DRl) DISCUSSION PAPER Report Noo DRD278 FISCAL POLICY AND DEVELOPMENT STRATEGY IN SOUTHERN ASIA by Gustav F. Papanek Boston University Development Research Department Economics and Research Staff World Bank The World Bank does not accept responsibility for the views expressed herein·· which are those of the author(s) and should not be attributed to the World Bank or to its affiliated organizations. The findings, interpretations, and conclusions are the results of research supported by the Bank; they do not necessarily represent official policy of the Bank. The designations employed, the presentation of material, and any maps used in this document are solely for the convenience of the reader and do not imply the expression o£ any opinion whatsoever on the part of the World Bank. or its affiliates conce-cning the legal status o£ any country, territory, city, area, or o£ its authorities, or concerning the delimitations of its boundaries, or national affiliation. FISCAL POLICY AND DEVELOPMENT STRATEGY ~N SOUTHERN ASIA Abstract The role of fiscal policy is studied in the context of two distinct development strategies in five countries (Indonesia, Bangladesh, India~ Pakistan, and Sri Lanka) over a period of 30 yearso Two distinct economic policy packages dominated this period. Dirigiste governments expanded the scope of physical controls and the range of publicly owned enterpriseso Market oriented regimes diminished the scope of controls and physical allocation while prices were given a greater role in influencing decisions. While all countries were mixed economies throughout their recent history, there were quite clear and distinct episodes when they tended more towards one end or the other of the spectrum of these two contrasting strategies. The market oriented strategy on averagB achieved a growth rate of 50 percent higher than the dirigiste, with no adverse consequences for income distribution and therefore was more successful in reducing poverty. The study reviews the evidence on the roles of taxation and public expenditure, and of overall fiscal incidence, and the effects of fiscal actions on savings. Some of the conclusions of the study are as follows. Tax structure and expenditure tended to be similar under both types of policy regimes. They differed among countries because of differences in economic structure and history more than differences in strategy. Fiscal policy tended to be less powerful than other economic policies. However, in some cases studied, the potential of fiscal policy in its impact on equity and growth is demonstrated. The tax system seems to have been as progressive under market oriented as under dirigiste regimes. But pro-poor expenditures were higher under the dirigiste if only because of the higher proportion of GDP devoted to social and welfare expenditures. However, this is to be matched against the relatively hfgher growth rate of the market oriented regime, which benefited also the poor. Cont:ents Comparative Analysis - Similar Economies, Different Policies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 A. Ta.xtation 1. Tax Rates and Development Strategy ............................. 5 2. Tax Structure ........................................ , . . . . . . . . . 11 3. Tax Reforms and the Movement to Greater Reliance on the Market ................................. 15 B. Government Expenditures 1. Expenditure Trends and Development Strategy~·············· ..... 22 2. The Structure of Government Expenditures1 .....•..•..•.•......... 25 C. Fiscal Incidence 1. Tax Incidence Under Different Strategies ....................... 33 2. Ex.pendi ture Incidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 3. Fiscal Incidence: The Effect of Taxes and Expenditures ........................... 54 D. The Fiscal System and Savings 1. Economic Strategy, Fiscal Policy and Savings in South Asia ...................................... 61 2. Fiscal Policy, Growth and Savings in Sri Lanka ................. 64 3. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 E. Why the Neglect of Fiscal Policy When Strategy Changes? ............ 71 References Tables 1. Growth Rates Under Dirigiste and Market Oriented Governments ....... 3 A 2. Total Revenue as Percent of GDP .................................... 5 A 3. Tax Revenue as Percent of GDP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 B 4. Tax Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 A 5. Total Expenditures as Percent of GDP ............................... 22 A 6. Structure of Expenditures ............................ ~ ............. 25 A 7. Social Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . • . 27 A 8. Effective Tax Rates by Income Groups-- Sri Lanka ..•.•.•............ 34 A 9. Tax Incidence on Alternative Assumptions-- Sri Lanka ............... 35 A 10. Effective Tax Rates in Sind, Pakistan- 1977/78 ................... 39 A 11. Income Distribtuion and Tax Incidence in Pakistan .................. 39 A 1.2. Incidence of Non-Oil Taxes in Indonesia- 1980 .................... 42 A 13. Social and Welfare Expenditures and Economic Strategy in Sri Lanka .................................. 46 A 14. Incidence of Social and Welfare Expenditures, Sri Lanka ............ 46 B lS. Per Capita Food Subsidy as Percentage of Per Capita Income, Sri Lanka .................................... 48 A 16. Effective Benefit Rates for Specific Public Goods, Sind Province. Pakistan ................ ··~·· ................ 49 A 17. Income Distribution and Benefits Incidence in Sind Province, Pakistan ' .............................. J • 50 A 18. Combined Tax And Government Expenditure Incidence, Sind Province, Pakistan ................................. 56 A 19. Fiscal Incidence by Income Groups, Sind, Pakistan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . • . . . . . . . . . . . . • . . . . 56 B 56 c 20. Combined Tax and Social/Welfare Expenditure Incidence, Sri Lanka, 1978 ......................................•.. 57 A FISCAL POLICY AND DEVELOPMENT STRATEGY IN SOUTHERN ASIA Fiscal policy provides one of the most powerful set of tools for economic strategy according to textbooks and teachers of economics or public finance. They provide an impressive list of objectives which governments try to attain through its use, including economic efficiency, more rapid growth, greater equity and stability, plus a host of more specific objectives from dealing with balance of payments problems to encouraging particular investments. As governments in southern Asia have changed their economic strategy one might therefore expect that fiscal policy would play a major role in achieving and reflecting that change. A 15 year study of development strategy in five countries -·- Indonesia, Bangladesh, India, Pakistan and Sri Lanka - and its effect on growth and equity provides the framework for this paper. The data are drawn mainly from that study and several Ph.D. dissertations which were an important element in it. The current paper looks at the work on development strategy over 30-odd years from a public finance perspective. Four of the five countries studied experienced radical shifts in economic strategy ~t least once in their history since independence. The shift was from an economy that might be called "more·dirigiste" (or one that uses more direct physical controls) to nmore market orientedn or vice versa. In the fifth country, India, strategy was more stable and changes were more modest. The role of fiscal policy in these strategy changes is the subject of this paper. ....., . ,..,~f. -2- Since the distinction between the two strategies is central to the subsequent discussion, the nature of the distinction needs to be clarified. Quite obviously none of the five countries during any period came even close to being.either a pure market/private enterprise or a dirigiste/centrally planned economy. All were mixed economies throughout their history since Independence. But there were quite clear and distinct episodes when they tended more towards one en.d or the other of the spectrum. There were two major policy packages which characterized the distinction. The dirigiste governments tended to expand the range of publicly owned enterprises and to extend the scope of various physical controls: licenses for imports, for investments, for exports; rationing for basic foodgrains and some other consumer goods; extensive controls over the prices of goods which were deemed essential to either consumers or producers of other goods; prices were generally set below the market equilibrium, thus requiring control over allocation. The more market oriented regimes maintained, or slightly reduced, the role of publicly owned enterprises, rather than expanding it. More important, during the market oriented periods the role of controls and physical allocation was reduced, while prices were given a greater role in the allocation process. The change in strategies occurred in nearly all cases as the result of a change in political regime and therefore was quite clearcut and pervasive (from SLFP to UNP in Sri Lanka, from Soekarno to SuHarto in Indonesia, from Ayub Khan to Bhutto and Mujibur Rahman in Pakistan ~nd Bangladesh and then to the two Zias). The distinction between the two strategies is really quite clear-cut if one compares strategies over time in four of the five countries under review, but no good index exists for measuring the extent of difference. One reasonably satisfactory quantitative measure for distinguishing the two -3- strategies was the share of public in total investment. It measures directly the role assigned the public enterprises in growth, one of the characteJ:istics which distinguishes the strategies and seems to be a reasonably good ..P~Coxy for the other (for further discussion see Papanek'. 1986). India'~ strates;y remained largely unchanged, and dirigiste, during the entire period stnce independence, but is slightly' more market oriented since 1978, and especially in the 1980s. The 1980s have therefore been identified-as .market oritented, although this identification ls by no means as c1ear-cut as in the other countries. Table 1 shows the growth rate for the five countries over 30 years and distinguishes the periods when governments were following a dirigisted from those when they were following a market oriented strategy. The pattern which emerges is quite clear-cut: the rate of growth of the economy was higher in the market oriented periods. The reasons for this difference and its consequences for income distribution are discussed elsewhere (ibid). This paper is concerned with the relationship to fiscal policy. Comparative Analysis -- Similar Economies. Different Policies Comparative analysis of these five countries can provide useful insights because they are relatively similar. Of course all countries and time periods are unique, but the five analyzed share a number of characteristics. They are all low income and labor abundant. Four share a colonial history, three were a .single country for over a century. These similarities make it easier to isolate and analyze the consequences of development strategy and of fiscal policy. Moreover~ when four of the countries made radical shifts in planned strategy, other circumstances within the same country, of course, remained basically unaltered, especially when one compares averages for 3-10 year -3A- Table 1 Gro~th Rates Under Dirigiste and Ma~ket Oriented Governments (Annual Compounded Percent Rate of Growth in GDP) Country '50s '60's Early 70's Late '70s 1981-83 India 4.2 3.4 3.5 3.8 4.9* Pakistan 3.6 6.7* 3.9 5.2* 6.3* Bangladesh 1.7 3.7* -0.2 4.7* -1.2* '60-'67 '68-'71 '72-'80 1981-83 Indonesia 0.5 9.5* 7.9* 4.9* '50-'60 '61-'65 '66-'ZO 'Z0-'77 '78-'83 Sri Lanka 3.1 3.8 5.7* 3.1 7.7* Source: Calculated from various official sources, except for 1950s Pakist8.n and Bangladesh, which are from Papanek (1967). *- Market oriented strategy followed. -4- periods to control for short-term fluctuations in weather or similar factors. ]n those cases one can analyze very effectively the consequences - c()·i;jt"s and benefits -- of shifting from a dirigiste regime of direct physical controls, with little use of prices to allocate resources e1nd provide incentives, to a more market oriented economy with planners manipulating prices rather than, quantities. Comparative analysis can be a powerful tool in isolating causal factors, and distinguishing them from correlations due to extraneous factors. It is particularly powerful if the social, politicm Table 11 are that the tax system in Pakistanr as in Sri Lanka, appears to be quite progressive and that the impact on income distribution in 1971 and 1978 was quite :Similar. 1971 was the year when the dirigiste. Bhutto goverrunet1.t came to power after <1. decade of a market oriented government from 1959 to 1970, while 1978 was a bit more than a year after another market oriented government took over, after 5 years o£ dirigiste and populist policies. Yet. if the data are to be believed; the overall impact of the tax system on income distribution was essentially the same in -42- the two years. Even with respect to its distributive consequences the tax system did not seem to change significantly with a radical change in economic strategy. c. For India the work on tax incidence is more readily available, unlike the unpublished Ph.D. dissertations in which the analyses of Sri Lanka and Pakistan were reported. Therefore the India work will not be discussed here. Four points are made by Acharya that are worth summarizing. First, there has been a sharp decline in the role of direct taxes and especially in personal income taxes in total tax revenue (from 1950/51 to 1984/5 direct taxes declined from 37% to 15%, income taxes from 21% to 5%). Since direct taxes, and especially the income tax, are usually a principal element in progressivity this development is of concern from a distributional perspective. Second, the recent abolition of the estate tax eliminates a further element of progressivity. However, the amounts collected were always small so the impact will be minimal. It is likely that it affected only the very wealthy, so may have some effect on their rate of taxation. Third, at least one study shows that indirect taxes were progressive, so that the overall incidence.has also remained progressive. Fourth, a later analysis which takes account of indirect effects of taxes/subsidies on inputs through inter-industry linkages finds much less progressivity for most taxes. For urban areas the tax burden is essentially uniform. d. For Indonesia Gillis also reports that non-oil taxes are progressive. But as one can see from Table 12, before the 1983 tax reform, the degree of progressivity was quite limited compared to other countries. No study of incidence subsequent to reform appears to exist. Indeed it would be difficult -42A- Table 12 Incidence of Non-Oil Taxes in Indonesia -- 1980 (income in Rp 000, per capita per month) Income group Incidence (%) Income group Incidence (%) 1 0-3 4.5 8.01-10 7.2 3.01-4 4. 7' 10.01-15 7.7 4.01-5 5.0 15,.01-20 9.0 5.01-6 5.8 20.01-30 9.3 6.01-7 6.3 30.01 + 13.6 7.01-8 7.1 Source: Gillis -43- to have completed such a study, since the last steps of the reform were phased in less than 2 years ago. In Indonesia; even more than in other countries, the usefulness of incidence studies is limited, since no meaningful analysis can be carried out of the largest source of government revenue, the taxes on oil and gas. The "incidence" of those taxes is on foreign stockholders, foreign tax collectors who provide tax credits and the owner of the resource, which is Indonesia collectively'through its government. Gillis points out that the Indonesian system was highly progressive in law, but much less so in actual practice. The actual average rate of income tax paid by the richest 5% was only 4% when the legal rate was 50%. A large part of the explanation was in various incentives and allowances, another part in evasion in a complex system administered by a civil service that had become very vulnerable during a period of abysmally low wages. Commodity taxes were also highly differentiated and progressive in law. Smuggling is always an especially serious problem in a nation of many islands like Indonesia and compounded the problems of complexity and enforcement that also plagued the direct taxes. e. Conclusions which compare the data from all the countries are difficult, for the reason mentioned earlier: incidence studies are only as good as the assumptions made -- many of which inevitably have an arbitrary element -- and the data available, inevitably flawed. Different analysts make different assumptions, about shifting of commodity taxes for instance, and s~me studies are not based on actual data on the income taxes paid by different ~ncome groupst but have to assume that there is close correspondence between legal and actual rates. Nevertheless, some comparative conclusions seem possible. All of the studies discussed above were carefully done and all included sensitivity analysis. So the conclusion reached by all of them that the tax -44- system was progressive in fact, as well as in design, is likely to reflect reality. Especially .in view of the fact that tax systems in all five countries relied heavily on commodity taxes that is an interesting and encouraging conclusion. Second, there appears to be no clear relationship between the economic strategy adopted and the progressivity of the tax system. In Sri Lanka and Pakistan observations were available for different years and under different strategies, and there is no evidence that the dirigiste governments achieved a more progressive system of taxation than the market oriented ones in the same country. The evidence from Sri Lanka is more useful in this respect than that for Pakistan, since in the case of the former both years were analyzed by the same person, using the same methodology and assumptions. But even the Pakistan comparison is useful because the approaches used by different analysts were quite similar. Evidence from cross-country comparisons are much less useful, because of the difficulty of comparing studies based on surveys of income that are often done quite differently. compounded by different incidence assumptions and the fact that income groups in the studies are quite disparate in size, But for what they at'e worth it appears that dirigiste India and market oriented Indonesia in the late 1970s both had a modestly progressive system. Pakistan and Sri Lanka had somewhat more progressive tax systems under both dirigiste and market oriented governments. Finally, the extent to which the systems are seen to be progressive seems to depend partly, on the structure of the economy, partly on the ability to collect substantial income taxes and partly on incidence assumptions. The substantial progressivity in Sri Lanka and Pakistan stemmed in part from their ability to collect en.ough personal income taxes to make this a major item of revenueJ while the systems in India and Indonesia seem to be less steeply -45- progressive in part because personal income taxes were such a small part of their tax revenue. Another large part of the progressive nature of the system in Sri Lanka was due to export taxes, whose cost impinged on the wealthy, while the implicit subsidy for the domestic consumer of exportables benefitted the poor disproportionately. Part of the progress~vity of the Pakistan system is due to the assumption that the incidence of import duties was on the importer, not the consumer. It could be that the Indian studies made a different assumption. Some of the points in the last paragraph are speculative. Equally uncertain is any firm statement on the country with the most progressive system, for the reasons cited earlier on the difficulty of cross country comparisons. But from an examination of e~fective tax rates for different groups it does seem that Sri Lanka had the most progressive system by a substantial margin. The reasons why this might be so are worth considering after examining the incidence of government expenditures, the other element in fiscal incidence. 2. Expenditure Incidence Analysis of expenditure incidence is even more difficult and the results more dubious than for tax incidence. As a result less has been done in this field. With taxes, and the usual household expenditure surveys, at least initial ineidence is usually known for most taxes. Only the shifting assumptions are often quite dubious. But little or nothing is usually known on the income of the households who benefit from government programs -- whose children go to public schools, who uses public hospitals, road transport $ systems and so on. Often it is even uncertain which income groups benefit from subsidized fertilizer, water, pesticides and agricultural extension -46- services, or even who consumes subsidized food. Expenditure incidence therefore usually has to be substantially·based on relatively arbitrary assumptions. The studies discussed below, with one exception -- the -analysis of Sind, Fakistan-- suffer from all these problems. a. Sri La~ka is widely regarded as the leading welfare state in the less developed world. All social indicators suggest that these measures have been successful in extending the life span, reducing infapt mo~tality and extending education and health facilities to most of the population.. But much less is known on the incidence of social and welfare measures and how that has changed as economic strategy has changed. Jayasundera, using the same household expenditure surveys for data as in his study of tax incidence~ has examined these questions. Unlike the case of taxes, there is a clear relationship between changes in economic strategy and changes in the rate of government social and welfare spending. The dirigiste goyerrunents increased, or at least maintained these expenditures, the market oriented reduced them as a group and especially cut back on food subsidies (Table 13). Since 1983, when the data presented end, the rate of food subsidies as percent of GDP has declined further. Using the income approach, in which benefits to recipients are equated wi·.:h costs to government, and based on the actual consumption of thll subsidized goods, Jayasundera allocates the benefits of food subsidies to different income groups. The 1978 survey used for this purpose also indicates whether children are in school and provides limited evidence on the health facilities used. It therefore permits the allocation of education and health expenditures by income group.s. These data are shown in Table 14. -46A- Table 13 Social and Welfare Ex~enditur~s and Economic Strateg~ in Sri Lanka (as percent of GDP) 1951-55* 1956-60 1961-65 1966-70* 1971-77 1978-83* Food subsidy 2.4 2.1 3.7 2.7 3.9 2.8 Education expen. n.a 3.3 4.0 3.6 2.8 2.1 Health expenditure n.a 2.0 2.0 1.8 1.4 1.1 Direct transfers 0.5 0.7 0.6 0.5 0.7 1.3 Other social serv 0.1 0.2 0.1 0.1 Total 7.1 8.1 10.5 9.0 9.1 7.6 Source: Jayasundera Notes: In the 1978-83 period "food subsidy" includes the implicit subsidy in food stamps. (* -market oriented strategy) Direct transfers -welfare payments, school lunches, kerosene stamps, etc. -46B- Table 14 Incidence of Social and Welfare Expenditures. Sri Lanka (as percent of income -- income groups identical with earlier tables) Income groups 1 2 3 4 5 6 7 8 9 Food subsidy 1973 3~3 2.4 5.2 5.2 3.2 0.9 0.1 -0.7 1978 6.2 10~2 10.6 9.3 7.5 4.6 2.3 1.3 0.3 Education-1978 6.0 3.7 3.6 3.4 3.0 2.2 1.5 1.1 0.4 Health-1978 4.4 3.9 2.5 1.6 1.5 0.9 0.4 0.3 0.1 Total 1978 16.6 17.8 16.7 14.3 12.0 7.7 4.5 2.7 0.8 Population 1973 0.3 2.5 21.7 49.0 21.0 2.1 1.5 1.0 1978 2.5 11.9 21.0 19.7 14.2 17.3 6.3 3.8 3.4 Source: Jayasundera -47- It appears that the social and welfare expenditures benefit the poor disproportionately, with the exception of the poorest group, but the latter conclusion is probably misleading. Both health and education expenditures in 1978 redistribute income to the poorer across the entire income spectrum. It is only with respect to food subsidies that the two lowest income groups appear to benefit less in both 1973 and 1978 than those slightly better off. But this is based on their reported consumption of rice and wheat flour, foods with a major subsidy. The household expenditure surveys actually show them consuming less of these foods than the subsidized ration to which they were entitled (Sri Lanka provided a limited amount of rice free to all participants in the ration system, which covered nearly the whole population, for much of its post-Independence period). What this indicates is that the poorest almost certainly obtained the full ration to which they were entitled and then sold some part of it to purchase other goods they needed even more. While this can . :) not be proved, it is highly logical and consistent with anecdotal evidence. That would mean that food subsidies provided even greater benefits to the poorest than to those with slightly higher incomes, since the base to which the benefits are compared is, by definition, lower. In that case the government expenditures on social and welfare services would redistribute very substantial income to the poor, adding as much as 20% to the income of the poorest groups. If the poor derive about 17-20% from these government expenditures. while the wealthiest income is increased by only 1-3%, then the reduc;:tion after 1977.of food subsidies and health and education expenditures would have affected the poor disproportionately. This is especially true bepause the reduction was concentrated in food subsidies, the most important source of additional income for the poor. The effect on the poor can be confirmed from -48- Table 15 which covers 1979, the first year in which the expenditures policies adopted after 1977 became effective. In order to analyze the effects of the cut in welfare expenditures after 1977 one can, unfortunately, not use the 1978 data as an indication of what happened. Food stamps were introduced only in 1979. The real value of these stamps was then allowed to erode with inflation. As a percent of GDP food subsidies were 4.8% in 1978, the year on which much of the analysis is based, but were down to 1.1% of GDP by 1983, and have continued to decline in value since. The data in Table 15 provide a better indication of the impact late in the yearJ although subsidies for the year as a whole were still high. By 1983 the value of these subsidies has shrunk to about a fifth (,f what it was then, so the impact on the very poorest was considerable. Clearly Sri Lanka developed a system of government social and welfare expenditures that transferred significant resources to the poor in the 30 years since independence. But the market oriented gove.rnments were less committed to these expenditures than the dirigiste ones and the poor did less well in this respect during the periods when they were in power. b. For Sind Province of Pakistan Husain has what may well be the most careful study of benefit incidence done in any less developed country. l'ure public goodsl by definition, had to be allocated on an arbitrary basis 1 but for most specific public goods he could base the allocation on the actual use by households of particular public services or the purchase of specific subsidized goods. The pure public goods (such as defense, general administration, the judiciary and police) are ignored in the further discussion because their allocation to income groups is essentially arbitrary and to include them adds little real information. -48A- Table 15 Per Capita Food Subsidy as % of Per Capita Income. Sri Lanka 1963 1973 1979 Income group (1 month) Below Rs. 100 38.8 40.1 21.2 101-400 14.6 17.9 9.8 401-800 3.3 4.6 1.2 Over 800 3.1 3.2 Source: Jayasundra Note: Incomes in 1963 prices -49- The effective benefit rates for specific public goods (Table 16) are remarkably pro-poor, especially in the urban area and especially for the poorest groups in the population. The extremely high rate for the very poorest group should really be ignored and even the estimate for the next group is probably t~o high. fo.r the reason discussed earlier: the inclusion of some retired people whose reported !~come is far below their reported expenditures on which the benefits calculation was based•· But even after such adjustment it would be reasonable to estimate that the po.or receive speci:f'ic benefits about equal to 15% of their income, a considerable sum. On the other hand the percentage of benefits for the richest are overstated, because of income under-reporting, especially by the urban rich. The true benefit rate may be around 2%. It is obviously important to determine what specific expenditures provide the progressive element. This is discussed below (see Table 19). For the urban areas the wheat flour subsidy is an important element. Other studies have shown that the flour sold through the ration shops is of such low quality that it is bought primarily by the poor, who can not afford higher quality foods which cost far more per calorie. By subsidizing an inferior good, the government achieved automatic targeting, with relatively little leakage. With respect t:o health care automatic targeting is again £tchieved, because the public health facilities are largely used by the poor. The wealthier households tend to use higher quality private doctors and to purchase drugs in the market 1 where there is less tro·~).ble and delay . .. Publicly supplied drinking water and sewerage (where available) in the rural areas are assumed to be largely used proportionately to the number of persons per household. These are generally communal facilities (a village well usually) 1 so the proportionality assumption is quite reasonable. If -1+9A- Table 16 ~ffective Benefit Rates for Specific Public Goods. Sind Province. Pakist~ ....,__ Income Groups Benefits as % of Income % of population Urban Rural All Sind Urban Rural 1. 1 - 200 (50.8) (37.5) (42.9) 6.5 11.0 2. 201 - 400 (18.3) 14.1 16.2 24.0 20.5 3. 401 - 600 12.3 11.6 11.9 17.5 18.3 4. 601 - 800 8.6 10.0 9.5 12.3 13.9 5. 801 - 1000 8.8 8.2 8.4 7.1 7.8 6. 1001 - 1500 7.4 7.1 7.3 12.7 10.9 7. a.SOl- 2000 5.0 4.5 4.7 6.3 7.4 8. 2000- 2.8 3.1 3.0 13.3 9.8 Source: Husain Notes: The income base for the calculation is adjusted income, including fringe benefits (pel:ks). Figures in parenthes1es are of dubious reliability, for reasons discussed in the text. -so- wealthy households had their own individual water supply, they were not cre.dited with benefits. from go,.;:a:s:mnent expenditures on drinking water. For public supply households of the same size are then credited with the same absolute benefits. Since the income base with which this benefit is compared is obviously much smaller for the poor households, this· gives an automatically . pro-poor distribution to these expenditures, on perfectly reasonable assumptions. (Note that other studies of expenditure incidence generally have no information on who benefits from public water supplies and have to use quite atbitray assumptions). The net egalitarian impac·t of goverrunent expenditure can be seen from Table 17. This again overstates the case, since the data include the unbelievably high incidence on the lowest income group, which is in very large part statistical artifact. (To a lesser extent this is also true of the second poorest group. The overstatment is offset i-,.-.. ,part by the overstatement of the benefits to the richest urban group, because of their under-reporting of income.) But ev:en ignoril.1g the poorest group and discounting the benefits to the next group, there is little doubt that on the whole the pattern of government expenditure is pro-poor. c. For Indonesia and India there are tlO studies of government expenditure incidence. However, both countriest as well as Bangladesh have major programs to provide employment and other benefits specifically targeted on the poor~ India has three major programs, all concerned with rural poverty and employment. Evaluations suggest that participants are the very poor (see Acharya for summary on IRDP), but that the impact in raising income has been quite smal~. The other tw-o countries have major employment programs, with compensation set at a level designed. to attract primarily or exclusively the poor. (This -SOA- Table 17 Income Distribution and Benefits Incidence in Sind Province, Pakistan (Gini coefficient) Pre-Fisc Post-expenditure % change Urban 0.522 0.471 -9.8% Rural 0.523 0.495 -6.3 All Province 0.505 0.482 -4.6 Source: Husain -51- is analogous to the self-policing feature of subsidizing inferior goods.) Evaluations indicate considerable success in providing both employment and developing infrastructure (irrigation, drainage, rc1ads in both countries and schools in Indonesia) , but do not quantify the incctme effects. The neglect of studies of benefits incidence, despite the importar.Lce of the subject, reflects the difficulty of carrying out such work. d. Comparative analysis and general conclusions are obviously difficult when observations are available on only two countries. Fortunately for Sri Lanka there are 2-3 observations and for Sind Province in Pakistan the coverage of benefits is more extensive and more careful than ir1 other studies. The first and most important conclusion is that in both countries the government expenditure pattern was significantly pro-poor, despite all the problems -- administrative and political -- such a policy usually faces. Even if one ignores some dubious data (the poorest group in Pakistan, with its understated income, most notably) both countries managed to transfer of the order of 16-18% of additional income to large groups of the poor through the expenditure system. The Sri Lankan estimates ignore any transfers via subsidies on agricultural inputs, on drinking water, sewage and miscellaneous items, and understate the food subsidies to the poorest through their sale of rationed food. It is therefore quite reasonable to estimate the transfer in that country at around 20% for the poorest third of the population. In Pakistan a small downward, rather than an upward adjustment, in the data are warranted even if one ignores the bottom group, because the estimates assume that th~ quantity and quality of some services received by the poor equals that for the wealthier. That assumption is not always justified. For instance, it is likely that the rich receive a higher quality of health and educational services per period of time spent in hospitals or schools than the -52- poor. But note that this study of expenditure incidence is far less plagued by arbitrary assumptions than most of the few other studies carried out, since it is based on a careful survey of actual use of government services. Even if one discounts for possible bias, transfers of 10% or more of income seem to have been effected in Pakistan. Second, a principal vehicle for these transfers was via subsidized basic food grains. A major point here is that Pakistan was able to apply a general subsidy (although for the urban areas only), yet keep leakage down by subsidizing an inferior good, which the rich did not consume in significant amount (and which remained too expensive to use for animal food). Apparently some of the same purpose was achieved in Sri Lanka in three ways: the rice distributed much of the time was a less favored variety, wheat flour was added to the ration in the 1970s and was widely considered less desirable .and the poorest seem to have sold some of their ration and with the proceeds may have bought foods that provided cheaper calories. Significant pro-poor expenditures were also achieved by expanding primary education and by parts of the public health system, not patronized extensively by the rich. On the other hand, expenditures on higher education and, most probably, on hospitals had a bias towards the wealthier. Third, the market oriented governments in Sri Lanka cut government social and welfare expenditures, quite sharply after 1977. Since the biggest cut came in food subsidies this greatly reduced the benefits to the poor and had serious effects on some of them. There was no necessary reason why a sharp cut in. food subsidies should hurt the poor. A substantial proportion of the total subsidy benefitted the rich, since everyone was entitled to an equal absolute amount of (free and) subsidized food. Indeed the scheme evolved by government in 1977 was precisely designed to protect the poor. Food stamps, --53- to be limited to lower income groups, were substituted for a general subsidy. However, the system proved difficult to police and was subject to political pressure, with the result that about half th~ population received food stamps. The benefits to the poor had to be cut, in order to pay for the much larger number of beneficiaries. When there was a general budget crunch the benefits were reduced further, since it proved political~y impossibl~ to adopt the more desirable approach of reducing the number of beneficiarie's instead. The essential requirement for an effecfive.pro-poor pattern of government expenditure is to find a politically and administratively feasible method of targeting. Pakistan achieved a substantial measure of.this in some programs by providing inferior gooJs and services which were available to all, but used mostly by the poor. In Bangladesh and in Indonesia in the 1960s, when poverty was much greater and the middle class was economically hard pressed, this approach did not work -well: much of the population at most income levels obtained the subsidized food and either sold it, used it to pay their servants, consumed it, or some combination of the. three (of course, when foods are very heavily subsidized there is the additional possibility of using them as livestock feed, apparently a problem in Egypt 't>Tith subsidized bread at one time). In Indonesia it has been proposed that production of a distinctly inferior food, (wet) cassava, be heavily subsidized, as an effective means for subsidizing the consumpcion of the poor with no need for a cumbersome administrative apparatus or the problem of political pressure to extend the benefits to ever wider groups of the population (Falcon, et al). Sri Lanka adopted an alternative approach, of quite deliberate and careful targeting of the poor. This was, in theory, a preferable system, since it could eliminate leakages. But in practice administrative and political problems proved insurmountable. The result was that the very poor were probably worse off. -54- It should be noted, however, that long before the change was made the total welfare system could no longer be supported by a stagnant economy, and had been cut even by the preceding dirigiste government. Fourth, another means for targeting government expenditures is an employment/public works program, with wages set to assure that only the poor participate. This has two further advantages: it can be used to increase the rate of investment, especially in infrastructure, and it provides employment rather than welfare payments, desirable for many reasons. Most notably it can help raise income for the poor throughout the economy, not just for those employed in the program (Papanek, 1986). This approach was used in Indonesia since the 1970s, in Pakistan in the 19·SOs, Bangladesh for much of the last 25 years-and in India in the last decade. But the real impact on poverty is only beginning to be measured. But it is important that programs to transfer resources to the poor through government expenditures have been effective to a significant degree in at least some of these countries and for some periods. There is some evidence that dirigiste governments were more effective at achieving such transfers, one of the few clear qifferences in the impact of the two strategies on the fiscal system. 3. Fiscal Incidence; The effect of taxes and expenditures. For the two countries where information on both tax and expenditure incidence exists, both the tax and expenditure systems were pro-poor, so their combined effect should also be progressive. Indeed so they were. a. For Pakistan one can trace in some detail what aspects of the system made for progessivityt with implications for egalitarian fiscal policy in other countries. The net benefits or costs of the system were quite different for -55- the same income group~ in rural and urban areas, but they were pro-poor in both and the system as a whole had a distinctly egalitarian impact. These data suggest that the system is more pro-poor than it is in reality: inco!Xle for poorest group is greatly, and for the next poorest somewhat, understated and ·net tran'sfers as a percent of income overstated, because of retirees. This can be seen more clearly in the details of table 19 below·. The data for the poorest group are best !gnored therefore. the arbitrary allocation to the poor of benefits from pure public goods makes the positive balance appear more favorable than it actually is. Table 19 excludes them and therefore provide a more accurate picture. income of the wealthiest is clearly understated. the poor are presumed to benefit proportionately from expenditures on drinking water and sewerage in their village, which means that in relation to income they benefit disporportionately. The effect of that can also be seen in Table 19. If one ignores data on the poorest group as hopelessly unreliable because it includes too many who belong among the upp~r income groups, and discounts results for the second poorest for the same reason, then one can draw some useful conclusions from these diaggragated data on the instruments which have sigr.ificantly affected income distribution: (i) in the urban areas despite tax evasion and avoidance, the personal and corporate income tax was the most significant vehicle for redistribution -56- next in importance and about equally significant were taxes on luxury goods, on intermediate products (including yarn and cloth, and import duties on metals and metal products), and a miscellany of other taxes, incb.!ding property on the expenditure side the three major items were education, health and the subsidy on wheat flour. More money is spent on the education of the rich, but since it is a smaller proportion of their income, the incidence is still pro-poor. (ii) in the rural areas the direct taxes are regressive 1 because agricultural income is not taxed, so a major element in the progressivity of the urban system is missing. The wealthy, mostly large landlords, pay no income tax, whose burden in mainly borne by the few middle income wage earners. - again, unlike in urban areas, luxury goods taxation did not add a strong progressive element, because of low consumption of these goods as in the urban areas, taxation of intermediate goods and other taxes were strongly progressive -- on the expenditure side education, health and drinking water show up as very pro-poor, the last in part because all users in a village are assumed to benefit equally agricultural subsidies benefit the middle income groups disproportionately; the poor do not have enough land (if any) and the rich probably do not earn their income primarily from farming (iii) in both areas taxes on alcohol, tobacco and beverages (soft drinks) are mixed, while the sales/consumption tax is :reg-,ressive. -56A- Table 18 Combined Tax and Government Expenditure Incidence. $ind Province, Pakistan (1978, percentages of broad incbme) Income Effective Fiscal Rates Ordinal Shares of Income Group Urban Rural All Sind Decile Pre-Fisc Post-Fisc % change 1 - 200 22.2 23.5 22.9 1st 1.2 1.7 +41.6 201 - 400 12.7 17.4 15.2 2nd 2.5 3.1 +24.0 401 - 600 13.3 15.4 14.5 3rd 3.3 3.5 + 6.0 601 - 800 7.7 13.0 10.9 4th 4.2 5.5 +30.9 801 -1000 2.4 13.6 9.3 5th 5.2 5.9 +13.4 1001 -1500 -1.0 10.2 5.2 6th 6.6 7.3 +10.6 1501 -2000 -36.9 5.8 -7.7 7th 8.4 8.7 + 3.5 1501 -2000 -18.4 5.0 -6.6 8th 11.4 12.8 +12.2 All Classes -8.7 9.5 2.0 9th 17.0 16.3 - 4.2 All Classes lOth 40.0 35.0 -12.5 Source: Husain -56B- Table 19 Fiscal Incidence by Income Groui2S 1 Sind 1 Pakistan - as ~ of income Urban Areas - TAXES Income Groups Direct Luxury Intmd Alcohol Consump- Other Total Taxes Goods Goods Tobacco tion Taxes Beverages 1. 1 ..... 200 8.2 7.1 0.6 3.2 13.0 32.2 2. 201 - 400 3.4 3.3 0.3 4.1 3.3 0.6 15.0 3. 401 - 600 1.7 1.2 0.1 3.5 3.1 0.5 10.1 4. 601 - 800 2.7 1.6 0.1 2.3 3.5 0.6 10.7 5. 801 - 1000 4.8 6.4 0.1 3.8 2.9 0.6 18.7 6. 1001 - 1500 6.9 3.9 3.4 2.1 1.6 2.9 .20.8 7. 1501 - 2000 18.6 5.9 8.5 4.7 2.2 6.8 46.7 8. 2000- 12.6 5.4 6.4 2.9 0.8 5.1 33.3 Urban Areas - GOVERNMENT EXPENDITURES Income Groups Wheat Subsidy Water/Sewer Education Health Total 1. 1 - 200 28.3 2.9 8.1 6.4 45.7 2. 201 - 400 4.5 1.5 6.,1 4.5 16.7 3. 401 - 600 2.6 1.0 4.4 3.8 11.7 4. 601 - BOO 3.3 0.7 1.6 2.3 7.8 5. 801 - 1000 2.6 0.5 2.2 1.9 7.1 6. 1001 - 1500 2.1 0.4 2.9 1.3 6.6 7. 1501 - 2000 2.0 0.3 0.9 0.7 3.9 8. 2000- 0.5 0.1 1.1 0.4 2.1 -56C- Table 19 Fiscal Incidence by Income Groups. Sind. Pakistan- as % of income Continued Rural Areas - TAXES Income Groups Direct Luxury Intmd Alcohol Gonsump- Other Total Taxes Goods Goods Tobacco tion Taxes Beverages 1~ 1- 200 3.1 2.2 4.3 6.8 0.3 16.7 2. 201 .... 400 1.4 0.5 1,6 1.7 0.3 5.5 3I 401- 600 1.4 0.4 2.4 1.0 0.5 5.6 4. 601 - 800 1.4 0.3 2.8 1,5 0 ,:> r 6.6 5. 801 - 1000 1.5 0.2 1.3 0.8 0.9. 4.7 6. 1001 - 1500 1~5 0.7 1.7 1.2 1.0 2.0 8.1 7. 1501 - 2000 0.9 0.9 3.3 2.1 0.8 3.6 11.7 8. 2000- 0.8 1.4 3.8 0.9 0.3 3.2 10.4 Rural Areas - GOVERNMENT EXPENDITURES Income Groups Wheat Subsidy Water/Sewer Education Health Total 1. 1- 200 9.7 1.4 10.5 9.3 30.8 2. 201 - 400 3.4 1.2 3.9 4.3 12.8 3. 401 - 600 2.1 1.2 3.8 3.1 10.1 4. 601- 800 1.5 2.0 3.2 1.7 8.4 5. 801 - 1000 1.2 1.9 1.5 2.0 6.5 6. 1001 - 1500 0.8 1.7 1.8 1.3 5.7 7. 1501 - 2000 0.6 1.0 0.6 0.8 3.1 8. 2000- 0.3 0.9 0.7 0.4 2.3 Source: Husain -57- In short, that the system is progressive depends l1eavily on the progressive structure of SQme commodity taxes, with food exempt, luxury goods heavily taxed and taxes on some widely consumed goods rising with quality; plus the impact of income taxes and the flour subsidy in the urban areas. b. Sri Lanka, was substantially more succesful than Pakistan, and probably than all other less developed countries, in transferring significant resources from the wealthy to the poor. The principal tools were income and export taxes, combined with food subsidies. Interestingly enough the principal beneficiaries of the system were the rural poor and those on plantations (estates). Their expenditure patterns were such that they were somewhat less heavily burdened by commodity taxes (especially excise taxes), but they still benefitted from food subsidies and from widely distributed social services. The rich of the same rural groups were also the largest net tax payers. They bore a heavier burden from import and export taxes. Sri Lanka is thus unusual not only in its success in transferring considerable resources from rich to poor through the fiscal system, but also in taxing the rich in the rural areas more heavily than in the urban areas. c. . . is not straightforward. Comparing both countries First, to make Pakistan data comparable with those for Sri Lanka one must exclude from the former pure public goods and a whole host of other specific goods and services provided and subsidized by goverrunent, most notably irrigation water, fertilizer, i .. nsecticides, transport, drinking water and sewerage. Second while the be~efits for the lowest income group in Sri Lanka are understated, they are overstated for Pakistan, for reasons discussed above. Third, given the political environment in the two countries in 1978, the rich had more reason to grossly understate their income in Pakistan. Taking all these factors into account one can say with some degree of confidence that the Sri Lankan system -57 A- Table 20 Combined Tax and Social/Welfare Expenditure Incidence. Sri Lanka 1978 (% - plus indicates income gain) Income grp 1 2 3 4 5 6 7 8 9 • Sri Lanka 12.0 12.2 9.8 5.7 1.5 -8.3 -13.5 -19.2 --25.1 Urban 2.5 1.6 -0.4 -1.6 -4.4 -7.7 -8.5 -10.0 -9.2 Rural 6.9 6.8 4.9 2.8 -4.6 -11.6 -15.5 -24.7 -28.5 Estate 14.1 10.2 11.5 13.4 7.9 1.6 -12.3 -22.0 -23.8 Memo-population in each income group (%) Sri Lanka 2.5 11.9 21.0 19.7 14.2 17.3 6.3 3.8 3.4 Source: Jayasundera -58- incorporates substanti,ally more redistributive elements. But even us~ng the broadest definition of income employed by Husain, and taking account of the caveats mentioned there is no dou.bt that the fiscal system in Pakistan (or at least Sind Province) redistributed income to the poor as well. That both systems were significantly pro-poor in total impact is surprising in view of the implicit assumption of governments in the region that the fiscal system was not a very effective tool of government policy. But it should also be noted that the incidence of the system could change quite readily, and probably already changed to a degree in Sri Lanka, if economic policies often advocated by international agencies are carried out: eliminate subsidies, especially on food 1 but also on urban transport. Ort the other hand, expansion of employment programs, more effective income tax collection and strengthening the progressive nature of commodity taxation can increase the progressive nature of the system in all the countties under study. d. The reasons why the Sri Lankan system has been so highly progressive, and why it has probably become considerably less so in the last few years, carry· useful lessons for other countries. Three factors explain the egalitarian nature of the Sri Lankan fiscal system: (i) Sri Lanka is the only country in Asia, and one of the few in the world, where there has been a democratic competition between two rather evenly matched political parties, which have tended to alternate in office. As a result, both parties had to pay careful attention to the demands of the majority of the population, a majority which, in Sri Lanka as els•a't'lhere, was relatively poor and demanded access ta1 food, education and health facilities. The elite could not give principal weight to its own needs and objectives without losing office; -59- (i.i) at Independence the economy was a relatively egalitarian one, with no large landlords controlling hundreds and even thousands of acres, or wealthy businessmen, as there were in other countries of the region. In the absence of highly concentrated economic power 9 egalitarian policies were easier to implement; (iii) much of the financi.ng for the welfare state came from squeezing the plantation sector, much of which was initially owned by foreigners and which was substantially staffed by non-voters (Indian Tamils), neither of them politically powerful. In the absence of these characteristics, not duplicated in many countries, it is not surprising that Sri Lanka is practically unique -- in the absence of revolutionary upheaval -- in its consistent ability to transfer substantial resources from rich to poor through the fiscal system. e. On the relationship between fiscal incidence and economic strategy it is not possible to draw any firm conclusions on the basis of the evidence so far provided. Comparisons over time are limited to Sri Lanka and cover only two years. As far as one can tell the system was about equally pro-poor in both years. But one year fell in a period when a dirigiste government was in power and the other came too soon after a market orient~d one had taken over to see the full conseqvenc of its economic policies. The one observation from Pakistan similarly came in the transition period between goverrunents with {. different strategies. But substantial indirect evidence was provided that the market oriented governments in Sri Lanka had reduced food Sl~bsidies ·' the central element in a pro-poor expenditure system, suggesting that fiscal policy had became less egalitarian. Financing the heavy Sri Lankan welfare measures also had costs, especially in terms of slow growth. Rising food subsidies became increasingly difficult, and ultimately impossible, to finance. This led to some cuts even -60- under the dirigiste/populist government of 1970-77 and greater reductions subsequently. The impact of the fiscal system on savings and on growth is the next issue to be analysed. --61- D. The Fiscal System and Savings. The fisc.'.al system can affect growth directly, by encouraging or discouraging risk taking, innovation and effort and indirectly, both by encouragin~ - or discouraging savings and by influencing whether decisions are socially efficient or inefficient. A full ej{amin.ation of these issues is beyond the scope of this paper, but a limited discussion drawing on '..lnpublished material is worthwhile. (The data for t\lis secti.on are again largely drawn from Ph.D. dissertations and other work~which were p~rt of the overall study of development strategy and have largely remained unpublished until now). Moreover, in growth models fiscal policy is usually not included as a variable. Therefore this pape.r concentrates on the· impact via the effect on savings. The fiscal system quite obviously directly determines public savings and indirectly influences private savings. In 4 out of the 5 countries public savings have been small, have tended to decline and frequently been negative. Indonesia is an exception only because oil revenues largely accrue to the public e~chequer. With respect to the non-oil accounts the Indonesian fiscal system would show the largest negative public savings. Despite its small absolute role, changes in the rate· of public savings still matter. Since different factors influence public, as distinguished from private savings, analysis should be carried out separately. But a growth model by Niazi which sheds light on the relationship between some aspects of fiscal policy and savings does not distinguish private from public savings. 1~ Economic Strategy. Fiscal Policy and Savings in South Asia. Like other gr.owth models, Niazi's does not include public finance variables in the factors that influence the rate of growth of GDP. In his -62- model, which covers India, Pakistan and Bangladesh, he includes the usual variables: the stock of capital and the level of education. In addition he gives importance to the external sector: the ability to import and the terms of trade and finds that both are powerful factors. The sectoral pattern of investment matters, while the role of government in the economy is of some (negative) significance. Finally chang~s in economic strategy, structural variables (differences among countries) and exogenous factors (like the breakup of Pakistan) all matter a good deal. The typical neglect of public finance reflects its neglect in economic strategy discussed earlier and will be the subject of the next section. In his savings function public finance variables are of some importance however. Increasing the rate of taxation raises savings somewhat mor!t~ than proportionately and the relationship is statistically very significant (t- statistic of 3.8). (The conclusions reported here are based on the OLS version, but 2 Stages Least Squares gives essentially the same results). Public savings are not important in these countries most of the time and the marginal.propensity to save of government does not appear to be high. Nevertheless it is probably not zero, so higher taxes increase public savings (or decrease negative savings, which amounts to the same thing). But the more significant impact of taxes on savings probably comes via their effect on private savings. For one mark-up pricing is widespread. Therefore any increase in taxes tends to increase the profit margin. It therefore shifts resources from low saving consumers to higher saving business groups. Probab~y more important is the impact via the relative cost of consumption. Close to three quarters of taxes are on consumption. To the extent that they raise the cost of consumption they amount to an indirect subsidy on savings.: the more is saved the lower the effective rate of taxation. (In theory high -63- consumption taxes need not increase the attractiveness of saving, if consumers expect that when they dissave in the future they will still have to bear the .. same tax costs~ But in these countries there are a number of perfectly good reasons why this expectation is unlikely to prevail. Moreover several other assumptions are not met which are needed to negate this effect of consumption taxes. Indeed in the late 1950s and early 1960s when consumption of the goods desired by the wealthy was difficult and costly, industrialists reached savings rates of 70-90% of profits (Papanek, 1967). For all three reasons an increase in taxes resulted in a commensurate:increase in savings. Conversely, an increase in defense expenditures, one form of government expenditure little discussed so far, has a negative effect on savings. Roughly one third of defense costs seem to have been met out of savings. The relationship is not statistically significant, but is quite consistent in various formulations of the model. One reason for the absence of statistical significance ( t~···~tatistic 0. 7) may be that there normally is little variance in defense expendit~re in the same country over time. Major changes occurred at the time when there were also major upheavals in the economy, and these have been captured in a dummy variable (e.g. Pakistan greatly increased defense as a percent ut' GDP in 1970-72, and there is a separate dununy variable for the 1970-73 period). Sacondt there is a great deal of variance among countries in the level of def~nse expenditure, but there are separate variables for each country which may capture some of the significance of defense. So the relationship between defense and savings may be more significant than appears from the regression analysis, but this is in the realm of plausibility, not clear evidence. Another variable not usually thought of in connection with fisc.;al policy is also significant: the consumption of luxury goods, proxied by cars per -64- capita. The argument is that the wealthy, with a higher propensit;: to save, are influenced in their decision on their actual rate of savings b;•' i~he relative cost of consumption and the returns from saving/investment. The more readily available and less costly the consumer goods and services on which discretionary income is likely to be spent -- consumer durables, housing and festivities mostly-- the less their savings. There are externalities here as well. If consumption is easy and inexpensive it can foster an atmosphere of competing in conspicuous consumption, if it is costlyand the desired goods are difficult to acquire it may foster an atmosphere of austerity and competition in investment, that is in the expansion of business and industrial empires (Papanek, 1967). Taxation of luxury goods and services clearly affects their costs and influences the extent to which a climate of austerity or consumerism prevails. The level of luxury consumption, as measured by the number of cars per 10,000 people, turns out to be a statistically significant (t of 2.0) factor in savings. (Of course, it can be argued that to the extent the expenditure on luxury goods was for consumer durables and housing, it represented investment, and to class it as consumption is an arbitrary accounting decisions. The exact demarcation between consumpt.ion and savings is always a bit murky. But for policy purposes it is not unreasonable to include consumer durables under consumption. Whatever the definition, it is of some interest that greater access to luxury goods does reduce the resources available for other forms of investment, a .relationship that has not been previously examined, as far as I know.) 2. Fi§c~tl Policy. Growth and Saving$ in Sri Lanka. Jayasundera's analysis of growth and equity in Sri Lanka has some variables in the growth regressions clearly related to public finance, -65- although the clearest effect of fiscal policy on growth is again through its impact on savings. A major direct effect was via government expenditures on infrastucture and on insti·tutions which provided important externalities 1 investments which only government could make, and in reducing the risk and uncertainty of some innovative activities~ The clearest case is the rapid irJ.crease in rice production, especially after 1977, which was a central element in growth of the economy. According to Jayasundera•s analysis the rate of fertilizer application and of irrigation account for more than half of growth in rice production. In turn, irrigated area depended almost wholly on government expenditure on this activity, while fertilizer use was greatly affected by the rate of government subsidies and by government-influenced distribution. While most growth models ignore this aspect of fiscal policy it would be almost universally accepted that the extent of government effort on such infrastructure investments as irrigation, drainage and roads, and of expenditure on such services as agricultural research and extension plays a crucial role in growth. Obviously these efforts need to be financed and are therefore crucially dependent both on raising adequate government resources and allocating them to these purposes, rather than to defense or oth~r activities which do not increase economic growth. (A study comparing growth in agricultural output in the two Punjabs attributes much of the higher average rate of increase in the Indian Punjab over the last 30 years to the Indian governments greater investment in infrastructure, research and extension; see Papanek, Mujahid and Kyn.) Another facet of government policy, the nationalization of tea plantati6ns, powerfully affected (negatively, with a t of 3.3) production of tett., the principal foreign exchange earner. Finally, exports - crucial to 'o,y,.,, -66- the capacity to import which in turn greatly affected growth -- were heavily influenced (statistically to the extent of 86%) by the relative price of exportables, in turn powerfully affected by export duties. Government policy to counteract international price movements by compensating changes in export duties can help even out exports and raise the average rate of growth, while a policy to maintain government revenues would exacerbate international price movements at the cost of growth, This suggests another important relationship, potentially.both positive and negative, between fiscal policy and growth. Most of these effects of fiscal policy on growth are usually ignored in growth models and are not further analyzed by Jayasundta, who concentrates on the relationship between growth and savings and the impact of fiscal policy on savings. He examines both total and government savings. Total savings is significantly affected by several public finance variables, as well as by per capita income and the income terms of trade, not readily influenced by government policy, at least in the short term. A particularly interesting result is that an incre:ase in government welfare expenditures reduces total savings by at least half of that increase (t of 2 .1). Through its effect on savings, an expansion of the welfare~· state therefore affects the rate of growth, and to that extent there is a trade-off between growth and equity, at least superficially. Some part of the low g1:owth rate of Sri Lanka in the 1960s and much of the 70s, especially when seen against its endowment of both human and natural resources, .is explained in the light of its outstanding achievements on meeting basic needs. Nationalization of the tea estates had two compensating effects on sav~ngs. The marginal propensity to cave increased (t <>£ 3 .1), presumably because the share of profits previously remitted to foreign owners now remain .-67- in Sri Lanka. At the same time there is a downward shift in the aggregate savings function due to nationalization (t of 2.7) which more than makes up for the increase in marginal propensity. The net result appears to be a, decline in savings as a result of nationalization, presumably .as a result of lower efficiency and profits.. The negative effect of nationalization on growth via the effect on savings needs to be added to the direct negative effect on production, discussed above. In Sri Lanka, as in the other South Asian co\intries, an increase in luxury consumption -- using the ~arne proxy of cars per thousand population -- reduces savings significantly (t of 1.7, significant at the 95% level). In this case a 10% increase in cars decreases savings by 6%, essentially the same as far the 3 ather countries, although slightly larger. This consistency increases ones confidence in the reliability of these results. The rise in luxury consumption also helps explain the stagnationt indeed slight decline, in savings between 1978 and 1983, despite a much more rapid rate of growth and a $harp reduction in welfare expenditures, both favorable to a hi_gher rate of savings. (Here again savings are defined, as they always are in the region, to exclude those used for expenditure on consumer durables.) This is an instance where the failure to use fiscal policy aggressively and imaginatively had costs for the Sri Lankan economy. 'l'he new government liberalized imports in 1977. Almost invariably one of the first consequences of such liberalization is a sharp jump in consumer goods imports, and especially in luxury goods, of which the market had been particularly starved by the previous import licensing system. The resulting increase in obvious conspicuous consumption carries with it political dangers and costs (Papanek, 1986). In addition it is now shown to redv.ce savings and with it to slow growth. The spurt in luxury imports could have been limited in Sri Lanka (as -68- in Egypt, Pakistan in the 1960s and elsewhere where liberalization was undertaken) if a sharp rise in luxury excise taxes had been substituted for controls over their impor~. This was not done, W'ith unfortunate economic (and probably political) consequences. Finally taxes had a large negative effect on national savings. For every 10% increase in taxes, total savings declined by 4% (tis 2.5). Presumably this reflects a lower savings propensity for gov·ernment than for the private sector. (Jayasundera cites an ESCAP study which shows a negative government savings propensity, which would only worsen this effect.) This result is plausible since taxes were usually raised by increasing the rate for exports) corporate profits and various forms of turnover taxes. Since these taxes are difficult to shift forward in the short term, they reduced profits, a form of income out of which savings tended to be high. Higher taxes of this kind also tended to encourage luxury consum,ption at the cost of savings, since they reduced the return on savings/investment. A separate government savings function confirmed some of these conclusions. The statistically significant variables W'ere export income and welfare expenditures. The coefficient for exports was so low that its practical impact is very limited (t of 2.3), while that for welfare expenditures suggest a 7% decline in government savings for every 10% increase. Presumably increases in welfare expenditures were usually financed by reduced government savings (or increased dissaving) and went very largely to increase private consumption. 3. Conclusions. Fiscal policy can have a very significant impact on the rate of growth, both directly through the effect on production and indirectly via savings. It -69- can also have an indirect effect by exacerbating or correcting for price and ~ncentive distortions. This last point has not really been examined here. There is some evidence on the direct impact of fiscal policy on growth from the analysis of Sri Lanka: government expenditure on infrastructure, irrigation; and production services substantially increased output; raising export duties, which lowered the price to the exporter; significantly discouraged production; -- nationalization reduced output. Fiscal policy also affected growth indirectly through its impact on the rate of savings. Evidence from a study of India, Pakistan and Bangladesh and from a separate study of Sri Lanka consistently supports conclusions on the following relationship: an increase in luxury goods consumption, for instance as the result of liberalizing imports without commensurate increase in taxes on these goods, very significantly lowers national savings; The three country study (four really, since Pakistan before the independenc:~~ of Bangladesh is treated separately from post-Bangladesh Pakistan) also shows: an increase in defense expenditure significantly lowers total saving~; for Sri Lanka with its ve~y small and relatively invariant defense expenditures this did not prove a significant variable. On the other hs.nd the Sri Lanka analysis indicates: nationalization lowers total savings, although it increases the marginal rat~, because it shifts the whole savings function downward; raising welfare expenditures very substantially reduces both government and total savings. The tw<"\ studies reach contradictory ¢onclusions on the effect of ta;K.es. In -70- Sri Lanka higher taxes meant lower national savings, in the three countries they resulted in higher national savings. These apparently contradictory results can be quite consistent. Increased Sri Lankan taxes mostly reduced profits, out of which savings tend to be high. Increased taxes in the three South Asian countries seem to have primarily raised the prices of goods and therefore encouraged savings. -71- E. Why the Neglect of Fiscal Policy When Strat'egy Changes? The story so far has involved an apparent contradiction. Very few aspects of fiscal policy changed even when there were quite radical changes in economic strategy, as though fiscal policy really was not an important instrument in the governments arsenal of weapons. But at the same time some elements of fiscal policy were shown to have considerable impact on the economy. The first aspect of the contradiction has run through much of the paper. With the exception of social and welfare expenditures in several instances, there were few if any significant changes in tax or expenditure levels or structure when quite radical changes in economic policy occurred. Over time both dirigiste and market oriented governments increased both tax revenues and government expenditures; maintained defense at the customary level, even though the market oriented were often also military dominated; and made few changes in the relative importance of direct and indirect taxes. The second strand of the contradiction, the power of fiscal policy, is less clear cut, but still quite evident. By the 1980s the governments in the region channelled one quarter to one third of national income through the fiscal system. The sheer magnitude showed its obvious potential for affecting economic decisions, including resource allocations. In studying tax and expenditure incidence it became clear that government could achieve a differentiated impact on different groups, with effective rates ior all taxes ranging from 5% to.25% and and for all expenditures from 4% to 15% for large income groups. Obviously even more highly differentiated systems could have been achieved with more clearly targeted fiscal measures and with particular -72- taxes or expGnditures. There are a number of plausible reasons for the relative neglect of fiscal policy despite the actual, and even greater potential, effectiveness of these instruments: 1. In all five of these countries there is one instrument that is clearly far more powerful than fiscal policy in affecting the functioning of the economy and determining its efficiency, and therefore rate of growth, and that is policy with respect to foreign trade and exchange. It also has a considerable impact. on equity. All of these countries, except Indonesia since the 1970s} have been foreign exchange constrained economies. How that constraint is dealt with fundamentally affects the extent to which market forces influence decisions. D~~.rigiste governments generally set an unrealistic exchange rate, exacerbating foreign exchange scarcities, and then dealt with the problem by controlling imports through licenses. They tended to neglect exports. This approach gave them a powerful tool to direct the economy, since most modern sector economic activities required at least some imported goods either for the original investment or for subsequent operation. By issuing, or denying, licenses government officials could determine what industries were expanded or contracted, where they were located, how labor or capital intensive they were and even how profitable they would be. Not only was the tool powerful, it was quite specific. Licenses could be granted or denied to particular individuals. The import licensing system coqld. be 1 and w~s, used to give windfall profits to some artd to destroy others. Only activities that were essentially non..,..traded escaped the all-pervasive influence of. exchange controls and licensing.. They were powerful tools foX' the altruistic planner, -73- civil servant and politician, or for their venal counterpart. A greater role for the market then inevitably meant a more realistic exchange rate, to stimulate competitive exports and restrict non-competitive imports and thus to reduce the need for rationing imports. Liberalization always, and really inevitably, involved a change in the exchange rate and trade regimes. Without that .change, no measure of fiscal reform would have made a significant difference in acc9rding a much greater role to market forces, with it the crucial step was taken even if there.Mere no fiscal reforms. 2. The second most powerful determinant of the role of the market in these countries was the role of the government owned enterprises. Indeed, in the 1970s in Bangladesh and in Indonesia in the early 60s the di~ect control exercised by government over the economy may have stemmed almost as much from its ownership and management of most o~ industry and much of trade as its control over the scarcest of the resources, foreign exchange. Fiscal policy influences economic decision by changing the effective prices facing decision makers, but if prices are largely irrelevant to decisions, as they often were for the managers of public enterprises, then its effectiveness can be negligible. Managers often were little concerned with profitability, but were evaluated on the basis of physical output, employment created and various perks they were able to provide. To the extent that the non-agricultural economy was dominated by the public sector, governments correctly understood the limited effectiveness of fiscal policy. A change in strategy to greater market orientation therefore usually meant stopping the expansion of the public sector, even selling off some public enterprises (privatization) and possibly shifting the criteria for '•·' w.· -74- success of the public enterprises from physica-l to economic measures. But these steps always took time to be effective, especially since privatization has always been limited by political and other constraints. In the interim fiscal measures were largely ineffective in influencing a major part of the economy. 3. Physical controls, other than those over foreign exchange and enterprises rationing, allocation, price controls -- over a wide range of activities were also seen by dirigiste government as more effective and sure than fiscal measures to achieve the same objectives. They therefore introduced such measures as compulsory procurement of some agricultural products and their allocation at controlled prices through the rationing system; the allocation of such producer goods as cement and fertilizer, and of scarce luxury consumer goods such as cars; and the control of prices over necessities; such as drugs and textbooks, even if they were not rationed. The market oriented governments then moved to reduce or even eliminate these controls, but often without substituting compensating taxes and subsidies. The reasons for the neglect of fiscal measures by these governments are addressed below. But they were clearly correct in assuming that before most fiscal reforms would be effective the physical controls that dominated decisions had to be dealt with. For instance, it would be of only limited effect to introduce a subsidy on fertilizer to encourage greater agricultural output, if that output was then subject to compulsory procurement at prices well below those prevailing in the market. 4. The dirigiste governments were influenced in their choice of instruments by ideological and historical considerations as well. The whole tradition of socialist tho~ght, influenced by Marxism, equated government intervention with -75- government ownership and management and was hostile to the exercise of ·influence over the ecqnomy by affecting prices prevailing in the market. As a consequence the dirigiste governments were often quite oblivious to the fiscal alternatives to direct controls and the possibility th~t these alternatives might actually be both more effective and efficient in achieving equity as well as.growth. As already mentioned, these prejudices are probably the result of historical circumstances. At the tim·e Marx developed his ideas fiscal instruments were indeed feeble and effectivJ government intervention might well have required government control over enterprises. Whatever the reasons, there was a clear ideological preference for ownership and direct controls over indirect fiscal measures in the dirigiste governments. Administrators and military officers, for quite different reasons also generally prefer direct controls. Not just becaus·e of self-interest, addressed below, but because they believe in, and are familiar with,. the issuing of instructions to accomplish their goals and naturally suspicious of a disembodied deus ex machina which supposedly will achieve the same ends in a way that seems uncertain and mysterious. 5. Another factor which has already been mentioned is that changes in the fiscal system often require considerable time to carry out. To cut imports by a licensing system can be do~@ quickly, with assured results in terms of reduce.d foreign exchange needs, even if the consequences are not necessarily efficient. To design a system of equivalent tariffs, that will cut imports to , the same degree is a more difficult and time consuming matter, since the impact of tariffs on the quantities imported is uncertain .. Because of tax shifting, tax evasion and avoidance, and the likelihood that gov·ernment -76- expenditures will benefit some for whom they are not intended, analyzing the economic consequences of fiscal changes is a complex matter. With a rationing system the beneficiaries, at least in the first instance, are clear. Admittedly the indirect consequences of controls are often as difficult.to foresee as those of the tax system, but often those imposing the controls are not aware of these difficulties. 6. Also already discussed was the fact that controls can be used to reward and punish, useful for any political leadership and indispensable for some. Here again, administrators often share the preferences of politicians, because a system of government ownership and controls gives them a power far greater than one based on more impersonal fiscal measures. As is ~1ell known, many leaders of business and industry, with access and influence, also prefer controls which assure them of profitability and even rents to the uncertainty and risk of a competitive market, especially if it is combined witK:~ progressive fiscal system that impinges heavily on their consumption. (The class interest of economists, in contrast, is in the use of the market and of fiscal measures, which they can best understand and interpret.) 7. Finally, and most important from the perspective of this essay, fiscal measures have been neglected, even by market oriented governments, because they doubted their efficacy. As the discussion of incidence has shown, it is quite possible to transfer significant resources from rich to poor through the fiscal mechanism and to deal with the administrative and other problems that usually plague such policies. The fiscal systems of these countries were not carefully designed to achieve purposes other than raising revenues. Like most fiscal systems they just evolved gradually. Even so, they were quite effective. The tax reforms carried out in most of the countries between 197 and 1986, especially far reaching in Indonesia and India should make them more efficient in attainirtg their multiple objectives. What emerges reasonably clearly from the analysis is that fiscal policy could ~~rve a much more effective role than it has, whatever the objectives of a government. Its neglect by both dirigiste and market oriented governments has involved some serious economic (and political) costs. References Acharya, Shankar (1987) "Fiscal Policy" in The Indian Econom;y.: Successes, Current Policie~. and External Links (tentative title), R. Lucas and G. F. Papanek (ed) Westview and Oxford University Presses, Boulder and New Delhi. Gillis, Malcolm (1985) 'iMicro and macroeconomics of tax reform: Indonesia", Journal of Development Economics, vol 19) No. 3. Husain, Ishrat (1981) "Impact of Government Expenditures on Income Distribution: An Empirical Study", Ph.D. dissertation, Boston University Jayasundera, P. B. (1986) "Economic Growth, Income Distribution and Welfare Expenditure: The Case of Sri Lankan, Ph.D. dissertation, Boston University Niazi, Muhannnad-Khan (1984) "Economic Growth in South Asia", Ph.D. dissertationm, Boston University Papanek, Gustav F. (1967) Pakistan's Development: Social Goals and Private Incentives, Harvard University Press, Cambridge, MA. -----------------, (1986) Lectures on Development Strategy. Growth. Equity and the Political Process in Southern Asia, Pakistan Institute of Development Economic.s, Islamabad. - - - - - - - - , Eshya Mujahid-Mukhtar, Oldrich Kyn (1986) ''Agricultural Development Strategy, Growth and Equity: Lessons from the Punjabs and the Bengals" Unpublished report to the Food and Agriculture Organisation. Uppal, J. S. (1986) Taxation in Indonesia, Gadjah Mada University Press, Yogyakarta. Some Recent DRD Discussion Papers 261. Taxation of Investment in East Asian Countries, by A.J. Pellechio, G.P. Sicat, and D.G. Dunn. 262. Effective Tax Rates Under Varying Tax Incentives, by A.J. Pellechio, G.P. Sicat, and D~G. Dunn. 263. A Model for Analysis of Taxation of Capital Investment in Developing Countries, by A.J. Pellechio. 264. The Role of Managerial ~bility and Equipment Quality in LDC Manufacturing, by I. Nabi. 265. Personal Income Taxes in Developing Countries: International Comparisons, by G.P. Sicat and A. Virmani. 266. Indirect Tax Evasion and Production Efficiency, by A. Virmani. 267. Protectionism and the Debt Crisis, by S. van Wijnbergen. 268. Inventories as an Information-Gathering Device, by S. Alpern and D. J. Snower. 269. Welfare Dominance: An Application to Commodity Taxation, by S. Yitzhaki and J .. Slemrod. 270. The Causal Role of Minimum Wages in Six Latin American Labor Markets, by M. Paldam and L. Riveros. 271. Tax Evasion, Corrumption and Administration: Monitoring the People's Agents Under Symmetric Dishonesty, by A. Virmani. 272 •. Characteristics and Operation of Labor Markets in Argentina, by C.E. Sanchez. 273. Growth and 3tructural Change in East Africa: Domestic Policies, Agricultural Performance and World Bank Assistance, 1963-1986, Part I, by U. Lele and R. Meyers. 274. Growth and Structural Change in East Africa: Domestic Policies, Agricultural Performance and World Bank Assistance, 1963-196, Part II, by U. Lele and R. Meyers. 275. Abstracts of Development Research Department Publications: April 1986 - April 1987. 276. Korea's Macroeconomic Prospects and Major Policy Issues for the Nex;t Decade, by V. Corbo and s.w. Nam. 277. The Pricing of Manufactured Goods During Trade Liberalization: Evidence v. ft"om Chile, Israel, and Kox-ea, by Corbo and P •.D. McNelis.