Reforming tax systems

Reducing the fiscal deficits analyzed in the pre-vious chapter will require some combination oflower public spending and higher public revenue.The following chapters discuss lowering and re-directing public spending. This chapter examinesthe scope for increasing and restructuring publicrevenues.

Ultimately public spending is limited by the abil-ity of the public sector to transfer resources fromthe private sector through taxes or charges on cur-rent economic activities, or to issue public debt se-cured by taxes or charges on future economic activi-ties. Other sources of finance are either temporaryor corrosive, as in the case of money creation inexcess of real economic growth, or of minor impor-tance, as is income from public property, licensesand fines, and other nontax revenues (see Figure4.1). This Report focuses, therefore, on taxes anduser charges (or public prices) as the primarymeans of financing public spending.

Taxes are unrequited, compulsory payments col-lected primarily by the central government. In con-trast, user charges are payments in exchange forspecific publicly provided goods and services andare collected primarily by state-owned enterprisesand local governments. The relative importance ofthese two sources of public revenue is difficult toestablish because the financial accounts of state-owned enterprises or local governments are rarelyavailable on an aggregate, nationwide basis.Rough estimates exist for a few countries, how-ever. In Thailand during 1977-83 the gross revenueof state-owned enterprises was estimated to be ofthe same order of magnitude as the central govern-

ment's tax revenue. In Bangladesh gross revenuefrom state-owned enterprises in fiscal 1985-86 wasestimated to be almost double central governmenttax revenue. The importance of user charges variesfrom country to country and depends on the num-ber of publicly provided goods and services out-side the budget. Nevertheless, user charges are animportant component of public revenueeventhough only a fraction of the funds they generate ispassed to the central government as income tax oras a transfer of the operating surplus of state-owned enterprises.

In principle the criterion for choosing betweentaxes and user charges is straightforward. Chargesshould be used wherever a publicly producedgood or service can be sold and should reflectsome measure of the costpreferably the incre-mental costof production (see Box 6.1). This is anefficient way to fund necessary public expendi-tures. Tax financing should be reserved for caseswhere user charges are not appropriate: to pay forpublic goods whose costs or benefits cannot be as-signed to individuals, to compensate for marketfailures (such as externalities), or to achieve a dis-tributional goal (such as alleviating poverty).

User charges can provide substantial revenue. Arecent study on Sub-Saharan Africa put the pro-ceeds from modest increases in charges at roughly20 to 30 percent of central government revenue or4 to 6 percent of GDP (see Box 4.1). In practice,though, taxes remain the principal source of in-come for central governments. Tax reform hastherefore become an increasingly important aspectof structural adjustment and stabilization. Indeed,

79

Figure 4.1 Share of tax and nontax revenuein central government current revenue,1975 and 1985

Property income

Percent100

Nontaxrevenue

Taxrevenue

to be successful, tax reform must serve both thesegoals at once. However, this may not always be thecase: lower international trade taxes in pursuit ofstructural adjustment can run afoul of revenue andother constraints; higher taxes aimed at reducingbudget deficits can hinder the efficient allocation ofresources or make the poor worse off.

This chapter examines the scope for reformingthe main central government taxes. User chargesare discussed in more detail in Chapters 6, 7, and8. Price ceilings, quantity restrictions, and otherdevices analogous in some ways to taxes are notcovered; nor are taxes that generate little revenue(poll taxes and stamp duties, for instance). Taxeson property, which are often important at the locallevel, are discussed in Chapter 7.

80

Patterns of taxation

Patterns of taxation differ from country to countryboth in level and composition. These are consid-ered in turn.

Tax levels

On average, taxes have risen slightly as a propor-tion of GDP since 1975 in all broad country groups(see Figure 4.2, top). However, this disguises widevariation from country to country. The tax-GDPratios for Botswana, Italy, and Yemen Arab Repub-lic grew much faster than the average of theirgroups, while the ratios for Sri Lanka, Venezuela,and Zimbabwe fluctuated sharply from year toyear, and those for Brazil, Canada, and Turkeywere lower in 1985 than in 1975.

Tax-GDP ratios appear to rise with per capita in-come, but the wide variation across countries sug-gests that income growth is only a partial explana-tion. For example, the average tax-GDP ratio forcountries in Sub-Saharan Africa, which are pri-marily low-income, is similar to that for countriesin Latin America and East Asia, which are primar-ily middle-incomeand higher, than the averageratio of South Asia's low-income countries (seeFigure 4.2, bottom).

Tax composition

Tax revenue is usually considered under two head-ings: direct taxes on individuals and firms, andindirect (commodity) taxes on goods and services.

Direct taxes include taxes on personal and com-pany income as well as other direct taxes, consist-ing mainly of social security contributions, payrolltaxes, and taxes on property and wealth. Indirector commodity taxes include domestic taxes, suchas broadly based taxes on turnover, value added,and sales, as well as excises on specific goods; andtaxes on international trade, namely import duties,export taxes, and cesses.

Difficulties over definitions and lack of datamake it hard to compare tax patterns across coun-tries. Nonetheless two important points seemclear. First, trade taxation is insignificant in indus-trial countries; second, developing countries relyvery heavily on commodity taxes (see Figure 4.3).Low-income countries collect almost three-fourthsof their tax revenue, and middle-income countriesalmost one-half, through commodity taxes. Excisesand import taxes account for approximately two-thirds of this.

1975 1985 1975 1985 1975 1985

Low- Middle- Industrialincome income

Notes: Figures are unweighted and represent the average patternfor countries in the sample. The low-income sample includesseventeen countries. The middle-income sample includes thirty-three countries; the larger role of nontax revenues in middle-income countries reflects the classification of oil royalties. Theindustrial sample includes seventeen countries.Source: IMF, Government Finance Statistics, 1987.

Box 4.1 Revenue and user charges

The growing experience with user charges in develop-ing countries suggests that their benefits have beenunderstated and their costs exaggerated (see Chapter6). The main advantages are efficiency, equity, and rev-enue, as discussed below.

Efficiency

Unlike taxes, most user charges do not involve a trade-off between revenue and efficiency. Setting the price ofa publicly produced good or service equal to marginalcost is often efficient (for some qualifications see Box6.1). Charging less than marginal cost leads to excessdemand and the need to generate funds from otheractivities, which can Create distortions elsewhere in theeconomy. These economic costs must be added to theefficiency loss associated with expanding underpricedpublic services. Setting prices correctly generates reve-nue while ensuring an efficient allocation of resources.

Equity

The tradeoff between efficiency and equity may beoverstated for user charges. At present there are manysubsidized services in developing countries that dis-proportionately benefit the better-off. Rationing is re-quired when production of subsidized goods is cur-tailed by the lack of financial resources. In thesecircumstances the poor often do not gain access to ra-tioned public goods and services. By charging marginalcost prices to most users or beneficiaries while target-ing limited subsidies to poor consumers (for example,through lifeline or multiblock pricing arrangements forwater and power services, as described in Box 6.1), it ispossible to improve efficiency and relieve poverty atthe same time.

Revenue

User charges are also a potentially important source ofrevenue. The public revenue aspect of user charges isnot readily apparent in standard fiscal statistics be-

cause this revenue is not transferred to the central bud-get directly. At best, net not gross, revenue is trans-ferred to the revenue account of the budget or subjectto profit taxes. More often the services concerned fail togenerate a surplus. In such cases higher user chargeswill reduce the need for borrowing or transfers fromthe budget to pay for such expenditures.

A recent study on Sub-Saharan Africa has linked thelimited use of user charges for infrastructure services,such as electricity, water, roads, and telecommunica-tions, to revenue shortfalls that worsen the central gov-ernment's budget deficit, undermine the quality ofservice, and restrict the provision of services to low-income groups and regions. The study estimates grossinvestments in infrastructure (water, electricity, tele-communications, and roads) at $6 billion in 1987 in theregion's oil-importing countries. Suppose the value ofthese assets is twelve times current investment, then a5 to 6 percentage point increase in the financial rates ofreturn on the capital stock could generate more than$3.6 billion, or approximately 20 to 30 percent of cur-rent central government revenue.

Raising prices and user charges to levels closer tomarginal supply costs could generate additional reve-nue to reduce, and possibly eliminate, deficits in theconsolidated enterprise accountsthe primary sourceof budget deficits in many Sub-Saharan African coun-tries. Greater reliance on user charges might also re-duce instability in public revenue, because demand forservices is much less volatile than revenue from tradetaxesparticularly on primary commoditiesa majorsource of current revenue. Finally, revenue from usercharges could finance an expansion of services. In thiscase the reduction in net deficits may be small, butwelfare would increase.

In spite of qualifications and limitations to the esti-mates above, the revenue potential is large enough tosuggest that user charges are worth exploring in otherdeveloping countries as well.

In industrial countries income and other directtaxes account for 69 percent of total tax revenue.The weight placed on personal income (27 percent)and social security taxes (31 percent) in industrialcountries is feasible because the necessary admin-istrative apparatus exists. (Even so, other factorsare evidently at work; among this sample of indus-trial countries the revenue share of personal in-come taxes ranges from Norway's low of 9 percentto Australia's high of 56 percent.)

Personal taxes are hard to collect in predomi-nantly rural, agricultural economies, where people

are widely dispersed. Taxes on company incomeincluding taxes levied on the profits of commodityexporting firms, especially mining and agriculturalestate operationspresent fewer administrativedifficulties. Company taxes are therefore relativelymore important in the revenue structure of devel-oping countries.

Cultural and historical factors also influence taxcomposition in developing countries. On averagelow- and middle-income countries raise roughly 10percent of their tax revenue through personal in-come taxes. Surprisingly, however, the richer Latin

81

82

Figure 4.2 Trends in ratios of tax revenues to GDP, 1975 to 1985(percent)

Industrial U Middle-income LI Low-income

Tax revenue/GDP (percent)32

By income

American countries raise a smaller share of reve-nue from personal income taxes than do the poorercountries of Sub-Saharan Africa, where the per-sonal tax base is limited to public employees andemployees of large firms, particularly multina-tional firms. In contrast, Latin American countriesare the dominant users, within developing coun-tries, of social security taxes, which fall primarilyon wage income (see Figure 4.4).

The revenue shares of general commodity taxes(that is, taxes on sales, value added, and turnover)are similar across three of the four developing

LI Middle East and LI Latin AmericaNorth Africa and Caribbean

LI Sub-Saharan Africa U East Asia South Asia

By region (developing countries only)

Tax revenue/GDP (percent)26

Notes: Figures are unweighted and represent the average pattern for countries in the sample. The sizes of the low-income, middle-income,and industrial samples are given in Figure 4.1. The Middle East and North Africa sample includes seven countries; the East Asia sample, sixcountries; the Latin America sample, twelve countries; the Sub-Saharan Africa sample, seventeen countries; and the South Asia sample, fourcountries. Data are for central government tax revenue. Inclusion of state tax revenue in federal systems, such as in Brazil and India, willchange the absolute magnitudes but not the trends or relative rankings.Sources: IMF, Government Finance Statistics, 1987, and World Bank data.

country groupings. Again this masks importantdifferences. General commodity taxes in LatinAmerica are usually value added taxes (VATs); inAfrica, Asia, and the Middle East they are usuallytaxes on turnover or manufacturer's sales. (In in-dustrial countries general commodity taxes aretypically retail sales or comprehensive VATs.) Gov-ernments in Asia and Latin America collect exciseson a wide variety of goods and services; in Africaand the Middle East excises apply to comparativelyfew products. Reliance on import taxes also variesby region. Sub-Saharan Africa depends on them

most, followed by the Middle East, Asia (particu-larly South Asia), and Latin America. Export taxesmatter more in Sub-Saharan Africa, Asia, andLatin America than in the Middle East; overall,however, their role is small and declining (see Fig-ure 4.4).

In summary, there is a clear difference in thecomposition of taxes between industrial and devel-oping countries and to a lesser extent betweengroups of developing countries. The differencesbetween industrial and developing countriesmainly reflect the difficulties of taxing informal sec-tors (such as subsistence agriculture, and informalproduction and distribution) with the limited ad-ministrative capacity available in the developingworld. The differences between groups of develop-ing countries are partly a matter of varying stagesof development and partly a reflection of historicaland cultural factors.

Objectives and constraints in tax reform

Governments attempt to use tax systems toachieve many goals; raising revenue is only one ofthem. To facilitate compliance and collection, how-ever, a tax system must be administratively feasi-ble. For the same reason, but also as an end initself, it must spread its burden equitably. To avoidmisallocating resources, it must not upset the pat-terns of production, trade, consumption, saving,and investment. All these aims can rarely be satis-fied simultaneously, so tax reform is a matter oftradeoffs.

The need for revenue

Over the long term, revenue cannot lag behindexpenditure. So unless public spending is ex-pected to grow at the same rate as national income,the government should ideally choose tax basesthat will expand in tandem with spending, notGDP. Since spending plans can change, tax reve-nue should be generated by a few broadly basedinstruments. Changes in a few tax rates will thenbe all that is required to adjust the revenue total.

It makes little sense to seek a norm for tax-GDPratios. The opportunity cost of raising more reve-nue, the benefits to be derived from extra publicspending, and the cost of servicing public sectordebt all change over time and differ across coun-tries. Decisions on public spending, borrowing,and revenue are highly interrelated; if they are tobe set, they must be set jointly.

Higher tax-GDP ratios may be necessary in some

Figure 4.3 Variations in tax composition, byincome group, 1975 and 1985(percentage of tax revenue)

Personal

Company

Other

Domesticincome taxes

Social security

Property

Other

Other directtaxes

Sales, VAT,turnover

Excise

Other

Domestic28 32commodity taxes

Import

Export

Other

Internationaltrade taxes

Low- Middle-income income

1975 1985 1975 1985

29 25 30 32 34 35

5 4 19 17 32 34

/13/13

Industrial

1975 1985

26 30 29 29

17,/

27

Notes: Figures are unweighted and represent the average patternfor countries in the sample. Totals may not add to 100 because ofrounding. The sample sizes are given in Figure 4.1.Sources: IMF, Government Finance Statistics, 1987, and World Bankdata.

countries where public deficits are high and unsus-tainable and where feasible public spending cutscannot reduce the deficit as required. What mattersis how any such increase is brought about. Experi-

83

38 38 25 19 4 2

17 16

3 12 12

2

2 2

84

Figure 4.4 Variation in taxes, by regionalgroup, 1985(percentage of tax revenue)

LI Asia Latin America El Middle East andand Caribbean North Africa

Personal

Company

Other

Domestic income

Social security

Property

Other

Other direct

Sales, VAT, turnover

Excise

Other

Domesticcommodity

Import

Export

Other

International trade

32 21 37 34 35

5 25 19 5 34

, 4/ //J1.38 36 21 26 29

23 17 22 35 2

Notes: Figures are unweighted and represent the average patternfor countries in the sample. Totals may not add to 100 because ofrounding. The sample sizes are given in Figure 4.1.Sources: IMF, Government Finance Statistics, 1987, and World Bankdata.

ence suggests that increases in tax-GDP ratiosshould be gradual. During the late 1970s and early1980s some countries (Kenya, Malawi, and Sene-gal, for example) increased their tax-GDP ratios by3 to 4 percentage points in the short span of five orsix years. The increases were soon eroded. Even iftax-GDP ratios can be increased, domestic savingmay fall if public saving rises by less than privatesaving falls, as happened in Senegal.

In the short run the urgency of deficit reductionwill generally necessitate the use of easily activatedtaxes. In developing countries this has often meantincreasing international trade taxes, as in Argen-tina, Kenya, the Philippines, and Thailand in theearly or mid-1980s (see also Chapter 3). However,these taxes are among the most damaging for theefficient allocation of resources. Since quick fixeshave a tendency to become permanent, the cumu-lative effect of repeated short-term measures canseriously distort the system of taxation. In suchcircumstances there is a strong case for fundamen-tal reform. Jamaica, Malawi, and the Philippinesimplemented such reforms in the mid-1980s.

The concern for efficiency and growth

Any intentional change in tax revenue will requirea change in the base or rate of some tax. Firms andhouseholds will then shift resources from heavilytaxed activities to lightly taxed ones. When marketprices reasonably reflect social costs and benefits,this poses a tradeoff between revenue and effi-ciency. Sometimes market prices may not reflectsocial costs and benefits. Taxes can then improvethe allocation of resources, but only if the marketimperfections can be quantified so as to guide thedesign of the tax structure. Such cases are rare. Asafer course is to aim for a tax structure that isrelatively neutral: one that generates the necessaryrevenue with the least effect on the allocation ofresources.

As a rule the economic cost of taxation increasesmore than proportionately with the rate of taxa-tion. In other words, the economic cost of a taxlevied at 15 percent is likely to be substantiallymore than three times those of a tax levied at 5percent. The narrower the base, the higher the taxrate will have to be to generate a given amount ofrevenue. This is one of the strongest arguments infavor of broadly based taxes.

Evidence on the efficiency losses resulting fromtaxation in developing countries is sparse. Studiesof the tax structure in India, Kenya, and Pakistanin the early 1980s suggest, however, that the effi-

LI Sub-Saharan Africa Industrial

ciency or economic cost of increasing trade taxes ishigher than that of increasing domestic taxes andthat the economic cost of taxes on all sales (that is,turnover taxes) is higher than that on the sale offinal goods only (that is, retail sales taxes andVATs). A recent study on the Philippines focusedon the economic cost of trade taxes versus domes-tic commodity taxes; its results are shown in Fig-ure 4.5. The study found that the marginal (incre-mental) economic cost of trade taxes is higher thanthat of domestic taxes and that this cost rises withthe rate of the tax. While the magnitudes are case-specific and reflect the prevailing structure of taxesand assumptions about their interactions, the di-rection and pattern of these findings are consistentwith those of other studies.

The pursuit of equity

Tax reform raises questions of equity. This hasmany dimensions. Equity in the distribution ofhousehold expenditure may matter more than inthe distribution of personal income. Attention hastraditionally focused on income distribution, how-ever, and on the distinction between horizontaland vertical equity. Horizontal equity asks howthose with similar incomes are treated: it is con-cerned, in other words, with fairness. Vertical eq-uity refers to the scope for reducing income in-equality by taxing the rich more heavily than thepoor.

Taxes in developing countries often fail badly interms of horizontal equity because the coverage oftax instruments is spotty and arbitrarily enforced.The tax net may capture income in some formalactivities, but not its equivalent in informal orhard-to-tax formal activities, such as professionalservices. This undermines the system's credibilityand the average taxpayer's willingness to pay.Even in terms of vertical equity, tax systems indeveloping countries are not notably successful,despite the fact that they would generally behighly progressive if their rate structures werefully applied. But that is rarely so. A 1978 study ofincome tax in Argentina found that 80 percent ofgross income was not reported and that only 30percent of 1.6 million people eligible to pay taxeson nonwage income did so.

In practice it seems that taxes do little to changethe overall distribution of income. Their importantrole in the pursuit of equity is to raise the revenueneeded to pay for distributive spending, particu-larly to alleviate poverty. So it is public financebroadly definedtaxes and spending together

Figure 4.5 Marginal economic costs of raisingrevenue from tariffs and domestic commoditytaxes in the Philippines

Marginal economic cost(pesos per peso of revenue raised)

2.25

Source: Clarete and Whalley, 1987.

Domestic commodity taxes

that matters for equity, not the structure of taxa-tion alone.

Consistency with administrative capacity

Lack of trained administrative personnel and theaccounting sophistication of taxpayers preventmany developing countries from applying broadlybased income or consumption taxes. Instead theyhave to rely on taxes on trade, production, andcompany income. These can be collected from rela-tively few sources. Given the staffing and resourcelimitations in developing countries, tax reformmust give preference to taxes that are simple andenforceable. But this preference is not unlimited.Sometimes simplicity can conflict with fairness be-cause it means that taxes pay no heed to the vary-ing circumstances of the taxpayers. Sometimes itcan lead to inefficiency, too.

For example, the administrative costs of trade andexcise taxes normally range from 1 to 3 percent ofrevenue collected. The corresponding figure forVATs can be as high as 5 percent; for personal in-come taxes it can reach 10 percent. However, the

85

0.00

-0.25

10 15 20 25

Tax or tariff rate (percent)

economic costs of trade and excise taxes are oftenhigher than those for income taxes and VATs. Re-form of the tax structure must try to weigh thesetwo types of cost. The resulting tradeoff may sug-gest different systems for different countries ac-cording to the existing tax structure, the effective-ness of the administrative apparatus, and thestructure of the economy. In Papua New Guineatrade taxes are low, and the administration of anincome tax or VAT would be extremely difficult, sothe government has been advised to increase reve-nue from trade taxation. In Thailand, by contrast,trade taxes are high and have created serious dis-tortions; the government has been advised to shifttoward a simple VAT

The next two sections examine the options forimproving the design of commodity and incometaxes. However, ultimately it is the interaction ofthe different taxes that determines revenue and in-fluences economic behavior. For example, increas-ing domestic or trade taxes on inputs used in pro-duction may reduce the revenue collected fromtaxes on company profits. It is important that taxreforms also take account of these interactions.

Commodity taxation

Commodity taxes are taxes on the transaction ofgoods and nonfactor services. They include the ar-ray of taxes on domestic production and consump-tion, as well as those on international trade. Re-ducing the distortionary effects of commoditytaxes can be important for two reasons. First, theycurrently account for 50 to 70 percent of all taxrevenue in most developing countries (see Figure4.3). Second, in the early stages of developmentgovernments often rely heavily on the least desir-able sort of commodity taxes, namely turnovertaxes on domestic production and taxes on interna-tional trade. These latter taxes are often used be-cause they generate revenue with limited adminis-trative costs. As economic and administrativeconditions change, however, it is useful to reassessthe tradeoff between the administrative and eco-nomic costs of these tax instruments.

Taxes on domestic production and consumption

Production taxes are levied on goods before theyenter the distribution chain. Often they fall ontransactions between producers, such as the saleof an intermediate good. As such they affect pro-duction decisions and feed through the system of

86

production to affect consumption decisions also.In contrast, taxes on the sale of final goods onlyconsumption taxesdo not generally affect the ef-ficiency of domestic production. They are there-fore a better way to raise revenue. Commoditytaxes, whether on domestic production or con-sumption, can be general or selective.

GENERAL COMMODITY TAXES. The most commongeneral tax on production is the turnover tax. Itsbase is every salewhether between firms or be-tween firms and consumers. As such it is a multi-stage tax, activated at every stage of theproduction-distribution chain. Turnover taxes arerelatively easy to administer because they do notrequire tax authorities to differentiate between dif-ferent kinds of transactions. This simplicity isbought at the expense of distorting transactionsbetween producers. In addition this tax "cas-cades": tax liabilities accumulate as each succeed-ing transaction adds tax to that already paid at pre-vious stages of production and distribution. Thisincreases the price of outputs that use taxed in-puts, as in exports, and it generates differentialtaxation of consumption even when the turnovertax is applied at a single rate.

Under a pure consumption tax all domesticallyconsumed goods, whether imported or producedlocally, would be taxed at the retail stage. Ratesmay vary for different consumer goods, but similargoods would be subject to the same tax rate inde-pendent of origin. All inputs into productionintermediate products, raw materials, and capitalgoodsand all exports would not be taxed. As aresult consumption taxes have some general ad-vantages over other broadly based taxes. Unlikeproduction taxes they do not interfere with pro-ducers' choices between intermediate inputs, orbetween the latter and factors of production (capi-tal, land, and labor). Consumption taxes also donot cascade through the production process anddo not create incentives for firms to avoid tax liabil-ities through vertical integration. In contrast totaxes on international trade, they do not favor pro-duction of import-substitutes or reduce incentivesto produce for export.

Commodity taxes on consumption are of twotypes. The first is a general sales tax on final goodsimposed at the retail level. This ensures that allconsumed goods are taxed, but leaves other goodstax free. The second is a value added tax (VAT). Inits most popular version the VAT is a tax on con-sumption. Applied to all transactions in theproduction-distribution chain up to and including

the retail stage, it has the same final tax base as aretail sales tax. Each intermediate purchaser in thechain is able to credit taxes paid on purchasesagainst taxes due on sales. All inputs are therefore,in effect, tax free. The final purchasertheconsumerhas no means of crediting, and thusall sales at this stage are taxed. The rate is set tozero for exports. Thus, both general sales taxesand comprehensive VATs have the economicallydesirable properties of commodity taxes onconsumption.

Retail sales taxes are rare in developing countriesbecause of the prominence of informal distributionnetworks. Instead, single stage manufacturer'ssales taxes are commonly used, as in the Philip-pines (before 1986) and Kenya. In some develop-ing countries the pattern of sales taxes and excisesresembles a turnover taxthe sales taxes in Zam-bia and Tanzania, for example, or the excises in theRepublic of Korea (before 1976) and India (before1986).

The VAT has made consumption-type taxes moreaccessible to developing countries (see Box 4.2).Some distortions between sectors will remain be-cause the VAT, although a high yield tax, can becostly to administer for producers in agricultureand services and for small enterprises generally.However, a movement toward a VAT is likely topromote efficiency while also generating a sub-stantial amount of tax revenue.

India is a case in point. Until 1986 its extendedsystem of excise taxes covered a wide range ofgoods, including intermediates. It thus resembleda turnover tax. Because of cascading, export pricesincluded a 5 to 7 percent tax. This was only partlyoffset by rebates. In addition the prices of goodssuch as cereals and edible oilsespecially impor-tant to the poorcontained a 5-to-lO-percent tax,although they were nominally exempt. Indiasharply reduced excises on intermediate goods in1986 by implementing a modified VAT through themanufacturing stage. A higher rate is required toraise the same revenue as before, because the taxbase shrinks from gross output to net output.However, the new tax will interfere less with pro-duction and trade decisions.

SELECTIVE COMMODITY TAXES. Some taxes, bytheir nature, cannot be broadly based. Taxes to cor-rect for specific market failures, such as external-ities, are best restricted to a few goods only, be-cause a great deal of information is needed todetermine the appropriate rate structure. Taxes tocover specific spending programsfuel taxes to re-

cover road use costs, for instanceare sometimesnecessary, but the rates cannot deviate signifi-cantly from taxes on close substitutes. Studies onthe Philippines, Thailand, and Tunisia found thatpetroleum taxes led producers and consumers toswitch to other fuels. Other selective taxes includetraditional excises on demerit goods such as alco-hol and tobacco, and luxury excises on goods suchas cars or jewelry.

Governments generally set the base and rates forthese selective taxes for ease of collection; thus thetaxes are often not well integrated with the broadlybased taxes. This is of greatest concern for excisetaxes, which are an important source of revenue inmost developing countries. In contrast to broadlybased taxes, many excise rates are specified perunit of quantity rather than as a proportion of theprice. Therefore excises can be regressive and notinsulate revenues against inflation. Where infla-tion proofing is desired, the tax rates should be setrelative to prices rather than quantities. For exciseson goods such as tobacco and alcohol, it is alsopossible to retain specific rates, provided there isperiodic adjustment for inflation and the rate struc-ture is differentiated to reflect distributional con-siderations.

The case for some progressivity in commoditytaxes is strengthened by the limited coverage ofpersonal income taxes and the scale of evasion bythe highest income groups. So, for example, gov-ernments that collect the bulk of their domesticcommodity tax revenue from a general tax, such asa single rate VAT, can supplement this with a selec-tive luxury or excise tax with a few rates. The baseof such a tax should be goods whose share inhousehold expenditure increases with incomemotor vehicles in Indonesia, for instance, or enter-tainment and recreational services in Korea. Sucha tax can be outside the VAT crediting system if it isrestricted to final consumer goods. This distin-guishes the combination of the VAT plus a luxuryexcise tax used in Korea and Indonesia from themultirate VAT used in the EC. The latter attemptsto promote equity within the rate structure of theVAT and therefore within the crediting system.This increases the administrative burden of theVAT and may be premature for many developingcountries.

International trade taxes

International trade taxes generate about a third ofthe tax revenue in developing countries and areamong the easiest taxes to administer.

87

IMPORT TAXES. In principle, taxes can be collectedfrom imports at the border without driving awedge between the price of imports and compet-ing domestic products, provided the tax on im-ports has a counterpart on domestic production. In

88

practice, however, import taxes are used not onlyto raise revenue but also to protect domestic pro-duction and to promote equity in consumption.

In view of these multiple objectives it is not sur-prising that the typical import tax regime is com-

Box 4.2 The value added tax in developing countries

In 1967 Brazil imposed the first comprehensive value high administrative costs, a VAT is often implementedadded tax (VAT) extending to the retail stage and apply- at a rate of at least 10 percent. Where the tax base ising to all states of the federation. It was designed to narrow, particularly when the tax does not go to theensure greater tax coordination among the states and retail stage, it is likely that the VAT will have to beto overcome the defects of the turnover tax. The Brazil- imposed at 15 percent or more to generate sufficientian VAT is based on the destination principle, which revenues. Despite initial skepticism about high rates,focuses on the use of the product. As such it is a tax on experience (in Brazil and Chile, for example) hasconsumption or final sales. It is this comprehensive shown that rates of about 17 to 20 percent can also beform of the consumption-type VAT that the European enforced, even for VATs that extend to the retail stage.Communities adopted in the late 1960s. An alternative A valuable feature of the VAT is its potential for self-type of VAT is based on the origin principle, which enforcement through a system of credits. However, anfocuses on the income generated by an activity. It is important requirement for successful VAT administra-used in Argentina and Peru, and some of its features tion is to minimize problems of implementation. Fromare found in the VAT introduced in Turkey in 1985. The an administrative point of view a single rate is prefera-consumption-type VAT is easier to implement and has ble to a multirate VAT. To reduce regressivity, the VATbecome by far the most popular version of the VAT in can be supplemented by a luxury tax with two to threedeveloping countries. rates. Exemptions complicate administration because

VATs generally replace a multitude of small taxes and the distinction between what is exempt and what iscan greatly simplify the system of commodity taxation. taxed is often tenuous or arbitrary. Nevertheless, distri-The consumption-based version has three main advan- butional objectives have led many countries to exempttages. First, by not taxing inputs used in production some basic commodities (such as some unprocessed(for example, through a system of credits), it simulta- foods and selected medical items). Zero-rating, a moreneously avoids the distortion of choices between in- complex form of exemption that requires refunds andputs; the cascading of taxes, which can lead to ineffi- therefore burdens the administration, has been limitedcient vertical integration; and the presence of multiple to exports by most countries. The need to provide spe-effective tax rates in consumer prices. Second, it does cial treatment for small businesses under a VAT is muchnot discriminate between imports and domestic pro- more pressing in developing than in industrial coun-duction in domestic markets. Third, exports are not tries. Various methods for dealing with small taxpayerstaxed. Together these provisions ensure that the tax are used, but all methods present technical and practi-does not interfere with production or trade. cal problems.

VATs have become an important source of revenue in Successful introduction of a VAT depends largely onmany countries. They yield more than 20 percent of tax whether the country has had previous experience withrevenue in about thirty industrial and developing multistage taxes or general sales taxes, the nature ofcountries. Some twenty developing countries, primar- the taxes that the VAT will replace, the lead-in time,ily in Latin America, now have comprehensive VATs and the structure of rates and exemptions, includingthrough the retail stage. Many others, including some provisions for small taxpayers. Indonesia and Koreain Sub-Saharan Africa, have taxes with VAT-like charac- introduced a VAT after two to three years of prepara-teristics through the manufacturer-importer stage. tion, whereas Turkey successfully implemented a VAT

VATs at the retail stage are more feasible in middle- within two months of its enactment, following a rela-income developing economies, such as the Republic of tively long period of analysis. Administrations withKorea, than in low-income ones, because the formal few resources often stress enforcement for large tax-distribution network is more developed in the former. payers and practice restrictive refunds. Such adminis-VATs through the manufacturer-importer stage are in- trative practices weaken the broadbased and neutralcreasingly common in lower middle-income countries features of an ideal VAT. Most successful tax reforms,such as Côte d'Ivoire and Indonesia, since they are however, have introduced some form of a VAT, both toeasier to implement. Even these VATs can cover large- reduce distortions in production and trade, and to gen-scale distributors, agricultural estates, and other activi- erate adequate revenues to compensate for revenuesties beyond manufacturing. Because of its relatively lost through rationalizing other tax instruments.

/

plex. For imported goods that have no domesticcompetition, tariff rates are determined by theneed for revenue or adjustments in the exchangerate rather than the need to achieve a desired de-gree of protection. But for competing imports,where protection is the primary concern, rates areoften differentiated, with goods for productionsubject to lower rates than goods for consumption.Rebates or duty drawbacks are often introduced toavoid increasing the cost of production for ex-porters and for firms that have been granted in-vestment incentives. For equity reasons some basicgoods are either exempt from tariffs or subject tovery low rates, whereas luxuries are subject tohigh rates. In some countries strategic or priorityimports, including government and parastatal pur-chases, are exempt from duties. Finally, wherehigh rates do not stem the volume of selected im-ports, quantitative restrictions or prohibitions areintroduced, but these entail a loss of tariff revenue.

The incentives generated by such complex sys-tems are often not transparent. It would be prefer-able to transfer as much as possible of the revenuefunction of tariffs to broadly based domestic con-sumption taxes, such as the VAT, and of the equityfunction to selective taxes. Quantity restrictions ontrade are best replaced by tariffs, and "specific"rates (per unit quantity) should be changed to advalorem rates (per unit price). These changeswould make it easier to rationalize the protectivefunctions of tariffs.

As noted above, a tax on domestic productioncan also be collected from imports at the border.This wifi fulfill the revenue function of a tariff andbe equally easy to administer without protectingdomestic producers. Similarly, if the objective is torestrict the consumption of imported luxuriesrather than to stimulate their domestic production,it is better to subject them to a domestic luxury taxrather than a higher rate tariff. In the absence of a

retail-level consumption tax, the luxury tax on im-ports can be collected at the border, with its do-mestic counterpart collected at the factory gateasin Indonesia.

In rationalizing tariffs there is a general consen-sus that protection should be reduced in the longrun because it penalizes consumers and promotesinefficient patterns of production. In practice it ishard to cut tariffs quickly because of revenuelosses and opposition from the protected sectors.

Moreover statutory tariff rates are a poor mea-sure of the protection provided to domestic pro-ducers because of interactions with other taxes. Ifthe domestic producer is subject to a domestic ex-cise or turnover tax and the competing import isnot, then the nominal rate of protection is not thestatutory tariff, but the difference between the stat-utory tariff rate and the domestic tax. If, in addi-tion, domestic production uses imported inputs,then the nominal rate of protection is unlikely to bea good measure of the protection afforded to do-mestic value added. A better measure is the effec-tive rate of protection (ERP). This takes into ac-count the interaction between tariffs on output andinputs. The dispersion of ERPs is often largelarger than for statutory ratesand can includenegative rates of protection (see Table 4.1). Whencalculating ERPs, taxes on domestic inputs mustalso be taken into account, so restructuring thepattern of protection generally requires a joint re-view of taxes and tariffs.

Import tariffs also implicitly tax exports. An in-crease in import tariffs can result in an exchangerate appreciation and the preferential treatment ofimport-substituting industries. This reallocates re-sources toward import-substituting industries andaway from all other industries, including exports.This is so even when imported inputs are not sub-ject to tariffs; where imported inputs also face tar-iffs, the distortion against exports can be greater

Table 4.1 Distribution of effective rates of protection for selected countries in East Asia

89

(percent)

ItemIndonesia,

1987

Republic ofKorea,1982

Malaysia,1982

Philippines,1985

Thailand,1985

Selected sectorsTextiles - 11-155 54 106 118

Intermediates 4-280 40-62 17 15-125 45-60

Machinery 75-82 31 37 116-201 18-37

Transport equipment 6-220 124 74 118 60-90

Summary measuresImport competing sectors 30-380 27 25

Export sectors 23-11 5 3

Box 4.3 Integrating trade and domestic taxes in Malawi

In the early 1970s the tax-GDP ratio in Malawi wasrelatively low (11 percent), and trade taxes applied pri-marily to consumer imports. By the late 1970s revenuepressures forced the government to introduce new taxmeasures annually to generate additional revenue andreduce its budget deficit. During the same period im-ports were constrained to reduce the trade deficit. Im-port priority was given to government, aid-financedprojects, and necessities-all of which were duty free.As a result the taxable import base shrank. To compen-sate, tariff rates were first increased on consumergoods, particularly luxuries, and then extended to in-termediate and capital imports; finally, taxes were im-posed on exports.

By 1984-85 the tax-GDP ratio had reached approxi-mately 20 percent, and it was clear that the ad hocapproach to generating revenue had relied too heavilyon easily administered instruments, even though theywere likely to have adverse incentive effects. Increas-ing tariffs and excises on intermediate goods raised thecost of exportsmaking Malawi less competitive, espe-cially in nontraditional exports such as textiles andeven in traditional agricultural exports. The tax rebatesystem was not functioning well because of administra-tive problems and restrictive interpretations of the defi-nition of inputs qualifying for rebates. In addition, in-creasingly high tax rates on imported luxuries andexemptions for imported necessities were creating aprotective structure inconsistent with the objectives ofindustrial development. Finally, increased import tar-iffs and excises on intermediate goods caused the taxcontent in consumer prices to cascade and to reducethe already limited progressivity of indirect taxes.

On the basis of a 1985 tax study the government initi-ated a comprehensive tax reform in 1986-87 to broadenthe tax base and simplify tax procedures. The firstphase was to eliminate the export tax and reduce taxeson intermediate goods. Revenue losses were to be off-set by increasing the rate of the surtax. However, thesurtaxessentially a consumption tax at the manufac-turer and importer level and less distortive of produc-tion and trade decisionshad to be increased by 5 per-centage points to 35 percent to offset revenue lossesassociated with the declining import tax base and theelimination of the export base. This unusually high ratedemonstrated the narrowness of the domestic tax base.

Expanding the base to include additional producersand distributors will take a few years and will include,among other things, the introduction of a crediting sys-tem within the surtax. This new feature will reduce taxpressure on exports, which results from the taxation ofinputs used in production. It will also indirectly tax theinformal sector producers and traders who would notbe eligible for a credit unless their output is taxed. Dis-tributional concerns in the reformed surtax are ad-dressed by introducing two or three luxury rates,which will apply equally to domestic and importedgoods. This will enable import tariff rates on luxuries tobe lowered and restructured so as not to inadvertentlystimulate their production relative to necessities. Thejoint determination of domestic and trade taxes willallow improvements in trade tariff incentives withoutlosing revenue. However, revenue needs will still setlimits on the extent to which the tax structure can berationalized in the short run.

still. Many developing countries have tried to dealwith this using export subsidies, export rebates, orduty drawback systems. Their record of success ismixed. When there is no paper trail of taxes andtariffs paid, it is difficult to avoid over- or under-compensating different exports, although well-administered schemes, as in Korea, have been rea-sonably successful. Linking information about tar-iffs and the VAT may improve such creditingschemes, because the VAT provides a fuller recordof taxed transactions. The standard design of aVAT automatically eliminates the need for a sepa-rate export rebate for taxes on domestic inputs.

During a fiscal crisis trade liberalization can falterfor revenue reasons. Import tariffs on inputs are,in effect, also a tax on export production. Thisweakens the case for increasing tariffs on inputs to

90

compensate for revenue losses when tariffs on out-put are cut. Joint reform of tariffs and taxes thenbecomes desirable, as in Malawi (see Box 4.3). Re-structuring trade tariffs and domestic taxes to pro-duce a broadly based consumption tax should bethe primary objective of tax reform in countriesthat do not already have one. Such a tax can be-come an important source of revenue.

Where rudimentary taxes on consumption arealready in place, their role as a source of revenueshould be increased at the expense of tariffs. Thiscould be achieved by an increase in the tax ratewith a compensating reduction in tariff rates. Inthe long run an increasing amount of revenue canbe generated from taxing domestic activities. Thedevelopment of the manufacturer's stage VAT inCôte d'Ivoire illustrates this. In 1960 the tax ac-

counted for 15 percent of total revenue, with 70percent of its contribution coming from the taxa-tion of imports. By 1982 the corresponding figureswere 30 and 40 percent. Thus the tax generatedmore revenue, with an increasing share comingfrom the taxation of domestic activities.

EXPORT TAXES. Many countries levy export taxeson primary products. The use of export taxes inmining is less frequent than in agriculture primar-ily because economic rents in mining can often becaptured through company taxation, such as theresource rent taxes in Papua New Guinea. Exporttaxes are on occasion used, as in Liberia and Zam-bia, to supplement the company tax. Such use ofexport taxes is justified to the extent they substi-tute for royalties. They should not, however, be

greater than the royalties, otherwise they are likelyto interfere with the time profile of extraction.

The use of export taxes is more common in agri-culture. A 1987 study of seventy-four developingcountries found that export taxes were used in atleast fifty-three of these countries. In general thesetaxes did not account for more than 5 percent of taxrevenue, but there were exceptions to this observa-tion in selected periods (see Figure 4.6). Exporttaxes are inadvisable because they reduce the in-centive to produce for export. This is inappropriatein view of the slow rates of growth in agricultureand the importance of trade in agricultural prod-ucts to many of the countries that use this form oftax. Under some circumstances these taxes can bejustified as imperfect substitutes for other forms oftaxation, but for a limited period only (see Box 4.4).

Box 4.4 Export taxes and agriculture

Export taxes are commonly used in agriculture becausetraditional taxes on income and profit are hard to ad-minister in this sector. In principle land taxes are anattractive alternative. Where land is in fixed supply, aland tax is collected from economic rent and leavesproduction decisions unchanged. However, with a fewexceptions, such as Ethiopia, Kenya, Paraguay, Peru,and Somalia, land taxes generate less than 1 or 2 per-cent of total revenue. The low yield reflects the inade-quacy of land registration and valuation. In many Afri-can countries and the Pacific islands it is difficult toestablish ownership because land tenure is based oncustomary arrangements. In other countries rural landtransactions are infrequent, which restricts the use ofmarket prices to determine the value of land. There arealso limitations on the use of presumptive measures tolink land values to the productivity of land, becausedata on land quality and the variations in productivitybetween seasons are generally inadequate.

Some export taxes are implicit and result, for exam-ple, from the price-setting activities of marketingboards, such as the Cocoa Board in Ghana and theAgricultural Development and Marketing Corporationin Malawi. These boards act as distributor and exporterof a few important smallholder crops and usually setfarmgate prices below border prices, thereby implicitlytaxing smallholders.

Evidence on the level of taxation suggests that insome countries producers of agricultural exports maybe overtaxed. If export taxes substitute for incometaxes, it is possible to compute a rate of tax on exportsthat will generate the same amount of revenue as a taxon the smallholder's income. A simple calculation for a

typical cocoa farmer in Ghana in the early 1980s revealsthat an export tax of 4 percent of the farmgate pricewould have yielded as much revenue as if the farmer'sprofits had been subject to income tax. The prevailingexport tax was more than 100 percent, which suggeststhat to the extent export taxes substituted for incometaxes, rates could have been reduced substantially.Even as a tax to capture excess profits, the export taxwould be only 12 percent.

More important, export taxes create an incentive toshift production to other crops. Given the ample em-pirical evidence that smallholders respond to prices,the economic costs of export taxes are likely to be sub-stantial. Where feasible, presumptive taxes on agricul-tural income may be preferable, as in Uruguay.

Other arguments favoring export taxes include thedesire to manipulate the terms of trade and theneed for revenue. The former should be treated withcaution.

Inelasticity of world demand in the short run canlead quickly to loss of markets in the long run becauseof changes in both world demand and supply. Thishappened to Ghana and Nigeria's world market shareof cocoa and to Nigeria's and Zaire's share of palm oilin 1961-63. Given the large budget deficits in manycountries, the need for revenue cannot be ignored inthe short run, especially if there is a case for exporttaxes as a cess or proxy user charge. In the long runextending broadly based commodity and income taxesto also include the agricultural sector is necessary toreduce and eventually eliminate agricultural exporttaxes.

91

92

Figure 4.6 Countries in which agriculturalexport taxes provide more than 5 percent of taxrevenue for selected years

Source: Strasma 1987.

/ El Salvador (1985)Ethiopia (1980)Ghana (1985)Rwanda (1980)Uganda (1985)

/ Costa Rica (1983)Côte d'lvoire (1980)Honduras (1981)Malaysia (1984)Sri Lanka (1984)

/Argentina (1984)Guatemala (1982)Peru (1982)Zambia (1984)

/ /

/

/

Income taxes have long been the principal meansof taxation in industrial countries. With relativelyfew distortions they can generate a great deal ofrevenue and leave scope for income redistribution.Experience in developing countries, however, sug-gests that personal income taxes are difficult to ad-minister, raise little revenue, are weak in redistrib-ution, and are often unfair. Recent reforms havetherefore stressed the role of commodity taxes.Nonetheless, the reform of taxes on personal andcompany income will often be necessary to en-hance the revenue and efficiency of a tax system.

Company income taxes

Reform of taxes on company income is especiallyimportant because they account for about a third ofrevenue in developing countries and havea greater potential for misallocating newinvestments.

BASE AND RATE STRUCTURE. Company taxes aredesigned to collect revenue from a firm's economicprofits. In practice, the tax base is net accountingprofits: gross revenue less operating costs and cap-

ital adjustments. Often a single statutory rate isused and is usually most desirable, particularlywhen there are administrative constraints. How-ever, a few developing economies use an explicitlyprogressive rate structure with two to three brack-ets and a moderate range of 15 to 35 percent.Fewer still use more than three brackets: Guatema-la's and Mexico's rates ranged between 5 and 42percent up to 1987. Finally, some have an implicitlyprogressive rate structure through the use of dif-ferentiated surcharges, as in Brazil.

The statutory rate of the company tax is often apoor indicator of its effect on revenue or invest-ment behavior. Rates apply to financial income noteconomic income; inflation, for instance, drivesthe two apart. For policy, therefore, effective taxrates are more important. The average effective taxrate (AETR) is the ratio of total revenues collectedthrough the company income tax to the company'seconomic profits. For revenue purposes this rateshould be high. By contrast, the marginal effectivetax rate (METR) measures the effect of taxes oninvestors' rate of return for an incremental addi-

Figure 4.7 Asset-specific marginal effective taxIncome taxes rates in Malawi, 1974 and 1984

] 1974 1984

Statutory rate

Marginal effective rate

Short-lived assets

Manufacturing

Nonmanufacturing

Long-lived assets

Manufacturing

Nonmanufacturing

Rate of privateinvestment (GDP)

Source: Chamley and others 1985.

Percent0 10 20 30 40 50 60 70

More than20 percent

More than10 percent

More than5 percent

tion to their activities. To avoid interfering withinvestment decisions, the METR should be low.The main tasks of company tax design should be toachieve a high AETR while keeping the METR low,or preferably zero, and to avoid large variations inMETRs across different types of investment.

Differences between METRs and statutory ratesarise from provisions that allow the recouping ofinvested capital, deduction of interest incurred oninvestment debt, credits for investment, correc-tions for inflation, and so forth. As a result a singlerate company tax can mean many different METRsacross assets and sectors.

A 1985 study of taxes in Malawi found that whenstatutory rates changed, METRs changed in thesame direction (see Figure 4.7). However, METRsin manufacturing were substantially lower than innonmanufacturing. At a subsectoral level the

Table 4.2 Marginal effective tax rates for a hypothetical project investment, circa 1985(percent)

METR of an investment project varied according tothe economic lifespan of its assets. Since METRsare so hard to observe, it is difficult to usecompany taxes to steer investment in a particulardirection.

It is possible to compare company taxes acrosscountries as well, as in Table 4.2, by positing ahypothetical standard project with a fixed assetcomposition and a common pretax rate of return,investment horizon, and other relevant parame-ters. As a result the figures do not show actualafter-tax rates of return, which will be affected byvariations in the asset composition of projects andby tax enforcement practices. However, they areuseful for highlighting on a comparable basis thewide variation between statutory rates andMETRs. As is apparent in the table, METRs areequal to statutory rates only by chance. Countries

Note: The asset composition of the hypothetical project consists of 40 percent building, 40 percent machinery and equipment, 10 percent vehicles,and 10 percent land. Replacement investment is at the rate of econonmic depreciation for ten years. Real rate of return before taxes is fixed at 10percent. Calculations are based on tax code provisions, not actual enforcement. Ireland is included as an example of a tax code with 100 percentdepreciation in the first year but no adjustment to nominal interest deductions.

Ranked by statutory income tax rates.Refers to the use of negative tax liabilities in the project to offset positive tax liabilities on income from other investments. This can arise either

from legal provisions in the tax code, allowing the filing of consolidated returns for a firm or holding company, or through transfer pricingschemes when consolidated returns are not allowed.Sources: Pellechio and Dunn 1987, and Pellechio and others 1987a and 198Th.

93

Statutory

All equityfinancing with loss

carried forward

50 percent debtfinancing with loss

carried forward

50 percent debtfinancing with full

loss offset'

tax 5 percent 50 percent 5 percent 50 percent 5 percent 50 percentEconomy' rate inflation inflation inflation inflation inflation inflation

(1) (2) (3) (4) (5) (6) (7)

Hong Kong 18.5 18.4 29.5 16.4 17.4 9.6 7.3Ecuador 20.0 13.5 27.9 10.1 12.8 10.1 9.4Yemen, Arab Republic 25.0 32.2 62.2 30.5 47.4 30.5 45.4Colombia 30.0 28.5 47.4 36.9 43.0 14.5 35.1Korea, Republic of 30.0 33.2 48.0 32.8 52.3 24.6 42.8Egypt 32.0 37.0 73.9 31.2 56.8 29.2 48.7

Argentina 33.0 31.7 51.0 29.7 42.5 11.2 29.8Jamaica 33.3 40.6 59.0 35.3 37.1 33.7 28.4Brazil 35.0 54.4 68.1 45.9 62.9 45.9 62.9Indonesia 35.0 41.6 81.4 36.0 63.1 34.1 54.1Philippines 35.0 40.5 81.0 40.2 66.1 31.9 53.3Thailand 35.0 24.9 68.6 20.0 48.9 18.6 42.6

Jordan 38.0 37.4 64.2 27.3 37.8 25.1 34.6Tunisia 38.0 24.5 23.0 19.8 20.1 4.9 -60.9Malaysia 40.0 31.7 62.7 24.2 34.0 20.5 20.9Portugal 40.0 45.5 79.1 28.7 51.4 28.7 46.5Singapore 40.0 29.5 46.5 23.2 20.5 15.2 1.9Guatemala 42.0 10.7 40.3 2.8 39.1 -13.6 39.1

Mexico 42.0 19.6 24.0 10.3 6.9 -20.5 -22.9Turkey 46.0 45.5 81.5 27.7 47.9 25.6 30.0Morocco 48.0 44.0 65.3 24.0 65.3 22.9 60.4Greece 49.0 20.0 68.3 10.6 40.5 10.6 34.1Ireland 50.0 5.8 11.5 5.5 5.6 -65.9 -54.0

with equivalent statutory ratessuch as Brazil, In-donesia, the Philippines, and Thailand at 35 per-cent or Malaysia, Portugal, and Singapore at 40percentcan have dramatically different METRsbecause of other company tax provisions. Equally,differences in statutory rates may not reflect differ-ences between METRs. For example, Ireland'sMETRs are lower than Hong Kong's despite amuch higher statutory rate.

In most cases debt financing lowers the METRfor a given level of inflation (columns 4 and 5against 2 and 3, respectively, in Table 4.2). Thiscreates a bias in favor of debt financingincreasingly so for higher rates of inflation. How-ever, the interaction of inflation and the mode offinancing can vary. In Ecuador high inflation in-creases the METR for equity financing relative tothe statutory rate and lowers it for debt financing;in Argentina, Brazil, and Colombia high inflationincreases the METR relative to the statutory rateregardless of the mode of financing and despiteindexing provisions. If the tax code allows nega-tive tax liabilities in a project to be offset againstpositive tax liabilities on income from other invest-ments, the METR wifi be lowered (columns 6 and7, Table 4.2). It can even become negative, as inMexico and Tunisia, which suggests an implicit in-vestment subsidy at the expense of the treasury foractivities submitting consolidated returns.

The treatment of depreciation, debt, and infla-tion greatly affects the METR. Valuing assets athistoric cost and spreading depreciation allow-ances over more than one year ensure that tax de-preciation wifi diverge from economic depreciationin the presence of inflation; the recouping of theinitial investment is understated and taxable in-come overstated. When this is combined with fulldeductibility of nominal rather than real intereston debt, it is likely that the company tax wifi skewthe firm toward debt, as the deductibility of nomi-nal interest rates overcompensates for the real costof borrowed funds. The firm's reduced capitaliza-tion may then increase its vulnerabifity to externalshocks.

There is no single answer to this interlocking setof problems. When inflation is low, the overcom-pensation of financing costs (due to nominal inter-est deductions) may just offset the undercompen-sation of depreciation values based on historiccost. The effect of inflation on revenues wouldthen be limited. Although the incentive in favor ofdebt finance remains, it is likely to be small andunlikely to justify the administrative complicationsof schemes to convert financial income into eco-

94

nomic income. In such cases the METR can be re-duced by lowering the statutory rate. This ap-proach, however, also lowers the AETR, whichmeans a windfall to past investments and a reve-nue loss for the treasury whether or not new in-vestments materialize. The revenue loss can bepartially offset by reducing asset-specific invest-ment incentives. The combination of a lower statu-tory tax rate and streamlined investmentincentivesas in Jamaica and Indonesia, as well asin the recent U.S. tax reformwifi reduce tax dif-ferentiation between taxed sectors. (It may reducethe difference between the taxed and untaxed sec-tors.) However, the METR remains positive.

When inflation is high, other measures may beneeded. Indexation of historic cost or periodic re-valuations are an important step toward bringingdepreciation allowances in line with economic de-preciation. Periodic revaluation of assets, as in Af-rica, or various indexing schemes, as in LatinAmerica, have a mixed record. Revaluations arecostly and infrequent; indexation is often insuffi-ciently comprehensive to avoid generating distor-tions between assets or sectors. For example, amove toward economic depreciation must be ac-companied by the use of real interest rates, yetnominal interest deductions are rarely adjusted forinflation. Such a correction was recently intro-duced in Mexico and has been proposed forTurkey.

A simpler alternative is "full expensing." Underthis approach, when calculating taxable profits,firms can treat investment expenses like otherbusiness costs at the time they are incurred. Thisrelatively new approach has not been applied fre-quently in practice, but it is similar to the treatmentof exploration and development expenses for min-ing in developing countries. It is also used for themanufacturing sector in Ireland and as an optionin the tax codes of Bangladesh and Zimbabwe. Fullexpensing eliminates the need for indexing, forspecial rules about inventories, and for estimatesof depreciation rates for different types of assets. Itwould also make it easier to withdraw explicit in-vestment incentives, many of which have the samepurpose, that is, to reduce the taxation of returnsto new investment.

If expensing is allowed, however, the cost ofdebt should not be allowed as a tax deduction. If aninterest deduction is granted, the firm would bereceiving a double deduction for assets financed bydebt. This can result in a negative METR, as inIreland. It is appropriate only if there is a strongcase to subsidize overall investment and if other

activities can generate the revenue to finance sucha subsidy.

Full expensing without interest deductions pro-vides, in effect, a zero METR and does not interferewith an investor's rate of return. It also reducesintersectoral differences in incentives and elimi-nates the bias toward debt finance and thin capital-ization. It can be difficult to introduce in somesectorsfinancial institutions, for instanceandinitially it can be costly in revenue foregone, be-cause the invested capital is recouped in the earlyyears rather than over the life of the asset. How-ever, income in later years will not be reduced bydepreciation allowances, and tax revenue will thenincreasealthough not to the levels associatedwith a positive METR. In the ore and hydrocarbonmining sectors in many countries (for example,Cameroon and Nigeria) revenue from companytaxes is high because of high economic profits. Thetaxes imply AETRs of 70 to 80 percent, even

though METRs may be negative because of thecombined immediate write-off of most invest-ments and the full deductibility of nominal inter-est. Transition problems make hybridswith par-tial expensing, positive METRs, and a lower initialloss of revenue, as in Malawiattractive. This is anarea that warrants further research.

INVESTMENT INCENTIVES. Governments oftenuse explicit investment incentives in addition tothose implicit in the tax treatment of depreciation,interest, and so forth. Where market failures canbe quantified, there may be a case to use tax instru-ments to promote efficiency. Special investmentincentives include exemptions, tax allowances, taxcredits, or special tax reliefs designed to assist par-ticular groups or activities in specified industries orlocations. These incentives serve either to reduceor defer tax liability; the latter corresponds to aninterest-free government loan over the deferment

Box 4.5 Reform of Indonesia's investment incentives

The government of Indonesia adopted a major tax re-form in late 1983. Here the focus is on only one aspectof the reformnamely the wholesale elimination of taxincentives for investment.

Before 1983 the tax structure was inordinately com-plex. Hundreds of ad hoc amendments had beenadopted, which created a law that was incomprehensi-ble to taxpayers and tax collectors alike. Many amend-ments resulted from changing trade and business con-ditions, and many more were for special nonrevenuepurposes, with predictably negative consequences forrevenues and unforeseen results on equity and devel-opment

The massive array of incentives in the investmentcode was designed to favor specific industries, pro-mote exports, develop remote regions, promote tech-nology transfers, strengthen the stock exchangeandeven to encourage firms to be audited by public ac-countants. The numerous and often contradictory taxincentives created an excessively complicated systemunable to fulfill its revenue function or to serve thespecial purposes originally intended.

Investors and the Investment Coordinating Board(BKPM) negotiated many incentives as part of an over-all package. These incentives, and the relatively rapidchange in both their design and their structure, meantthat firms in the same industry were taxed under dif-ferent rules and that the same firm faced a different taxregime at various times. Such incentives created effec-tive tax rates that varied both between and within sec-tors and thus misallocated the capital stock. For exam-

ple, the tax rules created incentives to change thecomposition of investment toward short-term projectsthat, in extreme cases, never paid taxes, such as "hit-and-run" projects, particularly in textiles and lightmanufacturing.

Because of a lack of communication with the BKPM,auditors in the tax department did not know what in-centives were available to firms, which resulted in au-dit conflicts. In addition some firms did not file returnsduring the tax holiday period, or simply filed blankreturns, which made it difficult to audit returns oncethe holiday period ended.

Finally, nonuniform tax holidays created the impres-sion of discrimination against certain industries, whichwould then seek extended tax holidays or alternativeincentives to offset the perceived discrimination. Prob-lems similar to these existed for every incentive. Taxincentives are difficult to administer, and thus the gainsfrom incentives must be weighed against the increasedadministrative costs.

The principles underlying the tax reform were ad-ministrative simplicity, transparency, and minimumdistortion of economic behavior. As a result all specialtax incentivestax holidays, investment allowances,and accelerated depreciation other than double-declining balancewere eliminated. The expected rev-enue gains from eliminating incentives allowed the taxrate to be reduced. The simplified incentive system isexpected to minimize tax-induced intersectoral prefer-ences, while the lowered company tax rate is expectedto benefit all investors.

95

96

Figure 4.8 Income level at which personalincome tax liability begins and thesubsequent structure of the marginaltax rates during 1984 and 1985

Marginal tax rate (percent)70

60

50

40

30

20

10

0

BrazilColombiaGhanaJamaicaMalaysiaMexicoSierra LeoneSri LankaTunisiaTurkey

BangladeshBurmaChileGuatemalaIndiaIndonesiaKenyaNigerPeruZaireZambia

period. Tax incentives for special purposes, how-ever, are often ad hoc and poorly integrated intothe overall tax structure,

In general the effectiveness of a tax is inverselyrelated to the number of goals it is meant toachieve. Tax incentives overload tax instrumentswith multiple objectives. They complicate compli-ance and prompt unproductive efforts to obtaintheir benefits. If the incentives are small, the eco-nomic gains are likely to be limited. If they arelarge, the erosion of the tax revenue base is likelyto be significant.

Investment incentives are also difficult to admin-ister. Consider tax holidays, for instance. To beconsistent, they would have to be granted to exist-ing firms making new investments as well as to

new firms. But costs and profits must then be di-vided between old and new operations, thus caus-ing problems of internal transfer pricing and costallocation. It makes sense to reduce the number ofinvestment incentives; Indonesia has eliminatedthem altogether (see Box 4.5).

Personal income taxes

Personal income taxes account for about a tenth oftotal tax revenue in developing countries. The lowyield reflects limited coverage and poor design.Improving the yield requires changes in the baseand rates to make the tax easier to administer,without adverse effects on incentives to work andsave.

BASE AND RATE STRUCTURE. The typical personalincome tax is levied on net taxable income, derivedby deducting allowances and exemptions fromgross personal income. A schedule of rates is ap-plied to determine tax liability. Tax credits are thensubtracted from this tax liability to generate thefinal tax obligation.

The design of personal income taxes varies con-siderably across countries. In some countries, suchas Ghana in 1984, very low levels of income arelegally subject to tax; in others, such as India, ex-emption rates are quite high. In some the marginalrate increases very rapidlyas in Jamaica beforetax reform. In others the rate schedule is relativelyflatas in Côte d'Ivoire. Finally, the highest mar-ginal rate and the level of income to which it ap-plies vary significantly.

Figure 4.8 shows two groups of countries, basedon their legal or intended tax structures, not thetax structures as actually enforced. In the group Acountries low levels of income are subject to tax,and the marginal tax rate increases rapidly. Thisstructure is difficult to administer since large num-bers of small taxpayers are caught in the tax netand subject to high rates. The higher exemptionsand more gradual increase in marginal tax rates ofthe group B countries are better suited to the ad-ministrative capacity of most developing coun-tries.

Figure 4.9 shows that many countries have maxi-mum rates above 50 percent. These rates often af-fect only a handful of individualsthose with in-comes in excess of fifty times per capita GDP. Highrates on narrow bases generate little revenue and,if not enforced, damage the credibility of thesystem.

0 3 6 9 12 15

Income level(multiples of GDP per capita)

Source: Sicat and Virmani 1988.

The revenue share of personal taxes has grownslowly in the past two decades. Their base hadbeen expected to expand more rapidly than GDPas more and more activities entered the formal sec-tor. The ability to fine-tune tax rates according toability to pay was another reason to expect theshare of personal taxes to rise. But these factorshave been outweighed by the difficulties of en-forcement and collection. In many countries per-sonal income taxes are collected from less than 15percent of the population; in South Asia and Sub-Saharan Africa the figure is less than 5 percent.Almost everywhere the potential revenue frompersonal taxes is further eroded by avoidancethrough loopholes and tax shelters, as well as out-right evasion. A 1981 study of Bolivia estimatedthat 75 percent of the revenue due from labor in-come was collected primarily because of withhold-ing taxes on wages, whereas the equivalent figurefor capital income was 20 percent.

Many of the same features that limit the revenueyield of personal income taxes also limit the equity

Figure 4.9 Maximum marginal tax rate (MTR) and the level of personal income at which it becomeseffective during 1984 and 1985

Greater than 70

50 to 70

Less than 50

Source: Sicat and Virmani 1988.

features of these taxes in practice. In developingcountries personal income taxes are not the masstaxes they are in industrial countries. The progres-sivity of the rate structure is therefore less impor-tant when 80 to 90 percent of the population, pri-marily the lowest income groups and those insubsistence or informal activities, are outside thepersonal income tax net. With the difficulty of en-forcing this tax on high-income recipients in agri-culture, trade, and the professions, plus the preva-lence of a multitude of allowances and provisionsbenefiting wealthier groups in society, it is not sur-prising that in many countries it is now recognizedthat the personal income tax does not significantlyimprove the distribution of income. However, aless ambitious distributive objective can be at-tained. Legally excluding the poor from the taxbase altogether is a more powerful way to protectthem than incorporating lower rates in a multiratestructure. Revenue lost from more exemptions atthe bottom of the income scale can be largely offsetby eliminating loopholes for those at the top. This

97

Portugal Republic of KoreaZambia

BurmaJr MoroccoCôte d'lvoire NigerEgypt TanzaniaEthiopia TunisiaLiberia

Ghana MaliGreece PakistanJamaica Sri LankaMalaysia Sudan

ArgentinaBrazilIndiaSierra LeoneZaireZimbabwe

Benin MexicoChile NigeriaChad PeruKenya SenegalMadagascar ThailandMalawi Turkey

Hong KongYemen Arab Republic

Burkina FasoEcuador

Colombia JordanGuatemala PhilippinesIndonesia Singapore

Maximum MTR Multiples of per capita GDP at which the ma mum rate becomes effective(percent)

Less than 30 30 to 50 Greater than 50

wifi also improve the equity features of the tax.Horizontal equity requires that all sources of in-

come (from agriculture, trade, manufacturing, andservices) and all types of income (wages, interest,rent, profits, and so forth) be treated equally. Thisfavors a global income tax over schedular taxes fordifferent sources or types of income. A global tax,however, entails a tradeoff between equity andsavings. Personal income taxes can affect the vol-ume of private saving by reducing both the incomeof would-be savers (usually higher income house-holds) and the returns to savings. The second ef-fect depends on the openness of capital marketsand the extent of financial intermediationthat is,the availability of nonbank institutions to attractsavings through insurance schemes, social securityschemes, pension plans, and so forth.

Some governments have tried to exclude the re-turns to savings from the income tax base. Theyhave exempted interest from certain types of de-posits, for example, small post office deposits inIndia and Malawi, or interest income up to a ceil-ing, as in Jamaica. In other countries schedular in-come taxes are used to tax different sorts ofincomesuch as interest from savings depositsat a lower rate. Such taxes are used, as in WestAfrica, because they are considered easy toadminister.

There is some evidence, however, that in devel-oping countries changes in the returns to savingsmay have a greater effect on the composition ofsavings than on the level. Taxes on the return tofinancial savings can reallocate savings betweendifferent types of assetsfor example, betweenstocks and bonds in middle-income countries (ifcapital gains and dividend income is treated differ-ently from interest income) or between financialand real assets in lower income countries. Theseswitches can disturb the efficiency of intermedia-tion between savings and investment. Some havetherefore argued that personal taxes based on ex-penditures are preferable to personal taxes basedon income; expenditure taxes do not tax incomethat is saved. However, such taxes applied to indi-viduals, as opposed to transactions, have not yetbeen implemented anywhere.

It makes better sense to ease the tasks of admin-istration and enforcement by simplifying personalincome taxes. Most allowances can be eliminated.Instead the threshold should be set high enoughsay up to incomes three times per capita GDPtoexclude most low-income earners, and the maxi-mum rate should be set low enoughsay 30 to 40percentto reduce the incentive for tax evasion.

98

Revenue would in any case be low from the verylowest income groups and from those subject toconfiscatory rates. A multitude of brackets can bereplaced by a few brackets. Even a single rate taxwith the fewest number of loopholes and a highthreshold can still be reasonably progressive, as,for example, in Jamaica (see Box 4.6).

PRESUMPTIVE INCOME TAXES. One way to im-prove the income tax is to supplement it with apresumptive taxa tax assessed not on income it-self but on indicators of incomefor evasion-pronegroups such as self-employed professionals andthose employed in agriculture and trade. Incometax assessment has evolved from presumptive toexact methods as indicators of income have gradu-ally been replaced by measures of actual incomereceived. In practice, however, income tax assess-ment for large numbers of taxpayers in both indus-trial and developing countries is still largely pre-sumptive.

The francophone countries of West Africa rely onpresumptive or "forfeit" taxes more than other de-veloping countries. However, this kind of tax isalso used elsewhere. In the early 1980s Turkey'stax authorities noted that 85 percent of taxpayersfiling income declarations claimed to be in thelowest tax bracket; audits of cases of suspectedevasion found that approximately 50 percent of in-come was undeclared. The government intro-duced a system of presumptive taxation in 1983.Indicators of living standards are used to assesstaxpayers filing regular tax declarations. A pre-sumptive assessment of certain minimum taxamounts is made for activities in agriculture, trade,and professional practices. Further, specifiedamounts of income are presumed to be associatedwith, for instance, ownership of residential prop-erty (both owner-occupied and rental), automo-biles, boats, airplanes, and racehorses; foreigntravel; and employment of personal servants. Taxis levied on the income determined by a presump-tive assessment or the taxpayer's declaration,whichever is greater. This system increased tax col-lections; 84 percent of those who filed declarationsin 1985 had their tax liability based on the pre-sumptive assessment.

These methods can also be applied to taxes ongoods and services or on wealth, where valuationis difficult. However, experience in countries asdifferent as Colombia and Korea suggests that aconsiderable administrative effort is stifi requiredfor any type of presumptive tax to ensure it isbased on realistic criteria and applied fairly.

Tax administration

Tax administration in industrial countries by andlarge carries out the intent of tax legislation; in de-

veloping countries tax administrators often maketheir own tax policy by selective administration.

As a result steps to simplify the task of tax ad-ministration are likely to make tax policy more ef-

Box 4.6 Reform of Jamaica's personal income tax

The government of Jamaica embarked on a comprehen-sive tax reform in 1985. It includes changes in the per-sonal income tax, the company tax, and indirect taxes.The reform of the personal income tax is unique. Acomplicated, narrowly based individual income tax lev-ied under a progressive statutory rate structurecommonly found in developing countrieswas re-placed by a broadly based, single-rate tax in 1986.

Before the reform the highest marginal tax rate of 60percent (including payroll taxes) was reached at therelatively low annual income level of less than threetimes per capita GDP. The provisions in the tax codewere complicated. There was no standard deduction,but taxpayers qualified for sixteen separate credits.These credits had been added to the system over theyears for purposes that ranged from personal allow-ances to stimulation of savings and home ownership,and even to employment of helpers in the home. Inaddition employers could grant nontaxable allowancesto employees. These allowances were negotiated be-tween employer and employee, and the results did nothave to be reported to the income tax commissioner.The ratio of nontaxable allowance to taxable wage wasestimated to average 40 percent.

The tax was difficult and costly to administer. Impor-tant disincentives were inherent in the rate structure.Capital gains and interest income were tax free, butdividends were taxed twice. The pay-as-you-earn(PAYE) tax ensured that formal sector labor income wastaxed at a high rate but that self-employed incomewent virtually untaxed. In addition Jamaicans withhigher incomes, many outside the pay-as-you-earnsystem, tended to avoid or evade a substantially higherpercentage of the tax liability than did lower incomefamilies. This evasion and avoidance all but negatedthe progressivity of the statutory rate structure. A tax-payer survey suggested that the tax net did not coverabout half of the potential individual income tax liabil-ity. The complexities of the system contributed to poorenforcement, which compounded the inequities.

The primary objective of the tax reform was to sim-plify the tax and minimize adverse incentives. This ledto several changes in its design: the sixteen tax creditswere replaced with a standard deduction equal to twotimes per capita GDP, the present rate structure wasreplaced with a single rate of 331/3 percent, most non-taxable allowances were incorporated into the tax base,and interest income was included in the tax base.

The tax reform was enacted after a committee of citi-zens from the private sector spent several months scru-

tinizing and amending the proposals before recom-mending its adoption. Another committee ofrepresentatives from union, business, and public inter-est groups also reached a consensus that the revisedand streamlined tax seemed fairer than the previoussystem. From 80 to 90 percent of the population wouldnot pay income tax as a consequence of the relativelyhigh standard deduction. By eliminating loopholes, thetax base would expand, which enabled the maximumstatutory tax rates to be lowered. This would in turnreduce the incentive for evasion or avoidance at higherincome levels and would facilitate enforcement andcollection.

It now appears that the combination of a higher standard deduction, a broadened base, and a lower flat rathas improved administration and increased the prcgressivity of the tax system. Revenues from companyand personal income taxes are also running 18 percenthigher in the twelve-month period after introduction ofthe reform, in part because of improved administrationof the streamlined tax.

Perhaps the best indicator that the program has beenaccepted is the lack of public discontent with the newreform. However, the reform has three significarproblems. First, the potential exists for abuse in the fewallowances that remain; if the numbers or significanceof these allowances grow, they could compromise thefairness of the new structure and necessitate a rate in-crease to compensate for the narrowing of the base.Second, the standard deduction has not yet been in-dexed, and this could become another policy problemif the rate of inflation were to increase again. Finally, toavoid burdening lower income savers, there is no taxon the interest from bank deposits below a certain ceil-ing, which could encourage higher income depositorsto split their holdings. All these problems can, in prin-ciple, be dealt with by a continuing policy review.

The Jamaican case shows that comprehensive tax re-form can occur in a weak economic setting. Jamaicarestructured its tax system at the same time it faced aserious exchange rate disequilibrium and a substantialgovernment deficit. The tax system had become soonerous, obviously unfair, and out of control adminis-tratively, that there was substantial public support for amajor overhaul. This support also reflected the realiza-tion that tax rates would have to be increased signifi-cantly under the old system because of the country'sserious fiscal and external imbalances. In many ways,therefore, the time was right for tax reform.

99

fective. Administrative reforms can improve thetax structure by bringing reality in line with inten-tions. But they can also magnify distortions thatwere dormant when the structure was badly ad-ministered. Setting goals for long-term tax policybroadening the base, say, or shifting the tax basefrom production and trade toward consumptioncan identify needed improvements in administra-tion. So, even though present administrative re-sources limit the scope for tax reform, thinkingabout reform helps to set administrative priorities.

In the 1960s and 1970s comprehensive tax re-forms focused on instruments rather than on thesystem as administered. Some of these compre-hensive reforms were only partially implemented,as in Colombia, and some not at all, as in Ghana.Some partial reforms paid attention to administra-tive difficulties (as in Korea, for instance) and weresuccessful. In the 1980s, in contrast, comprehen-sive reforms placing a greater weight on adminis-tration have become more common (such as thereforms in Indonesia, Jamaica, and Malawi). Ad-ministrative reforms must try to address the fol-lowing problems.

Compliance and enforcement

Poorly drafted tax forms, long queues, rude offi-cials, and cumbersome appeals procedures all re-duce compliance. Slowor norefunds of legiti-mate claims can foster reluctance to pay taxes inthe first place. High tax rates increase the benefitsof evasion, particularly if the tax authorities areknown to lack the resources to track down the of-fenders. In most developing countries the sanc-tions on fraudulent taxpayers are neglible.

For obvious reasons evasion is hard to measure.Defining avoidance and evasion also raises concep-tual problems. As a result there are few country-specific or comparative studies on the subject. A1980 study of the income tax in Indonesia beforereform found that, depending on the year, tax eva-sion ranged from 84 to 94 percent of taxes due onpersonal income tax, and 76 to 93 percent of taxesdue on corporate income. These high evasion rateswere blamed on rates too high to be enforceableeven by a relatively efficient administration. Real-ism in tax laws is important.

A 1985 study estimated India's black, or unre-corded, economy to account for roughly a fifth ofGDP. Not only was the treasury losing revenue,but evasion was also blunting the allocative anddistributive features of the system. For example,tax rates could not be lowered to reduce tax-related

100

distortions without a loss in revenue given that thetax base had been narrowed by evasion. Other,older studies in the 1960s and 1970s for Chile, Co-lombia, Kenya, and Nigeria all found similar highrates of evasion.

Poor system design promotes corruption. Re-form can reduce the opportunities for taxpayers tobribe rather than pay taxes. One way to do this isto reduce the number of discretionary elements inthe tax code (as was done in Indonesia and Ja-maica). Another partial solution is to separate as-sessment and collection (as in Malawi), while en-suring that assessments are not made withoutregard for what is collectible.

Other measures are also often needed, includingreasonable salary levels and more trained officials,particularly those able to audit company and per-sonal accounts and to design and operate com-puter procedures. A greater capacity to gather andprocess data would enable administrators to iden-tify assessment and collection problems more eas-ily. Ultimately, though, political backing is neces-sary for successful enforcement.

Improving collection

The revenue yield of the tax system cannot readilybe increased unless ways are found to improvecollection.

TAx AMNESTIES. It serves no purpose to have atax assessed but not paid. In some countries theproblem of tax arrears has become so critical thatgovernments have taken emergency measuressuch as tax amnesties and provisions for resche-duling tax payments. These may make it easier tocollect delinquent taxes, but they can also under-mine voluntary compliance if used frequently.

WITHHOLDING. The scant auditing resourcesavailable in the tax administrations of most devel-oping countries make it impractical to audit morethan a small percentage of taxpayers. A system towithhold money from current income is thereforeone of the most efficient techniques for preventingdeliquency and evasion. Withholding is most com-monly applied to wage income, as in the pay-as-you-earn (PAYE) systems in Jamaica, Malawi, andother countries. Withholding is also applied to in-terest and dividends in some countries, such asColombia and Indonesia. An effective withholdingscheme, however, requires a relatively small num-ber of easily identifiable payers of income. With-holding is hard to apply to rental income, profes-

sional income, and small business income, wherethere are as many payers as receivers.

INFORMATION EXCHANGE. Another approach isfor tax-collecting agencies to exchange iriforma-tion. In many developing, as well as industrial,countries import duties and taxes on domestictransactions are administered by separate depart-ments, with little or no exchange of information. Inother countries sales and income taxes are admin-istered by separate departments. The exchange ofinformation between these revenue departments ishighly advisable because gross sales figures are im-portant in determining income tax liabilities, andvaluations of sales for income tax purposes make iteasier to implement ad valorem excises and duties.

SELF-ENFORCEMENT AND CROSS-CHECKING. Theavailability of personal and minicomputers makesthe use of self-enforcing taxes-based on matchinginformation from different sources-more feasiblethan it was a decade ago. It is now possible forinformation furnished by one taxpayer to revealthe receipts and gains made by other taxpayers, as,for example, in a VAT. The ultimate goal of alinked, self-checking system of taxes, however, isstifi far away.

COMPUTERIZATION. Automated data processing(ADP) can improve the administration of taxes.ADP systems that perform multiple functions re-quire an integrated master file system. The useful-ness of master files depends mainly on a reliableand up-to-date system of unique taxpayer identifi-cation numbers to distinguish one taxpayer'srecords from another's. Despite the technical prob-lems, automation may eventually offer the mosteffective way to deal with expanded workloads incustoms departments (with the growing volumeand complexity of international trade), income taxdepartments (with the growing number of taxpay-ers), and treasuries (which need to forecast andmonitor revenues). Such systems are currently be-ing set up in Indonesia, Jamaica, Malawi, and Mo-rocco. They are already partially or fully opera-tional in Brazil, Ecuador, Honduras, Korea, andNigeria. Experience suggests that ADP can in-crease the efficiency of well-run operations, but itcan exacerbate problems if superimposed on badlyorganized administrations.

Tax analysis units

Better collection and administration can improvethe implementation of tax policy. However, it is

also important in most developing countries tostrengthen the treasury's ability to analyze reve-nue options. A tax analysis unit can support poli-cymakers by analyzing the revenue consequencesof changes in exchange rates, interest rates, andtrade and industrialization policiesall of whichaffect tax bases and interact with tax rates. It canalso weigh the implications of new revenue mea-sures for other policies and forecast revenue to as-sist in fiscal planning. Such units feature in manyof the comprehensive tax reform programs cur-rently under way.

The scope for tax reform

With fiscal deficits high and access to new borrow-ing limited, there is little scope for deliberate re-ductions in taxation in the near future. Whethertaxes can be raised to cut fiscal deficits depends onthe existing structure of taxes and the period overwhich the deficit reduction is to occur. Where taxbases are narrow, a rapid increase in revenue willcall for higher rates. But in some cases higher rateswifi erode the tax base through evasion. In othercases they will cause inefficiencies in economic be-havior, especially if the changes rely on adminis-tratively convenient measures such as trade taxes.In contrast, carefully designed tax reform can re-duce the cost of raising additional revenue and en-sure that tax policy complements other policies.Such reforms take time.

Even in the absence of fiscal deficits, tax reformmay be necessary, especially when price regula-tions and barriers to market competition are beingremoved, or when there is a case for rectifying anaccumulation of ad hoc distortionary tax mea-sures. Recent tax reforms in developing countrieshave focused on reducing tax-induced distortionsand on simplifying tax administration. Reforms arelong term but not permanent. Significant changeswill be needed periodically to take account of shift-ing external circumstances or internal needs. (SeeBox 4.7 on Colombia's reforms.)

No system of taxation can be perfectly neutralwith respect to allocation. Nor can tax policy ig-nore distributional concerns. The balance betweenthe various taxes is a matter of changing prioritiesand constraints. Where the growth of income isadequate, equity can be given greater weightthrough expansion of taxes on income. However,where slow income growth and limited adminis-trative capacity are of greater concern, taxes onconsumption may need to be given preference.

In spite of the complexity of these issues, some

101

Box 4.7 Periodic tax reform in Colombia

At the start of the twentieth century Colombia reliedalmost exclusively on trade tariffs for public revenue.The collapse of international trade in the 1930s sharplycurtailed revenues from customs duties and prompteda reform that established the basis of the present taxsystem, including full-fledged taxes on income, netwealth, and inheritance. This early reform was in-tended to strengthen public revenues weakened by theeffects of the Depression and to increase the impor-tance of direct taxes, especially on capital income.

Since then Colombia has had major tax reforms in1953, 1960, 1974, and 1986. The objective of the 1953reform was broadly the same as in 1930: to increaserevenue and taxes on capital income, both by raisingincome tax rates and by taxing dividends. The reformswere carried out in the face of strong opposition bysome political groups and were successful in large partbecause of support from other political groups. By theend of the 1950s taxes on income and wealth weremore important, and probably more progressive, inColombia than elsewhere in Latin America.

Though it set out to increase revenue and imposetaxes on capital income, the 1960 reform had the oppo-site effect in both respects because a wide range of taxincentives was simultaneously introduced to foster in-vestment in manufacturing and exports. These incen-tives were so heavily used, however, that both the rev-enue and the progressivity of the income tax declined.

A series of ad hoc reforms, mainly income tax sur-charges, were implemented in an effort to close therevenue gap; by far the most important reform was theintroduction of a general sales tax in 1963. As a result ofstrong opposition aroused by fears that the sales taxwould be regressive, its implementation was delayed

general prescriptions for tax design can be gleanedfrom recent experience. Clearly their applicationwill vary from county to country.

Simplify the design of tax instruments, withfewer rates and fewer adjustments to the base; inparticular, eliminate or streamline special tax in-centives for investment, production, and trade.

Strengthen tax administration to improve col-lection and facilitate the shift in the tax structurefrom reliance on higher tax rates to reliance onbroader tax bases.

Avoid taxing the poor.

Simplify the design of tax instruments

Simplification of tax instruments applies primarilyto the definition of the base and adjustments to it.

102

until 1965. Serious administrative problems sooncaused the tax to be transformed into a value added taxat the manufacturing level, and it came to be secondonly to the income tax as a revenue producer. Adminis-trative problems with income taxation, particularly taxevasion due to high and rising marginal tax rates, alsoresulted in the introduction of a system of wage with-holding and current payments in the late 1960s. Thesemeasures helped the income tax retain its importancein Colombia's fiscal system.

The next major tax reform in 1974 reflected Colom-bia's extensive experience with such reforms. Not onlywas this reform intended to strengthen revenue as inprevious years, but it represented a return to the pre-1960 emphasis on taxation as an instrument of socialpolicy rather than the use of tax incentives as an instru-ment of economic policy. Ineffective tax incentiveswere substantially reduced. Additionally, a minimumpresumptive income tax was introduced to ensuremore adequate taxation of income from capital. Otherchanges were made to reinforce the role of income (andwealth) taxation. At the same time, however, the ratesof the sales tax with value added features were sub-stantially increased, and its base was expanded.

The immediate effect of the 1974 reform was to in-crease income tax revenue substantially, largely as aresult of the new presumptive tax regime. These effectswere not permanent, however. The courts had decidedthat some critical administrative changes included aspart of the reform package were beyond the power ofthe legislative authority. This greatly weakened theability to enforce the minimum tax. Moreover, a seriesof rate reductions and amnesties in the late 1970s, in-tended in part to offset the effects of inflation, not only

This in turn has a bearing on the number of taxinstruments and their rate structure.

Commodity taxes could be consolidated intothree or four instruments with the following char-acteristics.

A shift from the taxation of production to thetaxation of consumption. This could be achievedwith two instruments. The first is a broadly based,general tax on consumption (such as a retail VAT ora manufacturer's VAT), which does not tax transac-tions between industries, does not differentiatecommodities by source of production (import ver-sus domestic), and does not tax exports (implicitlyor indirectly); this tax could have a single rate ifequity concerns can be met with a luxury tax. Thesecond instrument is a selective commodity tax fordemonstrable and quantifiable externalities and for

eroded the capacity of the income tax to keep up withinflation but, in effect, eliminated the capital gains tax.In 1983 the changes of the previous years were to someextent offset both by regularizing the system of infla-tionary correction and by strengthening the presump-tive tax. At the same time the sales tax was consider-ably altered and became in effect a full-fledged, valueadded tax through the retail level.

Tax reform continued to be high on the Colombianpolitical agenda, however, leading to significantchanges in income taxation in 1986. This reform low-ered tax rates on business income, freed dividendsfrom taxation at the individual level, and abolished theinheritance tax. To some extent this reform reversedthe 1974 reform, just as that in 1960 had reversed partof the 1953 reform. Unlike in 1960, however, the 1986reform was intended more to unify the marginal effec-tive tax rates on different types of investment than tofavor some types of investment over others. At presentstill further changes along these lines in business taxa-tion are being considered. An especially interestingfeature of the current wave of tax reform in Colombia,however, is that it is the first designed primarily toimprove the tax structure rather than to increaserevenue.

The more than fifty years of tax reform in Colombiapoint to several lessons.

Except for the reforms of 1986 and (to a lesser extent)1974, all have been motivated primarily by economiccrises. These in turn were frequently caused by exter-nal shocks and required new efforts to raise revenue.

The influence of changing intellectual fashions ontax reform is as obvious in Colombia as in most coun-tries. The reforms of the 1930s and 1950s, like those of

purposes of equity. In the latter case the baseshould be luxuries (defined as those final con-sumption goods whose share of expenditure in-creases with income). Again there should be nodistinction between sources of production (im-ports versus domestic producers); exports shouldbe excluded and the number of rates limited.

A shift from the taxation of international tradeto the taxation of domestic transactions. Domestictaxes on goods and servicesrestructured as de-scribed abovecan be collected at the point of im-port for administrative convenience without beingconfused with tariffs. It is then possible to restruc-ture taxes on international trade so that the leveland variation of protection rates are reduced. Ex-port rebates or duty drawback schemes wouldhave to be strengthened if production inputs faced

1974, reflected the dominant "progressive" view ofmost tax experts in this century, while the 1960 reformreflected the transitory popularity of "incentive-directed" growth policies, and the 1986 reform reflectsthe renewed interest in "market-directed" growth.

Despite the strong influence of outside forceswhether economic or intellectualColombia's fiscalsystem, and the timing and manner in which it hasdeveloped, is peculiarly its own. It reflects both thebalance of forces in its rather stable political system andwhat has been called "fiscal inertia," or the tendencyof fiscal institutions to persist and change graduallyrather than radically.

A continuing undercurrent has been the inability oftax administration to cope with direct taxes in a dis-torted inflationary environment; hence the growingimportance of both the sales tax and presumptive in-come taxes.

Above all, the Colombian experience suggest that taxreform is inherently neither a continuous nor a once-for-all process, but a periodic one. The almost-annualminor changes in tax bases and rates common in manycountries are not usually enough to accommodate fun-damental changes in the economic and political envi-ronment of developing countries. Circumstanceschange, and tax systems must change with them. TheColombian case shows that such adaptative efforts areinevitably affected by external circumstances, the polit-ical context, and administrative constraints. They arenot always successful. However, Colombia's relativesuccess in maintaining government revenue, and evena moderate degree of progressivity, in the face of con-siderable adversity also suggests that tax reform is notbeyond the reach of a developing country.

tariffs. Export taxes should be phased out or rede-signed in light of their primary functionfor exam-ple, as a proxy for taxing income, profits, or eco-nomic rent.

Income taxes could be simplified as follows.A restructuring of company taxes so that aver-

age effective rates are high for revenue purposesand marginal effective rates low for investmentpurposes. This can be achieved through somecombination of better approximations of annualeconomic depreciation rates, elimination of sector-and asset-specific allowances, lower statutory taxrates, and adjustments for inflation where inflationrates are high (say more than 10 to 15 percent).Eliminating double taxation of dividends and im-proving the links between personal and companytaxes is also desirable.

103

A restructuring of personal taxes to include allsources of income, with lower maximum rates,fewer brackets, higher exemptions, and the elimi-nation of most existing special allowances. Index-ing wifi be important where inflation is high.

Strengthen tax administration

Administrative procedures, capabilities for dataprocessing and analysis, and staff training must beimproved in all types of tax reform. A reform thateliminates multiple adjustments to the base of a taxand reduces the number of rates can go a long waytoward improving administration. Increased ad-ministrative costs, however, may sometimes bejustified to lower economic costs, for example, byshifting from trade taxes to a VAT

In the abstract there is no reason to prefer com-prehensive over partial reforms. Country-specificneeds wifi determine the reform required. Com-prehensive reforms need not be avoided on thegrounds of overloading administrative capacity.The elements of a comprehensive reform can beintroduced simultaneously or in stages in light ofrevenue and administrative constraints. New taxinstruments have been introduced successfully, asin Colombia and Korea. Implementing a taxchange, however, is likely to be easier if the reformbuilds on existing tax instruments, as in India andMalawi. The benefits of tax reform take time tobecome apparent.

Avoid taxing the poor

The equity characteristics of actual tax systems can

104

be improved by incorporating moderate progres-sivity in both income and commodity taxes and bysimplifying tax instruments to free administrativeresources for collection, auditing, and enforce-ment. The progressivity of the tax system can beenhanced by exempting the income and essentialpurchases of the poor from the tax net, by elimi-nating most of the income tax deductions and al-lowances that primarily benefit the rich, and bysubjecting luxury purchases to higher rates of taxa-tion. The revenue loss resulting from high exemp-tions in the personal income tax can largely be off-set by eliminating most allowances.

Improving the resource allocation aspects of atax can also improve its equity aspects. For exam-ple, eliminating taxes on production inputs en-sures that nominally exempt basic goods are notinadvertently taxed. Efforts at fine-tuning the taxstructure to achieve income redistribution objec-tives, however, are not likely to be successful inpractice. Poverty alleviation can better be servedthrough coordination with other policies, espe-cially on the spending side of the budget.

Careful reform of revenue instruments can en-hance their contribution to revenue and minimizetheir social and economic costs. But the remainingcosts of raising revenue, in effect, set a floor to thebenefits required from public spending. Accord-ingly, revenue should be planned jointly withspending: cost-benefit considerations apply toboth parts of the budget. The next chapter turns tothe expenditure side of this equation.