Workshop on 2 Economywide Policies and the Environment World Bank, Washington, D.C. 14-1 5 December 1993 National Economic Policy and Industrial Pollution The Case of Indonesia: 1975-89 David Wheeler and Paul Martin Pollution and Environmental Economics Division Environment Department World Bank NATIONAL ECONOMIC POLICY AND INDUSTRIAh POLLUTION The Case of Indonesia:. 1975-89 David Wheeler* Paul Martin October, 1993 World Bank: Respectively Acting Chief, Environment, Infrastructure and Agriculture Division, Policy Research Department and Consultant, Environment Unit, East Asia and Pacific Department III. Our thanks to Basanta Chaudhuri for research assistance. Support for this research was provided by the ENVPE program "Economywide Policies and the Environment," under the direction of Mohan Munasinghe, Jeremy Warford and Wilfrido Cruz. In addition, the following contributors are gratefully acknowledged: Center for Economic Studies, U.S. Census Bureau; Richard Calkins, Gunnar Eskeland, Lili Liu, Mala Hettige and Shakeb Afsah (World Bank); Raymond Hartman (Consultant); and Manjula Singh (Boston University) 2 I. INTRODUCTION During the coming decade, the scope and capability of many LDC regulatory institutions will be quite limited. Without appropriate regulation, macroeconomic policy reform may inappropiately stimulate some activities which are environmentally damaging. Countervailing policies may work if the consequences of reform can be foreseen, but this will require much better knowledge of linkages between the macroeconomy and the environment. This paper attempts to contribute by studying the impact of policy on Indonesian industrial pollution since 1975. Drawing on extensive documentation of sectoral differences in pollution intensity (or pollution per unit of output), we link policy to pollution damage by tracing the impact of the former on the sectoral and regional pattern of industrialization. The paper is, of course, a study of unintended consequences. Like most countries, Indonesia has made little attempt to set national economic policies with industrial pollution in mind. However, major environmental impacts are quite possible in a weakly-regulated economy. Damaging interventionist policies may include promotion of pollution- intensive heavy industry; restrictions on foreign investment in sectors where multinationals operate with OECD-standard clean technologies; and heavy taxation of imported equipment which may incorporate recent pollution-reducing or recycling technologies. Liberalization without regulation or countervailing policy may also lead to serious damage when local raw materials provide strong comparative advantage in pollution-intensive sectors or processes; or when private colocational economies lead heavy industry to grow rapidly in a few centers. The paper is organized as follows: Section II presents the best available evidence on sectoral pollution intensities, 3 highlighting the difference across sectors and pollutants. Section III provides a brief overview of Indonesian macropolicies during the period 1975-89. Drawing on the pollution intensities developed in Section II, Section IV examines the impact of policy on the estimated growth and regional distribution of industrial pollution. The discussion pays particular attention to the impact of the economic reforms which began in 1983, and is structured around three questions: (1) Did the economic reforms affect the pollution intensity of industrial production by altering the sectoral composition of manufacturing output? (2) Is there any evidence that the reforms exerted an influence on industrial location, thus altering the impact of the associated pollution? (3) Has the protection of industry shown a "Brown" bias, by favoring more pollution-intensive sectors ? Section V summarizes the findings and the implications for macroeconomic/environmental policy analysis. II. WHAT ARE "POLLUTING INDUSTRIES" Industrial processes generate hundreds of potentially harmful substances. Those which are actually discharged into the environment cause damage to human health, economic activity, and ecosystems in degrees which vary enormously by substance, timing and medium of discharge. Among water pollutants, for example, biological oxygen demand (BOD) may devastate local fisheries but have little direct impact on community health. Waterborne toxic metals, while showing little immediate impact, may cause serious long run health damage as they accumulate in the food chain. Among air pollutants, very heavy concentrations of suspended particulate matter can almost immediately kill people who suffer from severe asthma; carcinogenic compounds may cause severe long run damage, but their workings are largely invisible and much slower. 4 A general study of industrial pollution must therefore begin with construction of appropriate categories for analysis. Three will suffice for the present study: (1) Traditionally regulated water pollutants such as BOD and suspended solids, which damage aquatic life and downstream economic activities; (2) Traditionally regulated air pollutants such as suspended particulates and sulphur dioxide, whose primary impact has been on human respiration; and (3) Toxic pollutants, including bioaccumulative metals and carcinogenic compounds emitted in all media, whose long-run impact on human and ecosystem health has more recently been recognized. For this study, we have chosen biological oxygen demand (BOD), total suspended particulates (TSP), and total toxic emissions (TOX) as representative pollutants from the three categories. Using a sample of approximately 40,000 U.S. manufacturing facilities, we have estimated pollution intensities for thirty-eight sectors defined by the International Standard Industrial Classification (ISIC). These intensities are defined as [pollutant volume]/(manufacturing scale), where the latter could be indexed by physical output, employment, value added, or output value. We have chosen output value in this case because the corresponding Indonesian data are more readily accessible.1 Table 1 presents estimates of BOD, TSP and TOX pollution intensities for thirty-eight ISIC sectors. For analytical 1. Can U.S. pollution intensities be used to approximate Indonesian intensities? The existing evidence (still quite scanty) suggests that absolute sectoral intensities vary significantly across countries with differences in regulation and input prices. However, relative sectoral intensities seem much more stable across countries. To characterize crudely: wood pulping produces far more BOD than garment assembly; cement production produces far more airborne particulates than microchip assembly. Since this study focuses on relative intensities, we are confident that our basic results are robust. Table 1 SECTORAL POLLTJTION INTENSITIES ESTIMATED FROM U.S. MANUFACTORING DATA (Tons/$Million) ISIC SECTOR BOD PARTICULATES TOXICS 3110 Food Products 11.840 1.035 0.391 3130 Beverages 3.945 0.068 0.103 3140 Tobacco 0.000 0.000 0.244 3210 Other Textile Products 0.000 0.593 1.751 3211 Spinning, Weaving 0.435 0.506 1.553 3220 Wearing Apparel 0.000 0.000 0.872 3230 Leather & Products 0.000 0.339 7.690 3240 Footwear 0.000 0.000 1.139 3310 Wood Products 0.000 5.148 2.200 3320 Furniture, Fixtures 0.000 0.656 2.683 3410 Other Paper Products 0.000 2.349 4.371 3411 Pulp, paper 17.047 2.669 3.113 3420 Printing, Publishing 0.234 0.218 3.757 3510 Other Industrial Chemicals 0.042 1.005 26.130 3511 Basic Industrial Chemicals 0.828 0.993 16.127 3512 Agricultural Chemicals 0.042 1.005 26.130 3513 Synthetic Resins 0.189 0.556 7.001 3520 Other Chemical Products 54.281 0.258 1.782 3522 Drugs and Medicines 0.061 0.157 1.983 3530 Petroleum Refineries 0.129 0.513 1.879 3540 Petroleum & Coal Products 0.434 5.912 1.272 3550 Rubber Products 0.006 0.249 1.467 3560 Plastic Products 0.000 0.116 4.668 3610 Pottery, China, etc. 0.000 0.000 1.807 3620 Glass & Products 0.000 1.553 0.741 3690 Non-Metal Products n.e.c. 0.000 12.524 1.927 3710 Iron and Steel 0.015 1.774 3.821 3720 Non-Ferrous Metals 7.853 4.537 4.667 3810 Metal Products 0.496 0.278 2.296 3820 Other Machinery n.e.c. 0.000 0.783 0.798 3825 Office & Computing Machinery 0.000 0.000 0.152 3830 Other Electrical Machinery 0.006 0.142 0.899 3832 Radio, Television, etc. 0.000 0.010 D.904 3840 Transport Eguipment 0.005 0.108 0.504 3841 Shipbuilding, Repair 0.000 0.318 1.273 3843 Motor Vehicles 0.001 0.124 0.333 3850 Professional goods 0.000 0.001 0.444 3900 Other Industries 0.000 0.286 1.353 Source: World Bank, Industrial Pollution Projection System (Current estimates - See Wheeler et. al., 1993 for discussion of further work) 5 purposes, the information in Table 1 can usefully be summarized as follows: 1. Skewed distributions All three intensity distributions are sharply skewed, with relatively few dominant sectors. Unless output shares have a strong inverse correlation with pollution intensities, this guarantees heavy concentration of pollution in relatively few sectors. Table 2 shows that such concentration does indeed seem to characterize Indonesia's manufacturing economy. For analytical purposes, we therefore identify sectors in the top quartile of pollution intensities as "Dirty" and the others as "Clean." The "Dirty" sectors dominate the emissions problem (but, as noted below, these sectors differ substantially by pollutant). 2. Activity-based concentration of intensities Since pollutants are harmful residuals from manufacturing, it is reasonable to suppose that activities which are intensive in the processing of primary materials will also be more pollution intensive. To test the strength of this intuitive notion, we have divided the ISIC sectors into two groups. The first, or "Assembly," group consists of sectors which focus mainly on fabrication or assembly activities. As Table 3 indicates, this extremely heterogeneous group includes sectors as diverse as Apparel (3220), Radio/TV (3832) and Motor Vehicles (3843). The second, or "Materials Processing" group includes the sectors which focus mainly (although, of course, not exclusively) on the processing of primary raw materials. Again heterogeneity reigns, with constituent sectors including Food Products (3110), Pulp and Paper (3411), and Steel (3710). Table 3 indicates the utility of this decomposition for a Table 2 ESTIMATED POLLUTION FROM INDONESIAN MANUFACTURING, 1989: 5 LARGEST SOURCES IN 38 ISIC SECTORS ISIC SECTOR BOD CUM. tb 3110. Food Products 49 3520 Other Chemical Products 85 3411 Pulp, paper 97 3211 Spinning, Weaving 98 3130 Beverages 99 ISIC SECTOR TSP CUM. k 3310 Wood Products 41 3690 Non-Metal Products n.e.c. 62 3110 Food Products 74 3710 Iron and Steel 82 3411 Pulp, paper 86 ISIC SECTOR TOXIC CUM. % 3512 Agricultural Chemicals 28 3310 Wood Products 38 3710 Iron and Steel 48 3511 Basic Industrial Chemicals 56 3211 Spinning, Weaving 63 Sources: World Bank, Industrial Pollution Projection System; BPS, Indonesia Industry Survey, 1989 'Cumulative % of total estimated output of the pollutant by manufacturing in 1989 TABLE 3 'DIRTY' SECTORS* FOR THREE MAJOR POLLUTANTS ASSEMNLY SECTORS ISIC SECTOR BOD TSP TOX 3210 Other Textile Products 3220 Wearing Apparel 3230 Leather & Products DIRTY 3240 Footwear 3320 Furniture, Fixtures 3410 Other Paper Products DIRTY DIRTY 3420 Printing, Publishing 3520 Other Chemical Products DIRTY 3560 Plastic Products DIRTY 3810 Metal Products DIRTY 3820 Other Machinery n.e.c. 3825 Office & Computing Machinery 3830 Other Electrical Machinery 3832 Radio, Television, etc. 3840 Transport Equipment 3841 Shipbuilding, Repair 3843 Motor Vehicles 3850 Professional goods 3900 Other Industries NATERIALS PROCESSING SECTORS ISIC SECTOR BOD TSP TOX 3110 Food Products DIRTY DIRTY 3130 Beverages DIRTY 3140 Tobacco 3211 Spinning, Weaving DIRTY 3310 good Products DIRTY 3411 Pulp, paper DIRTY DIRTY 3510 Other Industrial Chemicals DIRTY 35L2 Basic Industrial Chemicals DIRTY DIRTY 3512 Agricultural Chemicals DIRTY 3513 Synthetic Resins DIRTY 3522 Drugs and Medicines 3530 Petroleum Refineries 3540 Petroleum & Coal Products DIRTY DIRTY 3550 Rubber Products 3610 Pottery, China, etc. 3620 Glass & Products DIRTY 3690 Non-Metal Products n.e.c. DIRTY 3710 Iron and Steel DIRTY DIRTY 3720 Non-Ferrous Metals DIRTY DIRTY DIRTY 'Dirty' defined as upper quartile in distribution of 38 ISIC intensities by pollutant Source: World Bank, Industrial Pollution Projection System 6 general analysis of industrial pollution. Almost all the Dirty activities are in the "Material" sectors: 7/9 for BOD; 8/9 for TSP; 6/9 for TOX. For most practical purposes, the aggregative Materials Processing sector can also be labeled the Dirty sector. 3. Low correlation of intensities across sectors While the Dirty sectors are almost all concentrated in Materials, they differ across pollutants. Table 4 displays generally low rank correlations for the three sets of pollutant intensities. Even the highest correlation, between overall TSP and TOX rankings, is around a modest .5. This finding introduces an important corollary to the basic Assembly/Materials Processing story. Highly aggregative analysis along these lines will probably capture the main story for all three pollutants, since sectoral representation within Processing is apt to be random with respect to pollutant intensity. However, the lack of strong intensity correlation across Processing sectors also suggests that significantly different stories may hold for each pollutant. III. INDONESIAN MACROPOLICY, 1975-89 Indonesia's macroeconomy passed through three distinct stages during the period 1975-1989. From 1975 through 1979 protection rates were generally high; dependency on oil revenues was heavy; subsidies were pervasive; public investment was a major component of industrial capital formation; and the emphasis was on production of staples for the domestic market. The result of this policy stance was an economy which was highly "inward- oriented" by international standards. Table 5 summarizes an analysis of four indices of relative TABLE 4 RANK CORRELATIONS (38 ISIC SECTORS): INTENSITIES FOR THREE MAJOR POLLUTANTS OVERALL BOD TSP TSP .28 - TOX .22 .50 ASSEMBLY SECTORS BOD TSP TSP -.03 - TOX .12 .54 MATERIALS PROCESSING SECTORS BOD TSP TSP .18 - TOX .16 .40 TABLE Sa COUNTRY OPENNESS RASKINGS MEDIN OF FOUR MEASURES: 1975-83 1 2 3 4 5 1 Colombia Cyprus Belgium Barbados Australia 2 Malta Germany Chile Brazil Austria 3 Pakistan Hong Kong Denmark Japan Benin 4 South Africa Korea Fiji Norway Burma 5 Sri Lanka Mexico India Sweden Chad 6 Thailand Netherlands Ireland Togo Costa Rica 7 Singapore Israel Turkey Iran 8 Switzerland Italy Panama 9 UK Jordan Trinidad 10 Malaysia Tunisia 11 Mauritius 12 Nepal 13 Papua New Guinea 14 Peru 15 Philippines 16 Portugal 17 Spain 18 Syria 19 Taiwan 20 Uruguay 6 7 5 9 10 1 Bangladesh Argentina Algeria Angola Bolivia 2 Botswana Cote d'Ivoire Burundi Honduras Ghana 3 Burkina Faso Dominican Republic Cameroon Iraq Guinea 4 Canada France Central African Rp Niger Nigeria 5 Ecuador Gambia Congo Rwanda Sierra Leone 6 Ethiopia Guyana Egypt Tanzania Somalia 7 Greece Haiti El Salvador Uganda 8 Guatemala Jamaica Gabon Zaire 9 Indonesia Lesotho Liberia 1D Kenya Paraguay Mauritania 11 Madagascar Senegal Mozambique 12 Malawi Swaziland Nicaragua 13 Mali Venezuela Sudan 14 Morocco Yemen Zambia 15 Suriname Zimbabwe TABLE Sb COUNTRY OPENNESS RANKINGS MEDIAN OF FOUR MEASURES: 1975-83 SOUTH AND EAST ASIA 1 2 3 4 5 6 Pakistan Hong Kong India Japan Burma Bangladesh Sri Lanka Korea Malaysia Indonesia Thailand Singapore Mauritius Nepal Philippines Taiwan 7 country openness during the period 1975-83.2 one hundred ten countries have been divided into deciles for each measure, with the median decile ranking reported in Table 5a. Internationally, Indonesia ranked in the low midrange during this period. In East and South Asia, however, it ranked at the bottom along with Bangladesh (Table 5b). After 1979, Indonesia was strongly affected by the world recession; a fall-off in its primary export markets; and a sharp decline in petroleum revenues. Faced with the consequences of overdependency on revenues from resource extraction, the government initiated policy reforms which were intended to promote the growth of private manufacturing. In 1983, the government began a sustained program of liberalization. Initial deregulation of the financial system was soon followed by reduction of non-tariff trade barriers; lower tariffs on many items; less discrimination against foreign investors; general reduction of red tape; and a strong emphasis on the promotion of manufactured exports from the private sector. In summary, the Indonesian macroeconomy went through three sharply-defined periods of change from 1975-1989: Inward-looking growth propelled by oil exports and public investment (1975-79); recession and "policy introspection" (1980-84); and export- oriented liberalization (1985-89). This abrupt transition from inward to outward orientation provides a classic policy experiment. 2Sources: Harrison (1991); Dollar (1991) ; Wheeler/Mody (1992); Dollar (1992) a irIV. POLICY REFORM, sDEUSTRIAL CEI3GR AND IOtLUTION _ndustrial growth has clearly accelerated under liberalization, and parricularly cn Java. From an environmental persmective, this does not look like a pretty story: At constant pollution intensity (pollution Der unit of output), relatively rapid acceleration of industrial crcw-th on Indonesi-a2s most populous island would i=mly a very rapid increase in health damage. From an overall welfare perspective, however, it would be rerverse in the extreme tc sugaest that rapid industrial growth in a poor country is a bad thing. The real cuestion in this case, then, is not whether 1lberalization acceierated growth l.n industry and employment (good things), but whether it had a narkedly perverse e-ffect on overall pollution intensity and the regional distribution of pollution. 1. Liberaliza-ion and Pollution intensity. :4ere we recall the fundamental lesson of Pat= 77 of the paper: industrial pollution is largely concentrated in the Materials Processing sec.ors, while most Assembly Sectors are relatively benign f_om an envircnmental perspective. An argunent from standard comparative advantage would suggest =cre rapid growth of labor-intensive assemblv in the wake of liberalization. Table 7a suggests a powerful effect: In the late _97C's, Materials Processing was growing at -2.3% annuall-y in Indonesia while Assembly was growing at 9.'%. After liberalizazion, the pattern was markedly differen.: ;'-sseb3bly 19.6%; Materials Processing 14.8%. The -mplication, of course, is _-at the relative acceleration of clean Droduction substantially reduced the aggregate pollution intensity -f new cutput.- 3This does not, of course, imply a diminution of output -f dangerous pollutants. 'As -he nunbers act'arlly show, Dirt production grew faster affter l-beralization =tan previously. 9 Unfortunately, the apparent power of this result is considerably diminished by a regression analysis of sectoral output against ICP income for 130 countries. This reveals a robust positive correlation between ICP income and the share of assembly in total manufacturing output. As Table 6 shows, the behaviour of the Indonesian economy in this regard has not differed significantly from world standards. No significance may be attached either to a general residual dummy variable (INDON) for Indonesia, or dummies for the periods 1975-79 (DIND7579) and 1985-89 (DIND8589), which allow us to see whether within-period deviations from the world trend were more or less than expected, relative to the average Indonesian deviation. The observation that the relationship between income and sectoral composition in Indonesia followed the world trend does not, of course, preclude the possibility that liberalization led to both higher incomes and a greater share of assembly in manufacturing. Unfortunately the supporting evidence isn't very strong here either. As Figure 1 shows, ICP income grew rapidly until the recession in 1980, and following liberalization continued to increase at a slower pace. Figure 1 also indicates both the actual share of assembly in manufacturing output, and the share predicted from income by Model 1 in Table 6. The close association of income and assembly share, and the lack of correlation between income and policy regime, leave little evidence of a policy impact on the share of assembly in total production. 2. Policy and the Regional Distribution of Pollution The regional distribution of Assembly and Material Processing growth rates, shown in Table 7b, suggests a pattern which is less environmentally benign. For environmental impact, the best outcome would be a strong shift toward Assembly growth on Java, with concomitant shift toward relatively faster growth Figure 1 - Assembly Share & ICP Income Indonesia, 1975-1989 28- 1900 -800 -1700 026 1600k 15000 .25 2 -14008 IL Year | _AMWal%Amsembly -+-PredEb % Assemb. IA-CP Incom9 Table 6 SHARES OF ASSEMBLY AND MATERIALS PROCESSING SECTORS IN TOTAL MANUFACTURING OUTPUT, AND ICP INCOME: REGRESSION RESULTS Dependent Variable: Log(Assembly Share/Materials Processing Share) Model 1 Model 2 Constant -17.096** -17.099** (-2.256) (-2.525) INDON -0.253* -0.250 (-1.877) (-1.083) DIND7579 -0.007 (-0.022) DIND8589 0.000 (0.000) LY 6.253** 6.254** (2.362) (2.360) LY2 -0.833** -0.833** (-2.433) (-2.431) LY3 0.038*** 0.038*** (2.629) (2.627) Sample Size 1406 1406 Adj. R-square 0.33 0.33 Regr. F 171.6 114.2 Key: *** significant at the 1% level ** significant at the 5% level * significant at the 10k level INDON Dummy for Indonesia DIND7579 Dummy for Indonesia, 1975-79 DIND8589 Dummy for Indonesia, 1985-89 LY Log (ICP Income) LY2 (LY)2 LY3 (LY)3 10 in Naterials Processing off Java. Such an outcome would significantly reduce human exposure to the total pollution load. In fact, our results show the opposite occurred. The growth rate of Assembly accelerated nore off Java than on Java, and the growth rate of Materials Processing accelerated more on Java. A happier story emerges from an analysis of pollutant- specific growth rates, underlining the importance of disaggregation by pollutant, as discussed in Part II. Table 7b shows the aggregate growth rates of the nine most pollution- intensive sectors in BOD, TSP and Toxic emissions. Little locational impact of liberalization can be detected for the BOD- intensive sectors, which exhibit approximately equal growth rates on and off Java, both before and after the recession. There is, however, a clear acceleration of growth in both the TSP and Toxic-intensive sectors off Java following liberalization, that is not matched by the growth rates on Java.4 3. The "Pollution Intensity" of Effective Protection Thus far, we have discussed the impact of broad policy reform on the growth and regional distribution of aggregative Clean and Dirty sectors in the Indonesian economy. The evidence does give a relatively Clean cast to liberalization, since it may have accelerated the growth of Assembly relative to Materials Processing throughout Indonesia (certainly, the converse didn't occur), and led to faster growth off Java of key pollution- intensive sectors. This dynamic result, however, finds no apparent reflection in an analysis of the pollution intensity of effective protection 4The apparent discrepancy between the general Materials Processing story and the pollutant-specific stories is explained by the disproportionate contribution of the 'Dirty' Assembly sectors to overall manufacturing growth. Table 7a SECTORAL GROWTH RATES (%) 1975-89 75-79 80-84 85-89 ASSEMBLY 9.1 7.8 19.6 MATERIAL 12.3 8.8 14.8 Table 7b SECTORAL AND REGIONAL GROWTH RATES (%), 1975-89 75-79 80-84 85-89 ASSEMBLY 9.1 7.8 19.6 Java 9.5 7.7 20.5 Off Java 2.8 10.5 11.5 MATERIAL 12.3 8.8 14.8 Java 9.2 7.6 13.3 Off Java 28.3 12.5 18.5 BOD Intensive 7.1 7.2 17.7 Java 7.1 6.9 17.8 Off Java 7.7 10.7 17.5 TSP Intensive 12.2 11.1 17.1 Java 12.4 8.0 15.5 Off Java 11.2 20.0 19.9 TOXIC Intensive 16.3 12.6 18.7 Java 17.4 8.9 17.8 Off Java 12.3 22.3 20.7 11 in Indonesia. In recent years, very partial evidence has been used (by the present authors as well as others) to assert that industrial protection should generally have a "Brownw bias because of a presumed correlation between pollution intensity and intensity in factors (capital, energy, skill) which are relatively scarce in developing countries.5 By this argument, more protected sectors (those least able to compete in world markets) should be, on average, more pollution intensive. Because effective protection has been extensively documented through the years in Indonesia, we have been able to test the ' Brown bias' hypothesis. To do so, we have combined our pollution intensity rankings with the results of three detailed studies of effective protection rates (EP' s) undertaken in 1971, 1980 and 19896. Differences in methodology prevent comparison of absolute EPR's over time, but it is still possible to examine levels and trends in the rank correlations of effective protection and pollution intensity. In Table Ba, rank correlations for EPR' s and intensity in BOD, TSP and TOX are presented for the three years: 1971, 1980 and 1989. They suggest quite strongly that, for Indonesia at least, there has been no Brown bias in protection. Indeed, the converse seems generally to have been true. In 1971, effective protection apparently had a modest Green bias for TSP and TOX, with near-neutrality for BOD. The same pattern is evident (although more modestly) in 1980. The results do suggest, however, that the remaining protection following the 1983 reduction had a pronounced Green bias. By implication, protection was reduced more sharply for Dirty sectors then for their Clean counterparts. 5See for example Birdsall/Wheeler, 1993 6Sources: Pitt (1981), Pangestu & Boediono (1984), Fane & Phillips (1991), Wymenga (1991) Table 8a RANK CORRELATIONS: SECTORAL EFFECTIVE RATES OF PROTECTION AND POLLUTION INTENSITIES, 1971-89 Intensities BOD TSP TOXIC ERP, 1971 0.08 -0.25 -0.21 ERP, 1980 -0.06 -0.07 -0.19 ERP, 1989 -0.27 -0.19 -0.32 Table 8b RANK CORRELATIONS: SECTORAL EFFECTIVE RATES OF PROTECTION AND GROWTH RATES, 1975-1989 ERP Growth Rate, 1975-79 0.12 Growth Rate, 1980-84 -0.16 Growth Rate, 1985-89 -0.02 12 Unfortunately, the logic of this entire discussion is called into question by an even more fundamental result. Presumably, the rationale for concern about Brown bias in protection is enhanced growth prospects for the protected industries relative to their status under open competition. To see whether this rationale has in fact been supported by history, we have looked at the rough pattern of relative protection and sectoral growth for the late '70' s, early '80' s and late '8O' s. our datasets were not perfectly matched, and the results are therefore not definitive: We have calculated the rank correlations of sectoral growth rates for 1975-79 with effective protection in 1971; sectoral growth for 1980-84 with effective protection in 1980; and growth for 1985-89 with effective protection in 1987 as measured in an additional World Bank study. The results are presented in Table 8b. For the 1970' s, there was apparently a small positive correlation between protection and sectoral growth, as hypothesized. In the 1980's, however, the relationship was reversed -- more protected sectors grew slower, not faster! Analysis of protection measures is not really our province at present, so we will leave it to others to ponder the larger significance of this perverse result. But from the perspective of environmental policy, our inference is straightforward. Relative protection seems to have had little or no consequence for sectoral growth, so it scarcely seems worrying much about the Brown bias of protection. And in any case, Indonesian protection seems to have had a consistent (and, indeed, increasing) Green bias through the years. V. SUMMARY AND CONCLUSIONS In this paper, we have developed and applied some tools for analyzing the probable impact of macropolicy change on industrial pollution in unregulated economies which have few local emissions 13 measures. Using a large U.S. database, we have developed a broad analytical distinction between Assembly ('Clean') and Materials Processing ('Dirty') sectors. In the latter case, we have shown that water, air and toxic pollution intensities are themselves not highly correlated, although all the relevant sectors are disproportionately concentrated in Materials Processing. Thus, we suggest a two-level breakdown for analysis: Between two broad sectors, and, within the Dirty sector, among air, water and toxic pollutants. Our application of these measures to the Indonesian case yields a mixed judgment on linkages between macropolicy change and industrial pollution. At the broadest level, we find mixed evidence on the impact of general liberalization on the share of Assembly industry in total industrial output. Our analysis of the locational impact of liberalization highlights the importance of disaggregation by pollutant. At the aggregate level of Assembly and Materials Processing, we find that the broad impact of liberalization seems to have been regional convergence of sectoral growth rates. This is not progressive from an environmental perspective, as it indicates an acceleration of Dirty sector growth in heavily populated regions. Happily, regional growth rates of the most pollution intensive sectors for emissions of BOD, TSP and Toxics reveal a different -picture. At this level of disaggregation, liberalization seems to have had no regionally differentiated impact-on the growth of BOD-intensive sectors, and appears to have encouraged faster growth of the TSP and Toxic-intensive sectors away from the heavily populated provinces of Java. Looking more closely at the sectoral level, we have found little to support the common hypothesis that protection has a Brown bias. The converse has generally been true in Indonesia, and, paradoxcally, protection seems to have had little or no impact on relative sectoral growth in any case. 14 Thus, for Indonesia at least, we conclude that macropolicy reform has been at least environmentally neutral (and perhaps benign)in its impact on sectoral growth rates; benign in its regional impact for certain pollutants; and benign at the level of reduction in sectoral protection. We cannot, of course, be sure that other country reform processes have had similar environmental overtones. We can suggest, however, that the analytical tools and measures developed for this study could readily be applied to sectoral data for other countries as well. In closing, we should note once again that our paper is about changes in structure and location, not really about scale. The evidence suggests that liberalization was a resounding success for manufacturing growth, and the resulting jump in scale has more than compensated for any countervailing influences on the sectoral structure of manufacturing. Without explicit regulation of pollution, Indonesia's booming industrial areas are rapidly becoming dangerously polluted. 15 REPERENCES Birdsall, Nancy and David Wheeler, 1993, "Trade Policy and Industrial Pollution in Latin America: Where Are the Pollution Havens?", Journal of Environment and DeveloRment,2,1, Winter Dollar, David, 1991, "Outward Orientation and Growth: An Empirical Study Using a Price-Based Measure of openness," Background Paper, World Development Report 1991 , 1992, "Outward-oriented Developing Economies Really Do Grow More Rapidly: Evidence from 95 LDCs, 1976-85," Economic Develo-ment and Cultural Chance Harrison, Ann, 1991, "Openness and Growth: A Cross-Country, Time- Series Analysis for Developing Countries," Background Paper, World Development Renort 1991 Thomas, Vinod, Nadav Halevi and Julie Stanton, 1991, "Does Policy Reform Improve Performance?," Background Paper, World DeveloRment R,eport 1991 Wheeler, David, Shakeb Afsah, Mala Hettige, Paul Martin, Raymond Hartman and Manjula Singh, 1993, "The Industrial Pollution Projection System: Estimation and Application of Pollution Intensities by Pollutant and Industry Sector" (Washington, World Bank) Wheeler, David and Ashoka Mody, "International Investment Location Decisions: The Case of U.S. Firms," Journal of International Economics, 3, 1992