主题：【文摘】PRICING INDUSTRIAL POLLUTION IN CHINA -- 梦晓半生

Policy Research Working Paper #1644

The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be used and cited accordingly. The findings, interpretations, and conclusions are the authors' own and should not be attributed to the World Bank, its Executive Board of Directors, or any of its member countries.

PRICING INDUSTRIAL POLLUTION IN CHINA:

AN ECONOMETRIC ANALYSIS OF THE LEVY SYSTEM

Hua Wang*

David Wheeler

PRDEI

September, 1996

* The authors are respectively Consultant and Principal Economist in the Environment, Infrastructure and Agriculture Division, Policy Research Department, World Bank. Our thanks to Susmita Dasgupta, Mainul Huq, C.H. Zhang, Nick Anderson, Lee Travers, Richard Newfarmer and Songsu Choi for many useful comments and suggestions.

Table of Contents

EXECUTIVE SUMMARY

1. Introduction

2. The Levy System in Theory and Practice

2.1 National Rules

2.2 Provincial Realities

3. A Model of Equilibrium Pollution

3.1 Environmental Demand

3.2 Environmental Supply

4. Data Sources and Estimating Equations

4.1 Data Sources

4.2 Specification: Environmental Demand

4.3 Specification: Environmental Supply

5. Econometric Results

5.1 Environmental Demand Equation

5.2 The Problem of Unobserved Regulatory Activity

5.3 Environmental Supply Equation

5.4 Summary of Results

6. Illustrative Shifts in Pollution Equilibria

7. Summary and Implications

8. References

--------------------------------------------------------------------------------

EXECUTIVE SUMMARY

This paper analyzes China's experience with the pollution levy, an emissions charge system which covers hundreds of thousands of factories. Operation of the system has been well-documented since the mid-1980's, affording a unique opportunity to assess the implementation and impact of a pollution charge in a developing country.

The levy experience has not been studied systematically, but anecdotal critiques have created the impression that the system is arbitrarily administered and ineffective as a pollution control instrument. Strictness of enforcement is thought to vary widely, so factories in different regions face very different penalties for polluting. In addition, it is commonly believed that the levy provides little incentive to control pollution because official rates are below marginal abatement costs. Critics have therefore tended to view the levy as a local financing mechanism which is ineffective as a regulatory instrument.

In this paper, we test the conventional critique of the levy system using a new province-level panel database for the period 1987 - 1993. We analyze the water pollution levy because its implementation and impact are well-documented in the information available to us. We have no basis for judging whether our results are also valid for air pollution, solid waste, or emissions from facilities which do not report to provincial and national regulators (particularly township and village enterprises). However, water pollution control has played a major role in environmental regulation, and our database incorporates information drawn from many thousands of factories over a seven-year period which witnessed great changes in China's economy. We therefore believe that our results provide some valuable new insights.

Our econometric analysis focuses on explaining patterns of covariation in two province-level measures: industrial emissions intensity (provincial emissions/output) for organic water pollution, and the effective water pollution levy rate (provincial levy collections per unit of above-standard wastewater discharge). To analyze these patterns of covariation, we develop a formal model of 'equilibrium pollution': In each region and period, the effective levy rate (or 'price of pollution') and pollutant discharge are jointly determined by the intersection of environmental demand (ED) and supply (ES) functions. The ED function relates industrial pollution intensity to the local price of pollution. It reflects the economics of cost-minimizing abatement by industry, and is formally equivalent to the marginal abatement cost (MAC) function in textbooks of environmental economics. The ES function specifies the pollution price imposed by the community as damage rises. While it reflects considerations of marginal social damage (MSD), ES is not equivalent to the textbook MSD function because it reflects the interplay of limited information, perceived self-interest, and differential ability or willingness to enforce community standards. Equilibrium pollution (at the intersection of ED and ES) is therefore not necessarily optimal pollution (at the intersection of ED and MSD).

Contrary to the conventional wisdom, our results suggest that the water pollution levy system is neither arbitrary nor ineffective. Across provinces and over time, variations in the effective levy rate are well-explained by proxies for local valuation of environmental damage and community capacity to enforce local norms. During the period 1987-1993, rapid development in many provinces led to sharp increases in the effective rate. Our results also suggest that the emissions intensity of Chinese industry was highly responsive to these increases, because marginal abatement costs were actually lower than levy rates in many cases. From 1987 to 1993, provincial pollution intensities fell at a median rate of 50% and total discharges declined at a median rate of 22%.

We conclude that China's levy system has been working much better than has been supposed, and that provincial variations in enforcement of the levy reflect significant elements of self-interest. However, it is important to qualify our results. First, they suggest that equilibrium pollution in China is not optimum pollution. The implied valuation of pollution damage seems inappropriately low, and variations in enforcement reflect local education and bargaining power. Secondly, we have no doubt that the current levy system could be significantly improved. As we explain in the paper, the levy is closer to a water discharge fee than a true Pigovian tax. If some form of the latter were instituted, pollution reduction would probably be greater because all units of pollution (not just 'excess' units) would be subject to a charge.

Our results also provide a number of suggestive lessons for regulators in developing countries:

Local enforcement of national standards will determine the effective 'price of pollution' in each area. Our China results suggest that local assessment of self-interest plays an important role in this process. Recommendations for regulatory reform should recognize such regional heterogeneity as natural and legitimate.

The locally-enforced price of pollution rises steadily with industrial development, because community damage valuation and enforcement capacity both increase. Important factors include changes in total pollution load, population density, income per capita, and education.

Early in the regulatory process, industrial emissions intensity is highly responsive to changes in the price of pollution. This is principally because marginal costs are often quite low in low-medium abatement ranges. In the Chinese case, provincial adjustments of effective levy rates and other regulatory instruments have been sufficient to induce sharp declines in emissions intensity and reductions in total emissions from registered factories during a period of very rapid industrial growth.

1. Introduction

Article 18 of China's Environmental Protection Law specifies that "in cases where the discharge of pollutants exceeds the limit set by the state, a compensation fee shall be charged according to the quantities and concentration of the pollutants released." A few areas began experimental implementation of the compensation fee, or pollution levy, shortly after passage of this law in 1979. In 1982, China's State Council began nationwide implementation by issuing the "Provisional Regulations for Collection of Compensation Fees for Pollutant Discharge." Almost all of China's counties and cities have now implemented the levy system, and approximately 300,000 factories have been charged for their emissions. More than 19 billion RMB yuan in levies have been collected. About 80% of the funds have been used to finance industrial pollution prevention and control, accounting for about 15% of total investment in these activities (NEPA, 1994).

China's pollution levy is one of the few economic instruments with a long, documented history of application in a developing country. In sheer magnitude, the current Chinese system may be without peer in the world. However, it has been criticized on two principal grounds. First, critics have claimed that there are marked differences in the degree of enforcement across regions. Their observations have created the impression that the current administration of the system, while improving, remains relatively arbitrary (Qu, 1991). Secondly, the incentive properties of the system have been called into question. Case studies have been used to support the claim that levy rates are generally below the marginal cost of abatement needed for compliance with Chinese emissions standards. Drawing on this evidence, some critics have asserted that the levy rates are too low to have significant effects on industrial emissions (Qu, 1991; NEPA, 1992 and 1994; Shibli and Markandya, 1995).

To test these propositions more systematically, we have constructed a panel of annual environmental, regulatory and socioeconomic data from 29 Chinese provinces for the period 1987-1993. Focusing on organic water pollution, we specify and estimate a model of regulation and industry response which addresses two basic questions:

What explains the existing variations in effective levy rates and enforcement?

Is the existing levy system effective as a pollution control instrument?

In Section 2, we describe the existing levy system, effective levy rates across provinces, and variations in industry's environmental performance. We develop the model in Sections 3 and 4, and present our econometric results in Section 5. Section 6 uses simulation to explore the implications of the results, while Section 7 provides a summary and conclusions.

2. The Levy System in Theory and Practice

2.1 National Rules

The levy system formally requires that a fee be paid by any enterprise whose effluent discharge exceeds the legal standard. NEPA regulations specify variations in effluent standards by sector and fees by pollutant. With the approval of NEPA, local areas may raise both standards and fees above the nationally-mandated levels (in the latter case, these are called overstandard fees). Levies are charged only on the 'worst case' pollutant from each source.Footnote1 To illustrate, national regulations would stipulate the following levy for a factory j in sector k, emitting N pollutants:

(2.1)

where

(2.2)

Ljkl = Total levy for pollutant l

pkl = National levy rate for pollutant l in sector k

njkl = Discharge concentration of pollutant l from plant j

nkl* = Legal discharge concentration standard for pollutant l from sector k

Wj = Wastewater discharge volume from plant j

This formulation differs from a conventional pollution charge in two significant respects. First, the Chinese system penalizes only worst-case pollution from each source.Footnote2 Secondly, it uses a 'markup' based on percent deviations from discharge standards for effluent concentrations: [p[(n - n*)/n*]W]. This is quite different from a tax on specific pollutants: [pnW]. In fact, equation (2.3) shows that the Chinese levy can be viewed as a charge on 'excess' wastewater (We):

(2.3)

2.2 Provincial Realities

Although supervised by the central government, China's pollution levy system is implemented by the provincialFootnote3 and local governments. As Table 1 shows, the critics are correct when they assert that there is significant variation in implementation. The table provides estimates of effective levy rates (or levies actually collected per unit of above-standard wastewater discharge), denominated in 1990 yuan. The data reveal striking differences, both across provinces and through time. In the space of six years, the effective pollution levy rate more than doubled in some areas (Tianjin, Fujian) and fell significantly in others (Shanghai, Jilin, Shanxi). In general, real effective provincial levies increased during the sample period. Cross-provincial variation in 1993 yielded ratios as high as 8:1 (Tianjin vs. Qinghai).

Inspection of these provincial differences suggests that variations in the effective levy rate are far from random. Figure 1 displays their geographic distribution. In 1993, many relatively affluent, heavily-industrialized coastal provinces had the highest effective levy rates (e.g., Tianjin (.24), Zhejiang (.17), Liaoning (.16), Shandong (.16), Shanghai (.15)), while many poorer interior provinces had levy rates at the bottom of the scale (e.g. Qinghai (.03), Ningxia (.03), Guizhou (.04), Gansu (.06)). It is also worth noting the experience of Guangdong, the site of China's fast-growing new economic zones. Since 1987, the ratio of rates in Guangdong and its neighboring province, Jiangxi, has jumped from 1:1 to 2.6:1.

The provincial data also reveal great variation in industry's environmental performance. For 1987 and 1993, Table 2 displays levels and changes of industrial emissions intensity (emissions/output) and total discharges of chemical oxygen demand (COD). Water pollution intensity varies greatly, with interprovincial ratios as high as 25:1. For most provinces, however, the data provide striking evidence of progress in pollution control. With two exceptions (Hainan, Hubei), pollution intensity fell rapidly during the sample period. In most provinces, these reductions were sufficient to outweigh rapid increases in output, yielding significant declines in total COD emissions.

To summarize, the stylized facts offer little support for the view that the levy system is arbitrary and ineffective. While effective levies do vary greatly in China, their geographic distribution is correlated with provincial rates of urbanization and industrialization. Furthermore, recent increases in effective levies have been accompanied by large reductions in water pollution intensities and loads.

3. A Model of Equilibrium Pollution

To analyze these patterns of covariation, we develop a formal model of 'equilibrium pollution'. In each region and period, the effective levy rate and pollutant discharge are jointly determined by the intersection of environmental demand (ED) and supply (ES) functions. The ED function relates industrial pollution intensity to the local price of pollution. It reflects the economics of cost-minimizing abatement by industry, and is formally equivalent to the marginal abatement cost (MAC) function in textbooks of environmental economics. The ES function specifies the pollution price imposed by the community as damage rises. While it reflects considerations of marginal social damage (MSD), ES is not equivalent to the textbook MSD function because it reflects the interplay of limited information, perceived self-interest, and differential ability or willingness to enforce community standards. Equilibrium pollution (at the intersection of ED and ES) is therefore not necessarily optimal pollution (at the intersection of ED and MSD).

In modeling environmental supply, we focus particularly on the role of endogenous enforcement. Across provinces, regulators may differ greatly in their ability or willingness to enforce the formal regulations. In practice, regulation is almost never a 'pure' administrative process in which violations are unambiguously observed and rules are uniformly enforced. Because industry is a prime generator of income and employment, regulatory enforcement is often subjected to political pressure for leniency.Footnote4 Even when they are unencumbered, regulators are generally reluctant to impose penalties which will bankrupt or shut down factories. They may, however, be tougher in relatively affluent regions where communities put more stress on environmental quality.

In most countries, the 'price of pollution' is not directly observable because regulation is based on enforcement of quantitative emissions standards.Footnote5 China provides a unique opportunity to test a structural model because the price of pollution -- the effective levy rate -- can be measured across provinces and over time.

3.1 Environmental Demand

For above-standard emissions, plants subject to a pollution levy will pay a price for polluting. Of course, abatement also has costs and cost-sensitive plants will seek to reduce pollution to the point where the expected levy is equal to the marginal cost of abatement. In a particular Chinese province, plants will form expectations about enforcement of the pollution levy by observing the average experience of industrial facilities in their region. From data published annually by NEPA, we have estimated the expected provincial levy rate, pe, as total above-standard discharge levy collections divided by total above-standard wastewater discharge. In the case of a single pollutant, the total expected levy for the jth plant in a particular province is therefore:

(3.1)

where Lj = Expected total levy payment

pe = Expected levy rate

uj = Effluent concentration (e.g., COD emissions/wastewater volume)

us = Concentration standard

Wj = Waste water

Recent econometric work on factory-level abatement costs in China (Dasgupta, Huq and Wheeler, 1996) suggests the following model for the case of a single pollutant such as COD:

(3.2)

where

and Aj = Total abatement cost

u0j = Influent concentration

uj = Effluent concentration

At the plant level, y1 is significantly less than unity (i.e., abatement is subject to scale economies). While total abatement cost rises less-than-proportionately with scale of waste water treatment, marginal abatement cost increases with percent reduction in pollutant concentration from influent to effluent.

Total pollution-related cost is therefore given by:

(3.3)

Formally, cost minimization implies choosing an effluent intensity u such that

(3.4)

Thus, for a plant j which adjusts so as to minimize pollution-related costs, optimal effluent intensity for a given region (r) is given by the solution to (3.4):

(3.5)

Given the levy and abatement cost equations (3.1, 3.2) and the assumption of cost minimization, equation (3.5) describes optimal effluent intensity under differing conditions. This can be converted to a pollution intensity equation with the introduction of the following definition:

(3.6)

where

wj = Wastewater intensity of output

Qj = Output

Pj = Pollution volume

nj = Pollution intensity of output

Substituting (3.6) into (3.5), we obtain:

(3.7)

Equation (3.7) is the environmental demand function (ED) for a given sector and region. Pollution intensity decreases as output scale Q increases (0<y1<1). Since h2 >0, pollution intensity increases less-than-proportionately with wastewater intensity (w), influent intensity (u0) and the legal concentration standard (us); and decreases as the expected pollution levy increases. The levy elasticity of pollution intensity is inversely related to y2, the cost elasticity of pollution reduction (i.e., responsiveness to the levy drops as the cost elasticity of abatement rises).

Of course (and particularly in China, where state enterprises dominate), the assumption of cost minimization remains just that -- an assumption. However, the log-log ED function remains a useful approximation to the underlying relationship even if the cost-minimization criterion is not strictly met. The econometric results provide an empirical test of price responsiveness.

3.2 Environmental Supply

In China, provincial and local regulators do not have much reliable information about actual pollutant concentrations, dose-response functions, or opportunity costs for damage valuation. However, they do have information on local emissions, population and income which reflect the underlying relationships. If regulators respond to these considerations, their enforcement practices (and the expected levy) should reflect the calculus in equations (3.8 - 3.9) below.

3.2.1 Damage Function

Equations (3.8) relate population and pollution in a given province to total pollution-related health damage. Total damage (D) is equal to population (N) times expected damage per individual (p); the latter increases at the margin with ambient exposure (c) to pollution (). Exposure is in turn a function of total emissions (P), normalized by provincial area (T).

Apart from transient disturbances, exposure should increase proportionately with pollutant discharge into a fixed volume of air or water (). Thus, total damage should increase more-than-proportionately with pollution per unit area (). Finally, the evaluation of total damage (V) in equation (3.9) should be positively related to provincial income per capita ():

(3.8)

where:

pr = Expected health damage to individuals in region r

Dr = Total health damage in region r

Pr = Total emissions in region r

Tr = Area of region r

Nr = Population of region r

(3.9)

where yr = Income per capita

3.2.2 Marginal Social Damage Function

Given the total damage function (3.9), the provincial price of pollution appropriately reflects marginal social damage in the following equation.

(3.10)

If provincial regulators are well-informed, sensitive to marginal social damage, and if no other considerations apply, equation (3.10) describes the pollution price schedule which they will choose to enforce. Provincial differences in pollution, population and income will be reflected in effective levies.

3.2.3 Environmental Supply Function

As we have noted previously, however, other considerations also apply: 'Regulation is negotiation,' and enforcement practice may also reflect differential information and bargaining power. For example, local perceptions of pollution problems may well be affected by average education levels. Education may also affect communities' ability to organize and bring pressure on local regulators to enforce desired levels of compliance. Local willingness to enforce more strictly may also be affected by perceptions of the potential economic consequences, particularly for communities with low levels of industrialization.Footnote6 More heavily-industrialized regions, which already enjoy a strong advantage from agglomeration economies, may have much less concern about the impact of stricter regulation on industrial activity.Footnote7

To incorporate the potential effects of provincial education and industrialization, we specify the environmental supply function as follows:

(3.11)

where er = Secondary school completion rate

ir = Industry share of provincial output

4. Data Sources and Estimating Equations

4.1 Data Sources

For the empirical analysis in this paper, we have constructed a province-level panel database from official yearbooks available in China: The China Environment Yearbook (1987-1993); China Statistical Yearbook (1987-1993), and China's Industrial Economy Yearbook (1987-1993). Specific sources of data and variable definitions are reported in Tables 5-6. Problems with missing data have limited the estimation panel to five years: 1987, 1988, 1989, 1992 and 1993.

4.2 Specification: Environmental Demand

Several adjustments to equation (3.7) have to be made for estimation using provincial data. At the province level, we use output shares by sector to control for the combined effect of wastewater intensities, influent intensities and regulatory standards. The estimated effect of provincial differences in sector shares will depend on the degree to which sectoral regulatory standards adjust for differences in sectoral wastewater intensities, influent intensities and abatement costs. If there is full adjustment (i.e., proportionately lower standards for wastewater- and influent-intensive sectors with high abatement costs; higher standards for the opposite case), sectoral shares would have no independent effect.

Secondly, while abatement scale economies are clearly present at the plant level, they do not affect province-level aggregates. Therefore, we cannot use provincial industrial output to capture scale effects. However, the available data do distinguish between production from small, medium and large industrial facilities. We therefore control for this factor by introducing the share of output produced by large plants.Footnote8

We also control for state ownership, since several considerations suggest that state-owned factories may be more pollution intensive than other facilities. First, state-owned factories may simply generate more waste residuals per unit of output because they are less efficient. Secondly, soft budget constraints may make them less sensitive to pollution levies. Finally, evidence from other Asian countries suggests that state-owned factories resist regulation more successfully than privately-owned plants (Pargal and Wheeler, 1996; Hartman, Huq and Wheeler, 1996). We control for this factor by introducing the provincial share of production in state-owned factories.

After these adjustments, the estimation equation for the cross-provincial data set is as follows (for the rth province):

(4.1)

Prior expectations:

where

CODI = COD intensity (COD discharge / Industrial output)

skr = The industrial output share of the kth sector

LARGEr = The industrial output share of large plants

STATEr = The industrial output share of state-owned plants

PLWr = Effective pollution levy per unit of excess wastewater discharge

= A stochastic error term incorporating provincial components

4.3 Specification: Environmental Supply

On the supply side, we have measures for the relevant variables at the provincial level. The appropriate estimating equation is therefore specified as follows:

(4.2)

Prior expectations: w1 > 1 (w1 - 1 > 0), w2 > 0, w3 > 0, w4 > 0

where

PLWr = Effective pollution levy per unit of excess wastewater discharge

CODDr = COD emissions per unit of provincial land area (or pollution density)

POPDr = Provincial population density

INCr = Provincial income per capita

EDUr = Provincial secondary school completion rate

INDr = Industry share of provincial output

In equations (4.1 - 4.2), PLW and CODI are jointly determined. In addition, the error terms in both equations are likely to incorporate provincial fixed effects. We have therefore estimated the two equations using two-stage least squares and the appropriate error components model. The use of fixed-effects also permits us to control for institutional and historical factors which may affect industrial pollution intensities and regulatory enforcement practices across China's provinces.

5. Econometric Results

We have fitted the ED and ES equations to provincial data for 1987-89 and 1992-93. Fixed-effects estimation has been used to test the basic model; a cross-section equation has also been fitted to provincial average data on the supply side, since education data are only available for one year. The results are presented in Tables 7.1 (ED) and 7.2 (ES).

5.1 Environmental Demand Equation

In 7.1.1 and 7.1.2, we report estimation results for regression equation (4.1) with and without statistically insignificant variables. The overall fits are quite good (Adj. R2's of .80 and .77, respectively). While provincial fixed effects are clearly important, we cannot reject the joint null hypothesis that all sector shares coefficients are zero. Thus, the results suggest that sectoral composition has no significant effect on equilibrium pollution intensity at the provincial level. A plausible inference is that regulatory standards are approximately adjusted for intersectoral differences in intensities and abatement costs.

Our results indicate that Chinese industry has been highly responsive to the pollution levy. After insignificant variables have been removed from the regression, the estimated elasticity for PLW is almost exactly minus one and highly significant (t = 5.1). Since there have been large variations in the levy during the sample period, our results imply that the levy has played a major role in reducing industrial pollution intensity. Thus, our results strongly contradict the conventional critique. They are supported by a recent econometric study of Chinese water pollution abatement costs (Dasgupta, Huq and Wheeler, 1996), which suggests that marginal costs are lower than the levy rate over a broad range of abatement intensities

Although there were good a priori reasons to assume that STATE would have a significant, positive impact on pollution intensity, this hypothesis is not supported by the results. The estimated parameter has the correct sign, but it does not pass classical significance tests. As Table 3 shows, there is great variation in state ownership across provinces (from 31% (Zhejiang) to 89% (Qinghai) in 1993) and over time. (changes in the range [-20% , +15%]). The insignificance of STATE in the face of this variation suggests that the determinants of state enterprise environmental performance under Chinese socialism differ significantly from their counterparts in mixed Asian economies.

By contrast, we obtain very strong results for the share of large factories in industrial output (LARGE). The estimated parameter is large in absolute value, has the expected sign, and is highly significant (t = 6.2). Table 4 shows that LARGE also varies greatly across provinces and through time. Large-plant shares in 1993 ranged from 15% (Zhejiang) to 62% (Heilongjiang). During the sample period, rapid liberalization of the Chinese economy was accompanied by a sharp rise in the output share of large factories. Across provinces, increases in the range [10-15%] were not unusual. The impact of abatement scale economies in large plants is clearly seen in the large, negative estimated impact on pollution intensity. Whatever its socioeconomic impact, increasing scale in Chinese factories has apparently been good for the environment.

5.2 The Problem of Unobserved Regulatory Activity

While our results for the pollution levy are quite strong, we recognize the possibility that the effective levy is also serving as a proxy for enforcement of quantity-based standards. If provinces with higher levy enforcement rates are also enforcing command-and-control regulation more effectively, the estimated levy elasticity is likely to be biased upward in our regression results. We have no independent measure of command-and-control enforcement. However, an additional, unobserved component of regulation should also be affected by the exogenous variables in the environmental supply regression: INC and IND (see the supply equation estimation results below). Inclusion of these variables in the environmental demand function should therefore provide a first-order test of the robustness of the estimated levy elasticity. Columns 7.1.3 - 7.1.5 report results for INC, IND and their first principal component (PRIN1). As previously noted, EDU is available for only one year and cannot be employed for fixed-effects estimation. In no case is any of the new variables even close to significant, and the estimated elasticity of the pollution levy is hardly affected. This reinforces our conclusion that much of the change in pollution intensity during the period 1987-1993 is attributable to the pollution levy.Footnote9

5.3 Environmental Supply Equation

In Table 7.2 (7.2.1 - 7.2.3), we present fixed-effects estimates for the environmental supply function. To correct for simultaneity bias, we instrument provincial COD discharge before constructing our measure of provincial pollution density (CODD). Equations 7.2.1 - 7.2.2 impose the theoretically-appropriate restriction (unity) on the population density parameter. The regression fits are quite good (Adjusted R2 = .96). The results suggest a pollution density parameter slightly above 1 (w1 - 1 = .13) and a highly significant income elasticity which is somewhat greater than .60. (t = 4.2).

In equation 7.2.2, we introduce the degree of industrialization (IND). The estimated parameter has the expected sign, but is not statistically significant at conventional confidence levels. In equation 7.2.3, we test the unitary restriction on the population density parameter. The results show that this restriction cannot be rejected at a high confidence level,Footnote10 and imposition of the restriction also addresses a clear problem of collinearity between population density and income.

We cannot estimate a fixed-effects equation which incorporates education, because our panel includes provincial data on secondary school completion rates for only one year. Equations 7.2.4 - 7.2.8 report estimation results for a cross-section of provincial averages which include the education measure. The estimated population density effect falls sharply in these results, while the pollution density and income elasticities are quite close to the fixed-effects estimates. With limited degrees of freedom in the cross section, collinearity of INC, IND and SECED is clearly a problem. In 7.2.7 - 7.2.9, we report the results when equality is imposed on the parameters of (INC, IND) and (INC, IND, SECED), respectively. The results suggest a significant impact for the pair (IND, SECED), but we cannot distinguish clearly between their effects.

5.4 Summary of Results

Our fixed-effects estimation of the environmental demand function has yielded clearly-defined results, which identify strong and plausible effects for the effective levy rate and the output share of large factories. However, neither state ownership nor sectoral mix have any significant effect on variations in provincial COD intensity. For environmental supply, the results are more mixed. Provincial characteristics are collectively important in determining effective levy rates, but collinearity problems prevent clear separation of estimated effects in many cases. Estimated income elasticities are in a range consistent with results from willingness-to-pay surveys in OECD countries,Footnote11 and we cannot reject the hypothesis that damage assessment rises in proportion with exposed population. However, the estimated effect of pollution density is much lower than we would have expected a priori.. Our proxies for differential information and bargaining power (education and industrialization) have estimated elasticities similar to those for income, but collinearity prevents a clear separation of effects.

6. Illustrative Shifts in Pollution Equilibria

To assess the relative impact of environmental demand and supply variables on effective levies and emissions, we have used the results in 7.1.2 and 7.2.2 to simulate changes in pollution equilibria for three pairs of provinces: [Beijing / Qinghai], [Liaoning / Sichuan], and [Guangdong / Gansu]. As a group, these provinces illustrate China's regional heterogeneity (Table 8). Beijing is a small, heavily-industrialized region with relatively high income, education, and pollution density (pollution discharge per unit area). Qinghai is at the opposite extreme -- a large interior province with very little industry, low income, low educational level and very low total COD discharge. Liaoning Province is in China's first industrial heartland (formerly Manchuria), with a total COD discharge approximately equal to that of the much larger Sichuan Province. The latter, in south central China, is considerably poorer, less literate, and less industrialized. Finally, rural, agrarian Gansu in China's northwest contrasts with Guangdong, the southeast coastal province near Hong Kong which is the site of China's fast-growing new economic zone.

Figures 2.1 - 2.3 display simulated supply-demand intersections in 1987 and 1993, while Table 9 presents simulated and actual data for comparison. The Y-axis in each figure is scaled for the effective pollution levy in 1990 yuan/ton. The X-axis is scaled for COD discharge per unit of land area (for comparability across provinces with very different areas). The ED schedule is produced by multiplying the econometric results (for COD/output) by provincial output per unit area. The slope of ED is the estimated levy elasticity in equation 7.1.2. The position of the schedule at each point in time is determined by three factors: Total output/area, the share of large plants in provincial industrial output, and estimated provincial fixed-effects.

The slope of the ES schedule is the estimated value of (w1 - 1) in the pollution levy equation (7.2.2). Over time, it shifts position with changes in pollution density, population density, income, education and industrialization.

Three things are immediately striking about the results. First, the simulated equilibrium levy rates and pollution densities track the actual data quite well. Secondly, the quality of the fit is consistent across regions whose characteristics are completely different. For example, Beijing and Qinghai are at opposite extremes in both pollution density (5.5 tons/sq. km. in 1993 for Beijing, vs. .005 for Qinghai) and realized pollution levy rates (.15 yuan/ton in 1993 for Beijing vs. .03 for Qinghai). Nevertheless, the simulated changes in pollution equilibria are about equally accurate in both cases. Finally, the simulation results suggest approximate parity in demand- and supply-side impacts on levies and COD densities during the sample period.

7. Summary and Implications

Discussions of China's industrial pollution problem have commonly favored disaster scenarios, with widespread environmental destruction accompanying the relentless growth of highly-polluting industry. By implication, China's pollution control system has been unable to cope with the challenge of rapid industrialization. A corollary view has held that administration of a key regulatory instrument -- the pollution levy -- has been arbitrary and ineffective.

Our analysis suggests that these views have been mistaken, at least for water pollution control in the regulated industry sectors whose discharges are reported to NEPA. Our results do not provide a complete view, because the available data do not incorporate information on discharges from many township and village enterprises. Nevertheless, they suggest that China's environmental regulators compiled an impressive record during the period 1987-1993. Despite very rapid output growth, total organic water pollution from state-regulated industries actually fell. The accompanying decline in water pollution intensity (pollution per unit of output) was very steep. Our results suggest that much of the decline was attributable to increases in the effective pollution levy, with industry exhibiting an approximately unit-elastic response..Footnote12

We conclude that China's levy system has been working much better than has been supposed, and that provincial variations in enforcement of the levy reflect significant elements of self-interest. However, it is important to qualify our results. First, equilibrium pollution is not optimum pollution: The current implicit valuation of pollution damage may be inappropriately low, as suggested by the very small estimated impact of pollution density on the effective levy rate. Variations in enforcement also seem to reflect local education and bargaining power. Secondly, we have no doubt that the current levy system could be significantly improved. As we have noted, the levy is closer to a water discharge fee than a true Pigovian tax. If some form of the latter were instituted, pollution reduction would probably be greater because all units of pollution (not just 'excess' units) would be subject to a charge.

In closing, we should note that our results provide a number of suggestive lessons for regulators in developing societies:

Local enforcement of national standards will determine the effective 'price of pollution' in different regions. Our China results suggest that local assessment of self-interest plays an important role in this process. Recommendations for regulatory reform should recognize such regional heterogeneity as natural and legitimate.

The locally-enforced price of pollution rises steadily with industrial development, because community damage valuation and enforcement capacity both increase. Potentially important factors include changes in total pollution load, population density, income per capita, and education.

Early in the regulatory process, industrial emissions intensity is highly responsive to changes in the price of pollution. This is principally because marginal costs are often quite low in low-medium abatement ranges. In the Chinese case, provincial adjustments of effective levy rates and other regulatory instruments have been sufficient to induce sharp declines in emissions intensity and reductions in total emissions from registered factories during a period of very rapid industrial growth.

8. References

Dasgupta, S., M. Huq and D. Wheeler, 1996, "Water Pollution Abatement by Chinese Industry: Cost Estimates and Policy Implications," World Bank, Policy Research Department Working Paper (forthcoming)

Florig, H.K., W.O. Spofford Jr., X. Ma and Z. Ma, 1995, "China Strives to Make the Polluter Pay," Environmental Science and Technology, Vol. 29, No 6

Florig, K. and W. Spofford, 1994, "Economic Incentives in China's Environmental Policy," (Washington: Resources for the Future), October (mimeo)

Gray, W., and R. Shadbegian, 1993, "Environmental Regulation and Manufacturing Productivity at the Plant Level," Center for Economic Studies, U.S. Census Bureau, Discussion Paper No. CES 93-6

Hartman, R., M. Huq and D. Wheeler, 1996, "Why Paper Mills Clean Up: Survey Evidence from Four Asian Countries," World Bank, Policy Research Working Paper (forthcoming)

Hettige M., M. Huq, S. Pargal and D. Wheeler, 1996, "Determinants of Pollution Abatement in Developing Countries: Evidence from South and Southeast Asia," World Development (forthcoming)

Mody, A. and D. Wheeler, 1992, "International Investment Location Decisions: The Case of U.S. Firms," Journal of International Economics, 3

NEPA, 1992, Pollution Charges in China (Beijing: National Environmental Protection Agency)

NEPA, 1994, The Pollution Levy System, (Beijing: China Environmental Science Press)

Oates, W., K. Palmer and P. Portney, 1993, "Environmental Regulation and International Competitiveness: Thinking About the Porter Hypothesis," Resources for the Future (mimeo.)

Pargal, S. and D. Wheeler, 1996, "Informal Regulation of Industrial Pollution in Developing Countries: Evidence from Indonesia," Journal of Political Economy (forthcoming)

Qu, Geping, 1991, Environmental Management in China, (Beijing: UNEP and China Environmental Science Press)

Shibli, A. and A. Markandya, 1995, "Industrial Pollution Control Policies in Asia: How Successful are the Strategies?" Asian Journal of Environmental Management, Vol. 3, No. 2, November

Wheeler, D., 1991, "The Economics of Industrial Pollution Control: An International Perspective," World Bank, Industry and Energy Department Working Paper No. 60, January.

Yang, J. and J. Wang, 1995, "The Pollution Levy System in China," Chinese Research Academy of Environmental Sciences, Beijing, China (mimeo)

Table 1. Effective Levy Rate by Province

(Levy per Unit of Above-Standard Wastewater Discharge: 1990 Yuan per Ton)

Province 87 88 89 92 93

tianjin 0.09 0.11 0.14 0.23 0.24

zhejiang 0.09 0.09 0.09 0.15 0.17

liaoning 0.10 0.10 0.10 0.17 0.16

shandong 0.15 0.14 0.15 0.15 0.16

shanghai 0.26 0.23 0.20 0.15 0.15

guangdong 0.07 0.07 0.07 0.13 0.13

jiangsu 0.11 0.11 0.09 0.15 0.13

beijing 0.12 0.16 0.12 0.15 0.12

shaanxi 0.08 0.07 0.06 0.10 0.10

xinjiang 0.11 0.12 0.10 0.08 0.10

fujian 0.04 0.04 0.04 0.10 0.09

anhui 0.05 0.05 0.05 0.07 0.09

heilongjiang 0.05 0.05 0.04 0.08 0.09

hunan 0.05 0.05 0.05 0.08 0.09

hebei 0.07 0.07 0.08 0.08 0.08

hubei 0.05 0.04 0.06 0.10 0.07

henan 0.06 0.07 0.08 0.08 0.07

yunnan 0.04 0.06 0.06 0.07 0.07

jilin 0.09 0.07 0.05 0.13 0.07

guangxi 0.04 0.04 0.05 0.06 0.07

shanxi 0.09 0.06 0.06 0.08 0.07

hainan 0.07 0.03 0.04 0.07 0.06

Inner Mongolia 0.07 0.06 0.06 0.06 0.06

sichuan 0.05 0.05 0.04 0.04 0.06

gansu 0.05 0.03 0.04 0.05 0.06

jiangxi 0.07 0.06 0.05 0.05 0.05

guizhou 0.02 0.02 0.04 0.04 0.04

ningxia 0.04 0.05 0.05 0.07 0.03

qinghai 0.03 0.03 0.01 0.03 0.03

Table 2. Provincial Pollution Intensities and Pollution Loads, 1987-1993

COD Intensity

(Tons/10 million yuanFootnote13)

Total COD Discharge

('000 tons)

Province 1987 1993 % Chg. 1987 1993 % Chg.

qinghai 51.7 6.9 -86.7 11 3 -72.7

guangdong 102.4 19.4 -81.1 379 328 -13.5

fujian 126.4 33.9 -73.2 228 183 -19.7

zhejiang 106.7 38.0 -64.4 417 279 -33.1

anhui 162.4 64.7 -60.2 402 282 -29.9

shanxi 79.7 35.1 -56.0 177 124 -29.9

shandong 94.9 42.5 -55.2 667 582 -12.7

jiangsu 53.5 24.2 -54.8 571 402 -29.6

guizhou 67.2 30.6 -54.5 77 60 -22.1

shaanxi 32.6 15.0 -54.1 71 51 -28.2

sichuan 78.8 36.3 -53.9 432 338 -21.8

shanghai 25.9 12.0 -53.8 269 168 -37.5

jilin 117.8 54.8 -53.5 295 243 -17.6

ningxia 69.0 32.8 -52.5 17 20 17.6

yunnan 144.7 72.3 -50.0 170 200 17.6

hebei 68.7 34.6 -49.7 279 218 -21.9

Inner Mongolia 94.1 47.8 -49.2 115 99 -13.9

hunan 119.0 61.4 -48.4 382 303 -20.7

gansu 37.7 19.6 -48.0 51 37 -27.5

tianjin 37.6 20.0 -46.7 144 96 -33.3

liaoning 56.3 30.5 -45.8 462 354 -23.4

heilongjiang 64.9 36.3 -44.0 291 220 -24.4

henan 91.6 53.3 -41.8 335 351 4.8

xinjiang 79.4 49.7 -37.4 71 83 16.9

guangxi 266.8 173.7 -34.9 317 513 61.8

beijing 25.4 17.5 -31.2 106 89 -16.0

jiangxi 84.0 61.9 -26.3 150 181 20.7

hainan 102.4 107.3 4.8 103 63 -38.8

hubei 30.5 47.3 55.3 210 352 67.6

Total

7199 6222 -13.6

Table 3: Industrial Output Share of State-Owned Factories by Province, 1987-93

Province 1987 1990 1993 Ch. 87-90 Ch. 90-93 Ch. 87-93

zhejiang 42.8 31.2 31.2 -11.6 0.0 -11.6

jiangsu 46.2 34.3 31.9 -11.9 -2.4 -14.3

guangdong 58.0 40.2 33.6 -17.7 -6.7 -24.4

fujian 65.6 45.1 39.6 -20.5 -5.5 -25.9

shandong 63.1 41.4 49.7 -21.7 8.3 -13.4

shanghai 78.6 68.2 54.5 -10.3 -13.7 -24.1

tianjin 78.4 59.5 58.9 -18.9 -0.5 -19.5

sichuan 75.8 63.7 62.7 -12.1 -1.0 -13.2

hebei 71.5 49.4 63.5 -22.1 14.1 -8.1

anhui 72.3 58.2 63.5 -14.1 5.2 -8.9

beijing 76.7 63.2 65.1 -13.4 1.8 -11.6

hunan 74.5 64.0 65.8 -10.6 1.8 -8.7

hainan 58.0 75.7 66.6 17.7 -9.2 8.6

liaoning 73.1 61.2 68.9 -11.9 7.6 -4.3

henan 77.5 55.2 69.2 -22.3 14.0 -8.3

hubei 73.3 62.3 69.3 -11.1 7.1 -4.0

guangxi 81.5 72.2 69.7 -9.3 -2.5 -11.8

jiangxi 79.9 65.3 71.0 -14.6 5.7 -9.0

shanxi 77.7 59.7 72.7 -17.9 13.0 -5.0

jilin 78.6 70.4 75.4 -8.2 5.0 -3.2

shaanxi 83.3 68.7 77.1 -14.6 8.4 -6.2

Inner Mongolia 82.9 77.3 82.3 -5.6 5.0 -0.7

heilongjiang 84.2 80.5 82.7 -3.7 2.2 -1.5

yunnan 81.2 76.7 83.4 -4.5 6.7 2.2

ningxia 82.1 78.6 83.6 -3.5 5.0 1.5

gansu 89.3 78.1 83.9 -11.2 5.8 -5.4

guizhou 86.9 77.3 84.9 -9.7 7.7 -2.0

xinjiang 85.6 80.3 86.0 -5.3 5.7 0.4

qinghai 86.5 84.1 88.9 -2.4 4.8 2.4

Table 4: Industrial Output Share of Large Factories by Province, 1987-93

Province 1987 1989 1993 Ch. 87-89 Ch. 89-93 Ch. 87-93

heilongjiang 48.2 38.2 62.2 -10.0 24.0 14.0

gansu 46.4 37.3 58.7 -9.1 21.4 12.3

liaoning 46.5 37.3 57.5 -9.1 20.2 11.0

beijing 48.9 38.5 56.4 -10.4 17.9 7.6

shanghai 38.8 33.5 56.0 -5.3 22.5 17.2

qinghai 24.0 26.1 55.7 2.1 29.6 31.7

jilin 35.6 31.9 51.0 -3.7 19.1 15.3

yunnan 35.6 31.9 50.9 -3.7 19.0 15.3

guizhou 33.6 30.9 48.6 -2.7 17.7 15.0

shanxi 37.9 33.1 46.5 -4.9 13.5 8.6

shaanxi 40.2 34.2 46.3 -6.0 12.2 6.2

Inner Mongolia 29.3 28.8 44.9 -0.6 16.1 15.5

ningxia 31.7 29.9 44.4 -1.8 14.5 12.8

xinjiang 30.2 29.2 43.8 -1.0 14.6 13.6

hubei 32.5 30.4 42.7 -2.2 12.4 10.2

henan 31.1 29.6 41.8 -1.4 12.2 10.8

hebei 24.2 26.2 41.0 2.0 14.8 16.8

sichuan 29.1 28.6 38.4 -0.4 9.8 9.4

hunan 27.4 27.8 36.8 0.4 9.0 9.4

anhui 22.9 25.5 34.3 2.6 8.8 11.4

shandong 26.5 27.4 33.1 0.8 5.8 6.6

tianjin 35.3 31.7 32.6 -3.5 0.9 -2.6

jiangxi 16.3 22.2 28.3 5.9 6.1 12.0

hainan 21.6 24.9 28.2 3.3 3.3 6.5

guangdong 21.6 24.9 27.9 3.3 3.0 6.2

guangxi 18.3 23.2 27.5 5.0 4.3 9.2

jiangsu 15.9 22.0 18.3 6.1 -3.8 2.3

fujian 14.1 21.1 16.9 7.0 -4.2 2.8

zhejiang 8.9 18.5 15.4 9.6 -3.1 6.5

Table 5. Variable Definitions

Variable Name Definition

Dependent Variables

PLW Total levy collected on excess wastewater discharge / Total amount of wastewater discharge that did not meet discharge standards

CODI Total COD discharge / Total industrial output

Independent Variables

Supply Equation

CODD Total COD discharge / Land area

POPD Population / Land Area

INC Per capita income (approximated by per capita consumption)

IND Industrial Output / Total Output

SECED Education level (Secondary school completion rate)

PRIN1 First principal component of [log (INC), log (IND)]

Demand Equation

STATE Share of state-owned factories in industrial output

LARGE Share of large factories in industrial output

BEVERAGE Beverage sector share in industrial output

TEXTILE Textile "

LEATHER Leather "

PAPER Paper "

CHEMICALS Chemicals "

RUBBER Rubber "

PLASTICS Plastics "

POWER Power "

FOOD Food "

FERROUS Ferrous Metals "

COAL Coal Mining "

FERR. MIN. Ferrous Met. Mining "

BUILDING Building Materials "

LIGHT Light Industry "

Table 6

Sources of Data for Twenty-Nine Chinese Provinces

China Environment Yearbooks, 1987-1994

(1) COD (Chemical Oxygen Demand) intensity

(2) Wastewater discharge in excess of standards

(3) Total levy collected on excess wastewater discharge

China Industrial Economy Yearbooks, 1987-1994

(4) Output share by industry sector: food, beverages, textiles, leather, paper, chemicals, rubber, plastics, ferrous metals, power, coal mining, ferrous metals mining, building materials, light industry

(5) Output share by large plants

(6) Output share by state-owned plants

(7) Total industrial output share

China Statistical Yearbooks, 1987-1994

(8) Total provincial output

(9) Population

(10) Provincial area

(11) Consumption per capita

(12) Secondary school completion rate

Table 7.1 Environmental Demand Estimationa

Dependent Variable: log (CODI)

7.1.1 7.1.2

7.1.3 7.1.4 7.1.5

log (PLW) -0.727

(-1.551)

-1.082

(-5.149)***

log (PLW) -0.944

(-3.818)***

-1.095

(-5.024)***

-1.001

(-3.864)***

STATE 0.700

(1.048)

log (INC) -0.341

(-0.973)

LARGE -4.377

(-3.819)***

-4.110

(-6.17)***

log (IND)

0.100

(0.249)

BEVERAGE -8.245

(-0.882)

PRIN1

-0.071

(-0.516)

TEXTILE -3.291

(-0.839)

LARGE -3.623

(-4.455)***

-4.023

(-5.321)***

-4.060

(-6.156)***

FUR -14.599

(-1.610)

PAPER -15.576

(-0.856)

CHEMICALS 0.500

(0.151)

RUBBER 3.213

(0.854)

PLASTICS -8.386

(-0.447)

POWER -12.409

(-2.253)**

FOOD -0.108

(-0.057)

FERROUS 0.550

(0.364)

COAL 10.115

(1.502)

FERR. MIN. -10.840

(-0.676)

BUILDING 8.720

(1.238)

LIGHT 2.764

(1.679)*

Adj-R2 0.803 0.768

0.785 0.764 0.777

N 145 145

145 145 145

a t-statistics are included in parentheses under the estimated parameters. Asterisks indicate the associated significance levels:

* .10

** .05

*** .01

Table 7.2 Environmental Supply Estimationa

Dependent Variable: log(PLW)

Variable Fixed-Effects Cross-Sectionb

Names 7.2.1 7.2.2 7.2.3 7.2.4 7.2.5 7.2.6 7.2.7 5.2.9

log(CODD) 0.135

(1.465)

0.112

(1.208)

0.118

(1.279)

0.128

(0.869)

0.0728

(3.643)***

0.0583

(2.809)***

0.0576

(2.841)***

0.0575

(2.891)***

log(POPD) 1 1 2.438

(2.631)***

0.006

(0.035)

0.0728

(3.643)***

0.0583

(2.809)***

0.0576

(2.841)***

0.0575

(2.891)***

log(INC) 0.621

(4.200)***

0.688

(4.461)***

0.342

(1.255)

0.688

(3.074)***

0.738

(4.135)***

0.407

(1.730)*

0.347

(2.387)**

0.381

(5.013)***

log(IND)

0.374

(1.391)

0.247

(1.884)

0.290

(1.266)

0.347

(2.387)***

0.381

(5.013)***

log(SECED)

0.422

(1.530)

0.450

(1.747)*

0.381

(5.013)***

Constant

-7.190

(-5.709)***

-7.249

(-5.900)***

-6.462

(-4.848)***

-6.123

(-7.437)***

-6.061

(-7.789)***

Adj-R2 0.949 0.950 0.795 0.618 0.631 0.665 0.677 0.689

N 145 145 145 29 29 29 29 29

a t-statistics are included in parentheses under the estimated parameters. Asterisks indicate the associated significance levels:

* .10

** .05

*** .01

b Parameters subject to equality restrictions in bold type

Table 8

Six Chinese Provinces:

Comparative Data for 1987

Province Area

('000 sq. km.)

Population

(million)

COD

Discharge

('000 tons)

Industry

Share

(%)

Cons. per

Capita

(yuan/yr.)

Illiteracy

Rate

(%)

beijing 16 11 106 62 1040 11

qinghai 735 4 11 35 550 41

liaoning 147 38 462 63 750 12

sichuan 570 105 432 36 430 21

guangdong 183 59 379 36 650 15

gansu 418 21 51 43 410 40

Table 9

Observed vs. Simulated Levies and COD Densities

Chinese Provinces, 1987 - 1993

Year

1987

1993

Province Observed Price Simulated Price Observed COD Density Simulated

COD

Density

Observed Price Simulated Price Observed COD Density Simulated COD Density

beijing 0.12 0.13 6.58 6.37

0.12 0.13 5.50 5.98

qinghai 0.03 0.02 0.01 0.01

0.03 0.03 0.00 0.01

liaoning 0.10 0.11 3.14 3.31

0.16 0.14 2.41 2.37

sichuan 0.05 0.04 0.76 1.28

0.06 0.05 0.59 1.14

guangdong 0.07 0.06 2.07 1.36

0.13 0.12 1.79 2.42

gansu 0.05 0.04 0.12 0.15

0.06 0.05 0.09 0.10

tianjin 0.09 0.13 11.63 10.16

0.24 0.17 7.76 10.54

hebei 0.07 0.08 1.45 1.48

0.08 0.08 1.13 1.02

shanxi 0.09 0.06 1.10 1.11

0.07 0.09 0.77 0.84

Inner Mongolia 0.07 0.06 0.09 0.10

0.06 0.07 0.08 0.07

jilin 0.09 0.07 1.48 1.65

0.07 0.09 1.22 1.21

heilongjiang 0.05 0.05 0.60 0.76

0.09 0.07 0.45 0.42

shanghai 0.26 0.16 40.65 49.07

0.15 0.24 25.32 21.42

jiangsu 0.11 0.10 5.58 5.58

0.13 0.14 3.93 5.54

zhejiang 0.09 0.09 4.18 3.27

0.17 0.13 2.80 3.22

anhui 0.05 0.05 2.87 2.93

0.09 0.07 2.01 2.20

fujian 0.04 0.04 1.84 1.51

0.09 0.08 1.48 2.15

jiangxi 0.07 0.05 0.88 1.07

0.05 0.06 1.06 0.85

shandong 0.15 0.12 4.23 3.68

0.16 0.17 3.69 3.89

henan 0.06 0.06 2.03 1.98

0.07 0.08 2.13 1.66

hubei 0.05 0.06 1.11 2.19

0.07 0.07 1.86 1.17

hunan 0.05 0.05 1.78 1.94

0.09 0.08 1.41 1.42

guangxi 0.04 0.04 1.35 1.51

0.07 0.06 2.18 1.66

hainan 0.07 0.05 2.97 4.96

0.06 0.06 1.83 1.91

guizhou 0.02 0.03 0.43 0.48

0.04 0.03 0.34 0.36

yunnan 0.04 0.05 0.46 0.51

0.07 0.08 0.54 0.39

shaanxi 0.08 0.07 0.34 0.31

0.10 0.09 0.24 0.29

ningxia 0.04 0.04 0.33 0.33

0.03 0.05 0.37 0.34

xinjiang 0.11 0.09 0.04 0.05

0.10 0.12 0.05 0.04

Figure 1

Figure 2. Changes in Pollution Equilibria, 1987-93

2.1 Beijing vs. Qinghai

2.2 Liaoning vs. Sichuan

2.3 Guangdong vs. Gansu

--------------------------------------------------------------------------------

Footnote1

For more discussion, see Florig and Spofford (1994).

Footnote2

In the short run, all other emissions are effectively 'free' for factories which discharge multiple pollutants. As 'worst cases' are successively cleaned up, however, the levy will shift across pollutants.

Footnote3

In this paper, the term "province" refers to provinces, autonomous regions and municipalities which are directly affiliated with the central government.

Footnote4

See Wheeler (1991).

Footnote5

Using reduced-form estimation, one of the authors and colleagues have found strong supporting evidence for the 'equilibrium pollution' hypothesis in Indonesia, India, Bangladesh and Thailand. See Pargal and Wheeler (1996); Hettige, Huq, Pargal and Wheeler (1996); Hartman, Huq and Wheeler (1996).

Footnote6

Stricter regulation will undeniably impose some costs on factories. Empirical analyses have reached different conclusions about the significance of these costs in OECD economies (Oates, Palmer and Portney (1993); Gray and Shadbegian (1993)). The issue remains largely unexplored for China and other developing countries. Without any evidence to go on, local regulators could scarcely be blamed for adopting a conservative stance in areas which have a strong need for more industrial employment and income.

Footnote7

See Mody and Wheeler (1992) for evidence that agglomeration economies play a powerful role in location decisions, at least for multinational corporations.

Footnote8

The Chinese Industrial Economy Yearbook uses two systems for categorizing industrial facilities as large, medium or small. For sectors with an identifiable main product, annual production capacity is used as the criterion. For example, cement plants are classifed as large if their annual production capacity exceeds 600,000 tons; medium in the range 200,000 - 600,000 tons; and small otherwise. For sectors with heterogeneous products, the value of fixed assets is used for size-class assignments. Plants are classified as large if their assets are valued at more than 50 million yuan; medium in the range 10 - 50 million yuan; and small otherwise.

Footnote9

We should note that NEPA has introduced additional economic incentives to encourage pollution reduction since 1990, including four additional penalty categories (NEPA, 1992). In 1993, NEPA introduced a fee levied on the total volume of industrial wastewater discharge. In the first year of implementation, collections of this fee amounted to about 10% of the collections of the overstandard fee (Florig, Spofford, Ma and Ma (1995)). Unfortunately, we have no evidence on these measures by province. We therefore have no way to test their independent effect. In any case, the proxy test which we have employed for other instruments should apply to them as well.

Footnote10

Unitary elasticity falls within the 95% confidence interval around the estimated value.

Footnote11

We are indebted to our colleague Maureen Cropper for this point.

Footnote12

This result is nearly identical to much earlier findings for industrial water polluters in the Netherlands. See Wheeler (1991).

Footnote13

1990 prices.

The economist， 8月21日那期，有个讲中国环境和卫生的专题，共有三篇文章，很有意思的。