Professional Documents
Culture Documents
Market Risk, Accounting Data and Companies' Pollution Control Records
Market Risk, Accounting Data and Companies' Pollution Control Records
Market Risk, Accounting Data and Companies' Pollution Control Records
In recent years, the issues of corporate social performance and the social
responsibilities of business (and businessmen) have become the subjects of a
serious, and at times highly controversial, debate at all levels of society. The
accounting profession has not’been overloaked in this ongoing debate. In the
past few years accountants have been regularly exhorted to measure and report
on various aspects of corporate social performance including such issues as
environmental protection, discrimination in hiring and promotion, product
safety, and consumer policies, etc. - Beams and Fertig (1971), Siedler (1975),
Abt (1972), Linowes (1968, 1973), Colantoni, Cooper and Dietzer (1972),
Brummet (1973). Of particular significance in this regard is the proposal of the
Study Group on the Objectives of Financial Statements (1973, pp.53-55) that:
In response, several serious attempts have been made to take the first steps
towards developing a normative theory of accounting for corporate social
performance - Churchill (1974), Estes (1975), Ramanthan, (1976) - and the
question of whether companies should disclose more information (than they
now typically do) on their social performances is currently confronting
accounting policy-makers.
This paper seeks to provide some empirical evidence that may aid accounting
policy-makers t o form judgements on the social performance disclosure issue. To
accomplish this objective the following question is investigated. Does information
on the pollution control records of companies in pollution-prone industries
convey information to investors about the risk attaching to investment in these
companies’ common stocks? There is some evidence to suggest that this might be
the case. A major survey of institutional investor’s views on the perceived
relationship between corporate social performance and investment policy was
conducted for the Ford Foundation by Longstreth and Rosenbloom (1973). The
responses to this survey (1 15 responses from a possible total of 196) revealed
the existence of a relatively widespread view amongst institutional investors that
a moderate to strong association exists between the risk/return profile of an
The paper is structured as follows. The first section provides the context for the
study by outlining the research interest of a number of recent studies that have
investigated the association between market-based measures of risk and. various
other firm-specific variables. The second section discusses the selection of the
sample of companies used in the study and the collection of data. A discussion
of the variables used, the methodology employed, and the results of its
application is contained in the third subdivision.The final section presents a brief
summary and conclusions and makes some recommendations for future research.
The attention now being given to this topic stems from the central role that an
individual security’s systematic risk Pi (defied as the contribution an individual
security makes to portfolio risk) plays in the theory of portfolio mipagement
and the probability that conventional estimates of systematic risk (Bi based
exclusively on historical price series) will contain significant measurement errors.
This latter possibility arises because the process generating a security’s returns
may change over time (i.e., Pi may vary) as investors react to the receipt of
information on events they perceive to have firm-specific impacts. Changes in
management, financial policy or product lines, or exogeneous events such as the
passage of legislation or the promulgation of government regulations that may
68 Barry H . Spicer
have significant economic impacts on certain classes of companies, are examples.
In these circumstances, pi may not provide an unbiased and consistent estimate
of the “true” underlying pi and, hence, may not alone provide a sound basis for
the prediction of future levels of systematic risk. Moreover, evidence on the
stability (stationarity) of systematic risk indicates that while relatively stable at
the portfolio level (where the interest is in the behaviour of averages of pi’s over
time), at the level of the individual security, pi tends to change over time.
One criticism that has been leveled at association studies of this type is that they
are basically “fishing expeditions” which have failed to identify the basic
characteristics that determine risk because they are based on inadequate
theoretical investigations of the operational and financial processes affecting a
firm’s risk. However, this criticism seems overly harsh. The fact of the matter is
that the process by which investors judge the riskiness of common stocks is not
well understood. In the absence of a complete underlying theory, some searching
for significant associations by researchers can be expected as an initial step in an
attempt to explain the risk levels assigned by the market to common stocks and
to improve the prediction of future levels of risk.
The sampte selected for the investigation consists of 18 firms in the pulp and
paper industry. This particular sample was chosen for two reasons. First, a major
issue of social concern arising out of the operation of firms in the pulp and paper
industry is the resulting pollution of air and water resources and is a major
problem for companies in this industry. This conclusion is supportable in that
the pulp and paper industry is included amongst those industries that spend
substantial proportions of their capital and operating and maintenance budgets
on pollution abatement. In 1973 the pulp and paper industry spent 18.8 per cent
of its capital budget on pollution abatement (McCraw-Hill(l970)) and was
“Of the clouds that overhang the industry the question of costs of pollution
control and preservation of the environment is perhaps the darkest ... Overall,
it may be said that the question of pollution control is going to be exerting
considerable pressure on management of paper companies for some time. On
the immediate horizon is the threat of a zero pollution bill that would
prohibit any pollution by 1985. Versions of the bill have been passed by both
the House and Senate. It is estimated that such zero-discharge would require
an expenditure of $5 billion to $8 billion by 1985. Considering that the
industry’s total net worth is only $10.7 billion, this is not a good prospect”.
The second reason for choosing this particular sample is that the firms that make
up the sample were included among the 24 firms in the pulp and paper industry
whose pollution control records were the subject of a major study by the Council
on Economic Priorities CEP (1970). In this study the CEP, on the basis of
extensive data, rated each of the 13 1 pulp mills owned by study firms on the
efficacy of their air and water pollution control systems. This data is essential to
the empirical investigation because it provides the basis for the construction of a
pollution control index that provides a measure of a firm’s relative pollution
control record at a point in time. The CEP rating system and the derivation of a
pollution control index are discussed in the following section.
All financial data on the various companies were taken from the annual
COMPUSTAT tape. Monthly stock price data were obtained from the quarterly
COMPUSTAT tape supplemented where necessary by other sources.
Of the 24 companies studied by the CEP, four were excluded because each had
less than 25 per cent of their sales in the paper industry. While the 25 per cent
sales requirement is admittedly arbitrary, the rationale of using some such cutoff
is to include in the sample only those companies whose record in pollution control
in the paper industry is likely to be considered by investors in making investment
decisions. Two companies were excluded because their financial data were not
available on the annual COMPUSTAT tapes. The remaining 18 firms making up
the sample are listed on the New York Exchange and appear to be generally
representative of the larger firms in the pulp and paper industry. These firms are
listed below:
70 Barry H.Spicer
NAME OF COMPANY
The following subsecson discusses how the pollution control records and market
measures of risk were quantified for each of the sample companies. Also discussed
is the choice of the time period over which market measures of risk were
quantified. The manner in which other non-price corporate variables used in the
empirical investigation were quantified is explained as necessary in the discussion
of methods and results.
The ratings provided by the CEP were used in this study to develop a pollution
index for each company in the sample. This pollution index was based on the
percentage of a company’s pulp and paper productive capacity (tons/day) with
“adequate” pollution controls and was determined by taking a simple average of
the percentages of productive capacity with: (a) adequate primary water pollution
controls; (b) adequate secondary water ,pollution controls; (c) adequate controls
over particulate emissions; and (d) adequate controls over gas and odour
emissions. For example, if a company had a productive capacity rated at 5,000
tons/day of which: (a) eighty per cent has adequate primary water pollution
controls; (b) forty per cent has adequate secondary water pollution controls;
(c) fifty per cent has adequate controls over particulate emissions; and (d) thirty
per cent has adequate controls over gas and odour emissions, a pollution index
of fifty was assigned to that company. A company with adequate controls in all
categories would be assigned an index of one hundred whereas a company with
inadequate controls in all categories would be assigned a zero index.
72 Barry H. Spicer
Market measures ofrisk - Two market measures of risk are used in this study.
The first provides a measure of rota1 risk of an individual security and was
computed as the standard deviation of periodic stock returns for each company
in the sample over the 1968-1973 time period. The reasons for choosing this
particular time period are discussed below. The second measure of risk employed
is a measure of security’s systematic or non-diversifiublerisk. It was estimated
from the market or diagonal model as developed by Sharpe (1963), which asserts
a linear relationship between the return on a security and a market factor common
to all assets:
where Rit !i the return on security i in period t; a i and pi are constants for each
security i; Rmt is the aggregate return on all securities in the market in period t
and is referred to as the market factor; and Eit is a random disturbance term
which reflects that portion of a security’s return in time period t that is not a
linear function of Rmt.It is assumed that E(Eit) = 0; 4 R m t , Zit) = 0 for all i; and
cr(Pit, Ejt) = 0 for all i # j. The tildes on Rit, Rmt and Zit denote random variables.
The contribution of the individual security to portfolio risk, i.e., the security’s
systematic risk, is given by pi and is the relevant parameter of interest to investors
concerned with the selection of an optimal portfolio. Estimates of pis for the
time period under study were obtained in the usual manner by running a time-
series least squares regression on monthly price data for the period 1968-1973.
The measure of general market conditions (the market factor) utilized was
Standard and Poor’s Composite Price Index.
A measure of total as well as systematic risk was included for two reasons. First,
most non-market indicators or measures of risk (such as those based on accounting
data) are seen as measures of total rather than systematic risk - Beaver, Kettler
and Scholes (1970, p.659), Bildcrsee (1975, p.82). If this is so, then an improved
understanding of the determinants of total risk may lead also to an improved
understanding of the determinants of each of the systematic and individualistic
components of total risk. Along these lines, Beaver, Kettler and Scholes (1970,
pp.659-660) argue that it is reasonable to view accounting measures not only as
surrogates for total risk of a company’s common stock but as surrogates for
systematic risk as well. They base this argument on evidence that indicates a
The choice of the 1968-1973time period - This specific time period was chosen
for several reasons. First, it spans a period of significant growth in public pressure
for environmental protection and pollution control (c.f., Coppock et al. 1972)
in which a number of major and far reaching pieces of environmental legislation
were enacted.’ Second, the time at which the CEP released its 1970 report on
the pollution control records of sample companies is centered at the midpoint of
the 1968-1973 time period. If information on companies’ pollution control
records is relevant to investors’ risk assessments, then it seems reasonable to
expect the impact of this information to be reflected in measures of market risk
over the 1968-1973 time period. Furthermore, as it is likely that at least part of
the information compiled by the CEP 1970 report was previously available to the
market via press reports, company news releases, etc., the assumption made in
this study is that market risk, if impacted, would be impacted over the entire
1968-1973 time period. No attempt was made to ascertain whether changes in
risk lagged the public availability of information on companies’ pollution control
records or vice versa. Rather the assumption is that the impact on market risk
occurred over the entire 1968-1973 time period.* In this context, a reasonable
interpretation of a company’s pollutipn control index based on the data
contained in the CEP’s 1970 report, is that it provides a measure of the
responsiveness of sample companies to public pressure for pollution abatement
over the study period.
74 Barry H. Spicer
do not allow rejection of a null hypothesis of no association at a .05level of
significance for a one-tailed test, they are, in each case, reasonably close to the
margin of significance.
TABLE 1
SPEARMAN RANK ORDER CORRELATIONS
To extend the correlational analysis, two sets of partial correlations also were
calculated between each measure of risk and the measure of companies’
pollution control records, i.e., the pollution index based on productive capacity.
In each case, a different set of control variables was used. These control variables
were drawn from previous studies that had shown them to be associated
significantly with market measures of risk. Removal of the effects of variation in
these variables from the measure of association between the variables of interest
here, should provide at least a preliminary indication of the potential contribution
the pollution control variable can make towards the explanation of the observed
variation in market measures of risk.
The first set of partial correlation coefficients reported in Table 2 control for
those accounting-based measures of risk used by Beaver, Kettler and Scholes
(1970) as instrumental variables in the prediction of future systematic risk on
the basis of past estimated systematic risk. These included earnings variability,
average asset growth and average payout. Computation of these variables for
the sample period 1968-1973 followed the definitions given by Beaver, Kettler
and Scholes (1 970, p.666). While seven accounting variables originally were
included in the regression equation by the researchers, due to multi-collinearity
problems the final instrumental equation contained only the three accounting
measures mentioned above. The seven accounting variables originally included
were payout, leverage, liquidity, size, earnings variability and an accounting beta
which measured the covariability of each firm’searnings with that of an overall
market average of earnings.
TABLE 2
PARTIAL CORRELATIONS BETWEEN COMPANIES’ POLLUTION
CONTROL RECORDS AND MARKET MEASURES OF RISK
aAccounting based variables used by Beaver, Kettler and Scholes (1970) as instrumental
variables in the prediction of future systematic risk on the basis of past estimates of
systematic risk.
hariables found by I,ev and Kunitzky (1974) to be significantly associated with the risk of
common stocks.
(1) For each measure of market risk,a stepwise multiple regression was run to
select from among a list of nine accountingbased variables those three variables
which explain the greatest proportion of’the observed variation in the dependent
variable (total risk or systematic risk) over the sample period.
The stepwise regression program used was that contained in Nie, Hull, Jenkins,
Steinbrenner and Bent (1975, Ch.20). In this program, the independent variable
chosen for entry at each step is the variable not yet in the regression equation
which has the largest square partial correlation with the dependent variable.
That is, the variable chosen to enter is the one which explains the greatest amount
of the variance in the dependent variable which is unexplained by variables
already in the equation. The nine accounting variables from which this selection
was made were:
(b) Size - average of the natural logarithms of the yearly total asset size.
(c) Leverage - average of the annual ratios of senior securities to total assets.
(d) Growth - natural logarithm of the ratio of total closing asset size in year
1 to the total closing asset size in year t.
(e) Coverage - average of the annual ratios of net income plus fmed charges
to fmed charges.
(f) Turnover - average of the annual ratios of net sales to total assets.
TABLE 3
MULTIPLE REGRESSION ANALYSIS: TOTAL RISK ON ACCOUNTING
DATA AND A MEASURE OF POLLUTION CONTROL
Stepwise Regression iddition of Pollution Addition of Pollution
to Find Best Three :ndex to the Accounting Ratio to
Accounting Ratios iegression the Regression
Independent
Variable Zoefficient t-value "efficient t-value
-
Constant .0584 1.94 ,0875 2.91
Earning variability .3846 4.01 .4123 3.81
Size -.0069 - ,0002 .05 -.0058 1.88
Leverage .0382 .0811 2.65 .0379 1.67
Pollution index - ,0004 1.86
(productive capacity)
Current ratio .0031 .61
Summary statistics
Coefficient of multiple
correlation ~2 ,677 .749 ,688
Corrected R2 ,631 .692 .6 17
F-rat io 9.08 8.98 6.63
Proportionate increase .225 ,035
in explained variation
due to addition of
independent variable
78 Barry H. Spicer
which of the two variables - the pollution index or the next best accounting
ratio - explained the greatest proportion of the variation in total and systematic
risk not explained by the regression containing the best three accounting variable
predictors.
The results of these various regressions are set out in Table 3 and Table 4. The
results in Table 3 are for the dependent variable total risk.
As can be seen from Table 3, the initial stepwise regression selected three variables:
earning variability, size and leverage. The addition of the pollution control
variable (which enters the re ression with the correct sign) results in an
improvement in corrected The squared partial correlation coefficient
between total risk and pollution control variable when earnings variability, size
and leverage are controlled for is .225. This indicates that the increase in
explained variation expressed as a proportion of unexplained variation is 22.5
per cent. This is significantly higher than the proportional increase in explained
variation of 3.5 per cent that results from the addition of the next best account-
ing variable (the current ratio), to the regression equation. It is interesting to
note that an unrestricted stepwise regression to find the best four predictors of
total risk from all nine accounting variables and the pollution control index based
on productive capacity, resulted in the same regression equation as in column 2
of Table 3.
TABLE 4
MULTIPLE REGRESSION ANALYSIS: SYSTEMATIC RISK ON
ACCOUNTING DATA AND A MEASURE OF POLLUTION CONTROL
Stepwise Regression 9ddition of Pollution Addition of Next Best
to Find Best Three Lndex to the Accounting Ratio to
Independent
Variable
Accounting Ratios
Coefficient
h e Regression
- -
h e Regression
Constant 1.3841
t-value
2.13
Zoefficient
1.2640
-
t-value 3efficient
1.91 1.412
-
t-value
2.13
Earning variability 4.SO41 1.97 4.869 1.99 3.519 1.18
Size -.lo28 1.44 -.0153 .5 8 -.0914 1.24
Current ratio .1568 1.34 ,1323 1.1 1 .2128 1.53
Pollution index -.0039 .98
(productive capacity)
Payout -.4317 .78
Summary statistics
Coet'ficient of multiple
correlation ~2 .433 .475 .461
Corrected R2 .352 .354 .336
F-rat io 3.3 1 2.12 2.56
Proportionate increase .075 .048
in explained variation
due to addition of
independent variable
- -
Market risk, accounting data and companies 'pollution control records '9
The results in Table 4 are for the dependent variable systematic risk.
In this case the initial stepwise selected three accounting variables: earning
variability, size and the current ratio. The addition of the pollution control
variable (which once again enters the regression with the correct sign), results in
only a marginal improvement in corrected R” The squared partial correlation
coefficient between systematic risk and the pollution control variable when
earning variability, size and the current ratio are controlled for, is .075,indicating
that the proportional increase in explained variation is 7.5 per cent. This may be
compared to the 4.8 .per cent increase which results when the next best account-
ing variable, the payout ratio, is added to the regression.
To summarize, the results of the multiple regression analyses, like the results of
the preceding partial correlation analysis, also provide some support for the
contention that information on companies’ pollution control records may be
relevant to investors for assessing both the total and systematic risk of common
stocks. However, one interesting aspect of the regression analysis should be
noted. As can be seen from both Tables 3 and 4 when the pollution control
variable is forced into the regression there is a sharp decline in the coefficient
and significance of the size variable in the regression equation. Statistically, this
is due to the fact that t h e e two variables are highly correlated. (Pearson’s
product moment correlation is .64). One plausible explanation of this result may
be that risk is closely related to size which in turn affects the availability of
resources to the company and hence its ability to finance the purchase and
installation of pollution control equipment.
One avenue for future research would be to determine whether the passage of
major environmental control legislation (which establishes a basis for judging
80 Barry H. Spicer
companies’ pollution control records) has an observable impact on the market’s
perception of the level of risk (total and systematic) attaching to the shares of
companies in pollution-prone industries. If it can be established that investors
(in aggregate) do react to information about events pertinent to the determination
of companies’ pollution control records (to the extent such information is
publicly available) then a stronger case can be made for increasing the content of
this information by (1) experimenting with alternative specifications, measures
and reporting formats for companies’ pollution control records and/or (2)
increasing the accessibility of this information by reducing its cost to investors.
NOTES
At the federal level major enactments include: The Clean Air Act of 1970, the Federal
Water Pollution Control Act Amendments of 1972, the Resource Recovery Act of 1970
and the National Environmental Policy Act of 1969.
It may be noted at this point that &s for sample companies were subjected to a relatively
simple test for stationarity. This was accomplished by an F test of the null hypothesis of
no difference for each company’s &s computed for equal time periods before and after
1970. It was possible to reject the null hypothesis for only one of the 18 sample
companies.
’ However, the F test for the contribution of the pollution control variable to the
regression equation was not significant at 01 = .05 for either the regression on total risk or
the regression on systematic risk.
REFERENCES
Abt, Clark. “Managing Money by Doing Good”. Innovation, No.27.1972.
Ball, Ray and Brown, Phiiip. “An Empirical Evaluation of Accounting Income
Numbers”. Journal of Accounting Research, Autumn, 1968.
Beams, Floyd A. and Fertig, Paul E. “Pollution Control Through Social Cost
Conversion”. The Journal of Accountancy, November, 1971.
- . Kettler, Paul and Scholes, Myron. “The Association Between Market Determined
and Accounting Determined Risk Measures”. The Accounting Review, October, 1970.
Bierman Jr., Harold. “The Implications to Accounting of Efficient Markets and the
Capital Asset Pricing Model”. The Accounting Review, July, 1974.
Breen, William J. and Lerner, Eugene M. “Corporate Financial Strategies and Market
Measures of Risk and Return”. The Journal of Finance, May, 1973.
Colantoni, C.S., Cooper, W.W. and Dietzer, H.J. “Accounting and Social Reporting”.
Joint Urban Sciences and Information Institute Research Report #8, Carnegie-Mellon
University, August, 1972.
Coppock, R., Dierkes, M.,Snowball, H., Thomas, J., “Social Pressure and Business
Actions”, paper presented to the Seminar o n Corporate Social Accounts, Battelle
Seattle Research Center, November 10-11, 1972.
- and Kunitzky, Sergius. “On the Association Between Smoothing Measures and the
Risk of Common Stocks”. The Accounting Review, April, 1974.
82 Barry H . Spicer
Decisions as Reflected in Common Stock Price Changes”. Empiricai Research in
Accounting: Selected Studies, 1971, Supplement to Vo1.9, Journal of Accounting
Research.
(25) Nie, N.H., Hull, C.H., Jenkins, J.G., Steinbrenner, K.,and Bent, D.Statistical Packclge
for the Social Sciences. 2nd ed. New York: McGraw-Hill Book Co., 1975.
(27) Rosenberg, Barr and McKibben, Walt. “The Prediction of Systematic and Specific Risk
in Common Stocks”. Journal of Financial and Quantitative Analysis, March, 1973.
(28) Shape, William. “A Simplified Model for Portfolio Analysis”. Management Science,
January, 1973.
(29) Seidler, Lee J. “Dollar Values in the Social Income Statement”. Social Accounting:
Theory, Issues and Cases. Edited by Lee I. Seidler and Lynn L. Seidler, Los Angeles:
Melville Publishing Company, 1975.