Mean Reversion in Profitability and Earnings: Evidence From India (2007-2020)

Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 22

Mean reversion in profitability and earnings: Evidence from India (2007-2020)

Abstract
The present study explores the projected mean reversion of profitability and profits, based on
the idea proposed by Fama and French (2000) that changes in profitability and earnings are
somewhat predictable. Using a sample of 445 BSE listed companies and 309 companies of
manufacturing industries in India, the study employed a simple Partial Adjustment Model.
The findings show that profitability of the overall sample is mean reverting for the years 2007
to 2012, for both BSE listed and manufacturing industries. There is no evidence of mean
reversion for the years post that.
Keywords; Mean reversion, profitability and earnings, Partial Adjustment Model.

1.Introduction
Forecasting or predicting the behaviour of firm’s profitability has been a central focus for
researchers for almost 40 years. Predicting profitability is important as it influences various
elements such as firm’s implicit value, investment decisions and much more. In economics it
is typically believed that profitability is mean reverting in a competitive environment.
Mueller (1986) described mean reversion in profitability as the “competitive environment”
hypothesis. Mean reversion is based on the argument that competitive forces will bring a
correction of unusually high or low profitability with time. Enterprises with great profitability
will face fierce rivalry from both new entrants and current firms that mimic their business
models. In the coming period, this will have a negative impact on profitability. Firms with
poor profitability, on the other hand, may try to better their condition. They might also go out
of business, either willingly or via bankruptcy, or be acquired by another company. This will
help to improve the situation for the surviving businesses. Firms with poor profitability that
make it through the following period are likely to enhance their profits. The earlier research
by by Lipe and Kormendi (1994) and Ramakrishnan and Thomas (1992) analysed mean
reverting behaviour of profitability by performing yearly regression using yearly estimates.
It has been extensively accepted that the concept of mean reversion is accurate. In simple
terms, mean reversion indicates that a given element will, over time, revert to the dataset's
average or equilibrium level. Profitability and earnings are the elements of interest for this
study. mean reversion in earnings, indicates that firms that have above-average profitability
in one year are likely to have lower profitability in subsequent years due to various reasons
mentioned above, and firms that have below-average profitability in one year tend to have
higher profitability in subsequent years due to various reasons mentioned above (Welc, J.
(2011)). Assessing mean reversion is significant as there is mean reversion in profitability, if
proven implies that fluctuations in a company's profitability and earnings are predictable to
certain level (Fama and French (2000)). Mean reversion in profitability may be studied at the
company, industry, or national level. The present study focusses on mean reversion at
Industry level.
Although there are many studies regarding this subject, for instance study by Fama and
French (2000) focused on forecasting profitability and earnings in the context of European
Union countries or Study by Welc, J (2011) focused on analysing mean reversion in Polish
Public
companies over the years from 2000 to 2009, the main contribution of the present study to
forecast profitability and earnings in Indian stock market. The initial purpose of this study
was to see if there was any mean reversion among all BSE-listed companies in India.
Following that, a separate analysis was conducted for the manufacturing industry, which is
not greatly influenced by regulatory reforms (Fama and French (2000)). Our findings reveal
that the profitability of BSE-listed companies and manufacturing industry is mean reverting
for the years 2007 to 2012, with no indication of mean reversion for the years beyond that.
When comparing the results from previous research, the study discovered that our results, in
later years differed from those of Allen and Salim (2005) for UK listed firms and Fama and
French (2002) for US enterprises. The study believes that the rationale for not mean reverting
may be ascribed to the country's economic growth and a political paradigm shift in 2014.
The paper is organised as follows. The second section contains the literature on the
predictability of earnings and profits. The third section covers data and methodology. The
fourth section explains the results and interpretations. The last section concludes the paper.
2.Literature Review

As stated in the finance theories that the dividend payment and the unexpected earnings
announcements highly influence the stock price of the listed companies. This connection
probably helps the market players to predict the future earnings of the firm, and somewhat the
returns of the firm’s stock (Altunbas and Karagiannis, 2008). Together with the idea of
behaviour of firm profitability and random walk, the further research explored the evidence
of market efficiency since Little (1962) and Little and Rayner (1966). The pioneer of market
efficiency, the Fama (1965), later has researched the profitability and earnings fluctuations
among US enterprises (Fama and French, 2000). The profitability of a business is assumed to
be mean reverting in economics. What is the relationship between the mean reversion and
market efficiency? Most capital holders are encouraged to diversify their investments into
more profitable industries. As a result of the successful allocation and utilisation of resources,
a uniform rate of return would be achieved within and across industries. In the case of
profitability, as unproductive industries are converted to profitable ones, the likelihood of
achieving anomalous profits decreases (Fama and French, 2000; Allen and Salim, 2005;
Altunbas and Karagiannis, 2008).

The focus of previous research on the predictability of profitability and earnings was mostly
on determining the rationality of the price-earnings connection, which was generally done
using time series analysis (Lev, 1969, and Freeman et al., 1982). Small sample testing,
survivorship bias for time series analysis, and omission of variables other than price for
forecasting predicted profits were all issues in these analyses. Even if accounting and finance
research such as Collins and Kothari (1989) and Basu (1997) focused on the same issues,
economics literatures found the same results. In contrast, a few other studies in the United
States (Mueller, 1977, 1986), Europe (Geroski and Jacquemin, 1988), and the United
Kingdom (Goddard and Wilson, 1999) that estimated profitability using firm specific or
cross-sectional observations produced better findings. Furthermore, cross-sectional study of
changes in profitability on lagged changes with more independent variables, as reported by
Freeman et al (1982), Collins and Kothari (1989), Basu (1997), and Fama and French (2000),
provides more consistent evidence on prediction. As a result, using cross-sectional regression
to verify predictability assures that a broad data sample is covered while meeting minimum
survival requirements, reducing survivor bias.

For forecasting profitability in a US sample, Fama and French (2000) used a basic partial
adjustment model using cross-sectional regressions. The profitability was calculated as the
ratio of earnings before interest but after taxes over assets in the study. The data showed a
yearly mean reversion of roughly 38 percent and were extremely nonlinear, confirming the
mean reversion assumption of profitability. Surprisingly, the study found that forecasting
profitability and earnings is easier when they are not close to their mean. Subsequently, Allen
and Salim (2005) adopted the technique of Fama and French (2000), which encompassed all
UK industries except banking and utilities, presuming that profitability is heavily influenced
by laws and regulations. The investigation revealed a lower rate of mean reversion of roughly
25% in a year, but the rate of adjustment is not asymmetric. As a result, the evidence from the
UK and the US are nearly identical, except for a few minor differences in the UK sample.

Altunbas and Karagiannis (2008) investigated the assumption of mean reversion of


profitability among European Union listed businesses (EU). Even though the study used the
same methodology as Fama and French (2000) and Allen and Salim (2005), namely cross-
sectional yearly regression with a basic Partial Adjustment Model (PAM), both the financial
and utility industries were included. Financial and utility companies were removed from
Fama and French (2000) and Allen and Salim (2005) analyses because they were heavily
influenced by industry regulations. According to Allen and Salim's (2005) findings in UK
listed companies, there was 27% mean reversion for the entire sample. Furthermore, the rate
of reversion varied from 5% to 40% for different industries and was 32% for the sample
excluding the finance and utilities businesses. Moreover, there was no nonlinearity or second
order effect on mean reversion of profitability because of the results. As a result, the study
determined that the explanation for the disparity in outcomes is most likely due to the higher
degree of competitiveness in the United States than in the United Kingdom.

Previous research suggested that the idea of the pattern of earnings and profitability would
always help to have a better understanding of the stock market. Hence, the purpose of the
present research is to check if there is a mean reversion of profitability and earnings across
Indian firms. The research explores mean reversion in profitability and earnings across BSE
listed companies in India from 2007 to 2020, using the approach proposed by Fama and
French (2000). The further research focuses on the mean reversion of profitability and
earnings in the manufacturing industries as an additional check. The present study thus
contributes to the existing literature in two ways. First, by examining the evidence of mean
reversion in the Indian setting by incorporating recent firm year observations from all BSE
listed firms. Second, by conducting a separate study on the manufacturing industry to ensure
that the findings are consistent with previous findings.

3. Data and Methodology

3.1 Data
The present study focused primarily on the listed companies of the Bombay stock Exchange
(BSE) in India. The data were obtained from CMIE ProwessIQ for the period of 2007-2020.
Initially, there were around 56000 companies on year-on-year basis. However, the final
sample comprised of only 445 companies after the missing data analysis. Following that, a
secondary analysis is conducted among industrial industries that are not heavily impacted by
regulatory changes to validate the findings of the first analysis with BSE listed companies.
The manufacturing industry had roughly 17635 companies, however after missing data
analysis, the final data only had 309 firms. Table1 provides the definition of study variables
as per the source.

Table 1 List of variable definition.


Variables Definition
Profit Before Interest and Tax (PBIT) Profit before interest and tax as reported by
company.
Total Assets Total of all current and non-current assets of
a firm on the last day of an accounting
period.

Market capitalization The total market capitalization of the


company.
Dividend Payout Ratio (Equity dividend as Earnings per share/ Dividend per share
% of PAT)
Source: Author’s calculation
3.2 Methodology:
The present study employed year-on-year regressions using the linear and nonlinear Partial
adjustment Models (PAM) for forecasting probability and earnings as in Fama and French
(2000), Allen and Salim (2005), and Altunbas and Karagiannis (2008). PAM takes into
consideration the estimation error that occurs due to correlation of regression residuals across
firms by using standard error of the average slope, which in-turn aids in determining mean
reversion of firms.
3.2.1 Regression analysis for profitability
a. Linear PAM
A two-stage attempt is used in the methodology. Initially, expected profitability is estimated
using the following cross-sectional linear regression,

(1)

where represents expected profitability ratio, is calculated as PBIT over total assets for
year t, means the market to book value of assets is computed as the current market
capitalization over total assets, dd is the dummy variable, which takes zero for companies
paying dividends and 1 otherwise, and is the dividend payout ratio during the year t.
The three variables, according to Fama and French (2000), provide information about
profitability on many fronts. This indicator is not profitability in the traditional sense, since it
measures the rate of return on assets invested. Instead, it serves as an indication of a firm's
potential to make profits for each unit of asset invested. Dividends will provide a measure of
the firm's constant projected earnings, while market to book value will provide the current
worth of all future cash flows not included in the dividend related factors. As a result,
regressing equation (1) will provide a complete picture of the projected profitability
disparities between enterprises.
Once the expected profitability has been determined, the next step will be to investigate
profitability behavior using PAM as follows.

(2)

where equals change in profitability from next year to the current year( ), CP
is the change in profitability from current year to previous year ( ), indicates
the deviation of profitability from its expected value of profitability ( ) for the year t. The
fitted values of regression equation 1 are used as proxy for .

b. Non-linear PAM
Although equation 2 implies that profitability behaviour is linear, the mean reversion in
profitability is largely influenced by predictability of profitability, as postulated by Fama and
French (2000) and Elgers and Lo (1994) using tests with the addition of some variables in
their functional forms as follows.

(3)

where equals change in profitability from next year to the current year( ),
is the deviation of profitability from its expected value of profitability ( ),
equals negative deviation of profitability, denotes the squared negative deviation of
profitability, indicates the squared positive deviation of profitability, is the
change in profitability from current year to previous year ( ), equals the
negative changes in profitability, denotes the squared negative changes in
profitability, indicates squared positive changes in profitability. Here,
measures nonlinearity in the mean reversion of the profitability and measures
nonlinearity in the autocorrelation of changes in the profitability.
3.2.2 Regression analysis for earnings:
The third phase of the study addresses the question of whether changes in earnings can be
predicted and, if so, to what degree these changes in earnings are attributable to mean
reversion. Furthermore, the linearity restriction is removed, and the change in earnings is
utilized as the dependent variable instead of the change in profitability.

(4)

where change in earnings from next year to the current year ( ), is the
deviation of profitability from its expected value of profitability ( ), equals negative
deviation of profitability, denotes the squared negative deviation of profitability,
indicates the squared positive deviation of profitability, is the change in earnings
from current year to previous year ( ), equals the negative changes in
earnings, denotes the squared negative changes in earnings, indicates squared
positive changes in earnings. Here, measures nonlinearity in the mean
reversion of the profitability and measures nonlinearity in the autocorrelation of
changes in the earnings.
4.Results and interpretation

4.1 Mean reversion in profitability among all BSE listed companies

Cross sectional regressions were carried out for individual firms for each year from 2007-
2019. Table 1 reports the results for expected profitability including means of regression
intercepts and slopes, and t-statistics (computed as the mean divided by the standard error)
for the means years.

Panel A of Table 2 shows the average slopes from the estimation of PR (with every
individual firm) on MA, dd, and DPR. For BSE listed companies the MA has a substantial
positive relationship with PR, with a coefficient value of 0.025 and a t-statistics of 11.07
showing that the fluctuation in predicted profitability that was overlooked by the dividend
factors may be captured. The measure of MA can be used as a proxy for Tobin's Q, according
to Fama and French (2000) and Allen and Salim (2005). Further, both dd and DPR gather all
the data about predicted profitability. The negative slope coefficient of -0.15 suggests a
nonlinear relationship between dd and profitability. Moreover, the expected profitability of
firms that do not pay dividend is 0.15 lower than predicted by the relation between MA and
DPR.

Part B of Table 2 explains five different regression results. The PAM must have a negative
PR coefficient and a positive EP coefficient with equal absolute values for mean reversion to
occur (Fama and French (2000), Allen and Salim (2005), and Altunbas (2008)). The
coefficients for PR and EP, on the other hand, are 0.105 and -0.053, respectively, indicating
that there is no evidence that profitability is mean reverting.

The lagged change in profitability, CP, is added as an additional variable to check whether
the mean reversion captured by the partial adjustment term, DP, is the only cause of
predictable difference in profitability, as suggested by Fama and French (2000), Allen and
Salim (2005), and Altunbas (2008). In the third regression equation, the slope of the lagged
profitability become strongly negative (-0.375) once CP is alone added to explain ∆PR. The
value is somewhat higher than 30 per cent in the Fama and French (2000) and Altanbas
(2008). However, the value of slope is reduced from 37.5 to 17.7 per cent, result does not
support the argument that the mean reversion is the major reason for predictability in
profitability as the PR is positive and EP is negative. The findings indicate that the role of
mean reversion in predicting the profitability is comparatively less than in the US and UK.

The second regression equation examines what might occur if those companies converged on
a single projected profitability threshold. The predicted rate of mean reversion is -0.242 per
cent, which is essentially same as (26.2) Altunbas (2008), and lower than (30 and 31) Fama
and Freanch (2000) and Allen and Salim (2005). Regression equations 4 and 5 explain the
nonlinearity models. Th e CP slope is negative at -0.348, and most of the nonlinearity
variables are not significant as in Fama and French (2000). Similarly, the fifth regression
equation shows that many nonlinearity variables are not statistically significant.

The overall result of change in profitability reveals that there is no significant evidence of
mean reversion in Indian firms. The mean reversion is not a significant reason for the change
in profitability during the period from 2007 to 2019. Hence, a further analysis is performed to
test whether the results are constant if there is any change in the sample period. The cross-
sectional analysis exhibit that the change in profitability is mean reverting over a period from
2007 to 2012.

Table 3 reports the results of cross-sectional regressions for individual firms for each year
from 2007-2012. Panel A of Table 3 also depicts that the MA has a positive relationship with
PR having a slope of 0.029 with strong significance (t-statistic – 10.289) suggesting that
variation in the predicted profitability can be captured. It further supports the argument that
the MA can be used as a proxy for Tobin's Q as opined by Fama and French (2000) and Allen
and Salim (2005). The slope of dd is -0.165 suggesting a nonlinear relationship between dd
and Profitability.

Remarkably, the regression results in Panel B of Table 3 reverses the findings of Panel B of
Table 2. Despite the small absolute differences, there is strong evidence of mean reversion
(from 2007 to 2012), as the slope of PR is negative and EP is positive, as predicted by PAM.
The coefficients of PR and EP in regression equation 1 are -0.422 and 0.248, respectively,
and both are highly significant with t-statistics more than 2.8 as in Allen and Salim (2005)
and Altunbas (2008).

The predicted mean reversion rate was 33.6 percent, substantially identical to Fama and
French (2000) with 30 per cent, Allen and Salim (2005) with 31 per cent, however, was
higher than Altunbas (2008) with 26.2 per cent.
The variable CP in regression equation three has a slope of -0.391, indicating very strong and
substantial negative outcomes. The change in profitability from year t to t+1 reverses 39.1%
of the lagged change. Further, it was found that mean reversion lowered the slope of CP from
39.1 percent to 19.4 percent) and similar significant results were found in Altunbas (2008)
and Allen and Salim (2005).
The slope of CP in regression equation four is found to be -0.442 with a t-statistic greater
than 3, which is higher than Allen and Salim (2005) but less than Altunbas (2008). The
significance of variables in regression equation five is similar to Fama and French (2000) and
Altunbas (2008).

4.1.2 Mean reversion in earnings among all BSE listed companies

The regression results on earnings in table 4 presents that, when lagged changes are only
considered for the years 2007-2019, the study finds that the t-statistic is insignificant with a
coefficient of -0.000000592. Though this conforms with Beaver (1970) and Ball and Watts
(1972) in the US that the linear autocorrelation of changes in profits is modest in random
walk behavior, it disagrees with Fama and French (2000) who claim that the evidence for
profitability.
When negative earnings changes are included in the model, the coefficient on earnings
changes increases to -0.000000517 with a t-statistic of -0.5909, implying that there is a
predictable component to earnings changes, whereas the coefficient on negative earnings
changes is high at -1.87 with a t-statistic of -0.907, implying that there is a predictable
component to earnings changes. This corroborates US findings that negative earnings
changes tend to revert to the mean.
Table 2. Cross sectional regression results reporting the level and change in profitability from 2007 to 2019
Panel A: Regression results showing the degree of profitability, PR
Mean of Coefficient and t-statistics for the yearly regression from 2007-2019
Intercept MA dd DPR R-Square
Coefficient 0.109 0.025 -0.156 -0.00028 0.381
t-statistics 19.341 11.031 -9.544 -2.686
Means and standard deviations of the regression variables
PR MA dd DPR
Means 0.122 1.417 0.082 32.558
Standard deviation 0.106 1.831 0.266 90.000
Panel B: Regression results showing the degree of profitability, ∆PR
Mean of Coefficient and t-statistics for the yearly regression from 2007-2019
Intercept PR EP NDP SNDP SPDP CP NCP SNCP SPCP R-Square
Coefficient -0.0119 0.105 -0.053 -0.177 0.269
t-statistics 0.9566 -1.490 -1.932 -3.469
Coefficient 0.020 -0.242 -0.247 0.242
t-statistics 4.309 -5.805 -4.501
Coefficient -0.0097 -0.375 0.169
t-statistics -0.863 -8.369
Coefficient -0.006 -0.348 0.257 0.781 -0.414 0.245
t-statistics 0.337 -2.420 1.009 1.129 -0.654
Coefficient -0.015 -0.339 0.4331 0.186 0.236 -0.543 -0.060 -0.116 -0.115 -0.042 0.349
t-statistics 0.815 -1.042 0.6438 -0.016 -0.197 -1.571 -0.968 0.339 -0.202 -0.128
Means and standard deviations of the regression variables
∆PR PR EP NDP SNDP SPDP CP NCP SNCP SPCP
Means -0.009 0.122 0.122 -0.025 0.0019 0.0037 -0.0017 -0.0195 0.0052 0.0046
Standard deviation 0.200 0.106 0.062 0.051 0.051 0.0256 0.0939 0.0614 0.0634 0.0529
Table 3. Cross sectional regression results reporting the level and change in profitability from 2007 to 2012
Panel A: Regression results showing the degree of profitability, PR
Mean of Coefficient and t-statistics for the yearly regression from 2007-2012
Intercept MA dd DPR R-Square
Coefficient 0.121 0.029 -0.165 - 0.359
t-statistics 21.241 10.289 -8.644 0.00034
-3.251
Means and standard deviations of the regression variables
PR MA dd DPR
Means 0.135 1.118 0.050 29.763
Standard 0.097 1.439 0.214 51.213
deviation
Panel B: Regression results showing the degree of profitability, ∆PR
Mean of Coefficient and t-statistics for the yearly regression from 2007-2012
Intercept PR EP NDP SNDP SPDP CP NCP SNCP SPCP R-Square
Coefficient 0.019 -0.422 0.248 -0.194 0.293
t-statistics 2.204 -8.070 2.801 -3.956
Coefficient 0.041 -0.336 -0.217 0.274
t-statistics 6.082 -7.701 -4.756
Coefficient -0.0039 -0.391 0.169
t-statistics -1.064 -9.326
Coefficient 0.0036 -0.442 0.354 0.346 -0.382 0.252
t-statistics 0.627 -3.135 1.357 1.688 -1.103

Coefficient 0.020 -0.299 0.161 0.138 0.059 -0.415 -0.189 0.102 -0.281 0.147 0
t-statistics 1.938 -1.877 0.945 0.643 0.195 -1.534 -0.994 0.139 -0.736 -0.118
Means and standard deviations of the regression variables
∆PR PR EP NDP SNDP SPDP CP NCP SNCP SPCP
Means -0.0036 0.135 0.135 -0.025 -0.0017 0.004 -0.00067 -0.017 0.0043 0.0039
Standard 0.089 0.097 0.053 0.042 0.029 0.032 0.089 0.056 0.046 0.036
deviation
The coefficient on earnings increases incrementally in the next regression, which covers
squared negative and squared positive earnings changes. For the years 2007-2019, the t-
statistics were negligible for majority of the variables, with coefficients as high as -
0.00000187.

In regression equation 4, PR and EP have coefficients of -0.322 and 0.375, respectively, with
significant t-statistics of -6.365 and 4.067, to see if autocorrelation of earnings changes can
be the cause of non-linear mean reversion in the level of profitability. The study can conclude
from these findings that mean reversion is the driving force for predictable variance in
profitability.

All variables that indicate non-linear mean reversion of profitability are included in the final
regression equation. The coefficient on CE rises to 0.043 with an insignificant t-statistic of
0.12, which contrasts with the coefficient and significant t-statistic of 0.34 and 2.36 in Allen
and Salim's (2005), as well as Fama and French's (2000), where the coefficient is 0.18 with a
significant t-statistic of 5.76.

To check for mean reversion, the same regressions were done by limiting the sample period
from 2007 to 2012. The results are satisfactory because mean reversion is observed. The
results are summarised in Table 5. When lagged changes are considered for the years 2007-
2012, the study finds that the t-statistic is insignificant with a coefficient of -0.000000832,
which agrees with Beaver (1970) and Ball and Watts (1972) in the US and contradicts Fama
and French (2000), while also converging with the results for 2007-2019.

The coefficient on earnings changes increases to -0.000000820 with a t-statistic of 0.6543


after including the negative earnings changes, implying that earnings changes have a
predictable component, whereas the coefficient on negative earnings changes is high at -2.7
with a t-statistic of -0.968, implying the same. As a result, the negative earnings changes will
revert to mean as examined from US findings.

The coefficient on earnings increases in the next regression, which includes squared negative
and squared positive earnings changes, for the years 2007-2012, comparable to 2007-2019,
where the t-statistics were negligible for the bulk of the variables, with coefficients as high as
2.7.

PR and EP exhibit coefficients of -0.208 and 0.270, respectively, in the regression equation 4,
with significant t-statistics of 3.11 and 2.00, showing that earnings autocorrelation can be the
cause of non-linear mean reversion. As a result, the study suggests that mean reversion is
responsible for the predictable variation in profitability.

After all variables are included in the final regression equation, the coefficient on CE rises to
0.094 with an insignificant t-statistic of 0.129, which contrasts with Allen and Salim (2005)
as well as Fama and French's (2000), however is similar to the results of 2007-2019.

4.1.3 Mean reversion in profitability among manufacturing companies

To double-check, the present study looked at the manufacturing sector as a separate


subsample, where regulatory initiatives or reforms do not have the same impact as they do in
the financial and utilities sectors (Fama and French (2000) and Allen and Salim (2005)).
Table 6 shows the manufacturing sector's predicted profitability outcomes, including
regression intercepts and slopes, as well as t-statistics (mean divided by standard error) for
the mean years. The average slopes from PR estimation (for each individual firm) on MA, dd,
and DPR are shown in Panel A of Table 6. With a coefficient of 0.029 and a t-statistics of
12.760, MA for the manufacturing sector indicates the significant positive connection with
PR, showing that it contains the variances in expected profitability that were missed by
dividend factors. Furthermore, estimated profitability for enterprises that do not pay
dividends is 0.13, which is lower than that indicated by the MA-DPR relationship. Panel B of
Table 6 explains five alternative regression outcomes. For the profitability to mean reversion,
the PAM must have a negative PR coefficient and a positive EP coefficient with equal
absolute values. However, the data for the manufacturing industry contradict the same, as PR
is 0.024 and EP is -0.161.

As indicated by Fama and French (2000), Allen and Salim (2005), and Altunbas (2005),
adding CP as an additional variable check whether mean reversion is covered by the partial
adjustment term, DP, which is the only cause of predicted differences in profitability (2008).
To explain PR, the third equation is regressed simply using CP, where the lagged profitability
slope becomes severely negative (-0.344). Though this figure is somewhat greater than 30%
when compared to Fama and French (2000) and Altunbas (2008), the slope is lowered from
34.4 to 16.4 percent, indicating that mean reversion plays a less role in predicting profitability
in the US and UK.

The second regression equation examines what might occur if those companies converged on
a single projected profitability threshold. The predicted rate of mean reversion is 23.4 per
cent, which is almost same as (26.2 per cent) Altunbas (2008), and lower than (30 per cent
and 31per cent) Fama and French (2000) and Allen and Salim (2005). The remaining two
regression equations 4 and 5 explains the nonlinearity models. The slope of CP is negative
21.4, and other variables are also not significant for non-linearity. Even the fifth regression
also provides similar results as fourth equation that the variables are not significant as in
Fama and French (2000).

Overall, the findings show that there is no strong evidence for mean reversion in profitability
in Indian manufacturing firms, meaning that mean reversion is not a significant factor for the
shift in profitability between 2007 and 2019. These results are in agreement with BSE listed
firms of 2007-2019, showing no mean reversion. As a result, the analysis is extended to see if
the results remain consistent when the sample period is reduced. The cross-sectional analysis
reveals that the change in profitability is mean reverting over the period of 2007 to 2012.

Table 7 shows the findings of cross-sectional regressions for individual manufacturing


subsample firms for each year from 2007 to 2012. In Panel A of Table 7, the MA exhibits a
positive connection with PR, with a slope of 0.035 and substantial significance (t-statistic –
12.234), implying that variance in expected profitability can be captured. It also confirms
Fama and French (2000) and Allen and Salim's (2005) arguments that the MA can be utilized
as a substitute for Tobin's Q. (2005). The slope of dd is -0.142, indicating that dd and
profitability have a nonlinear relationship.
Notably, the regression results in Panel B of Table 7 contradict those in Panel B of Table 6,
indicating that there is substantial evidence of mean reversion (from 2007 to 2012), with PR
slopes being negative and EP slopes being positive, as expected by PAM. According to Allen
and Salim (2005) and Altunbas (2005), the coefficients of PR and EP in regression equation 1
are -0.294 and 0.062, respectively, and both are highly significant with t-statistics more than
2.8. (2008).
The lagged change in profitability, CP, is added to check if the mean reversion is explained
by the partial adjustment term, DP, being the only reason for the projected difference in
profitability, as indicated by Fama and French (2000), Allen and Salim (2005), and Altunbas
(2008).
The second regression equation shows what would happen if all the companies agreed on a
single expected profit criterion. The mean reversion rate was predicted to be 28.4 per cent,
like Fama and French (2000) and Allen and Salim (2005).
By regressing the equation with CP in the third equation, the PAM's capacity to capture mean
reversion may be determined. With a slope of -0.385, the CP indicates very strong and
significant negative results. Year t to year t+1 profitability changes reverse 38.5 percent of
the delayed change. It can be noticed that the slope of CP has decreased from 39.1 per cent to
22 per cent, which can be attributed to mean reversion. Similar findings were observed by
Altunbas (2008) and Allen and Salim (2005).

Equations 4 and 5 then described non-linearity. Because of the negative slope and strong
significance, the slope of CP in equation 4 is -0.171, suggesting a reversion in profitability.
Similarly, to the model proposed by Fama and French (2000), many of the fifth equation
variables are not significant (2000). On comparing with BSE listed firms, we are getting the
same results where mean reversion is 33.6% slightly higher to a mean reversion percentage of
28.4% for manufacturing.

4.1.4. Mean reversion in earnings among manufacturing companies

The first equation demonstrates that when just lagged changes in earnings are incorporated,
the coefficient is -0.0000019 with insignificant t statistics when regressing earnings equations
for the manufacturing subsample, as shown in table 8. This agrees with Beaver (1970) and
Ball and Watts (1972) in the US, but not with Fama and French (2000), who contend that the
evidence for profitability is lacking.

The coefficient on earnings changes increases to -0.00000024 with a t-statistic of -1.409 in


the following regression, showing a predictable component in earnings, and the coefficient on
negative earnings changes is at -0.0000009 with a t-statistic of -0.1719, implying the same.
This backs with US findings that negative earnings changes tend to revert to the mean.

The next regression, which includes the squared negative and positive earnings changes,
shows a small increase to -0.0000006. Similarly to BSE results, the t-statistics for most of the
variables in the manufacturing industries were negligible, with coefficients as high as -
0.000009.

The PR and EP of the fourth equation show coefficients of -0.261 and 0.283, respectively,
with significant t-statistics of -4.338 and 2.981, confirming that non-linear mean reversion
profitability is caused by autocorrelation of earnings fluctuations. As a result, the study
claims that the predicted variation in profitability is due to mean reversion.
In the final regression equation, include all factors that suggest non-linear mean reversion of
profitability. The coefficient on CE becomes zero with an insignificant t-statistic of 0.0073, in
contrast to Allen and Salim (2005) and Fama and French (2000).

After lowering the sample period from 2007 to 2012, the same regressions were performed.
The results are satisfactory because mean reversion is evident. The findings are presented in
table 9. When lagged changes for the years 2007-2012 were evaluated, the coefficient and t-
statistics were 0.000003 and -1.239, respectively, converging with results for 2007-2019.

After accounting for negative earnings changes, the coefficient on earnings changes remains -
0.0000003 with a t-statistic of -1.348, showing minimal predictability. However, the
coefficient on negative earnings changes is high at -0.000013 with a t-statistic of -2.011,
signaling the same. As a result, the negative earnings fluctuations will revert to the mean.

The coefficient on earnings increases in the next regression, which includes squared negative
and squared positive earnings changes. For the years 2007-2012, the t-statistics were
insignificant for many of the variables, with coefficients as high as 0.13, lower than the
results for 2007-2019.

In regression equation 4, PR and EP have coefficients of –0.273 and 0.264, respectively, with
significant t-statistics of -3.572 and 2.123, indicating that earnings autocorrelation might
produce non-linear mean reversion.

After accounting for all variables in the final regression equation, the coefficient on CE rises
to 0.0000001 with an insignificant t-statistic of 0.211, which differs from Allen and Salim
(2005) and Fama and French (2000) but is consistent with the results of 2007-2019.
Table 4. Cross sectional regression results reporting the level and change in earnings from 2007 to 2019
Regression results showing the degree of profitability, ∆PR
Mean of Coefficient and t-statistics for the yearly regression from 2007-2019
Intercept PR EP NDP SNDP SPDP CE NCE SNCE SPCE R-Square
Coefficient 0.0167 -5.92E-7 0.0036
t-statistics 3.139 -0.763
Coefficient 0.0159 -5.17E-7 -1.1E-3 0.0052
t-statistics 2.993 -0.5909 -0.193

Coefficient 0.016 -1.87E-6 1.3E-5 0.016


t-statistics 2.934 -0.907 -0.157

Coefficient 0.008 -0.322 0.375 -1.5E-07 1.4E-5 -0.084 0.093 0.165


t-statistics 0.813 -6.365 4.067 -0.086 0.293 -1.913 0.811

Coefficient -0.023 0.1092 -0.253 -0.018 -0.438 0.052 0.0433 -0.067 0.013 0.001 0.271
t-statistics -0.261 0.512 -0.909 0.777 -2.548 0.341 0.125 -0.152 0.024 0.235

Table 5. Cross sectional regression results reporting the level and change in earnings from 2007 to 2012
Panel B: Regression results showing the degree of profitability, ∆PR
Mean of Coefficient and t-statistics for the yearly regression from 2007-2012
Intercept PR EP NDP SNDP SPD CE NCE SNCE SPCE R-Square
P
Coefficien 0.023 -8.32E-7 0.0019
t 4.047 -0.715
t-statistics
Coefficien 0.022 -8.20E-7 -2.454 0.0051
t 3.841 0.6543 -0.386
t-statistics
Coefficien 0.024 -2.760 0.000029 3.164 6.050 0.014
t 3.755 -0.968 -0.173 -0.329 0.764
t-statistics
Coefficien 0.012 -0.208 0.270 -1.19E-6 2.857E-5 -0.182 0.203 0.111
t 1.092 -3.119 2.002 -0.61286 0.444 -3.144 2.212
t-statistics
Coefficien -0.0073 0.114 -0.281 0.12 -0.535 0.112 0.094 -0.145 0.029 0.0067 0.172
t -0.541 0.554 -0.862 8 -2.108 0.417 0.129 -0.652 0.400 0.951
t-statistics 0.39
8

Table 6. Cross sectional regression results reporting the level and change in profitability from 2007 to 2019
Panel A: Regression results showing the degree of profitability, PR
Mean of Coefficient and t-statistics for the yearly regression from 2007-
2019
Intercept MA dd DPR R-Square
Coefficient 0.107 0.029 -0.136 -0.00033 0.439
t-statistics 19.224 12.760 -7.366 -3.069
Means and standard deviations of the regression variables
PR MA dd DPR
Means 0.128 1.414 0.061 33.302
Standard deviation 0.089 1.824 0.229 93.061
Panel B: Regression results showing the degree of profitability, ∆PR
Mean of Coefficient and t-statistics for the yearly regression from 2007-2019
Intercept PR EP NDP SNDP SPDP CP NCP SNCP SPCP R-Square
Coefficient 0.015 0.024 -0.161 -0.164 0.237
t-statistics 1.487 -0.162 -2.998 -2.632
Coefficient 0.027 -0.234 -0.209 0.213
t-statistics 3.920 -5.242 -3.473
Coefficient -0.0037 -0.344 0.139
t-statistics -0.872 -6.371
Coefficient -0.00014 -0.214 0.125 0.845 -0.980 0.239
t-statistics 0.055 -1.246 0.288 -0.041 -2.202
Coefficient 0.013 -0.248 0.133 0.079 0.188 -0.337 -0.00008 -0.136 0.488 -0.889 0.324
t-statistics 1.065 -1.389 0.814 0.244 0.281 -0.602 0.041 -0.580 -0.539 -0.776
Means and standard deviations of the regression variables
∆PR PR EP NDP SNDP SPDP CP NCP SNCP SPCP
Means -0.0034 0.127 0.128 -0.023 0.0016 0.0031 -0.0037 -0.018 0.0026 0.002
Standard deviation 0.074 0.086 0.062 0.032 0.0060 0.0195 0.068 0.047 0.018 5
0.018
3

Table 7. Cross sectional regression results reporting the level and change in profitability from 2007 to 2012
Panel A: Regression results showing the degree of profitability, PR
Mean of Coefficient and t-statistics for the yearly regression from 2007-2012
Intercept MA dd DPR R-Square
Coefficient 0.119 0.035 -0.142 -0.00042 0.381
t-statistics 20.624 12.234 -5.051 -3.530
Means and standard deviations of the regression variables
PR MA dd DPR
Means 0.139 1.051 0.036 29.437
Standard deviation 0.085 1.340 0.176 39.829
Panel B: Regression results showing the degree of profitability, ∆PR
Mean of Coefficient and t-statistics for the yearly regression from 2007-2012
Intercept PR EP NDP SNDP SPDP CP NCP SNCP SPCP R-Square
Coefficient 0.028 -0.294 0.062 -0.220 0.243
t-statistics 2.645 -4.883 0.705 -3.290
Coefficient 0.035 -0.284 -0.237 0.235
t-statistics 4.703 -5.944 -3.796
Coefficient -0.0046 -0.385 0.144
t-statistics -1.090 -6.924
Coefficient -0.0015 -0.171 -0.009 0.727 -1.826 0.216
t-statistics -0.128 -1.119 -0.206 -0.529 -2.335

Coefficient 0.026 -0.341 0.124 0.161 0.026 0.138 0.148 -0.352 0.624 -2.425 0.311
t-statistics 2.028 -1.748 0.781 0.570 0.143 -0.013 0.651 -1.173 -0.727 -1.752
Means and standard deviations of the regression variables
∆PR PR EP NDP SNDP SPDP CP NCP SNCP SPCP
Means -0.0037 0.138 0.139 -0.025 0.0016 0.0032 -0.0032 -0.019 0.0025 0.0027
Standard deviation 0.076 0.084 0.052 0.031 0.0036 0.019 0.068 0.047 0.016 0.018

The regression estimates of earnings in Table 8 show that when only lagged changes in earnings are used the coefficient is negative 0.00 with an
insignificant t-statistic. The fourth regression equation has the linear mean reversion variables PR and EP t. The results suggest that mean
reversion leads to predictable variation in profitability. The coefficient on Pt is negative 0.26 on average, with a t-statistic of -4.34, whilst the
coefficient on EP is 0.28 with a t-statistic of 2.98. The last regression equation includes all variables that show the non-linear mean reversion of
profitability. The coefficient on earnings increases to a positive number.

Table 8. Cross sectional regression results reporting the level and change in profitability from 2007 to 2019
Panel B: Regression results showing the degree of profitability, ∆PR
Mean of Coefficient and t-statistics for the yearly regression from 2007-2019
Intercept PR EP NDP SNDP SPDP CE NCE SNCE SPCE R-Square
Coefficient 0.016 -0.0000019 0.015
t-statistics 3.287 -1.559
Coefficient 0.017 -0.0000024 0.0000010 0.019
t-statistics 3.242 -1.409 0.295
Coefficient 0.018 -0.000009 0.000004 0.000 0.000 0.041
t-statistics 3.352 -1.719 0.441 -0.263 1.257
Coefficient 0.015 -0.261 0.283 -0.000006 0.000006 0.000 0.000 0.132
t-statistics 1.067 -4.338 2.981 -1.094 0.561 0.128 0.862
Coefficient -0.028 0.089 0.004 0.419 -0.972 -0.000005 0.000 0.000 0.000 0.011 0.219
t-statistics -0.259 0.497 -0.052 0.524 -2.287 -0.786 0.0073 -0.251 0.610 0.745

Table 9. Cross sectional regression results reporting the level and change in profitability from 2007 to 2012
Panel B: Regression results showing the degree of profitability, ∆PR
Mean of Coefficient and t-statistics for the yearly regression from 2007-2012
Intercept PR EP NDP SNDP SPDP CE NCE SNCE SPCE R-Square
Coefficient 0.021 -0.000003 0.0059
t-statistics 4.186 -1.239
Coefficient 0.022 -0.000003 0.000002 0.0069
t-statistics 4.105 -1.348 0.399
Coefficient 0.025 -0.000013 0.000004 0.000 0.000 0.019
t-statistics 4.195 -2.011 0.328 -0.565 1.622
Coefficient 0.026 -0.273 0.264 -0.000011 0.000008 0.000 0.000 0.079
t-statistics 1.627 -3.572 2.123 -1.538 0.525 -0.203 1.247
Coefficient 0.156 -0.090 -0.233 -0.125 -2.357 -0.000010 0.000001 0.000 0.000 0.016 0.136
t-statistics 0.345 -0.114 -0.255 0.116 -2.673 -1.542 0.211 -0.485 1.258 1.058
5. Conclusion

When firms design a new product and enter the market with the right positioning and
marketing strategy it leads to temporary competitive advantages over its competitors. These
competitive advantages lead to abnormal profits. But overdue course it is tough for
companies to sustain with these competitive advantages as there might be new product over
time or the competitors come with the same or a better product. This leads to a point that
companies cannot have extraordinary profits for a long period of time and will ultimately
revert to the industry average.
The present study explored the projected mean reversion of profitability and earnings,
following the idea suggested by Fama and French (2000) that changes in profitability and
earnings are somewhat predictable. The evidence presented in the paper is on the BSE listed
firms in India for the period from 2007 to 2020. The result shows that there is evidence of
mean reversion in profitability during the period from 2007 to 2012. A similar study was
done for the manufacturing sector with the same approach and the result found were similar.
Mean reversion was there from the year 2007 – 2012, however, there were no evidence of
mean reversion in profitability after 2012 to 2020. How these results are support or opposite
the earlier studies by Fama and French (2000), Allen and Salim (2005), and Altanbus (2008).
This can be related to the economic growth of the country. There was a political paradigm
shift from the year 2014 and the new policies and the ease of doing business for the firms in a
growing economy has resulted in no mean reversion during that period. The rate at which
these companies’ profitability reverts to average also depends on the intensity of the
competition. All the reversion has been from the higher end of the profitability spectrum;
reversal from the lower end has been negligible. This contradicts the notion that competition
forces businesses to restructure and improve their performance across the board. It's
reasonable to assume that the economy's much greater growth rate throughout the time
alleviated some of the competitive pressure on underperforming enterprises, even as
liberalization raised foreign competitive pressure on firms with supernormal profitability. To
persuade underperforming enterprises on the other end of the spectrum to improve their
competitiveness and performance is a major policy priority for India.
References

1. Allen*, D. E., & Salim, H. M. (2005). Forecasting profitability and earnings: a study
of the UK market (1982–2000). Applied economics, 37(17), 2009-2018.
2. Altunbas¸s, Y., Karagiannis, A., Liu, M. and Tourani‐Rad, A. (2008), "Mean
reversion of profitability: evidence from the European‐listed firms", Managerial
Finance, Vol. 34 No. 11, pp. 799-815.
3. Basu, S. (1997) The conservatism principle and the asymmetric timeliness of earning,
Journal of Accounting and Economics,24, 3–377
4. Collins, D. W. and Kothari, S. P. (1989) An analysis of inter-temporal and cross-
section determinants of earnings response coefficients, Journal of Accounting and
Economics,11, 143–81.
5. Fama, E. F. (1965). The behavior of stock-market prices. The journal of
Business, 38(1), 34-105.

6. Fama, E. F., & French, K. R. (2000). Forecasting profitability and earnings. The
Journal of Business, 73(2), 161-175.
7. Fama, E. F., & French, K. R. (2002). The equity premium. The Journal of
Finance, 57(2), 637-659.

8. Freeman, R. N.; Ohlson, J. A.; and Penman, S. H. 1982. Book rate-of-return and
prediction of earnings changes: An empirical investigation. Journal of Accounting
Research20(Autumn): 639–53.
9. Geroski, P. A., & Jacquemin, A. (1988). The persistence of profits: a European
comparison. The economic journal, 98(391), 375-389.
10. Goddard, J. A., & Wilson, J. O. (1999). The persistence of profit: a new empirical
interpretation. International Journal of Industrial Organization, 17(5), 663-687.
11. Lev, B. 1969. Industry averages as targets for financial ratios. Journal of Accounting
Re-search7 (Autumn): 290–99.
12. Lipe, R., & Kormendi, R. (1994). Mean reversion in annual earnings and its
implications for security valuation. Review of quantitative finance and
accounting, 4(1), 27-46.

13. Little, I. M. D. (1962) Higgledy piggledy growth, Oxford Bulletin of Statistics,24,


387–412.
14. MUELLER, D., 1977, 'The Persistence of Profits Above the Norm', Economica, 44
(November), pp. 369-380.
15. MUELLER, D., 1986, Profits in the Long Run (Cambridge University Press,
Cambridge). ORR, D., 1974, 'The Determinants of Entry: A Study of the Canadian
Manufacturing Industries', Review of Economics and Statistics, 61 (February), pp. 58-
66.
16. Ramakrishnan, R. T., & Thomas, J. K. (1992). What matters from the past: market
value, book value, or earnings? Earnings valuation and sufficient statistics for prior
information. Journal of Accounting, Auditing & Finance, 7(4), 423-464.
17. Rayner, A. C., & Little, I. M. D. (1966). Higgledy Piggledy Growth Again: An
Investigation of the Predictability of Company Earnings and Dividends in the UK,
1951-1961. Harvard University Press.
18. Welc, J. (2011). Is Sales Growth Of Companies Listed On The Warsaw Stock
Exchange Mean-Reverting?. International Business & Economics Research
Journal, 10(1).

You might also like