding Fama–French-adjusted returns are
0.71% and 1.59%, respectively. The raw return difference between the high
and the low inflationary periods is a statistically and economically significant
0.83% per month.
Panels B and C of table 5 report results from regression analyses that
examine the relationship between lagged inflation and the returns to PMN.
These regressions use overlapping returns measured over 3- to 12-month
periods following the portfolio formation month, and include the Fama
and French [1993] factors as control variables. To account for the overlap
in returns, t-statistics are computed using the Newey–West standard errors.
TABLE 5
Relationship between Payoffs to PMN t and Lagged Inflation
Panel A: Nonoverlapping returns to PMN in months sorted by lagged inf lation
Low Medium High
Inflation Inflation Inflation High – Low
INF t −2,t−2 (%) Mean 0.09 0.34 0.80 0.71
PMN t (%) Mean 0.48 0.95 1.31 0.83
t-stat. (2.38) (5.58) (4.89) (2.45)
Fama–French-adjusted Mean 0.71 1.02 1.59 0.88
returns (%) t-stat. (3.97) (6.51) (7.37) (3.13)
Panel B: Overlapping returns to PMN measured up to one year after formation month
Return-Measurement Period
3 months 6 months 9 months 12 months
Intercept 2.83 2.39 4.53 4.26 5.70 5.85 6.10 7.16
(4.91) (3.88) (4.38) (3.83) (3.74) (3.40) (2.99) (2.97)
INF t −4,t −2 0.98 1.74 1.94 2.18
(2.90) (3.37) (2.72) (2.11)
INF t −11,t −2 0.34 0.51 0.49 0.37
(3.67) (3.20) (1.96) (0.96)
MKT −0.07 −0.07 −0.13 −0.14 −0.12 −0.14 −0.14 −0.16
(−1.62) (−1.61) (−2.58) (−2.84) (−2.15) (−2.47) (−1.93) (−2.25)
SMB −0.29 −0.30 −0.27 −0.28 −0.28 −0.28 −0.30 −0.28
(−4.54) (−4.88) (−4.19) (−4.37) (−3.84) (−3.78) (−3.73) (−3.32)
HML −0.18 −0.19 −0.10 −0.12 −0.10 −0.11 −0.07 −0.08
(−2.73) (−3.00) (−1.44) (−1.70) (−1.19) (−1.39) (−0.76) (−0.87)
F -test (p-value) 0.00 (0.00) 0.00 (0.00) 0.00 (0.00) 0.00 (0.00)
Adj. R 2 (%) 20.43 21.82 22.78 22.59 22.35 20.93 25.46 22.34
T A B L E 5 — Continued
Panel C: Overlapping returns to PMN measured up to one year after formation month
Return-Measurement Period
3 months 6 months 9 months 12 months
Intercept 3.71 4.39 8.82 9.37 13.33 13.99 16.46 16.75
(2.47) (2.95) (4.37) (4.77) (4.59) (4.94) (3.84) (4.08)
INF t −4,t −2 1.17 2.41 3.51 4.42
(2.50) (3.41) (3.74) (4.56)
INF t −11,t −2 0.52 0.82 1.07 1.05
(3.70) (3.52) (3.40) (2.79)
MKT −0.06 −0.05 −0.11 −0.12 −0.11 −0.14 −0.13 −0.17
(−1.38) (−1.29) (−2.52) (−2.83) (−2.26) (−2.81) (−2.18) (−2.72)
SMB −0.29 −0.30 −0.25 −0.27 −0.23 −0.24 −0.24 −0.24
(−3.97) (−4.37) (−3.71) (−4.00) (−3.17) (−3.34) (−2.96) (−2.83)
HML −0.16 −0.17 −0.06 −0.08 −0.06 −0.09 −0.04 −0.08
(−2.40) (−2.64) (−0.84) (−1.18) (−0.77) (−1.22) (−0.57) (−0.98)
SUE pmn ,q 0.03 0.03 −0.08 −0.08 −0.12 −0.12 −0.14 −0.16
(0.82) (0.83) (−2.01) (−1.95) (−3.06) (−2.89) (−2.24) (−2.35)
SUE pmn ,q−1 −1.17 −1.64 −2.63 −2.98 −3.91 −4.16 −3.99 −3.79
(−1.70) (−2.36) (−2.89) (−3.27) (−2.60) (−2.85) (−2.18) (−2.08)
SUE pmn ,q−2 0.40 0.01 0.46 −0.04 0.94 0.28 −1.01 −1.36
(0.49) (0.02) (0.46) (−0.04) (0.75) (0.21) (−0.68) (−0.85)
SUE pmn ,q−3 0.93 0.81 1.20 1.33 −0.97 −0.51 −1.34 −0.42
(0.96) (0.85) (0.89) (1.00) (−0.56) (−0.30) (−0.63) (−0.19)
F -test (p-value) 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Adj. R 2 (%) 20.31 23.12 24.94 25.85 28.5 27.43 33.22 28.19
The standardized unexpected earnings (SUE) are defined in table 1. PMN is the zero-investment
portfolio that is long on the highest-SUE decile portfolio and short on the lowest-SUE decile portfolio.
MKT , SMB, and HML are the Fama and French [1993] factors corresponding respectively to market, size
and book-to-market ratios. Panel A presents summary statistics on raw returns and Fama–French-adjusted
returns of the PMN portfolio for calendar months classified into three groups (low, medium, and high
inf lation), based on lagged inf lation. The PMN returns for this panel are measured as nonoverlapping
monthly returns using the Jegadeesh and Titman [1993] approach. The Fama–French-adjusted returns
for any given month are obtained as the intercept from the regression of the payoffs to PMN on the Fama
and French [1993] factors over the entire sample period plus the residuals averaged across months in each
inflation group. Panels B and C present the coefficient estimates from regression of PMN returns on Fama
and French [1993] factors and lagged inf lation, which is measured either for the quarter ending in month
t−2 (INF t −4 ,t −2 ) or for the year ending in month t − 2 (INF t −11, t −2 ). Inf lation is lagged by two months
relative to the month from which PMN payoffs are measured to ensure that these variables are announced
by the portfolio formation month, t. The dependent variables and the Fama and French [1993] factors in
the regressions are measured as overlapping returns over intervals ranging from 3 to 12 months following
the portfolio-formation month, and t-statistics are corrected for the overlap in returns using Newey–West
standard errors. Panel C includes lagged SUE of the PMN portfolio as additional control variables. The
SUE for the PMN portfolio are computed as the difference in the average SUE of portfolios P 10 and P 1 .
The analyses use up to 360 observations over the period January 1972 through December 2001.
Our conclusions are unchanged if nonoverlapping returns (computed as in

Jegadeesh and Titman [1993]) are used instead in the analyses.
Panel B of table 5 shows that, regardless of the measurement interval for
inflation and for returns, there is a strong positive relationship between
the PMN returns and lagged inflation. For instance, in regressions of three-
month-ahead returns, the coefficient on the prior quarter’s inflation is 0.98
(t = 2.90). This coefficient suggests that a one standard deviation increase in
quarterly inflation of 0.88 would increase the payoffs to PMN by an econom-
ically significant 0.86% in the three months following the formation month.
This predictive ability of inflation is consistent with investors underreacting

to the information in inflation.12 The intercepts in these regressions are
always significantly positive, suggesting that underreaction to inflation does
not fully explain the drift anomaly.
Panel C of table 5 presents results from regressions that include lagged
SUE. Surprisingly, the coefficients on inflation in this panel tend to be
comparable to or even larger than those in panel B when lagged SUE were
not included as independent variables. The coefficients on lagged SUE,
although generally insignificant, are sometimes significantly negative in a
few of the regressions. This could arise because of our focus on portfolio-
level regressions, as opposed to firm-level regressions estimated in prior
studies. We revisit this issue in section 3.5.
Inflation levels were relatively high in the 1970s and decreased signif-
icantly in the 1980s and 1990s. Similarly, payoffs to the drift strategy are
known to be lower in the 1980s and 1990s than in the 1970s. Although these
patterns in inflation and post-earnings-announcement drift are consistent
with the inflation illusion hypothesis, we check whether exogenous time-
period effects cause a spurious relationship between inflation and PMN
returns. We divide our sample period into three roughly equal subperiods
(decades) and repeat the regressions separately for each subperiod. The
results from this analysis are presented in Table 6.
The results of the subperiod analysis are qualitatively similar to those pre-
sented in table 5 for the entire sample period. Inflation significantly predicts
PMN payoffs in all subperiods, although due to fewer observations in this
analysis relative to those in table 5, the magnitudes of the t-statistics tend
to be smaller. The coefficient on annual inflation in the 1980s subperiod is
the only one that is insignificant. The coefficients on the control variables
are almost always insignificant in the 1990s subperiod, but are qualitatively
similar to those in table 5 for the earlier two subperiods. These results in
general and, the results for the 1990s subperiod, which corresponds to a
low and stable inflationary period, in particular, suggest that the predictive
ability of inflation remains even when no period-specific effects exist for
inflation or for the post-earnings-announcement drift.
Overall, the ability of inflation to predict PMN payoffs is consistent with
the view that PMN reflects information in current inflation with a delay.
This delayed reaction occurs either because investors underestimate fu-
ture inflation or, as implied by the inflation illusion hypothesis, because
they underestimate the magnitude of the impact of current inflation on
12 To test whether the predictive ability of inflation is due to “inflation shocks” or, as im-
plied by the inf lation illusion hypothesis, “inflation levels,” we split current inflation into
“expected inf lation level” and “inflation shocks” using the Money Market Services survey data
on inflation. Only “expected inf lation level” is significant in this regression, confirming that
investors underreact to earnings information in inflation levels and not to “inflation shocks.”
In contrast to the predictive ability of inflation, lagged industrial production growth does not
have significant predictive power for PMN payoffs. This is consistent with our results in table 2
that earnings of SUE-sorted portfolios are related primarily to inflation and not to real output.
TABLE 6
Subperiod Analysis
Jul 1972–Dec 1981 Jan 1982–Dec 1991 Jan 1992–Dec 2001

Intercept 46.05 52.75 29.71 35.24 4.84 −11.16
(9.61) (10.14) (3.02) (3.57) (0.26) (−0.51)
INF t −4,t−2 3.48 2.78 5.58
(3.50) (1.94) (2.44)
INF t −11,t−2 1.51 1.19 3.45
(4.73) (1.42) (2.22)
MKT −0.15 −0.18 −0.08 −0.10 0.11 0.05
(−3.55) (−5.16) (−1.09) (−1.35) (1.36) (0.60)
SMB −0.28 −0.32 −0.53 −0.52 −0.10 −0.16
(−5.76) (−6.48) (−3.61) (−3.43) (−1.14) (−1.55)
HML −0.45 −0.46 −0.14 −0.13 0.16 0.05
(−5.81) (−7.38) (−1.19) (−1.08) (3.51) (0.78)
SUE pmn ,q −5.24 −7.78 −2.43 −3.11 0.62 1.33
(−3.58) (−4.34) (−3.33) (−4.04) (0.45) (0.89)
SUE pmn ,q−1 −0.18 −0.70 −4.19 −5.01 −2.90 −0.91
(−0.09) (−0.39) (−1.53) (−1.70) (−0.50) (−0.17)
SUE pmn ,q−2 −10.88 −7.68 −6.26 −7.56 −0.20 −1.41
(−3.01) (−2.30) (−1.57) (−1.93) (−0.04) (−0.28)
SUE pmn ,q−3 6.75 6.57 6.49 3.76 −7.44 −6.88
(2.34) (2.17) (1.57) (0.94) (−1.71) (−1.62)
F -test (p-value) 0.00 0.00 0.00 0.00 0.00 0.00
Adj. R 2 (%) 72.25 73.48 65.92 65.49 25.42 25.84
The table presents results from the regression of PMN payoffs on lagged inf lation, Fama and French
[1993] factors (MKT , SMB, and HML) and lagged standardized unexpected earnings (SUE), estimated
separately for three subperiods. The subperiods are July 1972 to December 1981, January 1982 to December
1991 and January 1992 to December 2001. PMN is the zero-investment portfolio that is long on the
highest-SUE decile portfolio and short on the lowest-SUE decile portfolio. Lagged inf lation is measured
either for the quarter ending in month t−2 (INF t −4 ,t −2 ) or for the year ending in month t−2 (INF t −11, t −2 ).
Inf lation is lagged by two months relative to the month from which the PMN payoffs are measured to
ensure that these variables are announced by the portfolio formation month, t. The dependent variables
and the Fama and French [1993] factors in the regressions are measured as overlapping returns over the
12 months subsequent to the portfolio-formation month, and t-statistics are corrected for the overlap in
returns using Newey–West standard errors. The SUE for the PMN portfolio are computed as the difference
in the average SUE of portfolios P 10 and P 1 .
future earnings. To mathematically illustrate the two alternatives, consider

the equations below for expected SUE for firm i (SUE it +1 ) and expected
inflation, INF t +1 , conditional on current inflation, INF t .
E [SUE it+1 | INF t ] = αi0 + αi1 E [INF t+1 | INF t ] (3)
E [INF t+1 | INF t ] = δ0 + δ1 INF t (4)

Using equation (4) to rewrite equation (3), we get:
E [SUE it+1 | INF t ] = αi0 +αi1 [δ0 +δ1 INF t ] = (αi0 +αi1 δ0 )+αi1 δ1 INF t (5)
From equation (5), we see that current inflation could predict future SUE
and future payoffs to PMN if investors either underestimate the impact of
current inflation on future earnings (underestimate α i1 ) or underestimate
the impact of current inflation on future inflation (underestimate δ 1 ).
Underestimation of future inflation on a consistent basis is not likely to be

an explanation. First, numerous studies in the macroeconomics literature
have shown that inflation expectations of economic agents are unbiased and
consistent with rationality (e.g., Rich [1989], Aggarwal, Mohanty, and Song
[1995], Grant and Thomas [2001]). Second, in untabulated results, we find
that financial variables that reflect future inflation, such as default spread
and yield on the three-month T-bill, also predict future payoffs to PMN.
If investors underestimate future inflation, then these financial variables
would also be affected and they would not predict PMN payoffs. Third,
underestimation of future inflation would affect market returns as well as
the returns on other portfolios, such as the PMN. This would result in PMN
and market returns being correlated. However, we do not find any evidence
of this.
These facts are, however, consistent with the inflation illusion hypothe-
sis that investors underestimate the impact of current inflation on future
earnings. For instance, the finding that lagged macroeconomic variables
predict payoffs to a portfolio that is essentially sorted on earnings sensitiv-
ity to inflation provides support for our contention that the delayed price
reaction has more to do with underestimation of earnings information in
inflation than with underestimation of inflation itself. In contrast, the ef-
fect of such underestimation on the market returns would be attenuated, as
the market portfolio includes firms with both positive and negative earnings
sensitivities. The market (PMN) portfolio should be a poor (good) proxy for
capturing mispricing of earnings information in inflation. Finally, although
Aggarwal, Mohanty, and Song [1995] show that forecasts of inflation are
consistent with rationality, they also show that forecasts of real output and
retail sales are not. Overall, the results support the argument that investors
misestimate the impact of current inflation on future earnings and returns,
and that this misestimation partly causes the post-earnings-announcement
drift.
3.5 INDIVIDUAL STOCKS

Thus far, our analysis is based on portfolio-level data, whereas most prior
studies of the post-earnings-announcement drift are based at the level of
the firm (e.g., Bernard and Thomas [1990], Ball and Bartov [1996], Soffer
and Lys [1999]). Although, compared with firm-level analyses, portfolio-
level analyses potentially reduce noise in regressions and do not suffer from
cross-correlation problems, Kothari, Lewellen, and Warner [2005] show that
results from analyses of portfolio-level SUE differ substantially from those
based on firm-level SUE. Further, Lo and MacKinlay [1990] argue that ana-
lyzing portfolios based on nonrandomly chosen characteristics could induce
data-snooping biases. Hence, in this section, we examine the sensitivity of
our prior conclusions to estimating regressions using firm-level data. Such
tests also allow us to better compare our results with those of prior studies.
To control for data errors and outliers, these regressions exclude the ex-
treme 1% of the SUE on either side.13
To test the conjecture that firms sorted on SUE vary systematically in their
exposure to inflation, we estimate a pooled regression of SUE i,q on inflation
measured in either the year or the quarter prior to the announcement
month for quarter q earnings. This regression is estimated separately for
SUE portfolios, formed by sorting stocks on SUE i,q −1 . By using lagged SUE to
form portfolios, we avoid the problem of estimating regressions on portfolios
sorted by the dependent variable, as in our analysis in table 2.
The regression results, reported in panel A of table 7, document an almost
monotonic relationship between SUE i,q and inflation that is qualitatively
similar to that reported in table 2. The coefficient on inflation, measured
over either a quarter or a year, is significantly negative for the lowest-SUE
portfolio, whereas it is significantly positive for the highest-SUE portfolio.
Interestingly, the magnitudes of the coefficients for all the portfolios are
comparable to those reported in table 2, indicating that the relationship be-
tween inflation and SUE is consistent across analyses based on portfolio-level
data and on firm-level data. For instance, the quarterly inflation coefficient
of portfolio P 1 is −0.10, and for portfolio P 10 it is 0.42, which are compa-
rable with the −0.09 and 0.39 reported for these portfolios, respectively, in
table 2.
We next analyze the impact of inflation on a firm’s future earnings by
estimating the following regression using firm-level data:

3
SUE i,q +1 = γ0 + γ1 INF q −k,q + γ2 j SUE i,q − j + νi,q +1 , (6)
j =0
where:
q = quarter in which SUE i,q is measured and
INFq −k,q = inflation measured over the quarter (k = 0) or year (k = 3)
ending one month before the earnings announcement month
for quarter q.
The regressions are estimated across all firms as well as for firms sorted
into portfolios based on SUE i,q .
The first regression of table 7 (panel B) replicates prior studies. In this
regression, the coefficient on inflation is restricted to be zero. Consistent
with the findings of Bernard and Thomas [1990] and Ball and Bartov [1996],
the coefficients on lagged SUE follow the expected sign, positive for the
first three lags and negative for the fourth lag. The coefficients on the first
to fourth lag of SUE are 0.23, 0.19, 0.09, and −0.18, respectively. These
coefficients are comparable to the mean autocorrelation patterns for SUE
reported in Bernard and Thomas [1990, table 1], which are 0.34, 0.19, 0.06,
and −0.24, respectively, for the first four lags.
13 The qualitative results are unaffected by whether or not the extreme values of SUE are
deleted.
TABLE 7
Firm-Level Regressions
Panel A: Regression of SUE i,q +1 on contemporaneous macroeconomic variables
Independent
Variable P1 P2 P3 P4 P5 P6 P7 P8 P9 P 10 F -test
INF q +1,q+1 Coef. −0.103 −0.006 0.007 0.028 0.021 0.051 0.078 0.142 0.185 0.421 0.00
t-stat. −8.52 −0.51 0.70 3.16 2.59 6.41 8.95 14.94 16.68 26.46
INF q −2,q+1 Coef. −0.025 −0.001 0.003 0.006 0.005 0.015 0.021 0.043 0.058 0.126 0.00
t-stat. −6.94 −0.40 1.17 2.16 1.96 6.52 8.47 15.58 17.99 27.19
Panel B: Regression of SUE i,q +1 on lagged inf lation

All Firms P1 P5 P 10 F -test P1 P5 P 10 F -test P1 P5 P 10 F -test P1 P5 P 10 F -test
Regression (1) (2) (3) (4) (5) (6) (7) (8) (9)
Intercept 0.051 −0.028 −0.951 0.010 0.001 0.00 −0.018 −0.936 0.015 0.002 0.00
(15.95) (−5.07) (−35.55) (0.74) (0.03) (−2.91) (−32.51) (1.04) (0.05)
INF q ,q −0.259 0.029 0.511 0.00 −0.079 −0.013 0.171 0.00
(−33.78) (4.15) (62.59) (−6.21) (−1.57) (10.62)
INF q −3,q −0.077 0.004 0.139 0.00 −0.023 −0.004 0.039 0.00
(−36.06) (2.34) (61.81) (−6.07) (−1.83) (8.09)
SUE i ,q 0.231 0.127 −0.028 0.982 0.159 0.00 0.114 −0.028 0.972 0.163 0.00
(134.43) (58.82) (−7.35) (22.13) (16.08) (50.28) (−7.51) (22.77) (16.49)
SUE i ,q−1 0.188 0.149 0.119 0.021 0.309 0.00 0.146 0.118 0.021 0.310 0.00
(76.93) (61.40) (16.12) (4.04) (28.28) (59.97) (16.04) (4.05) (28.40)
SUE i ,q−2 0.091 0.076 0.032 0.024 0.157 0.00 0.075 0.032 0.024 0.158 0.00
(36.17) (30.62) (3.87) (4.65) (12.65) (30.26) (3.87) (4.67) (12.71)
SUE i ,q−3 −0.181 −0.184 −0.212 −0.199 −0.022 0.00 −0.183 −0.212 −0.199 −0.019 0.00
(−74.74) (−77.25) (−24.82) (−39.70) (−1.83) (−76.73) (−24.84) (−39.71) (−1.63)
Adj. R 2 (%) 19.20 22.44 5.13 9.91 17.03 22.51 5.13 9.91 16.82
No. of obs. 182,251 182,251 18,104 18,488 18,612 182,251 18,104 18,488 18,612
Panel C: Regression of RET i,q +1 on lagged inf lation

INFLATION ILLUSION

Regression (1) (2) (3) (4) (5) (6) (7) (8) (9)
Intercept −0.677 −0.938 −2.448 −1.131 0.206 0.00 −1.147 −2.277 −1.256 −0.219 0.00
541
(−15.13) (−12.85) (−8.21) (−4.60) (0.67) (−14.24) (−7.13) (−4.74) (−0.68)

542
Panel C: Regression of RET i,q +1 on lagged inf lation
INF q ,q −0.661 0.089 1.159 0.00 −0.352 0.149 0.568 0.00
(−6.95) (1.03) (11.47) (−2.51) (0.98) (4.01)
INF q −3,q −0.183 0.058 0.398 0.00 −0.123 0.062 0.251 0.00
(−6.93) (2.50) (14.30) (−2.93) (1.45) (5.91)
SUE i ,q 0.480 0.205 −0.063 0.644 0.324 0.00 0.134 −0.064 0.729 0.292 0.00
(22.95) (7.66) (−1.56) (0.81) (3.80) (4.75) (−1.59) (0.96) (3.43)
SUE i ,q−1 0.124 0.014 0.167 −0.208 0.038 0.08 −0.006 0.164 −0.208 0.028 0.09
(4.18) (0.46) (2.11) (−2.27) (0.40) (−0.19) (2.07) (−2.27) (0.29)
SUE i ,q−2 0.088 0.045 0.233 −0.032 −0.069 0.02 0.037 0.234 −0.033 −0.078 0.02
(2.86) (1.46) (2.60) (−0.35) (−0.64) (1.20) (2.61) (−0.35) (−0.72)
SUE i ,q−3 −0.071 −0.072 0.004 0.100 −0.233 0.27 −0.074 0.006 0.100 −0.246 0.23
(−2.42) (−2.44) (0.04) (1.13) (−2.25) (−2.51) (0.07) (1.13) (−2.38)
MKT q +1 0.985 0.987 1.000 0.978 0.992 0.41 0.987 0.999 0.979 0.995 0.38
(169.43) (169.43) (53.49) (51.37) (54.75) (169.58) (53.48) (51.36) (54.95)
T. CHORDIA AND L. SHIVAKUMAR
SMB q +1 0.605 0.603 0.664 0.660 0.527 0.00 0.597 0.671 0.656 0.510 0.00
(80.56) (79.48) (26.58) (26.27) (22.54) (77.81) (26.49) (25.82) (21.51)
HML q +1 0.456 0.457 0.526 0.472 0.380 0.00 0.455 0.530 0.470 0.374 0.00
(68.54) (68.67) (24.56) (21.13) (18.75) (68.45) (24.76) (21.08) (18.49)
Adj. R 2 (%) 21.22 21.41 21.60 20.30 21.40 21.45 21.61 20.30 21.48
No. of obs. 181,752 181,752 18,023 18,431 18,515 181,752 18,023 18,431 18,515
The table presents results from the following regression:
3

VAR i,q +1 = γ0 + γ1 INF q −k,q + γ2 j SUE i,q − j + νi,q +1
j =0
where VAR i,q +1 is either SUE i,q +1 or RET i,q +1 . SUE i,q is the standardized unexpected earnings of firm i, based on earnings announced in quarter q. The SUE portfolios are formed by sorting stocks on
SUE i,q . RET i,q +1 is the stock returns in the three months following the earnings announcement month of quarter q. Pooled regressions are estimated for each SUE portfolio as well as for all firms. INF q − k,q
is inf lation measured over the past quarter (k = 0) or over the past year (k = 3). Panel A reports results from the regression of SUE i,q +1 on contemporaneous inf lation, i.e., in quarter q+1 or over the year
ending in q+1. Panel B (panel C) reports results from the regression of SUE i,q +1 (RET i,q +1 ) on lagged inf lation and lagged SUE. In Panel C, the Fama–French factors (MKT , SMB, and HML) are also
used as independent variables. In all panels, inf lation is lagged by one month relative to earnings announcements for quarter q to ensure that inf lation is announced prior to the portfolio formation
month. Regressions (2) and (6) of panels B and C report results from a regression that holds the coefficient on all variables, other than inf lation, constant across the SUE portfolios. This is implemented
by estimating the regression across all firms after interacting inf lation with dummy variables for each SUE portfolio. Regressions exclude the extreme 1% of SUE i,q + k (k = −3 to +1) on either side. The
column titled “F -test” gives the p-value for the F -test that the coefficient is the same across all SUE portfolios. The sample period is January 1972 through December 2001.
As our conjecture implies a differential impact of inflation on future earn-

ings growth across SUE-sorted stocks, we estimate regression equation (6)
separately for stocks in the different SUE portfolios. For brevity, we present
results only for portfolios P 1 , P 5 , and P 10 . In regression (2) of panel B,
table 7, we keep the coefficients on lagged SUE constant across the differ-
ent portfolios, as done in prior studies, while allowing the coefficient on
inflation to vary across the portfolios, as implied by our hypothesis. Lagged
inflation significantly predicts the quarter-ahead SUE i,q +1 for all 10 portfo-
lios. The coefficient on INFq,q is −0.26 for the P 1 portfolio, which mono-
tonically increases to 0.51 for portfolio P 10 . The null hypothesis that the
coefficients are the same across all portfolios is rejected at a less than 1%
significance level. The coefficients on the first three lags of SUE continue to
be significant in this regression, although their magnitudes are lower than
in regression (1).
Regressions (3) through (5) present the results when lagged SUE as well
as inflation are allowed to vary across the SUE portfolios. The coefficient
on inflation is qualitatively similar to those in regression (2), with lagged
inflation negatively (positively) impacting the future SUE of the P 1 (P 10 )
portfolio. These differences in coefficients across the portfolios suggest that
a portfolio that is long on stocks in P 10 and short on stocks in P 1 , such as
PMN, would have a positive coefficient of 0.25 [=0.17 − (−0.08)] on the
prior quarter’s inflation. This is comparable to the coefficient of 0.27 that
we obtain from the portfolio-level regressions in panel B of table 3. The
coefficients on lagged SUE follow the same pattern as before for portfolios
P 5 and P 10 , with the coefficients being positive for the first three lags and
negative for the fourth lag of SUE. However, for stocks in the lowest-SUE
portfolio, P 1 , the coefficient on the first lag of SUE is also negative. In these
regressions, the null hypothesis that the coefficients are the same across
portfolios is rejected for all coefficients including the intercept. The mono-
tonic increase in intercepts across the portfolios suggests that lagged SUE
and lagged inflation are insufficient to fully explain systematic differences
in quarter-ahead SUE across the portfolios. Regressions (6) through (9) re-
peat the above analyses using annual inflation instead of quarterly inflation
and obtain results that are qualitatively identical to those of regressions (3)
to (5).
Panel C of table 7 reports results from estimating equation (6), after re-
placing SUE i,q +1 with quarterly return, RET i,q +1 , as the dependent variable.
The Fama–French factors are also included as independent variables in this
regression. The returns in these regressions are measured over the three
months following the month in which quarter q’s earnings (i.e., earnings
used to form SUE portfolios) are announced. Regression (1) presents the
results from estimating the regression across all stocks. The coefficients on
lagged SUE are 0.48, 0.12, 0.09, and −0.07 for the first to fourth lag of SUE,
respectively. This pattern is the same as that obtained in panel B for regres-
sions of SUE, and supports the arguments of Bernard and Thomas [1990]
and Ball and Bartov [1996] that investors underreact to the autocorrelation
patterns in SUE, causing the post-earnings-announcement drift.
In regression (2) we introduce INF q,q as an independent variable that

is allowed to vary across the SUE portfolios. The coefficient on inflation
is significant for almost all the portfolios, and varies systematically across
the portfolios in the same manner as the corresponding coefficient in the
regression of SUE i,q +1 in panel A. This supports the argument that the post-
earnings-announcement drift arises partly from investors underreacting to
the impact of inflation on future earnings. Moreover, the differences in
inflation coefficients across SUE portfolios suggest that a portfolio that
is long on stocks in P 10 and short on stocks in P 1 , such as PMN, would
have a positive coefficient of 0.92 [=0.57 − (−0.35)] on the prior quarter’s
inflation. This is comparable to the coefficient of 0.98 (1.17) obtained from
the portfolio-level regressions in table 5, panel B (panel C). With regard to
the coefficients on lagged SUE, the coefficient on the first lag of SUE de-
clines by over 55% to 0.21, and the coefficients on the second and third lags
of SUE are no longer distinguishable from zero. These results indicate that
the results on underreaction to lagged SUE documented by Bernard and
Thomas [1990] and Ball and Bartov [1996] are weakened when inflation is
used as an independent variable.
In regressions (3) through (5), all the independent variables are allowed
to vary across the SUE portfolios. The qualitative results for the coefficient
on inflation are unchanged relative to those in regression (2). However,
the predictive ability of SUE documented in regression (1) is significantly
weakened in these regressions. Similar to the regressions of SUE i,q +1 , no par-
ticular pattern is consistently observed for these coefficients. The intercept
increases monotonically from −2.44 for portfolio P 1 to 0.21 for portfolio
P 10 . The null hypothesis that the intercepts are the same across portfo-
lios is rejected at a less than 1% level, indicating that lagged inflation and
lagged SUE are insufficient to fully explain the return differential across
SUE portfolios. These conclusions are consistent with those obtained from
portfolio-level regressions.
Regressions (6) through (9) report results from analyses based on annual
inflation. These results are qualitatively similar to those reported for quar-
terly inflation. For instance, the coefficient on inflation is significant for
almost all the portfolios and increases monotonically across the portfolios.
The coefficients of −0.18 for portfolio P 1 and 0.39 for portfolio P 10 im-
ply that a one standard deviation increase in annual inflation of 3.0 decreases
the three-month-ahead returns for portfolio P 1 by −0.5% while increasing
the three-month-ahead returns for portfolio P 10 by 1.2%. The intercepts also
vary monotonically across SUE portfolios, implying a three-month return dif-
ferential of 2.06% [=−0.22 − (−2.28)] across extreme SUE portfolios, after
controlling for the impact of lagged SUE and lagged inflation. The corre-
sponding figure in untabulated regressions that exclude inflation is 3.65%,
suggesting that over 40% of the return differential in PMN is explained by
lagged inflation.
In summary, the results based on firm-level regressions confirm our ear-
lier portfolio-level regressions. Unlike the return predictive ability for SUE iq ,
which Kothari, Lewellen, and Warner [2005] show to vary between firm-level
analyses and portfolio-level analyses, the predictive ability of inflation is ro-
bust to both firm-level and portfolio-level regressions. Moreover, we find
that lagged inflation continues to be statistically significant, even after con-
trolling for lagged SUE. Finally, we have also extended the above analyses
of SUE i,q +1 and RET i,q +1 to two- to four-quarter-ahead SUE i and returns.
The results from these analyses are very similar to those reported earlier
for portfolio-level data. Lagged inflation does have an impact on the future
returns of the SUE-sorted portfolios, even after controlling for the informa-
tion in lagged SUE.
4. Risk or Mispricing?
Although our results thus far are consistent with the conjecture that in-
vestors underreact to the impact of inflation on future earnings, in this
section, we consider alternative explanations for our findings. In particular,
we consider a risk-based explanation for our results, and conduct tests to
distinguish this explanation from the inflation illusion hypothesis.
4.1 INFLATION RISK FACTOR

Our finding of a monotonic variation in earnings sensitivity to inflation
across SUE portfolios raises the possibility that post-earnings-announcement
drift merely reflects compensation for inflation risk. However, Bernard and
Thomas [1989] find that the Chen, Roll, and Ross [1986] factors, including
unexpected inflation, do not explain the return predictive ability of SUE. In
particular, they show that the post-earnings-announcement drift is unrelated
to the current month’s inflation shock.14 However, since Vassalou [2003]
finds that the Fama and French book-to-market factor is mainly a risk factor
relating to news about future GDP growth, it is possible that shock to future
inflation rather than current inflation is the appropriate risk factor for SUE
portfolios. Hence we conduct asset-pricing tests to check whether a factor
relating to news about future inflation can account for the return predictive
ability of SUE in the cross section.
To construct a risk factor that reflects news about future inflation, we
closely follow the approach of Lamont [2001] and create an inflation-news
mimicking portfolio. The mimicking portfolio is created by regressing fu-
ture macroeconomic variables, in our case the month-ahead inflation, on
the returns to a set of base assets, an intercept, and a set of lagged predictor
variables.15 The base assets are the CRSP value-weighted market returns in
excess of the risk-free rate, excess returns on the eight industry portfolios
14 We replicate the results of Bernard and Thomas [1989] using Lamont’s [2001] mimicking
portfolio approach to capture the current month’s inflation shock.

15 Our conclusions remain unchanged when using the quarter-ahead and year-ahead
inflation variables in these tests.

identified in Lamont [2001],16 and excess returns on a long-term govern-

ment bond, an intermediate-term government bond, a one-year government
bond, a term structure bond portfolio, and a default bond portfolio.17 The
predictor variables are chosen to control for predictability in either the re-
turns to the base assets or the dependent variable. The predictor variables
are the term-structure premium, default bond premium, yield on a 3-month
T-bill, dividend yield on a CRSP value-weighted market portfolio, and prior
12-month inflation. The mimicking portfolio is then given as the predicted
component captured by the base assets returns.
We use the Brennan, Chordia, and Subrahmanyam [1998] methodology
in our cross-sectional asset-pricing tests. This methodology is very similar to
the two-stage Fama and Macbeth [1973] asset-pricing tests with the excep-
tion that the dependent variable in the second-stage cross-sectional regres-
sion is the factor-adjusted returns instead of excess returns. This modifica-
tion avoids the error-in-variables bias inherent in the Fama and Macbeth
asset-pricing tests.
For each stock, j, and each month, t, factor loadings are estimated from
a first-stage regression of stock returns on the Fama–French factors and the
inflation-news mimicking portfolio using the prior 60 months of data, with
a minimum requirement of 24 monthly observations.18 The factor-adjusted
returns for stock j and month t are then computed as:

L
R̃jt∗ ≡ R̃jt − RFt − βjk F̃ kt ,
k=1
where:
R jt = return on security j in month t;
Fkt = return on the kth factor in month t; and
β jk = estimated factor loading for the kth factor.
In each month, the factor-adjusted returns are then regressed on a set of
stock characteristics and SUE in a second-stage cross-sectional regression.
The time-series average and the t-statistics of the coefficients are then evalu-
ated to test whether the asset-pricing factors subsume the predictive ability
of these characteristics. Although we consider several firm characteristics
that prior studies have shown to be important in explaining cross-sectional
variation of returns, we are particularly interested in examining whether
the predictive ability of SUE is subsumed by the inflation-news mimicking
16 The industries identified in Lamont [2001] are: basic, capital goods, construction, con-
sumer goods, energy, finance, transportation, and utilities.

17 Data on industry returns are obtained from the CRSP, while the bond returns data are
from Ibbotson.
18 As an alternative, we estimate the factor loadings separately for each SUE portfolio and
assign these estimates to individual stocks constituting the portfolios. This modification leaves
the results qualitatively unchanged.
TABLE 8
Cross-Sectional Asset-Pricing Tests
Fama–French
+
Excess Returns Fama–French Inflation
(1) (2) (3) (4)
Intercept 1.32 −0.02 −0.06
(4.54) (−0.39) (−1.04)
SIZE −0.09 −0.04 −0.02
(−1.73) (−1.59) (−0.84)
BM 0.26 0.14 0.13
(3.68) (3.11) (2.76)
TURN −0.10 −0.15 −0.15
(−1.62) (−3.57) (−3.54)
SUE 0.37 0.36 0.36
(24.58) (24.84) (24.39)
RET2–3 −0.08 0.00 0.20
(−0.25) (0.00) (0.66)
RET4–6 0.24 0.28 0.41
(0.88) (1.24) (1.73)
RET7–12 0.83 0.80 0.79
(5.11) (5.75) (5.61)
This table presents the Fama–Macbeth estimates of monthly cross-sectional regressions. The dependent
variable in the second column is simply the excess return, whereas in the third it is the factor-adjusted return
using the Fama-French factors. In the fourth column, the dependent variable is the adjusted return using
the Fama–French factors along with a factor mimicking news about future inf lation. The independent
variables are the lagged firm characteristics, measured as the deviation from the cross-sectional mean
in each period. SIZE represents the logarithm of market capitalization in billions of dollars. BM is the
logarithm of the book-to-market ratio with the exception that book-to-market values greater than the 0.995
fractile or less than the 0.005 fractile are set equal to the 0.995 and the 0.005 fractile values, respectively.
TURN is the logarithm of share turnover. SUE represents the standardized unexpected earnings. RET2–3,
RET4–6, and RET7–12 equal the cumulative returns over the second through third, fourth through sixth,
and seventh through twelfth months prior to the current month, respectively. The sample contains all
NYSE-AMEX firms from January 1974 through December 2001. All coefficients are multiplied by 100 and
t-statistics are presented in parentheses.
factor. The stock characteristics are listed in Table 8. To conserve space, we

do not repeat them here.
Table 8 presents the results from the asset-pricing tests. Column 2 presents
regression results from the standard Fama and Macbeth [1973] tests that
use excess returns as the dependent variable. Consistent with prior results,
we find the book-to-market factor, SUE, and lagged returns to be significant
predictors of cross-sectional variation in returns. The coefficient on SUE is
0.37 in these regressions and is the most significant one in the regression.
These results continue to hold when returns are adjusted for the Fama
and French [1993] factors. From column 3, we find that the coefficient on
SUE is 0.36. This coefficient does not change when returns are adjusted
additionally for an inflation-news mimicking factor, suggesting that even if
inflation risk is priced, it does not explain the impact of SUE on returns.19
19 Our conclusions remain unchanged when a mimicking factor for news about industrial
production growth is included in the analysis.

Hence we conclude that neither the Fama and French [1993] factors nor
the mimicking factors for news about future inflation captures the post-
earnings-announcement drift anomaly.
4.2 UNDERESTIMATION OF IMPACT OF INFLATION ON FUTURE EARNINGS

If the ability of lagged inflation to predict PMN payoffs is due to investors
underestimating the magnitude of the impact of inflation on future earn-
ings, then these variables will also predict the payoffs to PMN at future
earnings announcements, which is when the earlier misestimation of cor-
porate earnings is likely to be corrected. Hence, we examine whether lagged
inflation predicts future earnings announcement returns. As earnings an-
nouncement returns are measured over a short time interval, differences in
risk premium across SUE deciles are likely to be trivial in these regressions.
Bernard and Thomas [1990], Sloan [1996], and Bernard, Thomas, and
Wahlen [1997] use a similar approach to distinguish risk-based explanations
from underreaction explanations for accounting-based market anomalies,
including the drift anomaly.
Table 9 presents the results of the following pooled regression:

3
ERNRET i,q +h = γ0 + γ1 INF q −k,q + γ2 j SUE i,q − j + νi,q +h (7)
j =0
where:
ERNRET i,q+h = earnings announcement return in quarter q+h for stock
i.
For each stock, we compute the earnings announcement returns for the
four earnings announcements subsequent to the announcement of earn-
ings used in forming the SUE portfolios (i.e., for quarters q+1 through
q+4, where earnings from quarter q are used to form SUE portfolios). The
earnings announcement returns are measured as the market-adjusted re-
turns in the three days centered on the Compustat earnings announcement
date. The regressions are estimated across all firms as well as for firms sorted
into portfolios based on SUE i,q . We do not include the Fama–French factors,
because it is unlikely that a three-day earnings announcement return is due
to risk.
Panel A of table 9 presents the results from regressing one-quarter-ahead
earnings announcement returns on the independent variables. Regression
(1) shows that the first two lags of SUE are significantly positive and the last
lag is significantly negative, a pattern that has been consistently reported in
prior studies (e.g., Bernard and Thomas [1990], Ball and Bartov [1996]).
When lagged quarterly inflation is included in regression (2), and its coeffi-
cient is allowed to vary across the SUE portfolios, the coefficient on inflation
is significantly negatively related to the earnings announcement returns of
the lowest-SUE portfolio, while being significantly positively related to the
earnings announcement returns of the highest-SUE portfolio. This pattern
TABLE 9
Regression of Earnings Announcement Return on Lagged Inf lation
Panel A: Regression of one-quarter-ahead earnings announcement (ERNRET i,q +1 ) on lagged inf lation
All
Firms P1 P5 P 10 F -test P1 P5 P 10 F -test P1 P5 P 10 F -test P1 P5 P 10 F -test
Regression (1) (2) (3) (4) (5) (6) (7) (8) (9)
Intercept 0.263 0.225 −0.118 0.172 0.352 0.01 0.234 −0.060 0.187 0.308 0.09
(22.70) (11.13) (−1.39) (2.53) (4.21) (10.37) (−0.66) (2.54) (3.51)
INF q ,q −0.257 0.041 0.222 0.00 −0.189 0.059 0.144 0.00
(−9.15) (1.63) (7.42) (−4.61) (1.35) (3.65)
INF q −3,q −0.077 0.007 0.066 0.00 −0.059 0.012 0.045 0.00
(−9.91) (1.05) (8.07) (−4.84) (0.98) (3.85)
SUE i ,q 0.090 0.016 −0.049 0.397 0.026 0.01 0.002 −0.050 0.459 0.025 0.00
(14.51) (2.08) (−4.10) (1.71) (1.08) (0.25) (−4.18) (2.06) (1.03)
SUE i ,q−1 0.083 0.056 0.061 0.000 0.051 0.10 0.052 0.060 0.000 0.051 0.11
(9.46) (6.26) (2.59) (−0.00) (1.91) (5.79) (2.53) (−0.01) (1.90)
SUE i ,q−2 0.002 −0.007 0.024 −0.031 0.033 0.10 −0.009 0.024 −0.031 0.032 0.10
(0.23) (−0.79) (0.89) (−1.13) (1.07) (−0.93) (0.89) (−1.14) (1.05)
SUE i ,q−3 −0.112 −0.108 −0.089 −0.107 −0.162 0.45 −0.107 −0.089 −0.107 −0.162 0.42
(−12.91) (−12.45) (−3.28) (−4.12) (−5.53) (−12.34) (−3.27) (−4.12) (−5.54)
Adj. R 2 (%) 0.30 0.50 0.33 0.12 0.22 0.52 0.34 0.12 0.23
No. of obs. 181,625 181,625 18,017 18,433 18,504 181,625 18,017 18,433 18,504
INFLATION ILLUSION
549
550
Panel B: Regression of two- to four-quarter-ahead earnings announcement return on lagged quarterly inf lation
ERNRET i ,q+2 ERNRET i ,q+3 ERNRET i ,q+4
Dependent Variable P1 P5 P 10 F -test P1 P5 P 10 F -test P1 P5 P 10 F -test
Regression (1) (2) (3) (4) (5) (6) (7) (8) (9)
Intercept 0.037 0.346 0.228 0.00 0.023 0.290 0.163 0.00 0.242 0.278 0.093 0.00
(0.42) (5.04) (2.76) (0.25) (4.15) (1.94) (2.60) (3.97) (1.10)
INF q ,q −0.160 −0.061 0.095 0.00 0.016 0.041 0.087 0.45 0.111 −0.039 −0.030 0.02
(−3.79) (−1.39) (2.45) (0.36) (0.92) (2.22) (2.50) (−0.87) (−0.75)
SUE i ,q −0.036 0.315 0.033 0.00 −0.012 −0.509 0.029 0.00 −0.028 −0.425 −0.014 0.11
(−2.91) (1.34) (1.40) (−0.95) (−2.15) (1.21) (−2.15) (−1.80) (−0.59)
T. CHORDIA AND L. SHIVAKUMAR
SUE i ,q−1 0.028 −0.009 0.078 0.08 −0.084 −0.077 −0.102 0.16 0.019 0.047 0.005 0.78
(1.15) (−0.31) (2.95) (−3.38) (−2.72) (−3.79) (0.73) (1.67) (0.17)
SUE i ,q−2 −0.087 −0.112 −0.109 0.00 −0.015 0.023 0.030 0.95 −0.061 −0.018 −0.062 0.12
(−3.17) (−4.02) (−3.62) (−0.53) (0.82) (0.98) (−2.10) (−0.63) (−2.03)
SUE i ,q−3 −0.020 −0.053 −0.031 0.10 −0.054 −0.026 0.006 0.40 −0.024 0.045 0.088 0.07
(−0.72) (−2.00) (−1.08) (−1.89) (−0.98) (0.22) (−0.80) (1.71) (2.99)
Adj. R 2 (%) 0.20 0.14 0.14 0.10 0.06 0.08 0.07 0.03 0.03
No. of obs. 17,549 17,961 18,100 17,104 17,522 17,785 16,580 17,106 17,367
Panel C: Regression of two- to four-quarter-ahead earnings announcement return on lagged annual inflation
ERNRET i ,q+2 ERNRET i ,q+3 ERNRET i ,q+4
Dependent Variable P1 P5 P 10 F -test P1 P5 P 10 F -test P1 P5 P 10 F -test
Regression (1) (2) (3) (4) (5) (6) (7) (8) (9)
Intercept −0.008 0.339 0.166 0.01 −0.024 0.312 0.150 0.00 0.121 0.292 0.128 0.02
(−0.09) (4.57) (1.92) (−0.25) (4.13) (1.70) (1.21) (3.86) (1.45)
INF q −3,q −0.033 −0.014 0.039 0.00 0.012 0.006 0.023 0.34 0.050 −0.012 −0.017 0.00
(−2.63) (−1.15) (3.37) (0.97) (0.48) (1.99) (3.83) (−0.99) (−1.43)
SUE i ,q −0.039 0.254 0.029 0.00 −0.013 −0.458 0.030 0.00 −0.029 −0.456 −0.011 0.01
(−3.11) (1.12) (1.23) (−1.00) (−2.01) (1.25) (−2.22) (−2.01) (−0.47)
SUE i ,q−1 0.027 −0.008 0.077 0.10 −0.084 −0.077 −0.101 0.16 0.020 0.047 0.006 0.77
(1.12) (−0.31) (2.90) (−3.37) (−2.73) (−3.77) (0.80) (1.67) (0.21)
SUE i ,q−2 −0.088 −0.112 −0.110 0.00 −0.015 0.023 0.030 0.95 −0.061 −0.018 −0.061 0.12
(−3.18) (−4.01) (−3.66) (−0.53) (0.81) (0.99) (−2.12) (−0.62) (−2.00)
SUE i ,q−3 −0.022 −0.053 −0.033 0.10 −0.055 −0.026 0.007 0.38 −0.025 0.045 0.089 0.06
(−0.77) (−2.01) (−1.14) (−1.91) (−0.97) (0.24) (−0.85) (1.70) (3.04)
Adj. R 2 (%) 0.16 0.14 0.17 0.11 0.06 0.08 0.12 0.03 0.04
No. of obs. 17,549 17,961 18,100 17,104 17,522 17,785 16,580 17,106 17,367
The table presents results from the following regression:
3

ERNRET i,q +h = γ0 + γ1 INF q −k,q + γ2 j SUE i,q −h + νi,q +h , h = 1 to 4
j =0
where ERNRET i,q + h is the earnings announcement return for firm i in quarter q + h. Earnings announcement returns are measured as the cumulative market-adjusted returns in the three days
surrounding the earnings announcement. SUE i,q is the standardized unexpected earnings of firm i, based on earnings announced in quarter q. INF q − k,q is the inf lation in the k+1 quarters prior
to the month in which earnings for quarter q (used to compute SUE iq ) are announced. Pooled regressions are estimated for each SUE portfolio as well as for all firms. The SUE portfolios are
formed by sorting stocks on SUE i,q . Panel A reports results from the regression of one-quarter-ahead earnings announcement returns on lagged inf lation. Panel B (panel C) reports results from
two- to four-quarter-ahead earnings announcement returns on lagged inf lation measured over one quarter (one year). Regressions (2) and (6) of panel A report results from a regression that
holds the coefficient on all variables, other than inf lation, constant across the SUE portfolios. This is implemented by estimating the regression across all firms after interacting inf lation with
dummy variables for each SUE portfolio. Regressions exclude the extreme 1% of the dependent variable and SUEi,q−j ( j = 0, 3)on either side. The column titled “F -test” gives the p-value for the
INFLATION ILLUSION
F -test that the coefficient is the same across all SUE portfolios.
551
is consistent with the coefficient patterns observed in table 7 for regressions

of SUE i,q +1 and of RET i,q +1 . Also, in line with the inflation coefficient in the
regression of future SUE, the coefficient on inflation in the current regres-
sion increases monotonically across the SUE portfolios, and the F -test that
the coefficients are the same across SUE portfolios is rejected at a less than
1% level. Relative to regression (1), the coefficients on the first and second
lag of SUE continue to be statistically significant, but decline significantly in
magnitude. The coefficient on the first lag, in fact, decreases by over 80%.
When the past year’s lagged inflation is used as an independent variable
(i.e., regressions (6) through (9)), qualitatively similar results are obtained
as those above. The coefficient on annual inflation is −0.059 for portfo-
lio P 1 , which is significantly different from the coefficient of +0.045 for
portfolio P 10 . These coefficients imply that a one standard deviation in-
crease in annual inflation of 3.0 would cause an economically significant
spread of 0.31% [=3.00 × (0.045 − (−0.059))] in the quarter-ahead earn-
ings announcement returns of the portfolio PMN. If one believes that earn-
ings announcement returns of PMN contain nontrivial risk premiums, and
consequently continues to interpret these regressions as inflation-capturing
time-varying risk premiums, then the above spread implies that a one stan-
dard deviation increase in inflation increases the annualized risk premium
by over 25%!
Interestingly, from regression (6), we observe that the coefficient on the
first lag of SUE is statistically insignificant. When the coefficients on lagged
SUE are allowed to vary across portfolios in regressions (7) through (9), the
coefficients on the first three lags of SUE are insignificant for almost all the
portfolios. This indicates that, after controlling for lagged inflation, lagged
SUE have little predictive ability for future earnings announcement returns.
Consistent with our earlier results from portfolio-level analyses, the inter-
cepts increase monotonically across the portfolios even after controlling for
lagged inflation and lagged SUE, indicating the inability of these variables to
fully explain the drift phenomenon. Nonetheless, inflation does have an eco-
nomically important role in explaining the post-earnings-announcement
drift. The difference in the intercepts across the extreme SUE portfolios, P 10
and P 1 , decreases from 0.86 in regressions with only lagged SUE as explana-
tory variables (results not reported) to 0.37 when lagged annual inflation
is included as an additional explanatory variable. This suggests that lagged
inflation captures over 55% of the difference in the quarter-ahead earnings
announcement return across the extreme SUE portfolios.
To study the ability of inflation to predict earnings announcement returns
more than one quarter ahead, we repeat the above regressions using two-
to four-quarter-ahead earnings announcement returns as the dependent
variable. The results from these regressions are reported in table 9, panel
B, for quarterly inflation, and in table 9, panel C, for annual inflation. For
brevity, these panels report only results from the most general specification
of equation (7). In both panels B and C, the differences in inflation co-
efficients across extreme portfolios are found to decrease over time. For
instance, the difference in the coefficient on quarterly inflation is 0.48 for

the one-quarter-ahead earnings announcement returns (from panel A of
table 9), which decreases to 0.26 for the two-quarters-ahead earnings an-
nouncement returns and turns insignificant by the three-quarters-ahead
earnings announcement. This decay in the spread of coefficients across
portfolios is consistent with a decreasing reaction to the information in past
inflation. With regard to the coefficients on lagged SUE, no particular pat-
tern emerges for these coefficients across the SUE portfolios. Moreover, very
few instances of significantly positive coefficients are observed on lagged
SUE in these regressions.
Our conclusions on the predictive ability of inflation for future earnings
announcement returns are not sensitive to inclusion or exclusion of lagged
SUE in the regression. Moreover, these results are also robust to analyses of
portfolio-level data. For instance, when quarter-ahead earnings announce-
ment returns for the PMN portfolio (computed by averaging earnings an-
nouncement returns of individual stocks in the portfolio) are regressed on
lagged annual inflation, the inflation coefficient is 0.11 (t = 3.58), which is
comparable to the coefficient of 0.104 implied from the analysis of firm-level
data.
4.3 INTERPRETATION OF RESULTS

Based on the inability of inflation risk to explain the drift and on the ability
of inflation to predict future earnings announcement returns, we conclude
that investors’ underestimating of inflation’s impact on future earnings bet-
ter explains the data. Apart from being consistent with the inflation illusion
hypothesis of Modigliani and Cohn [1979], this interpretation of the results
also finds support in prior studies of the post-earnings-announcement drift
(e.g., Bernard and Thomas [1990], Ball and Bartov [1996], Brown and Han
[2000]). For instance, Ball and Bartov [1996] show that the post-earnings-
announcement drift is attributable to investors underestimating the magni-
tude of serial correlation in seasonally differenced quarterly earnings. This
underestimation of the correlation pattern is also implied by an underesti-
mation of the impact of inflation on earnings growth, as inflation is highly
serially correlated. Consistent with this line of reasoning, the results in ta-
bles 8 and 9 show that controlling for inflation in the regression weakens
the predictive ability of lagged SUE. Furthermore, interpreting our findings
as investors underreacting to inflation information is also consistent with
the evidence from analyses of macroeconomic expectations of economic
agents that are obtained through surveys. Ball and Croushore [2003] show
that survey respondents systematically misestimate the effects of monetary
policies, which tend to be closely linked to inflation, on future output.
5. Conclusions
The predictable drift in stock prices following earnings announcements
is one of the longest-standing anomalies in the accounting and finance
literature. In this paper, we show that part of the drift anomaly is attributable
to the inflation illusion argument of Modigliani and Cohn [1979]. The idea
is that, whereas bond market investors understand the impact of inflation
on discount rates, stock market investors do not account for the impact
of inflation on future earnings growth. More specifically, inflation illusion
will cause firms with positive earnings sensitivities to inflation to be un-
dervalued and stocks with negative earnings sensitivities to inflation to be
overvalued. Subsequent correction of this mispricing implies that inflation
will be related to future returns, with the sign and magnitude of the relation
depending on the firm’s earnings sensitivity to inflation.
The findings of this paper are consistent with inflation illusion partly
causing the post-earnings-announcement drift. The main results are:
(1) sensitivity of earnings to inflation varies systematically across stocks
sorted on SUE; (2) both future earnings growth and future returns of SUE-
sorted stocks are positively related to past inflation; and (3) future earnings
announcement returns of SUE-sorted stocks are also positively related to
past inflation. Moreover, with regard to the last two findings, the predictive
ability of inflation for future earnings and returns is incremental to that
of past SUE. Although we check whether the drift could be interpreted as
compensation for inflation risk, we do not find evidence to support this ar-
gument. We conclude that delayed reaction to inflation better explains our
results, and the underestimation of the magnitude of the earnings impact
of inflation partly explains the post-earnings-announcement drift.
Why do investors underestimate the significance of current macroeco-
nomic conditions to future earnings? One possible reason for this underes-
timation is the complexity and the constantly changing nature of the eco-
nomic system within which market participants make earnings forecasts.
Given the frequent shifts in the earnings exposure of individual firms, the
various instances of regime changes in monetary policy, the relatively fre-
quent structural changes in the economy, and the large, but infrequent,
macroeconomic shocks, it may not be entirely surprising that investors, ei-
ther because of their limited processing ability or because of parameter
uncertainties, do not fully incorporate all macroeconomic-related earnings
expectations into prices. This could also potentially explain the persistence
of the post-earnings-announcement drift for over three decades after it was
first identified.
REFERENCES
AGGARWAL, R.; S. MOHANTY; AND F. SONG. “Are Survey Forecasts of Macroeconomic Variables
Rational?” Journal of Business 68 (1995): 99–119.
BALL, L., AND D. CROUSHORE. “Expectations and the Effects of Monetary Policy.” Journal of
Money Credit and Banking 35 (2003): 473–84.
BALL, R., AND E. BARTOV. “How Naı̈ve Is the Stock Market’s Use of Earnings Information?”
Journal of Accounting and Economics 21 (1996): 319–38.
BALL, R., AND P. BROWN. “Some Preliminary Findings on the Association Between the Earn-
ings of a Firm, Its Industry and the Economy.” Journal of Accounting Research 5 (1967): 55–
77.
BALL, R., AND P. BROWN. “An Empirical Evaluation of Accounting Numbers.” Journal of Account-
ing Research 6 (1968): 159–78.
BALL, R.; S. P. KOTHARI; AND R. L. WATTS. “Economic Determinants of the Relation Between
Earnings Changes and Stock Returns.” The Accounting Review 68 (1993): 622–38.
BERNARD, V. L., AND J. K. THOMAS. “Post-Earnings-Announcement Drift: Delayed Price Response
or Risk Premium.” Journal of Accounting Research 27 (1989): 1–35.
BERNARD, V. L., AND J. K. THOMAS. “Evidence that Stock Prices Do Not Fully Reflect the Im-
plications of Current Earnings for Future Earnings.” Journal of Accounting and Economics 13
(1990): 305–40.
BERNARD, V. L.; J. K. THOMAS; AND J. M. WAHLEN. “Accounting-Based Stock Price Anomalies:
Separating Market Inefficiencies from Risk.” Contemporary Accounting Research 14 (1997): 89–
136.
BOOTH, G. G.; J. KALLUNKI; AND T. MARTIKAINEN. “Post-Announcement-Drift and Income
Smoothing: Finnish Evidence.” Journal of Business Finance and Accounting 23 (1996): 1197–
211.
BRAV, A., AND J. B. HEATON. “Competing Theories of Financial Anomalies.” Review of Financial
Studies 15 (2002): 575–606.
BRENNAN, M. J.; T. CHORDIA; AND A. SUBRAHMANYAM. “Alternative Factor Specifications, Secu-
rity Characteristics and the Cross-Section of Expected Stock Returns.” Journal of Financial
Economics 49 (1998): 345–74.
BROWN, L. D., AND J. C. Y. HAN. “Do Stock Prices Fully Reflect the Implications of Current
Earnings for Future Earnings for AR1 Firms?” Journal of Accounting Research 38 (2000): 149–
64.
BROWN, P., AND R. BALL. “Some Preliminary Findings on the Association Between Earnings of
a Firm, its Industry and the Economy.” Journal of Accounting Research 5 (1967): 55–77.
CAMPBELL, J. Y., AND T. VUOLTEENAHO. “Inflation Illusion and Stock Prices.” Working paper,
Harvard University, 2003.
CHEN, N.; R. ROLL; AND S. A. ROSS. “Economic Forces and the Stock Market.” Journal of Business
59 (1986): 383–403.
FAMA, E. F. “Short-Term Interest Rates as Predictors of Inflation.” American Economic Review 65
(1975): 269–82.
FAMA, E. F., AND K. R. FRENCH. “Common Risk Factors in the Returns on Stocks and Bonds.”
Journal of Financial Economics 33 (1993): 3–56.
FAMA, E. F., AND J. D. MACBETH. “Risk, Return and Equilibrium: Empirical Tests.” Journal of
Political Economy 81 (1973): 607–36.
FOSTER, G.; C. OLSEN; AND T. SHEVLIN. “Earnings Releases, Anomalies and the Behavior of
Security Returns.” The Accounting Review 59 (1984): 574–603.
GONEDES, N. J. “Properties of Accounting Numbers: Models and Tests.” Journal of Accounting
Research 11 (1973): 212–37.
GRANT, A. P., AND L. B. THOMAS. “Supply Shocks and the Rationality of Inflation Forecasts.”
The Quarterly Review of Economics and Finance 41 (2001): 515–32.
HEW, D.; L. SKERRATT; N. STRONG; AND M. WALKER. “Post-Earnings-Announcement Drift:
Some Preliminary Evidence for the UK.” Accounting and Business Research 26 (1996): 283–
93.
JEGADEESH, N., AND S. TITMAN. “Returns to Buying Winners and Selling Losers: Implications
for Stock Market Efficiency.” Journal of Finance 48 (1993): 65–91.
KOTHARI, S. P.; J. LEWELLEN; AND J. B. WARNER. “Stock Returns, Aggregate Earnings Surprises
and Market Efficiency.” Journal of Financial Economics (2005): Forthcoming.
LAMONT, O. “Economic Tracking Portfolios.” Journal of Econometrics 105 (2001): 161–84.
LO, A. W., AND C. MACKINLAY. “Data Snooping Biases in Tests of Financial Asset Pricing Models.”
Review of Financial Studies 3 (1990): 431–68.
MAGEE, R. P. “Industry-Wide Commonalities in Earnings.” Journal of Accounting Research 12
(1974): 270–87.
MENDENHALL, R. R. “Arbitrage Risk and Post-Earnings-Announcement-Drift.” Journal of Business
77 (2004): 875–94.
MODIGLIANI, F., AND R. A. COHN. “Inflation, Rational Valuation and the Market.” Financial
Analyst Journal 35 (1979): 24–44.
RANGAN, S., AND R. SLOAN. “Implications of the Integral Approach to Quarterly Reporting for
the Post-Earnings-Announcement-Drift.” The Accounting Review 73 (1998): 353–71.
RICH, R. W. “Testing the Rationality of Inflation Forecasts from Survey Data: Another Look at
the SRC Expected Price Change Data.” The Review of Economics and Statistics 71 (1989): 682–6.
SLOAN, R. “Do Stock Prices Fully Reflect Information in Accruals and Cash Flows About Future
Earnings?” The Accounting Review 71 (1996): 289–315.
SOFFER, L. C., AND T. LYS. “Post-Earnings Announcement Drift and the Dissemination of Pre-
dictable Information.” Contemporary Accounting Research 16 (1999): 305–31.
VASSALOU, M. “News Related to Future GDP Growth as a Risk Factor in Equity Returns.” Journal
of Financial Economics 68 (2003): 47–73.
WATTS, R. L. “Systematic Abnormal Returns After Quarterly Earnings Announcements.” Journal
of Financial Economics 6 (1978): 127–50.

Justification - Inflation

Загружено:

Сведения о документе

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Justification - Inflation

Загружено:

Авторское право:

Доступные форматы

DOI: 10.1111/j.1475-679X.2005.00181.

Inflation Illusion and

Received 23 December 2003; accepted 5 December 2004

This paper examines the cross-sectional implications of the inflation illusion

2. Inflation Illusion Hypothesis and Its Implications

consequently, capitalize real earnings growth at a nominal interest rate

2.2IMPLICATIONS OF INFLATION ILLUSION HYPOTHESIS

[1989]. The results remain qualitatively unchanged with this modification.

5 The large standard deviations in SUE for the P and P

The alternative is to assume parameter stationarity at the portfolio level,

These results indicate that earnings sensitivity to inflation varies systemat-

irrespective of the period over which lagged inflation is measured. Further,

Given the evidence in previous studies of a positive serial correlation in

3.4 INFLATION AND FUTURE RETURNS

3.4.1. Summary Statistics on Payoffs to SUE Portfolios. Table 4 documents the

long on P 10 and short on P 1 in the 60 event-days subsequent to an earnings

3.4.2. Predictability of Future Returns to PMN . To test whether investors react

Our conclusions are unchanged if nonoverlapping returns (computed as in

This predictive ability of inflation is consistent with investors underreacting

Jul 1972–Dec 1981 Jan 1982–Dec 1991 Jan 1992–Dec 2001

future earnings. To mathematically illustrate the two alternatives, consider

E [INF t+1 | INF t ] = δ0 + δ1 INF t (4)

Underestimation of future inflation on a consistent basis is not likely to be

3.5 INDIVIDUAL STOCKS

Panel B: Regression of SUE i,q +1 on lagged inf lation

Panel C: Regression of RET i,q +1 on lagged inf lation

All Firms P1 P5 P 10 F -test P1 P5 P 10 F -test P1 P5 P 10 F -test P1 P5 P 10 F -test

(−15.13) (−12.85) (−8.21) (−4.60) (0.67) (−14.24) (−7.13) (−4.74) (−0.68)

As our conjecture implies a differential impact of inflation on future earn-

In regression (2) we introduce INF q,q as an independent variable that

4.1 INFLATION RISK FACTOR

portfolio approach to capture the current month’s inflation shock.

inflation variables in these tests.

identified in Lamont [2001],16 and excess returns on a long-term govern-

sumer goods, energy, finance, transportation, and utilities.

factor. The stock characteristics are listed in Table 8. To conserve space, we

production growth is included in the analysis.

4.2 UNDERESTIMATION OF IMPACT OF INFLATION ON FUTURE EARNINGS

is consistent with the coefficient patterns observed in table 7 for regressions

instance, the difference in the coefficient on quarterly inflation is 0.48 for

4.3 INTERPRETATION OF RESULTS

Вам также может понравиться