Does ISO14001 raise firms’ awareness ofenvironmental protection?—Case from Vietnam

Bin Ni∗ Hanae Tamechika† Tsunehiro Otsuki‡

Keiichiro Honda§

January, 2015

Abstract

Environmental protection is an inevitable issue that developing countries allhave to deal with during the process of inviting foreign direct investment (FDI).However, high correlation between FDI and pollution doesn’t necessarily indicatethat foreign firms are to blame. In this paper, we apply firm-level panel data inVietnam and unique information on waste discharge to show that foreign firmsare actually more active to acquire ISO14001, a voluntary environmental standard.And the adoption will in turn improves firms’ performance in waste control. It alsoincreases firms’ welfare as well as their productivity level. A theoretical dynamicgame model lends support to our empirical findings. This paper provides strongevidence that firms’ efforts towards corporate social responsibility will eventuallybenefit themselves as well.

JEL classification: D22, F21, F64, Q56Keywords: FDI, ISO14001, Vietnam, environmental protection

∗Corresponding author. The Graduate School of Economics, Osaka University (jiadaniel@hotmail.com)†The Graduate School of Economics, Nagoya City University‡Osaka School of International Public Policy, Osaka University§Department of Administration, Prefectural University of Kumamoto

1

1 IntroductionIt is known to all that foreign direct investment (hereafter FDI) has been a consider-able driving force to spur the economic growth in developing countries, especially innewly-emerging economies. But rapid growth usually comes with a price—pollution.Given the simultaneous rise in FDI and pollution level, it is natural to think that apositive correlation might exist between these two. Critics have accused foreign in-vestors of shifting their heavily-polluting activities to countries with lax regulations insearch of “pollution-haven”, however, empirical evidence consistent with this hypoth-esis is surprisingly rare (Cole, 2004). Actually foreign firms are found to be moreenergy efficient compared to state-owned firms (Eskeland and Harrison, 2003; He,2006). This might be due to the advanced waste-processing technology adopted byforeign firms and their awareness to achieve corporate social responsibilities (Lyon andMaxwell, 2008). Other stories include “protecting institutional reputation, appealing to‘green consumers’, deterring lobbying and boycotts by environmental groups, avoidingregulatory scrutiny, and preempting future regulation”1. Under certain circumstances,we highly doubt the “pollution haven” hypothesis and propose the opposite idea thatthe more foreign firms invest in the host country, the more likely they become self-restrained in the sense of environmental protection.

In this paper we would like to verify this hypothesis by evaluating firms’ partici-pation in a voluntary environmental program—ISO14001, in the context of Vietnam.ISO14001 is considered one of the most widely recognized voluntary standards2 forEnvironmental Management Systems, whose adoption is likely to be incurred by firms’spontaneous act3. Thus the possibility of acquiring ISO14001 is usually positively as-sociated with firms’ willingness to involve in the environmental protection. By quan-tifying firms’ efforts before and after attending this program, we hope to answer thefollowing questions: are foreign firms more active to be involved in ISO14001 thantheir domestic counterparts? How can ISO14001 improve firms’ overall performance,especially their efforts in waste control?

By applying various methods to clear these doubts, this study differs from the pre-vious literature in several ways: First, it will be the first paper to explore how firms’participation in voluntary program affects the pollution behavior in Vietnam, thus fillin the research gap in the developing countries (Arimura et al., 2014); Second, themeasurement will be based on multiple indexes, instead of single ones; Third, it com-bines empirical strategy with theoretical verification, which to our knowledge, is thepionerring one. After robustness check, we do achieve consistent results that supportour hypothesis. The coherence, on the other hand, also leaves us the room for policyimplications and future study.

This paper is organized as follows: in the next section we talk briefly about thepollution situation in Vietnam. Literature review then comes after. In section 4 wedescribe the data and estimation strategy, followed by robustness check and findings.Section 6 presents the theoretical model and the final section concludes.

1Quoted from Bui and Kapon (2012).2The environmental protection paradigm in developing countries is moving away gradually from com-

pulsory approach to more flexible and voluntary ones (Tambunlertchai et al., 2013).3Though some have argued that the adoption of ISO14001 is motivated primarily by domestic regulatory

and market pressures (Khanna and Anton, 2002; Lyon and Maxwell, 2004).

2

2 Overview2.1 About ISO14001

The International Organization for Standardization (ISO) was founded in 1946, whichhas 165 members by far, each representing a country. It is the most prominent developerof standards in the world. In 1980s, ISO introduced ISO9000 standards for qualitymanufacturing practices. Building on this system, ISO set up ISO14001 environmentalstandards in 19964. According to the definition by ISO, this standard enables firmsto adopt the policy following the legal requirements and provides them with updateenvironmental information. In other words, it forces the organizations to raise self-awareness of maintaining a healthy environment and assure themselves of conformitywith their stated policies. The benefits of ISO14001 include, but not limited to: reducedcost of waste management and distribution; savings in consumption of energy andmaterials; improved corporate image among regulators, customers and the public (ISOHomepage). Thus despite the considerable cost to acquire ISO14001, the nature of itsuggests sustainable benefit for the firms in the long run.

2.2 Why Vietnam?

In this paper we would like to focus on Vietnam, a typical example that witnesses bothfast growth in foreign investment and industrial pollution at the same time. Accordingto Central Institute of Economic Management Report (CIEM) 2007 and CIEM 2008,FDI in 2007, as a share of Vietnam’s GDP, has risen to more than 20%, to as muchas five times that in 2000. In a single year between 2005 and 2006, about 60% ofthe total FDI were in the industrial sectors while 66.7% of the capital flew into heavyindustries5. Figure 1 shows the structural change of the sectors in Vietnam. This is inline with Vietnam’s national policy of rapid industrialization and transition from ruraleconomy.

Meanwhile the pollution in Vietnam is considered to get even worse if the currentpattern of industrialization continues and no further control is activated. Take air pol-lution, for instance, nearly half of Nitrogen dioxide (NO2) emission is due to industrialdevelopment, and when it comes to Sulfur dioxide (S O2) industry becomes the majorcriminal (Vietnam: Air Quality Profile 2010). These two kinds of pollutants are detri-mental to both human health and the environment. Liquid (total suspended solids) andsolid wastes (chemical and metal) also constitute a large portion of the total industrialdischarge. Though Ministry and Departments of Natural Resources and Environment(MONRE) in Vietnam tries to record the pollution level in all aspects, the precise datais not available. Instead we approximate the amount of pollution in Vietnam by usingthe pollution intensity index constructed by World Bank’s Industrial Pollution Projec-tions System (IPPS). The same method has been applied in Mani and Jha (2006), andNgo (2010). From Figure 2, we can see that the total pollution has risen by nearly 150percent over five years. To decompose the contents, we find that most of the increasecomes from air and solid waste, suggesting a possible shift in the waste composition6.

3http : //www.iso.org/iso/about/iso member.htm Accessed on 2014/11/25.4In recent years, ISO22000 food safety standards, ISO26000 social responsibility standards, ISO36000

risk management standards, and ISO50001 energy management systems are also introduced.5Review and analysis of the pollution impacts from Vietnamese manufacturing sectors (2008), a report

conducted by World Bank. Hereafter referred to as Review (2008).6Review (2008) provides us with extra index: the most seriously polluted areas are those that experience

the fastest economic growth, such as Ho Chi Minh City and Hanoi. They have the highest overall ranking

3

Figure 1: Structural Chnage in Vietnam (percentage)

0  

5  

10  

15  

20  

25  

30  

35  

40  

45  

50  

1995   2000   2005  

Agriculture,  forestry  and  fisheries  

Industry,  construc=on  

Services  

Source:  Socio-­‐Economic  Development  Plan,  2006-­‐2010�

Given the fact that FDI keeps flooding into Vietnam, it should be important andinteresting to investigate what role foreign firms play in “contributing” to this wave ofpollution. By using ISO14001 as the benchmark, we would like to measure the aware-ness of foreign firms’ environmental protection and the post-adoption performance,compared with the case for domestic firms. If positive result is to be achieved for theforeign, then we can provide more evidence to refute the “polution haven hypothesis”,at least in the context of Vietnam.

3 Literature ReviewBy far several studies have investigated the direct relationship between FDI and pollu-tion level: Jiang et al. (2014), He (2006) as well as Eskeland and Harrison (2003) allreach the unanimous conclusion that FDI has a negative impact on the pollution levelin the host country. Based on this stylized fact but taking a step further, we would liketo make explicit of the mechanism behind the phenomenon. We will divide the pro-cess into two steps: to find out how FDI (or firm’s ownership in the micro level) affectsISO14001 adoption and the impact of ISO14001 adoption on firms’ polluting behavior.

As for the first step, there are mainly two kinds of theories: convergence and di-vergence (Prakash and Potoski, 2007). Convergence advocates claim that foreign sub-sidiaries usually conform to global standard rather than adapting to host country char-acteristics. In other words, if the subsidiaries come from a country with a high coverageof ISO14001 adoption, it is quite likely that these firms will also acquire the certificatein the host country. And within the convergence debate, the main stream opinion isthat foreign firms face greater scrutiny from local government, which gives them moreincentive to adopt ISO14001, or even encourage their suppliers. Thus FDI will have

on the National Pollution Index. As for sectors, heavy industries are undoubtedly responsible for the majorwaste discharge.

4

Figure 2: Pollution Level in Vietnam (tons)

0  

5000000  

10000000  

15000000  

20000000  

25000000  

30000000  

35000000  

40000000  

45000000  

2004   2005   2006   2007   2008  

Total  

Air  

Solid  

Liquid  

Source:  GSO,  Vietnam  and  World  Bank  IPPS  

a positive influence on firms’ adoption of ISO14001 in the target country. While di-vergence supporters hold the opposite opinion: foreign investors choose to locate indeveloping countries because they face less stringent environmental control. And theyno longer need to play by the rule as they do in the home country7. In the empir-ical verification, positive relationship is found between FDI and ISO14001 adoptionin Thailand (Tambunlertchai et al., 2013) as well as Malaysia (Arimura et al., 2014).Both studies applied firm level data. In the macro level, similar result is achieved(Prakash and Potoski, 2006). This paper will be aimed at re-evaluating FDI’s role inthe ISO14001 adoption preference.

The second step focuses on the relationship between ISO14001 adoption and firms’performance. There is a large theoretical literature to study the connection betweencompulsory regulations and firms’ polluting behavior, but few are found to investigatevoluntary program’s impact (e.g. Lyon and Maxwell, 2008). The conclusion has beenmainly drawn on empirical evidence. Previous studies have used single pollution mea-sure to assess the impact of ISO14001 (Potoski and Prakash, 2005; Turk, 2009) andthey all find ISO14001 reduces pollution discharge. Meanwhile, Arimura et al. (2008)verify ISO14001’s positive influence on reducing both solid and liquid waste in thecase of Japan. Furthermore, Arimura et al. (2011) find ISO14001 also improves firms’supply chain management. Apart from ISO14001, other voluntary enviromental pro-grams promote firms to curb pollution as well (Bui and Kapon, 2012; Kim and Lyon,2011).

Our methodology is closest to that used in Blackman et al. (2010). We will describethe estimation strategy and data in the following section.

7Political economists also argue that corruption plays its part in affecting environmental policy stringencyto attract FDI, such as Fredriksson et al. (2003), Cole et al. (2004).

5

4 Estimation Strategy and Data4.1 Estimation Strategy

For empirical verification we apply the firm level data in Vietnam. To account for thepotential self-selection endogeneity problem, we incorporate it into a two-step sampleselection framework.

IS O∗it = δit · Zit + Uit (1)

Yit = αit + β jt · IS Oit + βit · Xit + εit (2)

where IS O = 1 if IS O∗ >= 0 and IS O = 0 if IS O∗ < 0.Firstly we estimate the propensity of firms’ decision to adopt ISO on a series of

exogenous variables. ISO is a dummy variable which takes value of 1 if firm i adoptsISO14001 at time t, otherwise 0. It’s taken from the observed data. Zit in Eq. (1) isa vector of derterminants which lead to the adoption. It includes both objective andsubjective firm characteristics. The former consists of firm size (number of labor), FDI(foreign capital/total capital) and capital-labor ratio while the latter includes answersbased on the firms’ self-evaluation, such as whether they follow the environmentalregulations. Uit is the error term.

Xit in Eq. (2) has similar elements as Zit. But when the Heckman-style model isapplied, exlusion restriction has to be met. Thus we include some of the subjectivevariables in the selection equation (Eq. (1)), but exclude them in the outcome equation.Basically we choose the explanatory variables consistent with Tambunlertchai et al.(2013) and Arimura et al. (2014), but have to give up some (export status, ISO9000certification, etc) because of data availability. Concerning dependent variable Yit, it hastwo sets of indicators: waste discharge and non-environmental performance (turnover,average salary and total factor productivity). Each variable of interest will be estimatedseparately8. Year dummies and industry dummies are included in both equations. SeeTable 1 for detailed description of the variables.

As for TFP, it needs extra effort. Since traditional solow residue’s approach isunable to isolate the true productivity from statistical noise, recent literature has exclu-sively proxy methods (Olley and Pakes, 1996; Levinsohn and Petrin, 2003)9. Unfor-tunately, due to quality issue, neither is the ideal method to be used in our analysis,though we use them as robustness check tools. Another option is stochastic frontieranalysis (SFA). According to Kumbhakar and Lovell (2000), given the Cobb-Douglasproduction function, the model for the SFA is specified as10:

lnyit = Σβn · lnxnit + vit − uit (3)

where xit is a vector of inputs. vit is the noise component and uit is the nonnegativetechnical inefficiency component. Our objective is to obtain estimate of the technicalefficiency which is the proxy for TFP of a particular firm.

T Eit = exp{−uit} (4)

where it is assumed that ui ∼ iid N+(0, σ2u). The other options might include Normal-

Exponential Model, Normal-Gamma Model. We will use them alternatively.

8We use “treatreg” as our baseline estimation command and try other alternatives for confirmation.9Though De Loecker (2011) has criticized their methodology for ignoring the omitted price variable bias.

Since we don’t have the product information, we will not consider his method in this paper.10To distinguish from Eq.(1)-(2), we use lowercased letters.

6

4.2 Data

This paper uses a panel dataset, constructed from the Vietnam Enterprise Survey at firmlevel. The data was collected by the General Statistics Office of Vietnam for all sectorsand industries, as on March 1st annually. It covers all 22 manufacturing sectors out ofthe total 42. Company characteristics such as ownership, labor, capital stock, turnover,assets, FDI share, average wage rate, intermediate materials are also available. Apartfrom the above, GSO has taken a census of all multinational enterprises (MNEs), whichare defined firms that have foreign capital, regardless of the share. The advantage isthat investment behavior of these foreign capitalized firms can be captured over time.Census is also taken for firms with more than 10 employees. Each firm has an exclusiveenterprise code. We use it together with province code to identify the firms.

Another uniqueness of this dataset is that it collects information on firms’ engage-ment in enviromental protection, including the cost spent on enviromental protection,whether the firm carries out environmental management system, whether it followsthe clean manufacturing process, etc. And above all, whether the firm has ISO14001certification is recorded. Since it is relatively objective criteria which is free of mea-surement error, we use it to create our ISO adoption dummy. Unfortunately, the ISOinformation is only accessible from 2007 to 2009, we have to limit our analysis to thistime period. Finally, the detailed data on waste discharge categorized by form (air, liq-uid, solid) is complete. It differentiates between treated and untreated amount of wasteas well, which is important for us to conduct the second stage estimation.

Certainly we have to admit that the survey has its drawbacks. For example, theincomplete information about export and import, missing data for materials and othervariables, inconformity of units among different years, etc. As a result, we have to dealwith an unbalanced panel data here. We remove the missing observations necessaryto calculate TFP, and delete outliers. After these arrangements, the total number ofobservations for estimation is 28274 over three years.

Table 1 is inserted here

5 Results5.1 Baseline Results

As can be seen from Table 2, foreign direct investment, number of labor, waste depart-ment dummy and capital labour ratio, are statistically significant at the 1% level in thefirst stage. The estimated coefficient of FDI share is positive, which means that firmswith foreign capital actively adopt ISO14001. The positive sign of total labor indi-cates that the larger a firm is, the more likely it is to adopt ISO14001. One explanationmight be that larger firms have more capacity to partcipate in such voluntary programs.Meanwhile capital-labor ratio also plays a positive role, implying that capital-intensivefirms prefer ISO14001.

In the second stage, ISO14001 is positively significantly at the 1% level in all kindsof cases. As long as waste discharge is concerned, ISO14001 adoptions increase theshare of treated wastes in all three forms. The result provides evidence that once firmsacquire this environmental certificate, they raise the awareness of environmental pro-tection and try to control their polluting behavior.

7

On the other hand, the positive impact of ISO14001 on firms’ total salary, turnoverand total factor productivity is telling us another story. After firms adopt ISO14001, thecost of managing the waste is reduced and allow firms to have more resources (capitaland labor) to allocate. As a consequence, it benefits the firms by shifting up their totalperformance. This verifies our hypothesis by showing that firms’ devotion to the socialresponsibility can lead to a win-win situation.

Table 2 is inserted here

5.2 Robustness Check

Apart from applying waste department as the determinant of ISO14001 adoption,we use totalcost environ, environ system and enviromental standard alternatively11.Concerning TFP calculation, Levinsohn and Petrin’ method is also tried. We get simi-lar and consistent results in almost all cases12.

Next, we use propensity score matching (PSM) to re-confirm our findings. The pur-pose of our estimation is to find out the average treatment effect on the treated (ATT),which in this paper is the difference of performance between ISO14001 adopters andnon-adopters. The accurate measurement needs random experimental settings, how-ever, the counterfactual is usually unobserved. Rosenbaum and Rubin (1983) proposethe propensity score which can be used in our analysis to match the adopters with non-adopters. We will use the first stage equation as introduced in Section 5.1 to predict thelikelihood a firm adopts ISO14001.

The challenge is that firms didn’t report in which year they acquired ISO14001.Thus we use the information in year 2007 (the first year of the observation) to calculatefirms’ propensity score of ISO adoption. Then we match them with the firms in 2009(the last year) that have similar propensity and compare their difference. If the perfor-mance indicators in these two years are significantly different, then we can make thejudgement that ISO14001 has potentially promoted firms to improve. To proceed, wefurther asssume that by controlling the covariates, the error term will be uncorrelatedwith the decision of firms’ adopting ISO1400113

Our treatment sample (ISO14001=1) varies from 825 in year 2007 to 1201 in year2009. The average value of each control variable for the treated is higher than thatfor the control. For example, the average TFP for the treated group is 0.56 comparedto 0.49 for the control group. The estimation results are consistent with the statisticalintuition. In Table 3, we report results for both nearest 1-to-1 matching and kernelmatching in each model. ATT estimates are all significantly except for solid waste.This indicates that by adopting ISO14001, firms’ overall performance will be greatlyimproved.

Table 3 is inserted here

We conduct the balance tests (for matched) as well. The results of t test for eachcovariate does not reject the null hypothesis that the mean between treated and control

11We attempt different estimation models such as ivtreatreg and change the options to heckit, 2SLS, etc.The results vary a little, but don’t change considerably.

12To save space, we did not report all the results but they are available.13In reality this assumption can be violated. For example, a policy shock in some industry might promote

firms to apply for ISO, and an opposite scenario can also be considered.

8

are equal, meaning the models balance the covariates quite well. Another evidenceis that the standardized bias is 4.46% for kernel matching estimator while 1.38% fornearest 1-to-1 matching estimator. Based on the discussion in Caliendo and Kopeining(2008), standardized bias below 5% is enough to justify the balance.

Despite the strong evidence that ISO14001 improves firms’ competitiveness, andraise their awareness of environmental protection, the estimation might still suffer frombias due to data limitation, as previously explained. More acurate result can be achievedif we have more detailed information on the background of ISO14001 adoption, e.g.,why firms in some industries or areas have higher tendencey to acquire ISO, especiallyin the context of Vietnam. It also leaves us room for future study on whether ISO’simpact is temporary or not.

6 Theoretical Model6.1 General Setting

In this Section, we try to illustrate the mechanism behind ISO14001 adoption and firm’sperformance by applying a two stage dynamic games of complete but imperfect infor-mation, as described in Gibbons (1992, Chapter 2.2). Suppose there is an oligopolymarket with only 2 firms. Both firms produce a homogenous good in the domesticmarket. In order to be more competitive, now we assume that firm 1 has the option tochoose a strategy k in the first stage. In our model, to be specific, k means ISO14001adoption. Then k becomes publicly observable. In the second stage, both firms si-multaneously choose their output level q1 and q2 from R+ respectively. The profits foreach firm are π1(k, q1, q2) and π2(q1, q2), which are twice continuously differentiable.We make further assumption that π1 is strictly concave in k. In the following we willanalyze the cases with 2 profit-maximizing firms competing a la Cournot-Nash.

For firm 1 to realize maximum profit at stage one, firstly we total differentiate π1w.r.t. k:

dπ1

dk=∂π1(k, q∗1(k), q∗2(k))

∂k+∂π1

∂q∗1·

dq∗1(k)dk

+∂π1

∂q∗2·

dq∗2(k)dk

(5)

where the second term on the RHS of Eq. (5) reflects k’s influence on π1 throughq1. By envelope theorem, this term can be ignored. The third term is called “strategiceffect”, the sign of which depends on the several conditions. Given the fact the directeffect of k is always positve (i.e. ∂π1

∂k takes the positive sign), we will only need to focuson this last term.

Following the argument in Kerschbamer (2014), we define b1(q2|k) and b2(q1) aseach firm’s best response function at stage two. Thus b1 depends on b2 and k while b2only depends on b1. Note that bi and q∗i are basically different except for the equilibriumat k: q∗2(k)=b2(b1(q∗2(k)|k)). We differentiate this equation w.r.t. k:

dq∗2(k) =db2

db1· [

∂b1

∂q∗2(k)· dq∗2(k) +

∂b1

∂k· dk] (6)

⇒dq∗2(k)

dk=

db2db1·∂b1(q∗2(k)|k)

∂k

1 − db2db1·

∂b1∂q∗2(k)

(7)

9

By stability of Nash Equilibrium, the denominator of Eq. (7) is positive (Kohlberg andMertens, 1986). Thus the “strategic effect” actually is determined by three factors:

(i) ∂π1∂q∗2

(ii) db2(q∗1(k))dq∗1

(iii) ∂b1(q∗2(k)|k)∂k

Next we will discuss the cases and decide on the exact level of the “strategic effect”when the detailed forms of the production function are given.

6.2 Detailed Case Study

As outlined above, both firms compete over homogenous good, the price of which takesthe reverse demand function p = a−q, where q is total output and q =

∑ni=1 qi, i = f , d.

f and d stand for foreign and domestic firm respectively. For simplicity, we ignore themarginal cost of production and the the initial cost to acquire ISO1400114. Foreignfirm and domestic firm differ in their pollution abatement technology. It is reasonableto assume that foreign firm has better knowledge in the abatement and thus the marginalcost to deal with waste discharge for each firm has the following relationship: t f < td.We further assume that the total waste is proportionate to the output of the firm, and it isequal to δqi, δ ∈ (0, 1), i = f , d. There are two cases that interest us: the one when onlyforeign firm adopts ISO14001 and the other when only domestic firm adopts ISO14001.We compare them with the case when no strategy is introduced. Finally we will discusswhether the incentives for foreign and domestic firms to adopt ISO14001 are the same.

Case 1 (or base case): Neither firm adopts ISO14001.

Each firm’s profit function can be represented as:

π f = (a − q f − qd) · q f − t f · δq f (8)

πd = (a − q f − qd) · qd − td · δqd (9)

To maximize each firm’s profit, we take the first order condition and derive the bestreply function for each firm in Nash Equilibrium:

q∗f 1 =a − t f δ − qd

2(10)

q∗d1 =a − tdδ − q f

2(11)

The subscript number here indicates the equilibrium output for each firm in case 1.The same rule applies to the subsequent cases. Accordingly, we can get firms’ profitsat equilibrium as:

π f 1 = [a + (td − 2t f )δ

3]2 (12)

π f 1 = [a + (t f − 2td)δ

3]2 (13)

Case 2: Only foreign firm has adopted ISO14001 in the first stage.

14In reality ISO status needs to be renewed regularly and the update should incur extra cost. We will notinclude such cases in this paper but the detailed discussion is available upon request.

10

Now foreign firm becomes firm 1 described in the general setting. At this timewe assume that foreign firm can be rewarded with an extra consumer surplus becauseISO14001 tends to decrease the management cost of the waste handling. Let z be theweight assigned to consumer surplus, and z ∈ (0, 1). Thus foreign firm’s profit functionnow becomes:

π f = (a − q f − qd) · q f − t f · δq f +zq2

2(14)

meanwhile domestic firm’s profit stays the same:

πd = (a − q f − qd) · qd − td · δqd (15)

The best reply function for each firm in NE becomes:

q∗f 2 =a − t f δ − qd + zqd

2 − z(16)

q∗d2 =a − tdδ − q f

2(17)

and firms’ profits are:π f 2 = (a − q − t f δ)q∗f 2 +

z2· q2 (18)

πd2 = (a − q − tdδ)q∗d2 (19)

Then we substitute π f in case 2 into the conditions that have been drawn in Section6.1:(i) ∂π f

∂q∗d= z(q f + qd) − q f . Since ∂πd

∂q∗f= −qd < 0, to make the cross influence consistent,

we assume z(q f + qd) − q f < 0 ( q f

qd> z

1−z ). This is the so-called negative externality.

(ii) db2(q∗1(k))dq1

=∂2πd∂q f ∂qd

= − 12 < 0 (strategic substitute).

(iii) ∂b1(q∗2(k)|k)∂k =

∂q f

∂z . As the purpose of adopting the strategy is to increase q f , we canassume this term is positive (the sufficient condition is that a > tdδ).

Thus we get the result that dπ f

dz > 0. This indicates that to adopt ISO14001 promotesthe output of firm f . Such response is described by Fudenburg and Tirole (1984) as”Fat Cat Effect”.

Case 3: Only domestic firm has adopted ISO14001.

At this time the reward has been reversed. Now domestic firm gets the credit ofcost reduction (in the form of comsumer surplus gain). Foreign firm’s profit functionis the same as in case 1:

π f = (a − q f − qd) · q f − t f · δq f (20)

and domestic firm’s profit changes:

πd = (a − q f − qd) · qd − td · δqd +zq2

2(21)

The best reply function for each firm becomes:

q∗f 3 =a − t f δ − qd

2(22)

11

q∗d3 =a − tdδ − q f + zq f

2 − z(23)

Compared to case 2, the forms of firms’ profits have also been reversed:

π f 3 = (a − q − t f δ)q∗f 3 (24)

πd3 = (a − q − tdδ)q∗d3 +z2· q2 (25)

In this case, (i) ∂π f

∂q∗d=−qd + z(qd + q f ) < 0; (ii) db2(q∗1(k))

dq1=− 1

2 < 0; (iii) ∂b1(q∗2(k)|k)∂k < 0

(a > t f δ). And we obtain dπddz > 0. Similar to case 2, firm d also raises its output due to

ISO14001 adoption. Consequently, we have the following proposition.

Proposition 1: Both foreign and domestic firms have the incentive to adopt ISO14001.This can also be proved by using the graph of the best reply mapping. See Figure 1below.

Proof: In case 1, the slope of equation (10) is − 12 , and the intercept of foreign firm’s

response is a − t f δ. While in case 2, the slope of equation (16) is − 1−z2−z , which is not

as steep as − 12 . And the intercept is a−t f δ

2−z , which is above a−t f δ

2 on the y axle. Mean-while the response of domestic firm does not change. The mechanism is presented inFigure 3a. The intuition is that after adopting ISO14001, foreign firm’s output levelin equilibrium increases, which means the performance of foreign firm has improved(q∗f 2 > q∗f 1). If we apply the same methodology to case 3, we will have the sameconclusion for domestic firm as well, as shown in Figure 3b. �

Proposition 2: Foreign firm has more incentive to adopt ISO14001 than domestic firmdoes. And the incentive gap increases with 1) the initial capital level a; 2) the expansionof the abatement technology gap (|td − t f |) between the two firms.

Proof: Here in order to compare the incentive, we need to calculate the final benefitfor each firm given the condition of ISO14001 and measure their difference from thebase case. To get intuitive idea, we use numerical examples. On the premise that firms’output only take positive values, we try several combinations:

a) δ = 0.5, t f = 0.8, td = 1, z = 0.5, a = 1. In equilibrium,q f 1 ' 0.23, qd1 ' 0.13⇒ π f 1 = 0.054, πd1 = 0.0178.q f 2 = 0.38, qd2 = 0.06⇒ π f 2 = 0.1092, πd2 = 0.0036.q f 3 = 0.16, qd3 = 0.28⇒ π f 3 = 0.0256, πd3 = 0.0652.∴ |π f 2 − π f 1| > |πd3 − πd1| and the difference is 0.0078.

b) δ = 0.5, t f = 0.8, td = 1, z = 0.5, a = 2.1. In equilibrium,q f 1 = 0.6, qd1 = 0.5⇒ π f 1 = 0.36, πd1 = 0.25.q f 2 = 1.04, qd2 = 0.28⇒ π f 2 = 0.8308, πd2 = 0.0784.q f 3 = 0.38, qd3 = 0.94⇒ π f 3 = 0.1444, πd3 = 0.6988.∴ |π f 2 − π f 1| > |πd3 − πd1| and the difference is 0.022.

c) δ = 0.5, t f = 0.4, td = 1.1, z = 0.5, a = 2.1. In equilibrium,q f 1 = 0.75, qd1 = 0.4⇒ π f 1 = 0.5625, πd1 = 0.16.q f 2 = 1.21, qd2 = 0.17⇒ π f 2 = 1.1053, πd2 = 0.0289.q f 3 = 0.52, qd3 = 0.86⇒ π f 3 = 0.2704, πd3 = 0.6247.∴ |π f 2 − π f 1| > |πd3 − πd1| and the difference is 0.0781.

12

Figure 3: Best reply mapping

qd*

qf*

a − t fδ2 − z

0Case 1

Case 2 a − t fδ2

Figure 3a�

qf 2*

qf 1*

qd (qf )

qf (qd )

qd*

qf*

a − tdδ2 − z

0Case 1

Case 3

a − tdδ2

Figure 3b �

qd (qf )

qf (qd )

From a) and b), we can see that holding other elements constant, the higher a is,the more foreign firm will be motivated to adopt ISO14001 than domestic firm does.

From a) and c), we can see that holding other elements constant, the larger the

13

technology difference |td − t f | is, the more foreign firm will be motivated to adoptISO14001 than domestic firm does15. �

Apart from the propostions we have given, it is easy to realize firm’s performance interms of waste control will be improved after adopting ISO14001. The reason is that wehave made the assumption in the very beginning that the treated waste is proportionateto a firm’s output. Since ISO14001 promotes both firms’ outputs, the waste that hasbeen properly dealt will also increase. This model provides enough evidence for theempirical findings we have achieved in the previous section.

7 ConclusionWe use the firm level data from 2007-2009 in Vietnam to investigate the impact of theadoption of ISO14001, a voluntary environmental standard. In the empirical verifi-cation, two-stage selection model is applied to correct for the potential selection bias.The results show that foreign firms are more likely to adopt ISO14001, and the adoptiondoes affect firms’ overall performance, ranging from reduction on waste discharge toturnover and productivity. We use propensity score matching as the robustness checkand obtain consistent results. Apart from that, a dynamic game theoretical model isemployed to shed light on the mechnism behind the empirical result. This finding is inaccording with most of the existing literature16. It also provides evidence to refute thecritics of “pollution haven” and makes acknowledgeable foreign firms’ effort towardsenvironmental protection. In the meantime we are aware of the limitation of this study.With more detailed information, we would like to extend our analysis to investigationinto more specific industries and regions.

Vietnam is undergoing fast economic transition, however, it is important for Viet-namese government to realize its current situation of industrial pollution. We hope thispaper can offer decision-makers some clue to explore the real source of pollution andcome up with more efficient ways to protect the environment.

15We have tries other values of a as well as |td − t f |, and reach the same conclusion as the above one.16Blackman et al. (2010) do not find significant impact of Clean Industry Program on average environ-

mental performance.

14

References[1] Aseem Prakash and Matthew Potoski (2011) “ISO 14001 and Pollution Reduc-

tion: A Cross-Country Panel Study.” SSRN working paper: id1900139.

[2] Aseem Prakash and Matthew Potoski (2006) “Investing Up: FDI and the Cross-Country Diffusion of ISO 14001 Management Systems.” International StudiesQuarterly 51: 723-744.

[3] Cole, M. A. (2004) “Trade, the Pollution Haven Hypothesis and the Environmen-tal Kuznets Curve: Examining the Linkages.” Ecological Economics 48: 71-81.

[4] CIEM (2007) “Vietnams Economy in 2006, Central Institute of Economic Man-agement.” Hanoi, Vietnam.

[5] CIEM (2008) “Vietnams Economy in 2007, Central Institute of Economic Man-agement.” Hanoi, Vietnam.

[6] Drew Fudenberg and Jean Tirole (1984) “The Fat-Cat Effect, the Puppy-Dog Ploy,and the Lean and Hungry Look.” The American Economic Review: Papers andProceedings 70(2): 361-366.

[7] Elon Koulberg and Jean-Francois Mertens (1986) “On the Strategic Stability ofEquilibria.” Econometrica 54(5): 1003-1037.

[8] Eun-Hee Kim and Thomas P. Lyon (2011) “Strategic Environmental Disclosure:Evidence from the DOEs Voluntary Greenhouse Gas Registry.” Journal of Envi-ronmental Economics and Management 61: 311-326.

[9] Gunnar S. Eskeland and Ann E. Harrison (2003) “Moving to Greener Pastures?Multinationals and the Pollution Haven Hypothesis.” Journal of DevelopmentEconomics 70: 1-23.

[10] Hua Wang and Yanhong Jin (2002) “Industrial Ownership and EnvironmentalPerformance: Evidence from China.” World Bank Policy Research Working Pa-per No. 2936.

[11] Jie He (2006) “Pollution Haven Hypothesis and Environmental Impacts of For-eign Direct Investment: The Case of Industrial Emission of Sulfur Dioxide (SO2)in Chinese Provinces.” Ecological Economics 60: 228-245.

[12] Kanittha Tambunlertchai , Andreas Kontoleon and Madhu Khanna (2013) “As-sessing Aarticipation in Voluntary Environmental Programmes in the DevelopingWorld: The Role of FDI and Export Orientation on ISO14001 Adoption in Thai-land.” Applied Economics 45(15): 2039-2048.

[13] Liangliang Jiang, Chen Lin and Ping Lin (2014) “The Determinants of PollutionLevels: Firm-level Evidence from Chinese Manufacturing.” Journal of Compar-ative Economics 42: 118-142.

[14] Linda T.M. Bui and Samuel Kapon (2012) “The Impact of Voluntary Programson Polluting Behavior: Evidence from Pollution Prevention Programs and ToxicReleases.” Journal of Environmental Economics and Management 64: 31-44.

15

[15] Marco Caiendo and Sabine Kopeining (2008) “Some Practical Guidance for theImplementation of Propensity Score Matching.” Journal of Economic Surveys22(1): 31-72.

[16] Matthew A. Cole, Robert J. R. Elliott and Per G. Fredriksson (2004) “EndogenousPollution Havens: Does FDI Influence Environmental Regulations?” Scandina-vian Journal of Economics 108(1): 157178.

[17] Muthukumara Mani and Shreyasi Jha (2006) “Trade Liberalization and Environ-ment in Vietnam.” World Bank Policy Research Working Paper No. 3879.

[18] Paul R. Rosenbaum; Donald B. Rubin (1983) “The Central Role of the PropensityScore in Observational Studies for Causal Effects.” Biometrika 70(1): 41-55.

[19] Per G. Fredriksson, John A. List and Daniel L. Millimet (2003) “BureaucraticCorruption, Environmental Policy and Inbound US FDI: Eheory and Evidence.”Journal of Public Economics 87: 1407-1430.

[20] Potoski Matthew and Aseem Prakash (2005) “Covenants with Weak Swords: ISO14001 and Firms Environmental Performance.” Journal of Policy Analysis andManagement 24(4): 745-769.

[21] Robert Gibbons (1992) “Game Theory for Applied Economists.” Princeton, NewJersey: Princeton University Press.

[22] Rudolf Kerschbamer (2014) “Game theory lecture 8: Dynamic Games of Com-plete Information-Subgame Perfect Nash Equilibrium in Continuous Games withImperfect Information.” Retrieved from National Center for Econometric Re-search Archive.

[23] Subal C. Kumbhakar and C.A. Knox Lovell (2000) “Stochastic Frontier Analy-sis.” Published by the Press Syndicate of the University of Cambridge.

[24] Thi Xuan Hieu Ngo (2010) “Trade Liberalization and Environment in Vietnam.”Honor Essay.

[25] Thomas P. Lyon and John W. Maxwell (2008) “Corporate Social Responsibil-ity and the Environment: A Theoretical Perspective.” Review of EnvironmentalEconomics and Policy 1(0): 1-22.

[26] Toshi H. Arimura (2014) “Product-related Environmental Regulation and Volun-tary Environmental Actions: Impacts of RoHS and REACH in Malaysia.” IDEdiscussion paper No. 454.

[27] Toshi H. Arimura, N. Darnall and H. Katayama (2011) “Is ISO 14001 A Gatewayto More Advanced Voluntary Action? The Case of Green Supply Chain Manage-ment.” Journal of Environmental Economics and Management 61(2): 170-182.

[28] Toshi H. Arimura, A. Hibiki and H. Katayama (2008) “Is A Voluntary Approachan Effective Environmental Policy Instrument?: A Case for Environmental Man-agement Systems.” Journal of Environmental Economics and Management 55(3):281-295.

[29] Turk, A M. (2009) “The Benefits Associated With ISO 14001 Certification forConstruction Firms: Turkish Case.” Journal of Cleaner Production 17(5): 559-569.

16

Tabl

e 1

Sta

tistic

al S

umm

ary

Varia

ble

NM

ean

S.D

.M

inM

axD

efin

ition

air

2827

43.

1358

8212

.064

020

64.2

8571

treat

ed a

ir w

aste

/tota

l air

was

te (%

)liq

uid

2827

419

.954

2326

.671

610

100

treat

ed li

quid

was

te/to

tal l

iqui

d w

aste

(%)

solid

2827

423

.004

5624

.838

210

99.9

001

treat

ed so

lid w

aste

/tota

l sol

id w

aste

(%)

tota

l_sa

lary

2827

487

23.4

336

580.

520

1789

347

tota

l wag

es a

nd c

ompe

nsat

es o

f the

firm

(Mill

. Don

gs)

turn

over

2827

413

5829

.668

5380

.41.

0785

536

2000

00tu

rnov

er o

f goo

ds a

nd se

rvic

es (M

ill. D

ongs

)TF

P28

274

0.49

9957

50.

1377

992

0.00

0382

10.

8111

43to

tal f

acto

r pro

duct

ivity

usi

ng st

ocha

stic

fron

tier m

etho

dla

bor_

tota

l28

274

236.

2577

873.

1979

161

297

num

ber o

f em

ploy

ees a

t the

beg

inni

ng o

f the

yea

r (Pe

rson

)is

o140

0128

274

0.05

8216

0.23

4155

60

1w

heth

er th

e fir

m h

as IS

O14

001

fdi_

capi

tal

2827

417

.579

5437

.332

380

100

fore

ign

capi

tal/t

otal

cap

ital (

%)

cap_

lab

2827

416

7.09

8610

96.2

490

9862

3.28

capi

tal-l

abor

ratio

food

_mnf

2827

40.

1333

734

0.33

9983

90

1if

the

firm

is in

the

food

-man

ufac

turin

g in

dust

ry, t

he d

umm

y ta

kes 1

mnf

2827

40.

5108

934

0.49

9890

20

1if

the

firm

is in

the

food

-exc

lude

d-m

anuf

actu

ring

indu

stry

, the

dum

my

take

s 1to

talc

ost_

envi

ron

2827

411

99.9

7412

0352

.40

2010

0000

tota

l cos

ts fo

r env

ironm

enta

l pro

tect

ion

(Mill

. Don

gs)

was

te_d

epar

tmen

t28

255

0.27

2553

50.

4452

809

01

whe

ther

the

firm

has

env

ironm

enta

l pro

tect

ion

depa

rtmen

ten

viro

n_sy

stem

2826

80.

2772

747

0.44

7661

20

1w

heth

er th

e fir

m c

arrie

s out

env

ironm

enta

l man

agem

ent s

yste

men

viro

n_st

anda

rd28

265

0.26

1984

80.

4397

222

01

whe

ther

the

firm

mee

ts re

qure

men

ts o

f env

ironm

enta

l sta

ndar

dSo

urce

: Gen

eral

Sta

tistic

s Offi

ce, V

ietn

am

17

(1) (2) (3) (4) (5) (6)Stage1 ISO14001 ISO14001 ISO14001 ISO14001 ISO14001 ISO14001fdi_capital 0.00317*** 0.00317*** 0.00317*** 0.00317*** 0.00317*** 0.00317***

(0.000320) (0.000320) (0.000320) (0.000320) (0.000320) (0.000320)total_labor 9.89e-05*** 9.89e-05*** 9.89e-05*** 9.89e-05*** 9.89e-05*** 9.89e-05***

(9.42e-06) (9.42e-06) (9.42e-06) (9.42e-06) (9.42e-06) (9.42e-06)capital-labor ratio 3.36e-05*** 3.36e-05*** 3.36e-05*** 3.36e-05*** 3.36e-05*** 3.36e-05***

(7.88e-06) (7.88e-06) (7.88e-06) (7.88e-06) (7.88e-06) (7.88e-06)mnf 0.212*** 0.212*** 0.212*** 0.212*** 0.212*** 0.212***

(0.0349) (0.0349) (0.0349) (0.0349) (0.0349) (0.0349)food_mnf 0.267*** 0.267*** 0.267*** 0.267*** 0.267*** 0.267***

(0.0438) (0.0438) (0.0438) (0.0438) (0.0438) (0.0438)waste_department 0.956*** 0.956*** 0.956*** 0.956*** 0.956*** 0.956***

(0.0277) (0.0277) (0.0277) (0.0277) (0.0277) (0.0277)

Stage 2 air liquid solid total_salary turnover TFPiso14001 17.77*** 52.97*** 22.81*** 12,165*** 1.085e+06*** 0.209***

(1.102) (2.334) (2.193) (2,010) (57,421) (0.0123)fdi_capital -0.0211*** 0.0700*** 0.0160*** -4.230 165.4 0.000428***

(0.00223) (0.00496) (0.00435) (3.929) (116.0) (2.50e-05)total_labor -0.000174* 0.000365* 0.000599*** 33.31*** 229.7*** 3.81e-06***

(9.61e-05) (0.000212) (0.000188) (0.170) (4.993) (1.07e-06)capital-labor ratio 0.000146** 1.18e-05 0.000289** 0.732*** 86.54*** 3.95e-06***

(6.94e-05) (0.000154) (0.000135) (0.122) (3.606) (7.76e-07)mnf 1.378*** 2.620*** 13.06*** -3,277*** -55,078*** -0.00811***

(0.174) (0.388) (0.339) (305.6) (9,052) (0.00195)food_mnf 1.252*** 13.08*** 7.979*** -2,719*** 57,474*** 0.0117***

(0.243) (0.540) (0.472) (425.7) (12,606) (0.00271)lambda -7.705*** -24.61*** -10.63*** -4,665*** -388,262*** -0.0848***

(0.558) (1.164) (1.117) (1,028) (29,095) (0.00625)Observations 28,255 28,255 28,255 28,255 28,255 28,255Note: In the baseline estimation, waste_department is used in the first stage. Totalcost_environ, environ_system,

Standard errors in parentheses. Year dummies are included.

Table 2 Baseline Estimation Results

environ_standard and their combinations are used alternatively.

*** p<0.01, ** p<0.05, * p<0.1

18

Table 3 Propensity Score Matching Results(1) (2) (3) (4) (5) (6)

VARIABLES air air liquid liquid solid solidMethod Nearest Kernel Nearest Kernel Nearest KernelATT 2.695*** 2.695*** 5.211*** 5.211*** 1.685 1.685

(0.795) (0.795) (1.325) (1.325) (1.169) (1.169)Observations 2,910 2,910 2,910 2,910 2,910 2,910

(7) (8) (9) (10) (11) (12)VARIABLES total_salary total_salary turnover turnover TFP TFPMethod Nearest Kernel Nearest Kernel Nearest KernelATT 23,761*** 23,761*** 499,922*** 499,922*** 0.0541*** 0.0541***

(2,547) (2,547) (45,005) (45,005) (0.00593) (0.00593)Observations 2,910 2,910 2,910 2,910 2,910 2,910Standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1

19