1

ISO 14001 Certification and Environmental Performance in Quebec’s Pulp and Paper Industry

Philippe Barla* GREEN and Département d’économique, Université Laval

* Département d’économique, Université Laval, Québec, Canada, G1K 7P4, E-mail : philippe.barla@ecn.ulaval.ca, tel : +1 418-656 7707, fax : +1 418 656 7412.

2

ISO 14001 Certification and Environmental Performance in Quebec’s Pulp and Paper Industry

Abstract

This paper tests whether adopting the international norm ISO 14001 significantly impacts environmental performance in Quebec’s pulp and paper industry. Using monthly data collected from 37 plants between 1997 and 2003, we show that: i) ISO certification does not lead to a reduction in total suspended solid emissions or the total quantity of rejected process water; ii) discharge of biological oxygen demand appears to be significantly lower in the first year following certification; iii) this latter impact does not appear to last beyond the one-year window. We further show that, contrary to the group of plants that did not adopt the ISO norm, the adopting plants did not experience a significant negative trend in emissions over our sample period. Key words: Environmental Management Systems, ISO 14001, Environmental Performance. JEL Classifications: Q50, Q52, Q58

3

ISO 14001 Certification and Environmental Performance in

Quebec’s Pulp and Paper Industry

1. Introduction

Environmental policy has considerably evolved over the last decades. While in the seventies,

command and control regulations were the norm, incentive-based mechanisms such as tradeable

permits became increasingly popular in the eighties. More recently, governments and industry

lobby groups have favoured recourse to voluntary approaches when dealing with environmental

challenges. The basic idea is that businesses would adopt efficient pro-active environmental

initiatives in the wake of increased public pressure and the threat of more stringent governmental

regulations. In addition to lowering abatement costs, such voluntary measures would also help

reduce environmental policy monitoring and enforcement costs. This of course begs the question

as to whether these voluntary actions do or do not reduce pollution.

The growing popularity of the international standard ISO 14001 relating to

implementation of Environmental Management Systems (EMSs) is an excellent illustration of the

trend toward voluntary self-regulation. By December 2003, over 36,000 organizations

worldwide had voluntarily established an ISO 14001-certified EMS. An EMS may be viewed as

a set of management rules and procedures designed at reducing the environmental impacts of an

organization. It involves, among other things, reviewing and documenting an organization’s

activities that negatively impact the environment, developing an environmental policy statement

and a plan to achieve environmental objectives. However, these targets are, for the most part,

decided internally by the organization. In other words, the ISO norm does not prescribe any

specific objective except to respect existing regulations. Therefore, opponents have criticized

4

ISO 14001 as a potential marketing device for improving corporate image without significantly

impacting environmental performance [18]. In this paper, we test the impact of ISO 14001 on the

effluents of 37 Quebec pulp and paper plants using monthly data over the period 1997-2003. We

compare the environmental performances of ISO 14001 plants before and after certification and

the performance of plants that are not (yet) certified. Our results suggest that while ISO 14001

certification does not significantly impact total suspended solids (TSS), it does reduce plant

discharges of biological oxygen demand (BOD) in the first year following certification.

However, this impact does not appear to last beyond the one-year window. Furthermore, plants

that did not adopt the ISO norm during our sample period experienced an overall significant trend

in emissions reduction (TSS and BOD) contrary to plants that adopted the norm. Moreover, the

ISO certification is not associated with a reduction in the total quantity of process water rejected.

From a methodological point of view, our analysis stresses the importance of properly controlling

for plants heterogeneity when testing for the impact of voluntary initiatives.

The empirical literature on corporate environmentalism is still quite limited and has

primarily focused on the adoption determinants of self regulations (see for example [9], [10],

[15]). Nevertheless, a few recent works have examined the impact of voluntary initiatives, and

particularly EMS, on environmental performance. Using a sample of about 150 US companies,

Anton et al. [2] find that the level of comprehensiveness of an EMS has a significantly negative

impact on toxic release emission rates. The reduction is particularly important in the case of

firms with a history of high pollution intensity. A survey of 236 Mexican manufacturing plants

conducted by Dasgupta et al. [4] shows that plants with a high degree of conformity to ISO

guidelines are more likely to report (self-assessed) compliance with environmental regulations.

However, as both studies used data preceding official publication of the norm, they do not

directly test the impact of ISO certification. Therefore, our main contribution is specifically to

5

test whether the ISO 14001 certification of an EMS significantly impacts environmental

performance. Our analysis also avoids comparing very heterogeneous organizations by

focusing on plants belonging to the same industry. Moreover, the panel structure of our dataset

allows us to control for unobservable plant-specific factors that may be correlated with the

adoption decision, thereby reducing missing variable biases.

This paper also draws from the literature studying the determinants of environmental

performance. Several studies have examined the impacts of monitoring and enforcement

activities (inspections, legal proceedings, fines) on environmental performance

([5],[8],[12],[13],[14]). More recently, [7] compares these traditional enforcement strategies with

those based on information disclosure (i.e. publication of “worst” polluters’ lists). Of particular

relevance to our analysis is a study by Laplante and Rilstone [13] showing that inspections

significantly reduce emissions in Quebec’s pulp and paper industry.

Section 2 provides a general background on the ISO standard describing how it may

affect environmental performance and a brief depiction of Quebec’s pulp and paper industry.

Section 3 contains a description of the data and some preliminary evidence. The empirical model

and results are presented in Sections 4 and 5, respectively. We conclude in Section 6.

2. Background

2.1 The ISO 14001 standard

ISO 14001 requirements are set forth in a short document entitled Environmental management

systems – Specification with guidance for use published by the International Organization for

Standardization (ISO). It was officially introduced in 1996 after having been developed for a few

years by a technical committee composed of “environmental experts” representing over 50

national standards organizations. The ISO 14001 standard is part of a set of guidelines that

6

together constitute the ISO 14 000 series. However, ISO 14001 is the only mandatory guideline

that can be audited for certification (or registration) by an accredited third party. These

guidelines are applicable to any kind of organization of various nationalities and size.

We now briefly describe the main requirements contained in the ISO 14001 standard (for

a detailed description see [19]). To be ISO 14001 certified, an EMS must be established and

operated using a five steps approach. First, the organization’s senior management must adopt an

environmental policy statement containing an explicit commitment i) to comply with all

applicable regulations and other obligations and ii) to continuously reduce and prevent pollution.

The statement should also provide a framework for establishing the organization’s environmental

targets. Second, at a planning stage, the organization reviews the significant environmental

impacts of its activities, identifying all legal, contractual and voluntary environmental obligations

and establishes procedures for meeting them. On the basis of these reviews and the policy

statement, the organization then defines objectives and targets to be achieved through an

implementation plan. This plan must assign responsibilities within the organization and describe

resources dedicated to the plan as well as an implementation schedule. Third, the plan is

executed through such a means as worker training programs or development of operational and

communication procedures designed to prevent accidental pollution. Procedures must also be set

up to properly document EMS activities. Fourth, procedures and routines must be designed for

controlling and monitoring the organization’s environmental impacts and corrective procedures

must be established to deal with cases of non-conformity. The EMS should also develop detailed

procedures (frequency, methodology) for regular audits. To be certified, an accredited third party

must regularly audit the organization’s EMS verifying its adherence to ISO requirements. Fifth,

senior management periodically re-evaluates the EMS operations, modifying them as necessary

to ensure continuous improvement of their effectiveness.

7

Therefore, an ISO-14001 certified EMS may improve environmental performance

through: i) respecting all applicable environmental regulations; ii) documenting and analyzing the

plant’s environmental impacts; and iii) systematic, written and standardized checklist-type

procedures towards reducing and preventing pollution. In our analysis, aspect i) is likely of little

importance since, as shown below, the regulatory compliance rate is particularly high in

Quebec’s pulp and paper industry.

Several potential benefits may justify a decision by businesses to adopt the ISO 14001

norm. First, it may improve a firm’s corporate image among consumers, investors and the

surrounding communities affected by environmental externalities. Empirical studies (see [2] and

[9]) and managers’ surveys [21] suggest this is a significant factor favouring implementation of

an EMS and its ISO certification. Being certified may also help firms reduce liability risks by

demonstrating “due diligence”, thereby reducing insurance costs. For example, [2] and [9] show

that US companies with a large number of Superfund sites (a proxy for liability cost awareness)

are more likely to adopt comprehensive EMS. Evidence also suggests that EMS adoption may

result from regulatory pressures in the wake of regulator inspections, for example (see [4] and

[9]). In some cases, regulations may provide direct incentives for adopting an EMS and its ISO

certification (see [11]). Moreover, adoption of voluntary measures by a significant portion of an

industry could pre-empt more stringent environmental regulations. The process required by the

ISO standard may also lead to cost savings through improved input productivity or reduced waste

disposal and pollution abatement costs (for case study see [3]). Lastly, certified companies may

pressure suppliers to become certified.1

1 So far, empirical evidence suggests this effect is not particularly strong (see [4], [15] and [21]). However, it may become increasingly important as the number of certified plants continues to rise. For example, Ford Motors is now requiring ISO 14001 3rd party certification from its suppliers with manufacturing facilities.

8

Implementing and maintaining an ISO 14001 EMS involve internal costs, chiefly related

to administrative, external auditing and marketing expenditures. Szymanki and Tiwari [18]

reports costs of up to US$ 100,000 per year for US companies. In Canada, Yiridoe et al. [21]

found that, on average, organizations spend roughly 2% of total expenditures on obtaining and

maintaining ISO 14001 EMS. Their estimates also suggest significant economies of scale in

EMS adoption costs, with the initial implementation cost varying from approximately CND$

40,000 for small organizations (with fewer than 100 employees) to CND$ 75,000 for larger

organizations (more than 500 employees). In fact, scale is a very significant factor for explaining

the probability of adopting an EMS. Other factors increasing the likelihood of EMS adoption

through reduced implementation costs are: i) past experience with the ISO 9000 quality

management standard and ii) the average level of worker education [15]. On the other hand, a

firm’s level of debt reduces the probability of adoption.

2.2 Quebec’s pulp and paper industry

Over the past century, a thriving pulp and paper industry has developed in Quebec fuelled by

access to abundant forests and inexpensive energy. On December 2003, 54 plants belonging to

18 companies employed over 35,000 employees. Quebec produces approximately 30% of the

Canadian and 3% of the world paper and cardboard production. In terms of newsprint, Quebec is

responsible for 10% of worldwide production. Close to 60% of Quebec’s paper production is

exported to the US.

Besides its economic importance, this sector is also a major source of pollution,

particularly effluent discharge. Given the availability of data (see Section 3), our analysis is

focused on two conventional water pollution indicators for this industry, namely TSS and BOD.

TSS is a direct measure of the quantity of solid waste (wood fibbers, ashes etc.) rejected in

9

production water, while BOD is an indirect indicator of the environmental impact of effluent.2

Primary treatment using gravity is useful to control TSS, while a secondary treatment based on

biological processes is needed to reduce BOD. Both aspects may also be controlled up stream by

reducing effluents through the recycling of process water. The industry is also responsible for

other water pollutants such as BPC, dioxins and furans as well as air emissions (particles, SO2,

sulphates, greenhouses gases) and solid wastes.

Since the early seventies, the Quebec industry has been submitted to progressively more

stringent environmental federal and provincial regulations leading to significant abatement

investments. For example, between 1992 and 1999, the industry spent over 1.5 billions dollars on

improving environmental performance (62% was directed toward water pollution control). For

the most part, regulations have taken the form of industry specific performance standards. The

latest revision of norms pertaining to BOD and TSS dates back to 1992. The norms are

established as kilograms per ton of production and involve limits on both the average monthly

and daily maximal discharge. These limits are more stringent for plants built after 1992 and do

not apply to plants releasing effluents into municipal waste water systems (for an overview see

[13]). During our sample period, the compliance rate with regulations for BOD and TSS always

exceeded 95%.

Interestingly, in 1993, the provincial government introduced a gradual process of plant-

specific environmental certification involving some of the steps required by ISO 14001. All

plants have now received their first certification, which is valid for five years. This first

generation of attestations gathers in one document all environmental regulations plants must

respect and describes a process for characterizing plant effluent over a one-year period using

2 BOD is an indicator of the total quantity of oxygen required over a five-day period to decompose the organic matters contained in waste water. High BOD may damage the environment by asphyxiating aquatic life.

10

detailed indicators. This evaluation should eventually be used to define new plant-specific

environmental targets to be included in the second-generation attestation. These new regulatory

requirements have provided strong incentives for plants to actually implement an EMS. In fact,

the Quebec Forest Industry Council [16] reports that in 2001 all Quebec plants were operating an

EMS. Unfortunately, no information on non-ISO certified EMS was available. Therefore, in

this paper, we specifically test the impact of the ISO certification of an EMS rather than the effect

of adopting an EMS. This precision is important to bear in mind when interpreting our results.

3. Data and Preliminary Evidence

The main source of the data used in this paper is the Quebec Ministry of the Environment, which

collects information to assess industry conformity with existing environmental regulations [17].

Under the Règlement sur les fabriques de pâtes et papiers, plants must continuously measure

their effluent and report daily discharges of TSS and BOD. Obviously, the validity of these self-

reported discharges could be questioned. However, there are a number of reasons to believe

these data are relatively accurate [13]. First, the Ministry realizes each year five inspections to

validate the accuracy of the reported data and five inspections to review plants’ pollution

monitoring equipments. During the three year period for which information was available (2001-

2003), the conformity rate was 100%. The Ministry also conducts twenty toxicity tests on an

annual basis, to validate the data reported by plants (conformity rates were 90%, 95% and 100%

for 2001, 2002 and 2003, respectively). Second, the technologies used by these plants as well as

their production capacities are relatively well known by the regulator, which makes reporting

grossly inaccurate figures difficult. Third, in the past, regulatory conformity rates have not

always been as high as those for our sample period suggesting that plants do report non-

compliance. The Ministry database also includes information on production processes, types of

11

output produced and production for each plant. Unfortunately, this last information is

confidential. However, based on other variables, we were able to determine annual output, which

combined with the monthly water flow, helps us construct a proxy for monthly production (see

Section 4).

The date of ISO certification was obtained from the ISO 14001 Registered Company

Directory Database produced by the QSU Publishing Company. The information was updated

and validated by directly contacting the plants.

Our sample includes all Quebec plants in continuous operation during the 1997-2003

period. In other words, we eliminated plants that closed during our sample period.3 We

eliminated plants that produce paper-based building materials as they operate in a rather different

market. We also excluded plants that release their effluent into municipal water systems, as these

plants are not in charge of pollution control of their effluent and thus not subject to the same

regulations. Therefore, our sample consists of observations pertaining to 37 mills over 84

months.4

Figure 1 presents the pattern of ISO 14001 certification over our sample period. The first

two plants were certified in 1998 and, by December 2003, 18 plants were operating an ISO 14001

EMS. Our sample includes 548 ISO 14001 observations from a total of 3,108. Figures 2 and 3

compare average emission rates (i.e. BOD and TSS emissions per output unit) for the group of

plants that became ISO during our sample period (henceforth referred as adopters) and the group

that did not (non-adopters).5 Two noteworthy points at this stage: i) adopters appear to have

higher emission rates; and ii) while for TSS, the adopters appear to be catching up with the

3 However, we included plants with considerable output reduction caused by strikes. Also, for nine observations, we used a regression on 12-month lags to complete the missing information. 4 We also use the 1996 data when lags are required (see Section 4). 5 The average for the adopter group includes observations for plants not yet ISO accredited, but for which certification was obtained during our sample period.

12

environmental performance of non-adopters, the opposite appears to be true for BOD.

Obviously, several factors other than ISO certification may determine emission levels.

4. The Empirical Model

Our empirical specification is inspired by former empirical analyses of environmental

performance determinants, especially [13] and [14]. It is also guided by the theoretical framework

proposed by Dasgupta et al. [4] to explain variations in the environmental performances of plants

submitted to similar regulations. In this context, the cost-minimizing level of a plant’s emissions

is determined by comparing the Marginal Abatement Cost (MAC) with the Expected Marginal

Penalty (EMP). Figure 4 illustrates the optimal level of pollution. The MAC is quite standard

and reflects the increasing marginal cost associated with reducing emissions. EMP captures the

expected price of emissions, which increases with emission levels reflecting the increasing

pressure on the plant as pollution augments even if it fully complies with official regulations.

Indeed, as emissions increase, not only may regulatory scrutiny intensify (e.g. marginal price

increases as a result of more frequent inspections by the regulator) but also unofficial pressure

from consumers, investors and local communities is likely to intensify. Therefore, the

equilibrium level of emissions will depend upon the various factors affecting the MAC and EMP.

Based on this framework, our baseline reduced form emission equations take the

following form:

)1(

___

)()()(

,12

1,

3

1,

8

1,,3,2,1

11

1,,3,2,112,0,

tir

irl

ill

jijjtititi

ktkktititititi

REGIONQTYPE

PROCESSSTRIKEADOPTERSTADOPTERSNONT

MONTHQLogINSPISOELogELog

ηθ

λγγγ

δβββφα

+++

++++

+++++=

∑∑

∑

∑

==

=

=−

13

The variable Ei,t measures discharges of either BOD or TSS (in tons) by plant i over month t. We

also run alternative specifications based on emission rates (i.e. emissions per production unit).

Following [13], [14], the emission level is regressed on its 12-month lag to capture the underlying

inertia associated with installing new equipment or adjusting the production process. The

variable ISO is a dichotomous variable taking the value 1 if plant i is ISO 14001 certified at time

t, and 0 otherwise. As mentioned above, implementation of an ISO certified EMS implies

reviewing the production process and developing systematic procedures, which may cause a

downward shift in the MAC curve, thereby reducing the pollution equilibrium level.

Certification could also lead to a downward shift of EMP by reducing liability exposure, thereby

inducing an increase in emissions.

Several studies have demonstrated a significant impact of official regulatory pressure on

environmental performance (see [7],[13],[14]). Increased regulatory scrutiny is likely to cause an

upward shift in EMP, thereby leading to pollution reduction. In keeping with these research

findings, we introduce the INSP variables measuring the number of toxicity inspections over the

past six months at plant i. Each year, the Ministry of the Environment conducts twenty toxicity

inspections, which consist in immersing 10 rainbow trout in a plant’s undiluted effluent for a 96-

hour period. The test fails if more than 50% of the trout die. These inspections may trigger an

overall improvement in plant performance including BOD and TSS emissions.6 The absolute

level of discharge also depends upon plant production Qit. As previously mentioned, plant

production data are unfortunately confidential. However, we used other environmental data to

6 We tried several alternative specifications for capturing the impact of these inspections (e.g. current inspection plus lags). The results of these alternative specifications are very close to those reported here, especially for the variable of interest in this paper, namely ISO. We also tried specifications that include the other type of inspections (accuracy of self-reported emissions and inspections of pollution-monitoring equipment). These inspections did not significantly impact emissions.

14

deduce annual production. To construct Qit, we combine annual production data with monthly

level of water flow, which should reflect variation in production. Specifically, for a given year,

∑=

= 12

1,

,,

ssi

tiiti

W

WYQ

with Yi plant i’s annual production and Wi,t the amount of water rejected by plant i during month

t.

Seasonal effects are captured by introducing month-specific dummies (MONTH).

Temperature variations may affect efficiency of pollution control equipment (shifting the MAC)

and public pressures (thus shifting EMP), as water pollution may be more problematic during the

warm season when rivers are used for recreational activities. In light of the preliminary evidence

reported in Figures 2 and 3, we allow the time trend to differ for adopters and non-adopters.

More precisely, we define the binary variable P_ISO, which takes value 1 (for all t) if plant i

eventually becomes ISO during our sample period and zero otherwise. T_ADOPTERS is equal to

TREND x P_ISO while T_NON_ADOPTERS = TREND x (1-P_ISO). Ignoring this possibility

could bias our results since the variable ISO may be picking up trend differences between the two

groups (ISO takes the value 1 mostly at the end of our sample period). If there is no difference,

the coefficients on these two variables will not be statistically different. The production of a few

plants was disturbed by strikes that lasted several weeks or months. We control for these events

by including a binary variable STRIKE. To control for type of production process, we include

eight dummies (PROCESS) that are further defined in Table 1. We distinguish four types of

outputs (QTYPE), namely newsprint (reference group), pulp, cardboard and other paper.7

Controlling for technology and nature of output produced is important as they may affect

7 Note that most plants produce several output types, implying several QTYPE being set to one.

15

emissions by shifting the MAC. The variables REGIONS indicates the plant’s area of operation.

It is based on the 17 administrative divisions of Quebec. The plants in our sample are located

throughout 13 regions. These variables may capture variations in EMP. For example, plants

located in densely populated areas are likely subject to more intense pressure to control

emissions.

Finally, ti,η represents the error term. Given the panel structure of our data, we assume

that tiiti ,, εµη += i.e. is composed of a plant-specific component iµ and a plant-time varying term

ti,ε that is assumed, at this stage, IID Normal ti,ε ~N(0, 2εσ ). iµ accounts for any plant-specific

unobservable factor. It also accounts for unobservable factors that varied little during our time

period but that may be quite different across plants; for example, the socio-economic

characteristics of the surrounding communities (education or income levels), the importance of

R&D activities, the average age or education level of plant workers.8 These variables may affect

a plant’s environmental performance by shifting the EMP or MAC. In the absence of these plant

effects, the model could be estimated by OLS. If they are present, the appropriate estimation

procedure depends upon whether these effects are correlated or not with the included variables.

In the latter case, they may be treated as random effects (RE) and the model may be estimated by

GLS. Otherwise, they should be treated as fixed effects (FE) (for details see [20]).9 In the FE

specification, the coefficients on variables that are time-invariant, such as REGION, cannot be

identified. If we allow for the possibility that plant effects are correlated with the included

8 See [2] and [4] for evidence on the impact of these variables using cross-section data. 9 In short panels, the introduction of lag-endogenous variables (e.g. Log(Ei,t-12)) as an explanatory variable could lead to biased estimators (see [20]). However, this should not be a major issue here as T is relatively large (T=84) (see [20]). Furthermore, in our case, the cure proposed in the literature could be worse since it is based on first differencing the model in order to eliminate plant-specific effects and then proceeding with an instrumental-variable procedure. First differencing may be problematic since our variable of interest (ISO) is a binary variable. Indeed, (ISOi,t-ISOi,t-1) will only be one when the plant becomes certified, and zero otherwise. If the impact of ISO on emissions does not occur precisely at the time of certification, we will be unable to detect its impact.

16

explanatory variables, we maintain that all explanatory variables, including ISO, may be treated

as strictly exogenous.10 We come back on this hypothesis in Section 5.2.

Before turning to the estimation results, it is useful to examine some descriptive statistics.

Table 1 reports the variables mean and standard deviation both for the whole sample and for the

two groups of plants in our sample, i.e. adopters and non-adopters. These figures confirm that

ISO adopters are polluting somewhat more than non-adopters. However, this could be linked to

differences in the two groups’ characteristics. Adopters are larger (both in terms of production

and water flow) and tend to be producers of pulp and newsprints, whereas non-adopters are more

frequently involved in cardboard and other paper production. Relating to these differences are

variations in the type of technology used. These observations strongly suggest the variable ISO

is correlated with several included variables likely making RE specification inadequate.

5. Results

5.1 The base case

Table 2 reports the estimation results for BOD and TSS emission equations. Beside the RE and

FE estimates, we also report the OLS results. As expected, several coefficient estimates are

affected when plant-specific effects are taken into account. For both equations in the RE

specification, the Breusch-Pagan test strongly rejects the hypothesis of 02 =ασ . Similarly, in the

FE specification, the hypothesis of no plant-specific effect is rejected for both cases.

Furthermore, for several variables including ISO, the RE and FE estimates are quite different,

suggesting correlation between plant effects and included variables. Using a Hausman-type

specification test, we strongly reject the null hypothesis of no correlation for both BOD and

10 In other words, for all included variables, we assume that stXE siti ,,0),( ,, ∀=ε .

17

TSS.11 Therefore, our analysis is focused on FE results. As in [13] and [14], the coefficients on

12 month lag emissions are positive and statistically significant. For BOD, ISO certification is

associated with a statistically significant reduction in emissions of roughly 8.9%. However, ISO

certification does not appear to significantly impact TSS discharges. Interestingly, without

properly controlling for plants heterogeneity (OLS results), we would have concluded that ISO

has a major negative impact on both BOD and TSS emissions.

Inspections (INSP) significantly reduce BOD (by 8.4%), but not TSS. These last results

are very much in line with those obtained by [11]. The elasticity of emissions with respect to

production is close to 1 particularly for BOD. However, when the between plant variations is not

completely swept out (as in the FE model), the coefficient on this variable is less than 1 (OLS and

RE estimates). Large plants appear to proportionally pollute less than smaller plants. The

coefficient on the trend variable differs according to whether the plant is among the ISO adopting

or non-adopting group. For adopters, there is no significant trend, whereas non-adopters showed

an annual decline of approximately 7% and 5% per year in BOD and TSS emissions respectively.

Technological and output mix differences between the two groups could explain this major

difference. Not surprisingly, plants that experience a strike see their emissions drop

considerably. Emission levels are also significantly lower during warmer months.

5.2 Alternative specification and further results

In this section, we explore our results’ robustness to changes in specifications and estimation

methods, and provide complementary results. First, the nature of results is very similar if we use

11 The test is based on the RE model augmented by the within-transformation of the explanatory variables (see [20]). In the absence of correlation, these additional variables will not be jointly statistically significant. We obtain respective F-values for BOD and TSS of 138.35 and 64.38 with (7, 3060) degrees of freedom, clearly rejecting the null hypothesis of no correlation.

18

emission rates as the endogenous variable instead of absolute emission levels. The same hold

true if the two equations (BOD and TSS) are estimated as a seemingly unrelated regressions

model.12 Second, our specification so far assumes that ISO certification has a one-time effect.

The impact of ISO may be more gradual and may actually start before certification. Therefore,

we estimate a model where the variable ISO is replaced by the three binary variables PRE_ISO,

YEAR1_ISO and POST_ISO, which are set to 1 if, at time t, a plant is respectively within 12

months of obtaining certification, in its first year following certification or has been ISO for more

than one year. The FE results for the main variables are reported in column (1) of Tables 3 and 4.

It appears that the reduction in BOD emissions mostly takes place during the first year following

implementation of the ISO EMS. This impact does not seem to last as the coefficient on

POST_ISO is close to zero and not statistically significant.

We have assumed so far that the variance of the residuals is similar across plants. This

may not be the case as suggested by Figures 2 and 3. Furthermore, serial correlation may be

present. We re-estimate the model allowing the variance of the residuals to differ across each

plant and allowing for an AR(1) serial correlation structure (i.e. tititi ,1,, ξρεε += − with

2, )( itiVar σξ = ). Our main conclusions are unaffected by these changes (see column 2 Tables 3 and

4).

Next, we address the question of the potential correlation between the ISO variable and

the error term (i.e. ISO would not be strictly exogenous). To evaluate the relevance of this issue,

we first re-estimate the model using only observations for adopting plants thereby limiting the

potential for selection bias. The results are reported in column 3 of Tables 3 and 4. The results

are very similar to those obtained using the two groups. We also estimate the model on the same

12 Results available from the author.

19

sub-sample using a two-stage least squares approach.13 As instruments, we use: i) the number of

years since the plant received its first ISO certification relating to quality management (ISO

9000) (it is zero, if the plant is not ISO 9000), ii) the share of competitors’ plants that are ISO

14001 certified14 and iii) the total number of paper mills worldwide operated by the company

owning the plant.15 The results, reported in Tables 3 and 4 (column 4), are very much in line

with the above-reported results.

BOD and TSS emissions have long been regulated in the pulp and paper industry and

compliance has been relatively high. Therefore, it may well be that an ISO – EMS places greater

emphasis on newly-regulated pollutants or even contaminants that are not yet legally controlled.

Unfortunately, data for other pollutants are either unavailable or measured far less frequently,

making an econometric analysis more perilous. However, the total quantity of rejected water

may provide an indication of the plant’s general efforts toward reducing water pollution.

Moreover, keep in mind that pollution prevention is a very important principal underlining the

ISO philosophy. Table 5 reports the results for the impact of ISO on the total quantity rejected

water for the sub-sample of adopting plants (results are very similar when non-adopting plants

are also included). Results are somewhat different depending on whether the ISO variable is

instrumented or not, but, in any case, the total quantity of rejected water does not appear to

decline following ISO certification.

13 Note that we use a linear model in the first stage equation even if ISO is not a continuous variable. Indeed, the second-stage estimates are consistent even with a dummy endogenous variable (see [1]). However, this is not the case when a nonlinear procedure is used (e.g. logit or probit) in the first stage. Consistency is then conditional upon having an exact functional form in the first stage. In any case, we also used the Anderson-Rubin method to validate the confidence set for the ISO variable (see [6] and references cited therein). This method is robust to the first stage functional form and is valid even in the presence of weak instruments (or weak identification). This method provides a confidence set for ISO that is very close to the one obtained with the 2SLS method (results available upon request). 14 These shares are computed using our plant sample. 15 We use several sources to construct this variable including companies’ annual reports and websites.

20

6. Discussion and Conclusion

Despite growing interest in voluntary measures to address environmental externalities, thus far,

there is scarce empirical evidence of the effectiveness of these initiatives. One good example is

the international norm ISO 14001 that is becoming increasingly popular but whose environmental

impact has not yet been extensively evaluated. In this paper, we address this issue by testing

whether ISO certification of an EMS significantly impacts two traditional indicators (BOD and

TSS) in Quebec’s pulp and paper industry. Using a panel data set covering 38 plants over the

1997-2003 period, we find rather mixed evidence on the effectiveness of the ISO certification.

While it appears that BOD emissions significantly decline as a result of certification, we find

evidence that this decline is not permanent as emission levels return to normal after twelve

months. ISO certification is not associated with any significant change in TSS emissions or with

a reduction in the quantity of rejected process water. We also find that over time, non-adopting

plants experienced more significant emission reductions than plants that eventually adopt ISO.

Moreover, our analysis underlines the importance of properly controlling for plant-specific

effects stressing the importance of using panel data.

While one could conclude that the ISO norm is ineffective, several limits of our analysis

could temper this conclusion. As previously mentioned, ISO certification may have more

significant impacts on pollutants that are newly or not yet regulated (e.g. green house gases).

Also, bear in mind that our analysis specifically tests the impact of ISO certification of an EMS

rather than its implementation per se.16 Lastly, the impact of ISO may vary according to

industries. In conclusion, further research is needed before a definitive verdict that be rendered

on the effectiveness of these norms.

16 However, our ISO variable should be correlated with the adoption of an EMS.

21

05

1015

20N

umbe

r of I

SO

140

01 p

lant

s

1996 1997 1998 1999 2000 2001 2002 2003year

Figure 1. Number of ISO 14001 plants per year

.000

5.0

01.0

015

.002

.002

5N

on a

dopt

ers/

Ado

pter

s

Jan 1996 Jan 1998 Jan 2000 Jan 2002 Jan 2004date

Non adopters Adopters

Figure 2. Average BOD emission per unit of output (ton/ton)

.001

5.0

02.0

025

.003

.003

5.0

04N

on A

dopt

ers/

Adop

ters

Jan 1996 Jan 1998 Jan 2000 Jan 2002 Jan 2004date

Non Adopters Adopters

Figure 3. Average TSS emission per unit of output (ton/ton)

22

Figure 4. Determination of the Equilibrium Level of Emissions.

EMPMAC

Emissionse*

$

23

Table 1. Descriptive Statistics Variable Sample Adopters Non-Adopters

TSS emissions (tons) 51.02 (89.30)

66.44 (116.8)

36.42 (46.55)

TSS rate (emission/Q) 0.0021 (0.0024)

0.0024 (0.003)

0.0019 (0.0016)

BOD emissions (tons) 27.28 (57.34)

39.10 (77.58)

16.09 (21.11)

BOD rate (emission/Q) 0.0011 (0.0014)

0.0013 (0.0019)

0.0009 (0.0007)

Quantity of Water Rejected (W) (cube meter) 1257 (1117)

1590 (1134)

942 (998)

ISO 0.17 (0.38)

0.36 (0.48)

0 (0)

Number of Inspections in the last six months (INSP)

0.19 (0.40)

0.19 (0.40)

0.20 (0.40)

Production (Q) (ton) 21674 (14071)

25679 (12657)

17880 (14294)

Strike (STRIKE) 0.006 (0.078)

0.007 (0.088)

0.005 (0.070)

PROCESS 1 (1 = Mechanical Pulp) 0.078 (0.27)

0.05 (0.22)

0.10 (0.30)

PROCESS 2 (1 = Chemithermomechanical Pulp) 0.21 (0.41)

0.27 (0.44)

0.15 (0.36)

PROCESS 3 (1 = Kraft Pulp) 0.21 (0.41)

0.22 (0.41)

0.21 (0.40)

PROCESS 4 (1 = Other Chemical Pulp) 0.12 (0.33)

0.05 (0.22)

0.18 (0.38)

PROCESS 5 (1 = Recycled Pulp) 0.36 (0.48)

0.33 (0.47)

0.39 (0.48)

PROCESS 6 (1 = Pulp Bought) 0.39 (0.49)

0.33 (0.47)

0.44 (0.49)

PROCESS 7 (1 = De-inking) 0.21 (0.40)

0.27 (0.44)

0.15 (0.36)

PROCESS 8 (1 = Bleaching) 0.21 (0.40)

0.27 (0.44)

0.15 (0.36)

PROCESS 9 (Base: Thermomechanical Pulp) 0.38 (0.48)

0.50 (0.50)

0.24 (0.42)

QTYPE 1 (1 = Pulp) 0.27 (0.44)

0.33 (0.47)

0.21 (0.40)

QTYPE 2 (1 = Other Paper) 0.46 (0.49)

0.33 (0.47)

0.58 (0.49)

QTYPE 3 (1 = Cardboard) 0.21 0.11 0.32

24

(0.41) (0.31) (0.46) QTYPE 4 (Base: Newsprint) 0.38

(0.48)

0.44 (0.49)

0.32 (0.46)

Number of observations 3108 1512 1596

25

Table 2. Results for the Base Case. BOD Emissions TSS Emissions Explanatory Variables

OLS RE FE OLS RE FE

Constant -3.213*** (0.3507)

-4.016*** (0.3932)

-5.886*** (0.5751)

-3.8823*** (0.3244)

-4.586*** (0.3832)

-4.684*** (0.5970)

Log( )12, −tiE 0.5078*** (0.0145)

0.2808*** (0.0151)

0.1172*** (0.0150)

0.4155*** (0.0153)

0.2817*** (0.0157)

0.2054*** (0.0160)

ISO -0.146*** (0.0427)

-0.0266 (0.0400)

-0.0888** (0.0450)

-0.1225*** (0.0390)

-0.0142 (0.0386)

-0.0262 (0.0467)

INSP -0.087*** (0.0294)

0.0863*** (0.0262)

-0.0845*** (0.0233)

-0.0284 (0.0269)

-0.0244 (0.0252)

-0.0210 (0.0242)

Q 0.3768*** (0.0243)

0.5908*** (0.0334)

0.9662*** (0.0604)

0.5458*** (0.0237)

0.6892*** (0.0332)

0.8758*** (0.0627)

T_NON_ADOPTERS 0.0004 (0.0005)

-0.0015*** (0.0005)

-0.0060*** (0.0005)

-0.0007 (0.0005)

-0.0018*** (0.0004)

-0.0041*** (0.0005)

T_ADOPTERS 0.0005 (0.0005)

-0.0014*** (0.0005)

0.0009 (0.0008)

-0.0008 (0.0005)

-0.0020*** (0.0005)

-0.0009 (0.0008)

STRIKE -1.707*** (0.1542)

-1.4874*** (0.1443)

-0.9421*** (0.1571)

-0.7791*** (0.1417)

-0.6340*** (0.1397)

-0.3688** (0.1632)

MONTH 1 0.0251 (0.0564)

0.0584 (0.0500)

0.07405** (0.0445)

-0.0089 (0.0517)

-0.00000 (0.0481)

-0.0008 (0.0461)

MONTH 2 0.0283 (0.0563)

0.0492 (0.0499)

0.0738* (0.0444)

-0.0402 (0.0516)

-0.0374 (0.0481)

-0.0313 (0.0461)

MONTH 3 0.0250 (0.0563)

0.0341 (0.0499)

0.0286 (0.0444)

-0.0615 (0.0516)

-0.0709 (0.0481)

-0.0820* (0.0461)

MONTH 4 -0.0166 (0.0563)

-0.0224 (0.0498)

-0.0361 (0.0443)

-0.0536 (0.0516)

-0.0650 (0.0480)

-0.0760* (0.0460)

MONTH 5 -0.0420 (0.0563)

-0.0645 (0.0499)

-0.1041** (0.0447)

-0.0600 (0.0516)

-0.0782 (0.0481)

-0.0996** (0.0464)

MONTH 6 -0.0468 (0.0564)

-0.0931* (0.0500)

-0.1540*** (0.0449)

-0.0923 (0.0516)

-0.1158** (0.0482)

-0.1420*** (0.0466)

MONTH 7 -0.068 (0.0565)

-0.1278** (0.0503)

-0.2184*** (0.0460)

-0.0989* (0.0518)

-0.1321*** (0.0485)

-0.1734*** (0.0477)

MONTH 8 -0.090 (0.0564)

-0.1513*** (0.0503)

-0.2423*** (0.0459)

-0.1203** (0.0517)

-0.1537*** (0.0484)

0.1951*** (0.0477)

MONTH 9 -0.065 (0.0563)

-0.1114*** (0.0499)

-0.1720*** (0.0447)

-0.1026** (0.0516)

-0.1288** (0.0481)

-0.1565*** (0.0465)

MONTH 10 -0.0525 (0.0562)

-0.0895* (0.0498)

-0.1382*** (0.0445)

-0.0712 (0.0515)

-0.0870* (0.0480)

-0.1063** (0.0462)

MONTH 11 0.0020 (0.0562)

-0.010 (0.0497)

-0.0278 (0.0442)

-0.0480 (0.0515)

-0.0585 (0.0479)

-0.0684 (0.0459)

PROCESS 1 -0.1296** (0.0650)

-0.1174 (0.0855)

- 0.0385 (0.0596)

0.0597 (0.0832)

-

PROCESS 2 0.4806*** (0.0606)

0.8026*** (0.0900)

- 0.3804*** (0.0545)

0.5807*** (0.0872)

-

26

PROCESS 3 0.4199*** (0.1240)

0.7667*** (0.1987)

- -0.3138*** (0.1122)

-0.2587 (0.1935)

-

PROCESS 4 0.1425* (0.0844)

0.0982*** (0.1203)

- -0.4459*** (0.0784)

-0.4097*** (0.1180)

-

PROCESS 5 0.1272** (0.0619)

0.2804*** (0.0947)

- -0.1392** (0.0565)

0.2552*** (0.0925)

-

PROCESS 6 0.1408*** (0.0452)

0.2442*** (0.0647)

- 0.1852*** (0.0414)

0.3069*** (0.0631)

-

PROCESS 7 0.0787 (0.0630)

-0.0074 (0.0998)

- -0.1374** (0.0577)

-0.1926** (0.0977)

-

PROCESS 8 -0.824*** (0.1915)

-1.4554*** (0.2954)

- -1.0327*** (0.1753)

-1.1802** (0.2882)

-

QTYPE 1 1.2035*** (0.1474)

1.839*** (0.2263)

- 1.9077*** (0.1383)

2.2092*** (0.2227)

-

QTYPE 2 0.0649* (0.0377)

0.0772 (0.0603)

- -0.0674** (0.0346)

-0.0788 (0.0590)

-

QTYPE 3 -0.196*** (0.0542)

-0.2565*** (0.0850)

- -0.2512*** (0.0498)

-0.3342*** (0.0833)

-

R2: 0.81 ^2ασ =0.006

^2εσ =0.251

H0: 2ασ =0

2χ (1)=5718

R2 (within): 0.25 H0: all iα =0 F(36,3053)= 77.37

R2: 0.84 ^2ασ =0.008

^2εσ =0.271

H0: 2ασ =0

2χ (1)=2513

R2 (within): 0.18 H0: all iα =0 F(36,3053)= 54.89

27

Table 3. Alternative Specification Results – BOD Emissions. Explanatory Variables

(1) (2) (3) (4)

Constant -5.846*** (0.5855)

-5.5574*** (0.5918)

-9.9717*** (0.8180)

-9.8523*** (0.9648)

Log( )12, −tiE 0.1175*** (0.0150)

0.0808*** (0.0155)

0.1085*** (0.0201)

0.1086*** (0.0201)

ISO - -0.1082** (0.0523)

-0.0746* (0.0411)

-0.0501 (0.1124)

PRE_ISO -0.0195 (0.0469)

- - -

YEAR1_ISO -0.1424** (0.0569)

- - -

POST_ISO -0.0248 (0.0707)

- - -

INSP -0.0851*** (0.0233)

-0.0383 (0.0241)

-0.0450 (0.0303)

-0.0440 (0.0306)

Q 0.9744*** (0.0604)

1.1229*** (0.0519)

1.2553*** (0.0819)

1.2582*** (0.0829)

T_NON_ADOPTERS -0.0060*** (0.0005)

-0.0052*** (0.0007)

- -

T_ADOPTERS 0.0004 (0.0010)

0.0006 (0.0009)

0.00005 (0.0007)

-0.0002 (0.0015)

R2 (within): 0.25 ^ρ =0.54

R2 (within): 0.36 R2 (within): 0.36

28

Table 4. Alternative Specification Results – TSS Emissions. Explanatory Variables

(1) (2) (3) (4)

Constant -4.515*** (0.6064)

-7.446*** (0.5883)

-7.094*** (0.8325)

-6.621*** (0.9833)

Log( )12, −tiE 0.2048*** (0.0160)

0.1167*** (0.0165)

0.1363*** (0.0225)

0.1361*** (0.0226)

ISO - -0.0457 (0.0524)

-0.0213 (0.0418)

0.0754 (0.1145)

PRE_ISO 0.0337 (0.0486)

- - -

YEAR1_ISO -0.0567 (0.0590)

- - -

POST_ISO 0.1117 (0.0733)

- - -

INSP -0.0212 (0.0241)

-0.0192 (0.0242)

-0.0015 (0.0007)

0.0025 (0.0312)

Q 0.8854*** (0.0627)

0.9704*** (0.0530)

1.0832*** (0.0832)

1.0949*** (0.0844)

T_NON_ADOPTERS -0.0041*** (0.0005)

-0.0023*** (0.0007)

-

-

T_ADOPTERS -0.0022** (0.0010)

-0.0001 (0.0009)

-0.0012 (0.0007)

-0.0024 (0.0016)

R2 (within): 0.18 ^ρ =0.50

R2 (within): 0.19 R2 (within): 0.19

29

Table 5. Results for the Total Quantity of Rejected Water. Explanatory Variables

FE FE-IV

Constant -6.6803*** (0.1852)

-6.5659*** (0.2055)

Log( )12, −tiE 0.0467*** (0.0091)

0.0466*** (0.0091)

ISO 0.0192*** (0.0071)

-0.0043 (0.0194)

INSP -0.0027 (0.0052)

-0.0037 (0.0053)

Q 0.7669*** (0.0141)

0.7640*** (0.0143)

Trend -0.0021*** (0.0001)

-0.0017*** (0.0002)

R2 (within): 0.88

30

References [1] J.D. Angrist and A.B. Krueger, Instrumental variables and the search for identification:

from supply and demand to natural experiments, Journal of Economic Perspectives 15 (4) 69-85, (2001).

[2] W. R. Q. Anton, G. Deltas and M. Khanna, Incentives for environmental self-regulation and implications for environmental performance, Journal of Environmental Economics and Management 48 (2004) 632-654.

[3] O. Boiral and J.-M. Sala, Environmental Management: Should Industry Adopt ISO 14001?, Business Horizons / January-February 1998.

[4] Dasgupta S., H. Hettige and D. Wheeler, What Improves Environmental Compliance? Evidence from Mexican Industry, Journal of Environmental Economics and Management, 39, 39-66 (2000).

[5] C. Dion, P. Lanoie and B. Laplante, Monitoring of Pollution Regulation: Do Local Conditions Matter?, Journal of Regulatory Economics; 13:5-18 (1998).

[6] J.-M. Dufour, Presidential address: identification, weak instruments, and statistical inference in econometrics, Canadian Journal of Economics, Vol. 36 (4), 767-808 (2003).

[7] J. Foulon, P. Lanoie and B. Laplante, Incentives for Pollution Control : Regulation or Information?, Journal of Environmental Economics and Management 44, 169-187 (2002).

[8] E. Helland, The Enforcement of Pollution Control Laws: Inspections, Violations, and Self-Reporting, The Review of Economics and Statistics, Vol. 80, No. 1, 141-153 (1998).

[9] M. Khanna and W.R. Q. Anton, Corporate Environmental Management, Land Economics, Vol. 78 (4), 539-557, (2002).

[10] M. Khanna and L. A. Damon, EPA’s Voluntary 33/50 Program: Impact on Toxic Releases and Economic Performance of Firms, Journal of Environmental Economics and Management 37, 1-25 (1999).

[11] K. Kollman and A. Prakash, EMS-based Environmental Regimes as Club Goods: Examining Variations in Firm-level Adoption of ISO 14001 and EMAS in U.K., U.S. and Germany, Policy Sciences; March 2002; 35: 43-67, 2002.

[12] P. Lanoie, M. Thomas and J. Fearnley, Firms Responses to Effluent Regulations: Pulp and Paper in Ontario, 1985-1989, Journal of Regulatory Economics; 13:103-120 (1998).

[13] B. Laplante and P. Rilstone, Environmental Inspections and Emissions of the Pulp and Paper Industry in Quebec, Journal of Environmental Economics and Management, 31, 19-36 (1996).

[14] W. A. Magat and W. K. Viscusi, Effectiveness of the EPA’s Regulatory Enforcement: The

31

Case of Industrial Effluent Standards, Journal of Law and Economics, Vol. 33, No. 2 (Oct., 1990), 331-360.

[15] M. Nakamura, T. Takahashi and I. Vertinsky, Why Japanese Firms Choose to Certify: A Study of Managerial Responses to Environmental Issues, Journal of Environmental Economics and Management 42, 23-52 (2001).

[16] Quebec Forest Industry Council, Environnemental Performance, Quebec (2001).

[17] Quebec Ministry of the Environment, Bilan de conformité environnementale – secteur des pâtes et papiers, (1996-2002).

[18] M. Szymanski and P. Tiwari, ISO 14001 and the Reduction of Toxic Emissions, Policy Reform, March 2004, Vol. 7(1), pp. 31-42.

[19] E. Wall, A. Weersink and C. Swanton, Agriculture and ISO 14000, Food Policy 26 (2001) 35-48.

[20] J.M. Wooldridge, Econometric Analysis of Cross Section and Panel Data, The MIT Press, Cambridge MA, 2002.

[21] E. K. Yiridoe, J. S. Clark, G. E. Marett, R. Gordon and P. Duinker, ISO 14001 EMS Standard Registration Decisions Among Canadian Organizations, Agribusiness; Fall 2003; 19, 4; 439-457.