Printer Friendly

Pollution, production, and sectoral differences in a transition economy.

INTRODUCTION

Recent economic studies examine the factors driving corporate environmental performance in transition economies (eg, Bluffstone, 1999). Our study analyses firm-level environmental performance, as measured by the level of air pollutants emitted by large stationary sources, in the transition economy of the Czech Republic during the years 1993-1998. In particular, our study assesses whether firms face economies and/or diseconomies of scale with respect to pollution control by evaluating the effects of production on emission levels. As important, this paper assesses whether certain sectors face different scales of production--economy or diseconomy--than other sectors. Specifically, this paper assesses whether the effects of production on emissions differ across various sectors.

This close examination of production scale effects, including their variation across sectors, stands in stark contrast to previous studies. Certainly, no study of a transition economy scrutinises the effects of production and sector. This omission extends to studies of any economy. While previous studies include production and sector as control variables in their empirical analysis (Khanna and Damon, 1999; Magat and Viscusi, 1990; Foulon et al., 2002; Earnhart and Lizal, 2006), these studies fail to scrutinise the important relationship between pollution, production, and sector.

This paper explores the relationship between pollution and production within the following format. The following section develops a simple framework for understanding production scale effects. Next section describes the database on firm-level air pollutant emissions and production. Next one estimates and interprets the effects of production scale and sectoral differences on air pollutant emissions. The final section concludes.

SCALE OF PRODUCTION: ECONOMIES AND/OR DISECONOMIES OF SCALE

This paper analyses the effects of production scale on environmental performance as measured by air pollutant emission levels. In particular, this paper analyses whether or not firms face economies and/or diseconomies of scale, possibly depending on the level of production. The analysis assesses the noted possibilities by constructing emissions as a polynomial function of production. Specifically, the level of air emissions depends on production (linear), production-squared (quadratic term), and production-cubed (cubic term). Let p denote the level of pollution and y denote the level of production. The following equation captures the relationship between pollution and production:

p = [alpha] + [beta]y + [tau][y.sup.2] + [delta][y.sup.3] (1)

where [alpha] denotes a constant term. (1) Based on Equation (1), the two following equations capture the first and second derivatives with respect to production, denoted as p' and p", respectively:

p' = [beta] + 2 [tau]y + 3 [delta][y.sup.2] (2)

and

p" = 2[tau] + 6[delta]y (3)

A firm faces economies of scale if p" < 0 and faces diseconomies of scale if p">0. As displayed in Equation (3), the quadratic and cubic production parameters--[tau] and [delta]--and the production level, y, collectively determine whether a firm faces economies or diseconomies of scale. If the quadratic parameter is negative ([tau]<0) but the cubic parameter is positive ([delta]>0), the sign of p" and thus the production scale effect depends on the level of production. As production increases, a firm first faces economies of scale then later diseconomies of scale, as p" shifts from negative to positive once production becomes sufficiently high for the cubic term to dominate. If the quadratic parameter is positive ([tau]>0) but the cubic parameter is negative ([delta]<0), the sign of p" and the production scale effect again depend on the level of production. However, in this case, the opposite conclusion follows; as production increases, a firm first faces diseconomies of scale then later economies of scale. If both the quadratic and cubic production parameters are negative ([tau]<0, [delta]<0), then p" is unambiguously negative and a firm faces economies of scale regardless of the production level. If both parameters are positive ([tau] > 0, [delta] > 0), then p" is unambiguously positive and a firm faces diseconomies of scale regardless of the production level.

When the cubic parameter equals zero ([delta] = 0), which applies when the estimated cubic parameter is insignificantly different from zero, the quadratic parameter ([taul]) alone dictates whether a firm faces economies if scale ([tau] < 0] or diseconomies of scale [[tau]>0). Consequently, the identified scale effect is independent of the production level.

Most likely, the general tendency is for emissions to rise with production. If true, the linear production parameter is positive: [beta]>0. In this case, at sufficiently low levels of production, the first derivative with respect to production must be positive: p'>0. However, if present, economies of scale eventually become so strong that at sufficiently high levels of production the first derivative becomes negative: p' < 0, indicating that emissions are actually falling in production. As a means for assessing the strength of any identified economies of scale, the analysis identifies the level of production that serves as the transition from p'>0 to p' <0, that is, level of production that sets p' = 0.

Below this paper empirically assesses whether or not firms face economies and/or diseconomies of scale, possibly depending on the level of production. It also assesses whether these productions scale effects vary across different sectors.

DATA ON EMISSIONS AND PRODUCTION

Czech republic as study site

To examine the effects of production scale and sectoral differences on firm-level air pollutant emissions, we exploit data on firms in the Czech Republic between 1993 and 1998, which is an excellent site and time period for our study. First, the Czech Republic had a substantially degraded environment; in particular, poor ambient air quality and air pollution were large environmental problems of public concern in the Czech Republic (World Bank, 1992). In response to public concern, the Czech Republic's government authorities took substantial and effective steps to decrease air emissions dramatically during the period 1991-1998 (Czech Ministry of Environment, 1998). Figure 1 of the introduction to this symposium displays the trend of economy-wide air emissions over this period. Most likely, an overall decline in economic activity at least partially explains this decline; nevertheless, firms' pollution control efforts (eg, switching to low-sulphur coal) may also explain much of this decline (World Bank, 1999). Second, consistent with this focus on pollution control efforts, investment in environmental protection was most important during period between 1992 and 1998. As a percentage of Czech gross domestic product (GDP), investment rose dramatically after 1991 from a level of 1.3% to a peak of 2.5% in 1997 and tailed off after 1998 back to a pre-transition level of 1.1% by 2000; in 1990, investment was 1.1% of GDP. (2) Third, the Czech Republic was attempting to enter the European Union (EU) during this period and was required to reduce its industrial emissions in order to qualify for membership.

Panel data on production and emissions

To examine production at Czech enterprises, we gather data from a database provided by the private data vendor Aspekt. The database provides information drawn from firms' balance sheets and income (profit/loss) statements. This database also identifies a firm's primary sectoral classification. We gather balance sheet and income statement data for the years 1993 to 1998. The Aspekt database includes all firms traded on the Prague Stock Exchange, publicly traded firms (ie, firms registered for trading on the RMS (Registracni misto system) secondary market), and a majority of the remaining large Czech firms (plus the key trading partners of these large firms). This comprehensive database has been used by previous studies of Czech firm-level performance (eg, Claessens and Djankov, 1999; Weiss and Nikitin, 2002; Kocenda and Svejnar, 2002; Djankov, 1999). Production is measured as production value in terms of Czech Crowns. In order to compare properly across the 6 years of the sample period, the analysis adjusts the production value data according to the Czech Consumer Price Index so that all values are denominated in 1998 Czech Crowns. (3,4)

We also gather data on air pollutants emitted by facilities located in the Czech Republic during the years 1993 and 1998. The included pollutants are carbon monoxide (CO), sulphur dioxide (S[O.sub.2]), particulate matter (PM), and nitrous oxides (N[O.sub.x]), which represent the main and most heavily regulated pollutants in the Czech Republic, similar to other industrialised nations. The Czech Hydrometeorological Institute maintains two databases on air emissions from stationary sources. The REZZO-1 database records emissions for large, stationary sources. The REZZO-2 database records emissions for medium-sized, stationary sources. We use the REZZO-1 database for our analysis, which, by its nature, should have the largest overlap with the production data we possess. The REZZO-1 database records emissions at individual units of individual facilities. The Czech Hydrometeorological Institute aggregates the air emissions to the level of each facility before public release of the data. We further aggregate air emissions across all facilities associated with a single firm. Initially, we add the four pollutants into one composite measure of air emissions, similar to previous studies of environmental performance (Arora and Cason, 1995, 1996; Konar and Cohen, 1997, 2001; Khanna et al., 1998; Khanna and Damon, 1999). Later, we examine each pollutant separately.

In order to generate the largest sample possible and to avoid a sample selection bias due to attrition, we create an unbalanced panel of firm-year observations for the time period 1993-1998. The overlap between the production data set and the air emissions data set is quite limited. The two data sets only hold 4,688 observations in common. (5) Then we screen for meaningful data by applying the following criteria: non-missing emissions and positive production value. (6) This screening and restriction generates an unbalanced panel of 2,632 observations. (7) The data set contains 651 firms.

Descriptive statistics

Table 1 presents a statistical summary of the data. As shown in Table la, our data are sufficiently spread across the 6 years of our time frame. Table 1b presents a statistical summary of emissions and production value. As demonstrated by the standard deviation measures, our data set contains much variation in production value and emissions, which facilitates our estimation.

Table 2 demonstrates that emissions differ dramatically across the variety of sectors. In particular, per firm emissions vary dramatically across the sectors. On the low end, each firm in the finance, real estate, rentals, business, research, and public administration sector emits only 14t on average within a given year. On the high end, each firm in the electricity, gas, and water supply sector emits a substantial 6,677 t on average within a give year. (8) These tabulated differences indicate that sectoral effects on environmental performance may prove quite meaningful. Table 2 also reveals that production value differs substantially across the variety of sectors. At one extreme, each firm in the agriculture, hunting, forestry, and fisheries sector generates only 134 million Czech Crowns of production value on average within a given year. At the other extreme, the coke and refined petroleum sector generates a substantial 10,775 million Czech crowns of production value on average within a given year. By dividing emissions by production value, the last two columns of Table 2 provides a potentially better view of variation in emissions across sectors. Even after adjusting for production, emissions differ meaningfully across the variety of sectors. The electricity, gas, and water supply sector emits the most amount of pollution in a relative sense at 2.6 t per one million Czech crowns of production value. In contrast, the construction sector emits the least in a relative sense with only 0.02 t per one million Czech crowns of production value.

ECONOMETRIC ANALYSIS OF AIR POLLUTANT EMISSION LEVELS

In this section, we use the described data to explore the variation in Czech air pollution emissions across a broad range of production scales and numerous sectors.

Econometric structure

We estimate the relationship between air pollution levels and key explanatory variables. Specifically, we estimate air pollutant emissions in absolute levels for two reasons. First, this form is relevant for the Czech legal framework since Czech government regulators impose quantity-based limits (eg, tonnes per month), which relate directly to absolute levels. (9) Second, this form cleanly connects production levels to emission levels. (10)

To construct the econometric models, we define the following notation. As the dependent variable, Pit denotes the amount of pollution emitted by firm i in time period t. Emissions depend on the level of production, which is denoted as [y.sub.it]. The level of production enters in three terms: linear ([y.sub.it]), quadratic ([y.sup.2.sub.it]), and cubic ([y.sup.3.sub.it]). To capture sector-specific variation, we include a sectoral indicator for each sector displayed in Table 2. These indicators are collectively denoted as vector [X.sub.i]. Inclusion of sectoral indicators allows the intercept of the relationship between emissions and production to differ across sectors. Of course, the three production effects may also differ across sectors, a point which we address below. To control for variation over time with the legal framework controlling air emissions, we include individual year indicators for the years 1994 to 1998, with 1993 serving as the benchmark year. These indicators are collectively denoted as vector [T.sub.t].

Given this notation, we formulate the following econometric model:

[p.sub.it] = [alpha] + [beta][y.sub.it] + [tau][y.sup.2.sub.it] + [delta][y.sup.3.sub.it] + [gamma][T.sub.t] + [theta][X.sub.i] + [e.sub.it] (4)

where [alpha] denotes the constant term and [e.sub.it] denotes the error term. As the regression equation contains both a constant and an indicator for each sector, the estimation must restrict the sum of the coefficients associated with the sectoral indicators to equal zero (Suits, 1984). By including all the sectoral indicators, while restricting the sum of coefficient values to zero, each sectoral coefficient is calculated relative to the average sector rather than a specific sector, which applies when a single sectoral indicator is excluded from the regressor set. Specifically, each sectoral coefficient represents a particular sector's deviation from the sample-wide mean, as captured by the constant term.

To accommodate the panel data structure, we estimate Equation (4) using a between-group estimator. This estimator calculates the mean value of the dependent and independent variables for each firm and then estimates the model based on these mean values. In essence, a between-group estimator accommodates the panel data structure by collapsing the panel data set into a quasi-cross-sectional data set. The modified regression equation becomes the following:

[p.sub.i] = [alpha] + [beta][y.sub.i] + [tau][y.sup.2.sub.i] + [delta][y.sup.3.sub.i] + [theta][X.sub.i] + [e.sub.i] (5)

where [p.sub.i] and [y.sub.i] represent the firm-specific mean values for emissions and production, respectively. While the between-group estimator is not able to capture the experience of a single firm, use of this estimator appropriately frames our analysis. We are not interested in capturing the experience of a single firm since no single firm can operate at multiple levels of production.

Instead, we are interested in capturing (and are able to capture) the experience of an entire industrial sector since an entire industrial sector, as a set of different individual firms, can operate at multiple levels of production. In addition, we use other standard panel estimators--pooled OLS, fixed effects, and random effects--to estimate Equation (4); however, we neither report nor interpret these estimation results for reasons described below. Pooled OLS estimates frequently suffer from omitted variable bias by excluding firm-specific intercept terms. This study tests for this bias by implementing a F-test of fixed effects. If the F-test indicates significant firm-specific effects, the pooled OLS estimates are biased. Based on the estimated test statistics, the pooled OLS estimator suffers from the noted omitted variable bias. (11) Random effects estimates might not be consistent. Based on estimated Hausman test statistics, the random effects estimator is inconsistent. (12) In contrast, the fixed effects estimator is always consistent and unbiased. However, by construction, the fixed effects estimator examines deviations from each firm's individual mean. Thus, the fixed effects estimator identifies the effects of production on emissions based exclusively on within-firm variation. (Of course, the fixed effects estimator calculates the relevant coefficients using information drawn from across firms.) In contrast, the between-group estimator identifies the effects of production on emissions based exclusively on cross-sectional variation or across-firm variation, which arguably better captures production scale effects. The between-group estimator effectively captures the production scale effects experienced by an entire industrial sector and examines variation across multiple sectors.

Estimation results: sum of individual pollutant emissions

As its primary objective, this paper examines production scale effects. To examine fully the economies and diseconomies of scale associated with pollution control, the econometric analysis first considers a regression equation that includes three production terms: linear, quadratic, and cubic.

The estimation results are shown in the first column of Table 3. These results indicate that the cubic production term does not significantly affect the level of air emissions. Consequently, the analysis also estimates a regression equation that excludes the cubic production term. These estimation results are shown in the second column of Table 3. Both sets of results indicate that the linear production effect is significantly positive, while the quadratic production effect is significantly negative; in other words, emissions are generally rising in production but at a declining rate. As the cubic production term is statistically insignificant, the quadratic product term alone determines that the average firm faces economies of scale for all levels of production. When the cubic production term is included as a regressor, the quadratic production term is only marginally significant at a P-value of 0.09. However, when the insignificant cubic production term is excluded, the quadratic production term is highly significant at a P-value of 0.0001. With the exception of this one point, the two sets of results are both qualitatively and quantitatively highly similar.

As the average firm faces economies of scale, at sufficiently high levels of production, emissions will start to decrease in production, that is, the first (partial) derivative with respect to production is negative. This result follows since the negative quadratic term eventually overwhelms the positive linear term. The first derivative equals zero when the level of production equals -[beta]/2[tau]; denote this threshold level as [Y.sup.*] = -[beta]/2[tau]. (Manipulation of Equation (2) identifies this relationship when [delta] = 0.) The units for [beta] are emission tonnes per millions of Czech Crowns; the units for r are emission tonnes per trillions of Czech Crowns. Based on the model that excludes the statistically insignificant cubic production term, the estimated values for [beta] and [tau] are 0.951 and -[1.42.sup.-5], respectively. Consequently, [Y.sup.*] = 33,442 million Czech Crowns. This value lies between the 99.4 and 99.5 percentiles of the sample distribution for production value; only four observations possess a production value above [Y.sup.*]. Thus, emissions are rising in production at almost every sample-relevant level of production.

Next, this sub-section considers the variation of emissions across individual industrial sectors. Jointly, the set of sectoral indicators significantly affect emission levels; a F-test statistic of 5.80 is significant at the 0.0001 level, given 19 and 610 degrees of freedom for the F-test's numerator and denominator, respectively, based on the estimation that excludes the insignificant cubic production term. As shown in Table 2, certain sectors deviate from the average sector. In particular, the coke and refined petroleum sector generates significantly lower emissions than the average sector. This result may seem surprising at first blush. However, since the analysis controls for the scale of production, it may seem more believable that this potentially pollution-intensive sector actually controls emissions better than average. In contrast, the chemicals sector generates significantly greater emissions than the average sector. In other words, even after controlling for the scale of production, this sector controls emissions worse than average. Similarly, the electricity, gas, and water supply sector generates significantly greater emissions than the average sector. Lastly, the construction sector generates significantly lower emissions than the average sector. However, this effect is only marginally significant with a P-value of 0.09 when the cubic term is excluded from the estimation; the P-value equals 0.11 when the cubic term is included.

As both production and sectoral classification affect the level of emissions, it is important to examine whether production scale effects vary across sectors. With this goal in mind, as an alternative specification, we also estimate a regression equation that includes the interactions between each individual sectoral indicator and the three production terms--linear, quadratic, and cubic--as additional regressors. The estimation results indicate that each and every interaction term is statistically insignificant. (Given this insignificance, we do not report these additional results but offer them upon request.) This conclusion of insignificance is robust to the exclusion of the cubic term and its interactions with the set of sectoral indicators. Thus, it appears that each individual sector's linear, quadratic, and cubic production effects do not differ from the average sector's production effects. In other words, no individual sector deviates from the average sector with regards to the shape of the emissions--production relationship. In contrast, four sectors deviate from the average sector in the one remaining dimension of the emissions--production relationship: its intercept.

To demonstrate visually the relationship between emissions and production and the effect of sector on this relationship, we generate predicted emission values based on the estimated coefficients and various production levels. In total, we generate five relationships. One for the average sector and one for each of the four sectors that significantly deviate from the average sector. Then we relate the five sets of predicted emission values to production scale by graphing the predicted values against a broad range of production levels. Based on Equation (5), emissions depend on a constant term, production, and sectoral indicators. As the estimation restricts the sum of sectoral indicator coefficients to equal zero, the predicted values for the average sector do not depend on the sectoral coefficient estimates. Instead, the predicted values apply to the average sector without any further adjustment. The four sector-specific curves represent parallel shifts to the curve of the average sector by altering the intercept term.

Figure 1 displays the predicted emission values graphed against production levels. As shown, emissions rise in production at a declining rate until the very end of the sample when emissions fall in production. (13) Moreover, four sectors face different levels of emissions at comparable production levels even though they face similar marginal trade-offs between emissions and production.

[FIGURE 1 OMITTED]

Estimation results: emissions of individual pollutants

The previous sub-section examines the level of air pollutant emissions summed across the four pollutants: CO, S[O.sub.2], PM, and N[O.sub.x]. This sub-section examines the level of emissions for each pollutant separately. This separate treatment permits the analysis to assess the need to qualify the preceding results. As important, this treatment should allow the analysis to refine its examination of sectoral differences since the importance of particular pollutants most likely varies across sectors.

Our examination of individual pollutant emission levels excludes the cubic production term for two reasons. First, its exclusion tightens the estimate of the statistically significant quadratic production term since the cubic production term proves statistically insignificant. Second, the estimation results are otherwise qualitatively and quantitatively robust to the exclusion of the cubic production term. Estimation results are shown in Table 4.

First, we consider the effect of production on emission levels. As with the sum of air pollutant emissions, for each air pollutant type, the estimated linear production term is significantly positive while the estimated quadratic production term is significantly negative. Thus, Czech firms face economies of scale with respect to air pollution controls regardless of the air pollutant type. This point notwithstanding, the linear and quadratic production terms may quantitatively differ across the four pollutants, implying that the degree of scale economies varies across the four pollutants. To test these possibilities, we construct and estimate a multi-equation seemingly unrelated regression (SUR) system that includes a separate regression equation for each individual pollutant. While this implementation does not alter the coefficient estimates, since the set of regressors is identical for all four equations, it permits proper testing by allowing correlation across the four error terms associated with the regression system's four equations. Based on two F-tests of equal effects, we conclude that both the linear and quadratic production terms vary across the four pollutants, that is, we reject the joint null hypothesis of equal linear production coefficients (F-test statistic = 44.66, P-value = 0.0001) and the joint null hypothesis of equal quadratic production coefficients (F-test statistic = 22.24, P-value = 0.0001). As the linear and quadratic production terms vary across pollutants, it is worthwhile to examine the variation in the degree of scale economies across the four pollutants.

To compare the strength of the scale economies across the pollutant types, we assess the production level where the negative quadratic production term just starts to overwhelm the positive linear production term, which implies emissions decline in production. This point is identified by [Y.sup.*] = -[beta]/2[tau], as demonstrated in the preceding sub-section. Based on the estimated values of [beta] and [tau], as shown in Table 4, the values of [Y.sup.*] for CO, S[O.sub.2], PM, and N[O.sub.x] are respectively the following:

[Y.sup.*] for CO = 37,420.1 million Czech Crowns;

[Y.sup.*] for S[O.sub.2] = 31,268.0 million Czech Crowns;

[Y.sup.*] for PM = 34,346.3 million Czech Crowns; and

[Y.sup.*] for N[O.sub.x]-33,997.4 million Czech Crowns.

These values indicate that the economies of scale are strongest in the control of S[O.sub.2] and weakest in the control of CO. These values fall at certain points within the sample distribution of production value. With the strongest economies of scale, the [Y.sup.*] value for S[O.sub.2] lies at the 99.4 percentile. In contrast, with the weakest economies of scale, the [Y.sup.*] value for CO lies at the 99.6 percentile. In between, the [Y.sup.*] values for PM and N[O.sub.x] lie at the 99.5 percentile. Thus, the different estimates do not seem to vary meaningfully in relation to the sample distribution. Regardless of the pollutant, emissions are rising in production at almost every level of production.

As an alternative means for comparing the strength of scale economies across the pollutant types, we again generate predicted emission values based on the estimated coefficients and various production levels. In total, we generate four relationships: one for each pollutant based on the average sector. As shown in Figure 2, the shapes of the four emission-production curves differ substantially across the four pollutants. Based on these shapes, S[O.sub.2] emissions appear to possess the strongest economy of scale, while CO emissions appear to possess the weakest economy of scale. PM emissions and N[O.sub.x] emissions lie in between these two extremes.

[FIGURE 2 OMITTED]

Next, we examine the variation of individual pollutant emissions across the various sectors. As shown in Table 4, certain sectors deviate from the average sector in their effect on emissions. First, consider CO emissions, shown in the first column of Table 4. Relative to the average sector, the mining and quarrying sector, coke and refined petroleum sector, and chemicals sector each generates significantly lower CO emissions. In contrast, the basic metals and fabricated metal products sector generates significantly higher CO emissions than the average sector. Second, consider S[O.sub.2] emissions, as shown in the second column of Table 4. Again, the coke and refined petroleum sector generates lower S[O.sub.2] emissions than average. In contrast, both the chemicals sector and the electricity, gas, and water supply sector generate higher S[O.sub.2] emissions than average. Third, consider PM emissions, as shown in the first column of Table 4. Again, the coke and refined petroleum sector generates lower PM emissions than average, while the chemicals sector, metals sector, and electricity, gas, and water supply sector each generates higher PM emissions than average. Lastly, consider N[O.sub.x] emissions, as shown in the second column of Table 4. Both the coke and refined petroleum sector and the construction sector generate lower N[O.sub.x] emissions than average. The mining and quarrying sector, chemicals sector, and electricity, gas, and water supply sector each generates higher N[O.sub.x] emissions than average.

Collectively, this set of four estimation results provide a richer understanding of each sector's abilities to control air pollution with one exception: the coke and petroleum sector. This exceptional sector generates below-average emissions for all four individual pollutants and the sum of individual pollutants. The chemicals sector controls CO emissions, relative to other sectors, better than it controls the other three air pollutant emissions: S[O.sub.2], PM, and N[O.sub.x]. This sector's above-average ability to control CO emissions is not demonstrated by estimating the sum of individual pollutants. Instead, the pollutant-sum result reveals only the sector's below-average ability to control S[O.sub.2], PM, and N[O.sub.x] emissions. The mining and quarrying sector's above-average ability to control pollution is concentrated in CO emissions, while its below-average ability to control air pollution is concentrated in N[O.sub.x] emissions. These contradictory abilities might explain why the estimation of summed individual pollutants does not identify a significant effect for this sector. The metals sector's below-average ability to control pollution is concentrated in CO emissions and PM emissions. Perhaps this sector's average ability to control S[O.sub.2] and N[O.sub.x] emissions explains why the estimation of summed individual pollutant emissions does not identify a significant effect for this sector. The electricity, gas, and water supply sector generates above-average emissions for three of the four pollutants--S[O.sub.2], PM, and N[O.sub.x]. Not surprisingly, this same below-average ability to control emissions is reflected in the estimation results for the sum of individual pollutants even though this sector controls CO emissions with an average ability. The construction sector generates below-average emissions for only one pollutant--N[O.sub.x]--yet this effect is also identified in the estimation results for the sum of individual pollutants. Thus, the pollutant-sum-result masks the construction sector's average ability to control the other three pollutants--CO, S[O.sub.2], and PM.

SUMMARY

Based on our analysis of Czech firms in the years 1993 to 1998, we conclude that the average Czech firm faces economies of scale with respect to emission controls. As production rises, the effect of production on emissions falls. In other words, larger firms--in terms of production--are better able to manage their emissions, after making an adjustment for the level of production. This conclusion is robust to the estimation of all major air pollutants as a composite or each individual air pollutant separately. In addition, our analysis indicates that certain sectors are better able to control air emissions than the average sector, while other sectors are less able to control air emissions. Nevertheless, each sector seems to face the same form of scale economies. Based on our analysis, the linear and quadratic effects within the emission-production relationship do not differ across the various sectors. The same conclusion is drawn for the cubic production effects.

Use of these results is constrained. First, due to the scope of this analysis, these results need not generalise to firms operating outside of the Czech Republic. Future analysis of the economies or diseconomies of scale in pollution control for firms operating in other countries is needed before any generalisation is possible. Second, given the general nature of this analytical investigation, it would be difficult for any policymaker to modify environmental protection laws or protocols based on the analytical results. In particular, the analysis does not examine why larger firms are better able to manage their emissions. This point notwithstanding, the analytical results should at least prompt policymakers to explore more fully this point when examining policy changes.

Acknowledgements

We acknowledge the financial support of a COBASE grant from the National Research Council. Dietrich Earnhart acknowledges the financial support of the College of Liberal Arts and Sciences at the University of Kansas in the form of General Research Fund grant # 2301467.

REFERENCES

Arora, S and Cason, T. 1995: An experiment in voluntary environmental regulation: Participation in EPA'S 33/50 program. Journal of Environmental Economics and Management 28(3): 271-286.

Arora, S and Cason, T. 1996: Why do firms volunteer to exceed environmental regulations? Understanding participation in EPA*s 33/50 program. Land Economics 72(4): 413-432.

Bluffstone, R. 1999: Are the costs of pollution abatement lower in central and Eastern Europe? Evidence from Lithuania. Environment and Development Economics 44: 449-470.

Claessens, S and Djankov, S. 1999: Ownership concentration and corporate performance in the Czech Republic. Journal of Comparative Economics 27(3): 498-513.

Czech Ministry of Environment. 1998: Statistical Environmental Yearbook of the Czech Republic. Czech Ministry of the Environment: Prague.

Djankov, S. 1999: Ownership structure and enterprise restructuring in six newly independent states. Comparative Economic Studies 41(1): 75-95.

Earnhart, D and Lizal, L. 2006: Effects of ownership and financial performance on corporate environmental performance. Journal of Comparative Economics 34(1): 111-129.

Foulon, J, Lanoie, P and Laplante, B. 2002: Incentives for pollution control: Regulation or information. Journal of Environmental Economics and Management 44: 169-187.

Khanna, M, Quimio, WR and Bojilova, D. 1998: Toxics release information: A policy tool for environmental protection. Journal of Environmental Economics and Management 36: 243-266.

Khanna, M and Damon, L. 1999: EPA's voluntary 33/50 program: Impact on toxic releases and economic performance of firms. Journal of Environmental Economics and Management 37: 1-25.

Kocenda, E and Svejnar, J. 2002: The effects of ownership forms and concentration on firm performance after large-scale privatization William Davidson Institute Work Paper No. 471, May.

Konar, S and Cohen, M. 1997: Information as regulation: The effect of community right to know laws on toxic emissions. Journal of Environmental Economics and Management 32: 109-124.

Konar, S and Cohen, M. 2001: Does the market value environmental performance? Review of Economics and Statistics 83(2): 281-289.

Magat, W and Viscusi, K. 1990: Effectiveness of the EPA's regulatory enforcement: The case of industrial effluent standards. Journal of Law and Economics 33: 331-360.

Suits, D. 1984: Dummy variables: Mechanics v. interpretation. Review of Economics and Statistics 66(1): 177-180.

Weiss, A and Nikitin, G. 2002: Effects of ownership by investment funds on the performance of Czech firms Chapter 10. In: Anna, M and Anjan, T. (eds). Designing Financial Systems in Transition Economies: Strategies for Reform in Central and Eastern Europe. MIT Press: Cambridge, MA, 187-214.

World Bank. 1992: Czech and Slovak federal republic joint environmental study Report No. 9623-CS.

World Bank. 1999: Country economic memorandum--Czech Republic, report prepared by Julia Bucknall, Rita Cestiti and Gordon Hughes. World Bank: Washington, DC.

(1) For the purposes of this conceptual framework, we assume that [alpha] is non-negative. For the econometric analysis, we place no restrictions on the value of [alpha] since we do not know the exact shape of the polynomial function. By not restricting the value of [alpha], the econometric analysis is able to estimate at least an approximation of any unknown higher-order polynomial function.

(2) Source: Czech Statistical Office, Czech Ministry of the Environment.

(3) Estimation of values that are adjusted according to the Czech producer price index would generate nearly identical results since the consumer price index and producer price index are strongly and significantly correlated (D = 0.997, P = 0.00001).

(4) While our data measure production in monetary terms rather than physical terms, we argue that this difference does not alter the conceptual framework provided in the section 'Scale of Production: Economies and/or Diseconomies of Scale' for the following reasons. First, as long as firms are sufficiently price-takers, production sales are linearly proportional to physical production. Second, the analysis deflates the currency-denominated values into real terms. Third, since the analysis controls for any sectoral variation in general, it also controls for sectoral variation in prices in particular.

(5) While unfortunate, this limited overlap does not indicate a problem with the data. Instead, it may simply indicate that firms included in the Aspekt database own medium rather than large stationary air emission sources or own facilities that lack the need to measure air emissions because such emissions are non-existent or extremely low. In this way, the Aspekt database need not completely represent large stationary air polluters. Therefore, our results may not generalise to all or most large stationary air polluters. The opposite concern is not relevant. The REZZO-1 database is fully comprehensive of all large polluters.

(6) We also apply these additional screening criteria: positive total assets and positive fixed assets.

(7) Missing values, not inconsistent values, cause most of the reduction in sample size. The final sample size is highly comparable to previous studies of firms in the Czech Republic (eg, Kocenda and Svejnar, 2002).

(8) As in other industrialised economies, the Czech electricity sector emits a dramatically large amount of air pollutants.

(9) Even though air protection laws use concentration units to define air pollution standards, the Czech Environment Inspection--Air Division translates concentration-based standards into quantity limits when writing facility-specific air pollution permits.

(10) Alternative analysis estimates air pollutant emissions relative to production ('relative emissions'). To generate relative emissions, this alternative analysis divides the absolute emissions level by the production level. This alternative form may also plausibly capture firm-level environmental performance. However, this alternative form has two disadvantages. First, it does not permit a general analysis of economies and diseconomies of scale. Second, estimation of relative emissions generates much less meaningful and significant results, as judged by a priori coefficient sign expectations, estimated individual coefficients, and overall explanatory power.

(11) The relevant F-test statistics are 26.29 and 26.43 for the specifications including and excluding the cubic production term, respectively. Both statistics are significant at levels < 0.0001.

(12) The relevant Hausman test statistics are [chi square]-distributed with seven degrees of freedom. The estimates statistics are 24.76 and 26.71 for the specifications including and excluding the cubic production term, respectively. The related P-values are 0.0017 and 0.0004, respectively.

(13) Two features of Figure 1 are worthy of elaboration. First, since the quadratic polynomial function possesses a negative quadratic term, the curves eventually must tend towards zero. At exceptionally high production levels, the model predicts emissions near zero. Thus, the model does not predict these outliers well. Second, the model predicts a negative intercept for the coke and refined petroleum sector. As the intercept represents a situation that does not occur in the sample, it is difficult to interpret. Moreover, as described in footnote #1, by allowing the intercept to take any value (including a negative value), we are able to approximate any higher-order polynomial function. Perhaps, the negative intercept reflects this approximation. As possible extension of this approximation, the negative intercept may indicate that the model does not predict well at very low production levels.

DIETRICH EARNHART (1) & LUBOMIR LIZAL (2)

(1) Department of Economics, 213 Summerfield HALL, University of Kansas and William Davidson Institute (WDI), Lawrence, KS 66045, USA. E-mail: Earnhart@ku.edu

(2) Holds Citicorp Professorship at CERGE-EI, a joint workplace of Charles University and the Academy of Sciences of the Czech Republic
Table 1: Descriptive statistics

Year Frequency Per cent

(a) Year distribution of data

1993 356 13.52
1994 469 17.81

1995 468 17.77
1996 484 18.38
1997 457 17.36
1998 398 15.14
Total 2,632 100.00

(b) Statistical summary of production value and emissions

Variable Mean s.d.

Production value (000s Czech Crowns) (a) 1,618,320 4,618,679
Emissions--all pollutants (t) 962 4,060
Emissions of carbon monoxide (CO) (t) 127 1,189
Emissions of sulphur dioxide [S[O.sub.2] (t) 515 2,440
Emissions of particulate matter (PM) (t) 121 621
Emissions of nitrous oxides ([N[O.sub.x]) (t) 200 849
N = 2,632

(a) Production value is adjusted to 1998 real
Czech Crowns using the Czech consumer price index.

Table 2: Sector-specific statistics for emissions and production value

Industry Emissions

 Mean (t) Sum (t)

Agriculture, hunting, forestry, 16.1 322
fisheries

Mining and quarrying 3,621.6 119,513

Manufacturing of food products, 150.2 59,635
beverages, and tobacco

Manufacturing of textiles, textile 265.5 57,338
products, leather, and leather products

Manufacturing of wood, wood products, 1,116.7 99,388
pulp, paper, PAPER products, and
publishing and printing

Manufacturing of coke and refined 1,107.6 15,506
petroleum

Manufacturing of chemicals, chemical 2,732.2 344,256
products, and synthetic fibres

Manufacturing of rubber and plastic 92.9 4,922
products

Manufacturing of other non-metallic 542.3 128,526
mineral products

Manufacturing of basic metals and 1,702.5 524,359
fabricated metal products

Manufacturing of machinery and 165.6 49,846
equipment n.e.c.

Manufacturing of Electrical and 83.5 9,766
Optical Equipment

Manufacturing of transport equipment 151.5 29,397

Manufacturing n. e. c. 144.8 13,320

Electricity, gas, and water supply 6,677.0 1,068,312

Construction 42.0 5,001

Wholesale and retail trade, motor 17.8 888
vehicle repair, hotels and restaurants,
transport, postal service, storage, and
telecommunications (a)

Finance, real estate, rentals, 14.4 1,048
business, research, public
administration

Education, health, and veterinary 27.1 895
services; other public and social
services

Industry Production

 Mean Sum
 (million CZK) (million CZK)

Agriculture, hunting, forestry, 133.9 2,678.04
fisheries

Mining and quarrying 5,631.5 185,838.82

Manufacturing of food products, 1,045.2 414,946.51
beverages, and tobacco

Manufacturing of textiles, textile 701.1 151,438.33
products, leather, and leather products

Manufacturing of wood, wood products, 1,539.3 136,998.38
pulp, paper, PAPER products, and
publishing and printing

Manufacturing of coke and refined 10,775.5 150,856.81
petroleum

Manufacturing of chemicals, chemical 3,313.7 417,522.35
products, and synthetic fibres

Manufacturing of rubber and plastic 868.6 4,634.83
products

Manufacturing of other non-metallic 1,095.8 259,694.91
mineral products

Manufacturing of basic metals and 2,814.8 866,944.65
fabricated metal products

Manufacturing of machinery and 905.7 272,628.82
equipment n.e.c.

Manufacturing of Electrical and 615.1 71,970.48
Optical Equipment

Manufacturing of transport equipment 2,740.6 531,684.35

Manufacturing n. e. c. 529.0 48,668.08

Electricity, gas, and water supply 2,534.1 405,462.94

Construction 1,850.3 220,185.21

Wholesale and retail trade, motor 654.2 32,712.22
vehicle repair, hotels and restaurants,
transport, postal service, storage, and
telecommunications (a)

Finance, real estate, rentals, 510.3 37,251.69
business, research, public
administration

Education, health, and veterinary 178.8 5,901.66
services; other public and social
services

Industry Emissions/
 production
 ratio

Agriculture, hunting, forestry, 0.1202
fisheries

Mining and quarrying 0.6431

Manufacturing of food products, 0.1437
beverages, and tobacco

Manufacturing of textiles, textile 0.3786
products, leather, and leather products

Manufacturing of wood, wood products, 0.7255
pulp, paper, PAPER products, and
publishing and printing

Manufacturing of coke and refined 0.1028
petroleum

Manufacturing of chemicals, chemical 0.8245
products, and synthetic fibres

Manufacturing of rubber and plastic 0.1069
products

Manufacturing of other non-metallic 0.4949
mineral products

Manufacturing of basic metals and 0.6048
fabricated metal products

Manufacturing of machinery and 0.1828
equipment n.e.c.

Manufacturing of Electrical and 0.1357
Optical Equipment

Manufacturing of transport equipment 0.0553

Manufacturing n. e. c. 0.2737

Electricity, gas, and water supply 2.6348

Construction 0.0227

Wholesale and retail trade, motor 0.0272
vehicle repair, hotels and restaurants,
transport, postal service, storage, and
telecommunications (a)

Finance, real estate, rentals, 0.0281
business, research, public
administration

Education, health, and veterinary 0.1517
services; other public and social
services

(a) These disparate sectors are combined because individually
they represent too small a portion of the sample to facilitate
estimation. This sectoral category also includes 17 observations
(0.65% of sample) from the sector of 'Other n.e.c'

Table 3: Between-group estimation of air
emissions summed across individual pollutants

 Include cubic
RHS variable production term

Production (a) 0.8658 *** (0.1018)

Production-squared (a) -8.814E-6 * (5.176E-6)

Production-cubed (a) 6.343E-11 (5.865E-11)

Agriculture, hunting, forestry, fisheries 373.8 (952.9)

Mining and quarrying -83.2 (968.6)

Manufacturing: food products, beverages, -187.8 (317.8)
and tobacco

Manufacturing: textiles and leather 135.8 (417.3)

Manufacturing: wood, pulp, paper, 156.9 (576.5)
publishing, printing

Manufacturing: coke and refined petroleum -5,109.6 *** (1,591.2)

Manufacturing: chemicals and synthetic 1,131.8** (525.5)
fibres

Manufacturing: rubber and plastic products -168.3 (814.6)

Manufacturing: other non-metallic mineral 120.8 (396.3)
products

Manufacturing: metals and fabricated metal 332.0 (344.1)
products

Manufacturing of machinery and equipment -122.6 (363.4)
n.e.c.

Manufacturing of electrical and optical 43.7 (519.9)
equipment

Manufacturing of transport equipment -240.2 (431.8)

Manufacturing n.e.c. 160.6 (616.1)

Electricity, gas, and water supply 4,507.2 *** (464.2)

Construction -854.5 (536.5)

Trade, motor vehicle repair, hotels and -473.7 (780.9)
restaurants, transport, postal service,
storage, and telecoms

Finance, real estate, rentals, business, -45.4 (577.3)
research, public administration

Education, health, and veterinary services, 322.7 (855.4)
other public and social services

Adjusted [R.sup.2] 0.386

No. of observation 631

 Exclude cubic
RHS variable production term

Production (a) 0.9510 *** (0.0646)

Production-squared (a) -1.422E-5 *** (0.135E-5)

Production-cubed (a) NA

Agriculture, hunting, forestry, fisheries 417.1 (952.1)

Mining and quarrying -209.3 (961.7)

Manufacturing: food products, beverages, -197.8 (317.7)
and tobacco

Manufacturing: textiles and leather 136.6 (417.4)

Manufacturing: wood, pulp, paper, 115.4 (573.3)
publishing, printing

Manufacturing: coke and refined petroleum -4,856.0 *** (1,574.0)

Manufacturing: chemicals and synthetic 1,067.6 ** (522.3)
fibres

Manufacturing: rubber and plastic products -180.7 (814.7)

Manufacturing: other non-metallic mineral 103.2 (396.0)
products

Manufacturing: metals and fabricated metal 347.8 (343.9)
products

Manufacturing of machinery and equipment -132.2 (363.3)
n.e.c.

Manufacturing of electrical and optical 51.7 (520.0)
equipment

Manufacturing of transport equipment -290.5 (429.4)

Manufacturing n.e.c. 175.3 (616.0)

Electricity, gas, and water supply 4,506.6 *** (464.2)

Construction -901.9 * (534.8)

Trade, motor vehicle repair, hotels and -483.8 (781.0)
restaurants, transport, postal service,
storage, and telecoms

Finance, real estate, rentals, business, -25.3 (577.1)
research, public administration

Education, health, and veterinary services, 356.1 (854.9)
other public and social services

Adjusted [R.sup.2] 0.386

No. of observation 631

s.e. are noted inside parentheses; P-values are noted inside
parentheses. *, **, and *** indicate statistical significance
at the 10, 5, and 1% levels, respectively. Each regression
also includes an intercept term. Sum of coefficients from
individual sectoral indicators is restricted to zero.

(a) Units for production are millions of Czech crowns
([10.sup.6] CZK); units for production-squared are trillions
of Czech crowns (1012 CZK); units for production-cubed are
quintillions of Czech crowns ([10.sup.18] CZK). 30 CZK = 1
euro.

Table 4: Between-group estimation of individual air pollutant emissions

 Sulphur dioxide
RHS variable Carbon monoxide (CO) (S[O.sub.2)

Production (a) 0.2170 *** (0.0159) 0.4171 *** (0.0426)

Production- -2.900E-6 *** (0.333E-6) -6.670E-6 *** (0.891E-6)
squared (a)

Agriculture, 241.2 (235.2) 42.4 (628.7)
hunting, forestry,
fisheries

Mining and 522.1 ** (237.6) 127.2 (635.1)
quarrying

Manufacturing: 72.9 (78.5) 205.2 (209.8)
food products,
beverages, and
tobacco

Manufacturing: 141.1 (103.1) 42.1 (275.6)
textiles and
leather

Manufacturing: 66.7 (142.1) 6.1 (379.9)
wood, pulp,
paper, publishing,
printing

Manufacturing: 1,334.1 *** (388.8) 1,741.4 * (1,039.4)
coke and refined
petroleum

Manufacturing: 207.3 * (129.0) 865.2 *** (344.9)
chemicals and
synthetic fibres

Manufacturing: 95.5 (201.3) 219.4 (538.0)
rubber and
plastic products

Manufacturing: 130.0 (97.8) 189.8 (261.5)
other non metallic
mineral products

Manufacturing: 396.2 *** (85.0) 136.9 (227.1)
metals and
fabricated
metal products

Manufacturing of 98.2 (89.8) 188.7 (239.9)
machinery and
equipment n.e.c.

Manufacturing of 146.6 (128.5) 111.1 (343.3)
electrical and
optical equipment

Manufacturing 28.9 (106.1) 233.4 (283.5)
of transport
equipment

Manufacturing 175.7 (152.2) 53.0 (406.8)
n.e.c.

Electricity, gas, 130.2 (114.7) 3,120.3 (306.6) ***
and water supply

Construction 58.4 (132.1) 537.0 (353.2)

Trade, motor 27.8 (192.9) 346.0 (515.7)
vehicle repair,
hotels and
restaurants;
Transport, postal
service, storage,
and telecoms

Finance, real 129.8 (142.6) 145.5 (381.1)
estate, rentals,
business,
research, public
administration

Education, health, 241.2 (211.2) 11.6 (564.5)
and veterinary
services, other
public and social
services

Adjusted [R.sup.2] 0.298 0.276

No. of 631 631
observations

RHS variable Particulate matter (PM) Nitrous oxides (NO)

Production (a) 0.1326 *** (0.0091) 0.1843 *** (0.0130)

Production- -1.930E 6 *** (0.190E-6) -2.710E-6 *** (0.273E-6)
squared (a)

Agriculture, 86.2 (134.3) 47.3 (192.4)
hunting, forestry,
fisheries

Mining and 196.9 (135.6) 382.5 ** (194.4)
quarrying

Manufacturing: 7.1 (44.8) 72.6 (64.2)
food products,
beverages, and
tobacco

Manufacturing: 54.4 (58.9) 16.8 (84.4)
textiles and
leather

Manufacturing: 14.0 (81.1) 28.5 (116.3)
wood, pulp,
paper, publishing,
printing

Manufacturing: 825.4 *** (222.0) 955.1 *** (318.2)
coke and refined
petroleum

Manufacturing: 223.0 *** (73.6) 186.7 * (105.6)
chemicals and
synthetic fibres

Manufacturing: 6.4 (114.9) 63.1 (164.7)
rubber and
plastic products

Manufacturing: 44.9 (55.8) 118.1 (80.1)
other non metallic
mineral products

Manufacturing: 81.3 * (48.5) 7.2 (69.5)
metals and
fabricated
metal products

Manufacturing of 22.0 (51.2) 63.6 (73.4)
machinery and
equipment n.e.c.

Manufacturing of 37.9 (73.3) 21.7 (105.1)
electrical and
optical equipment

Manufacturing 0.7 (60.5) 86.6 (86.8)
of transport
equipment

Manufacturing 60.0 (86.9) 7.4 (124.5)
n.e.c.

Electricity, gas, 434.2 (65.5) *** 821.8 (93.8) ***
and water supply

Construction 99.5 (75.4) 207.0 (108.1) **

Trade, motor 42.0 (110.1) 123.5 (157.9)
vehicle repair,
hotels and
restaurants;
Transport, postal
service, storage,
and telecoms

Finance, real 23.4 (81.4) 33.1 (116.7)
estate, rentals,
business,
research, public
administration

Education, health, 68.2 (120.6) 58.2 (172.8)
and veterinary
services, other
public and social
services

Adjusted [R.sup.2] 0.348 0.376

No. of 631 631
observations

s.e. are noted inside parentheses; P-values are noted inside
parentheses. *, **, and *** indicate statistical significance
at the 10, 5, and 11 levels, respectively. Each regression
also includes an intercept term. Sum of coefficients from
individual sectoral indicators is restricted to zero.

(a) Units for production are millions of Czech crowns
([10.sup.6] CZK), units for production-squared are trillions
of Czech crowns ([10.sup.12] CZK); units for production-cubed
are quintillions of Czech crowns ([10.sup.18] CZK).
30 CZK = 1 euro.
COPYRIGHT 2006 Association for Comparative Economic Studies
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2006 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Earnhart, Dietrich; Lizal, Lubomir
Publication:Comparative Economic Studies
Geographic Code:4EXCZ
Date:Dec 1, 2006
Words:8623
Previous Article:The adoption of environmental management practices in a transition economy.
Next Article:Voluntary environmental agreements when regulatory capacity is weak (1).
Topics:

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters |