Taxation and Development Author(S): Robin Burgess and Nicholas Stern Source: Journal of Economic Literature, Vol

American Economic Association

Taxation and Development Author(s): Robin Burgess and Nicholas Stern Source: Journal of Economic Literature, Vol. 31, No. 2 (Jun., 1993), pp. 762-830 Published by: American Economic Association Stable URL: http://www.jstor.org/stable/2728515 . Accessed: 09/09/2013 09:19

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to Journal of Economic Literature.

http://www.jstor.org

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Journal of Economic Literature Vol. XXXI (June 1993), pp. 762-830

Taxation and Development

By ROBIN BURGESS and NICHOLAS STERN STICERD,London School of Economics

We are grateful to Charles Bean, Willem Buiter, Nigel Chalk, David Coady, Francisco Ferreira, Haris Gazdar, Stephen Howes, Jenny Lanjouw, Robin Sherbourne, and Jim Thomasfor advice and assistance, and to Christine Wu and Vito Tanzi of the Fiscal Affairs De- partment of the International Monetary Fund (IMF) for help with the data in Section 3. We have benefited greatly from the helpful comments on an earlier draft of Ehtisham Ahmad, Tony Atkinson, Richard Bird, Sherman Robinson, and anonymous referees. This research was supported by the Suntory-Toyota International Centre for Economics and Related Disciplines at the London School of Eco- nomics.

1. Introduction expenditure.' We may see the problem of tax design as one of finding a way of THIS PAPER CONCERNS the theory and raising these resources which is adminis- practice of taxation in poor countries. tratively and politically feasible and We should ask at the outset whether the which promotes equity and efficiency as analyses of taxation for developed and far as is possible. There must clearly be less-developed countries are, or should some trade-offs among revenue, adminis- be, essentially similar, or whether tration, political acceptability, equity, the objectives, responsibilities, and con- and efficiency. The problem of tax reform straints are so sharply at variance that is to find an improvement, with respect wholly different approaches are required. to these criteria, of an existing system. We shall argue that there are many prob- Equity and efficiency, in much of the for- lems and questions in common and mal, normative part of the analysis, will that, broadly speaking, similar analytical be considered in terms of the well-being methods can be used. At the same time of households. These criteria and con- there are some issues which arise in less- cerns may be common to governments developed countries which are not faced in both developed and developing coun- in developed countries, and some prob- tries. lems which are much more acute. The While at a general level the objectives analysis of taxation in less-developed countries may go badly wrong if it fails ' Other importantobjectives are redistributionand to take these into account. the control of externalitiesor other marketimperfec- tions; these are embodied in what follows. The main The main purpose of taxation is to purpose of negative taxes (subsidies)is to redistribute raise resources to finance government or to encourage certain activities. 762

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 763 of taxation for developed and developing why the role of government should be countries may be alike there may be im- more than minimal in developing coun- portant intermediate objectives or em- tries (see Jean P. Dreze and Amartya K. phases which are different. Some com- Sen 1989; Burgess and Stern 1991; Stern mentators have argued, for example, that 1991a; and Section 2.1 below). These fo- there should be a particular concern in cus attention on the need to raise suffi- developing countries with raising the cient revenue and in Section 2.2 we pro- rate of savings, with incentives for invest- vide reasons as to why the bulk of ment, or with self-sufficiency. We would revenue should be raised through taxa- suggest, however, that such intermediate tion. Further, it will be a major theme objectives, if they are to be regarded as of this paper that taxation and expendi- persuasive, should be derived from a ture should be analyzed together. The broader concern with standards of living problem of taxation should not be seen of individuals of current and future gen- simply as the use of taxes to fund a fixed, erations. exogenously given, expenditure. We must also ask whether the per- Arguments concerning the overall ceived responsibilities of government level of government activity and the bal- are, or should be, different for develop- ance of its finance between taxation, bor- ing countries. More broadly, should rowing, and money creation are provided there be a different role for the state in in Section 2. In Section 3 we examine these countries? It may be argued that tax structures which are actually ob- market failure is more prevalent in back- served. These vary widely and we com- ward economies so that there is a greater ment briefly on reasons for the variations. justification for government intervention The basic principles of taxation for devel- using, for example, corrective taxes and oping countries, derived from formal the- regulatory instruments. Widespread vul- orizing, are considered in Section 4. The nerability to extreme deprivation may formal theory abstracts from many diffi- also point toward state intervention. In cult conceptual and practical problems, recent years, however, many have some of which are discussed in Section stressed the problems of government fail- 5. From -the perspective of taxation, ure, associated with, for example, rent these problems constitute the most im- seeking, corruption, and inefficiency, portant difference between developed which should be set beside those of mar- and less-developed countries. Some ex- ket failure in assessing the desirability periences of tax reforms in less-devel- of government action. It may even be oped countries are described in Sec- argued that government failure is so se- tion 6. In the final section (Section 7) vere in developing countries that the we pull together the lessons from theory level of government action should be and from experience for the design of (proportionally) lower than in developed tax reform and provide guidelines for pol- countries. The judgment of the relative icy. severity of the problems of market failure and government failure is not our main purpose here (see Stern 1989 and 1991a, 2. Taxation, the Role of the State, and the for a discussion). We would argue, how- Macroeconomy ever, that the very low living standards of some groups, missing markets and the The first part of this section (Section comparative advantage of government in 2.1) examines the arguments concerning some areas (e. g., infrastructure, social the appropriate nature and scale of gov- sectors) provide strong arguments as to ernment involvement in the economy,

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 764 Journal of Economic Literature, Vol. XXXI (June 1993) i.e., it considers the role of the state in ment and analysis of the problems of gov- economic activity. A view of the appro- ernment action, or potential "govern- priate type and level of government ac- ment failure." tivity influences one's judgment not only (i) Market failure: including, notably, on the level of taxation but also its struc- externalities, public goods, and missing ture, as we shall see in Section 4. The markets, in addition to conditions leading problems and potential of taxation (dis- to the violation of perfectly competi- cussed in Section 3 and Section 5) will, tive behavior such as imperfect infor- at the same time, influence judgments mation, increasing returns, and entry bar- as to the appropriate scale and balance riers. of expenditure. (ii) Poverty and income distribution: Government activities are not, how- outcomes, whether efficient or other- ever, financed only by taxation. There wise, may be such as to leave some mem- are, as we describe in Section 3.2, many bers in situations of extreme deprivation other sources of public or government or result in a distribution regarded as un- revenue: public-sector profits, mineral acceptable or unattractive. royalties, net revenues of marketing (iii) Rights to education, health, nutri- boards, and so on. These sources of gov- tion, and housing: many would argue ernment income should, however, be ap- that "equality of opportunity" implies a praised in much the same way as taxes state responsibility to provide literacy (see Section 4) and, for the moment, we and to ensure basic nutrition, health can include them under the heading of care, and shelter, without which an indi- taxation. In addition to taxes, defined vidual's participation in the economy and broadly, the government can meet its ob- society are severely limited. ligations by borrowing domestically or in- (iv) Paternalism: the state may decide ternationally, or by printing money. it has a superior view of individuals' own These other sources of finance have their self interest than individuals have them- own costs and disadvantages and in the selves and further that it should, in some second part of this section (Section 2.2) cases, override individual preferences; we consider the balance between taxa- relevant examples may include the re- tion, borrowing, and money creation. quirement to attend school, limitations Taking Section 2.1 and Section 2.2 to- on the use of certain drugs, and compul- gether, this section assesses the role of sory pension schemes. taxation in the economy as a whole. (v) Future generations: private individuals acting in their own self interest 2.1 The Role of the State and and that of their descendants may not Government Expenditure make decisions which take the welfare or rights of future generations appropri- Analyses of, and judgments on, the ately into account; relevant examples appropriate scale and balance of expendi- may include global warming, air and wa- ture should begin with the reasons for ter pollution, conservation of rain forests, government intervention. We may dis- protection of species, and so on. tinguish five groups of arguments. The All of the above arguments provide co- force of these arguments, some but not gent reasons for government action. all of which concern "market failure" in They point fairly directly, through the terms of departures from the standard examples given, to particular areas of conditions of welfare economics, can government expenditure, notably infra- then be examined alongside an assess- structure, social security, education,

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 765 health, pensions, and the environment, politics are honest and constructive, one as well as to the government's role in has to take into account the cost of keeping the economy competitive. Mar- raising resources. But the problems are ket failure arguments are especially more fundamental than this. Govern- persuasive concerning infrastructure ments may be corrupt, manipulative or (power, communications, and so on), manipulated, and inefficient. Indeed the where increasing returns, public goods, self-interest of participants, -inside and and externalities can all be of considera- outside government, may make such out- ble importance. In addition, legal and comes likely. Many have argued, follow- regulatory structures that, for example, ing the political economy approaches of ensure that property rights are well de- those such as Hayek and Buchanan that fined and respected, contracts enforced, the problems are so severe that the size and illegal activity contained are essential of government should be kept as small for the competitive functioning of the as possible. Some moral philosophers economy. To these tasks one has to add such as Robert Nozick (1974) have come basic administration, law and order, and to similar conclusions on ethical grounds defense.While this brief review of the concerning the rights of individuals arguments for state action has identified against the state. Over the 1970s and important (and costly) areas for state ac- 1980s many writing on less-developed tivity, it does not provide a convincing countries (see Anne 0. Krueger 1990, rationale for direct state involvement in for a recent expression) have emphasized the production of ordinary producer and the problems of rent-seeking and corrup- consumer goods, at least from the per- tion during economic development. It is spectives included here (for elaboration, argued therefore that the extent of gov- see Stern 1991a). ernment action should be substantially In Table 1 we present figures on the limited, and the types of measures em- actual pattern of expenditures in indus- ployed by the state be designed with the trial/developed and developing coun- problems of manipulation and corruption tries. Governments in developed coun- in mind. tries (i) spend more as a percent of GDP These arguments are of considerable (31.5 percent compared to 25.4 percent) importance but so also are the arguments and (ii) have expenditures that are more for government intervention. In less- concentrated on social security (particu- developed countries both the difficulties larly) and health, whereas in developing of and needs for state action would ap- countries these areas exhibit a smaller pear to be stronger than in their indus- share with the bulk of expenditures going trial counterparts. In the light of the to economic services, general public ser- scope and magnitude of problems facing vices, and defense, though education also the developing world, and given strong obtains a substantial share. In the light arguments for state intervention in cer- of the arguments for state action identi- tain areas, resort to "government failure" fied, attention should be paid to both arguments to justify minimal state activ- the level and composition of expendi- ity would, however, appear to represent tures. an unreasonable and inadequate re- The identification of important and sponse. We would suggest that the gov- costly areas for government action does ernment call on resources is likely to be not by itself justify extensive interven- major and therefore the magnitude of tion. Even in a system where govern- the task of taxation likely to be substan- ment tasks are efficiently discharged and tial.

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions " o 010 'C10 0

lq Ittc ;0 -4

C)>

I~~~~-. ~ t Q c- o t-: --c o N A:~~~~~~~t-C6 C's6 CO 0

U =

H SW 0 = Si .

0 i

S~~~~~~~_ n L-L- 6 46t o 0;..t o o~gG>.c 8o C) ~kfQCDX ~ - Iq C-0 C

I! -~co. - c.

~~oo,- Q.5: 8 @z ~ 6 cor--~~. C,~ ~ 8 oo ct-4 t6 c, r-iO

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 767

2.2 Financing the Spending countries, comparing the effects of taxation and borrowing on aggregate de- How is the expenditure to be fi- mand. They conclude that full Ricardian nanced? We have essentially four equivalence can be rejected for 15 of the sources: government revenue, internal 16 countries. Problems of model specifi- borrowing, external borrowing, and cation, measurement of variables, and printing money. Our emphasis in this pa- data quality in this area should warn us per is on government revenue and we against the general applicability of results consider only briefly in the remainder from empirical studies.3 What remains of this section some effects of other clear and intuitive is that due to greater sources relevant to an assessment of the problems of distorted capital markets, re- appropriate balance. We shall argue that, pressed or underdeveloped financial sys- in the long run, a healthy tax system must tems, and the haphazardincidence of dif- be at the heart of the public finances. ferent taxes, conditions (i) to (iii) are The adjustment experiences of the less- likely to apply with less force in develop- developed countries over the last two de- ing than in industrial countries. Given cades has served to elevate fiscal issues the weak empirical evidence for each of to the head of the policy agenda. Fiscal (i) to (iii), taken individually, we would, correction is now seen as essential for a priori, not expect Ricardian equiva- long-term macroeconomic stability (see lence to apply in the majority of develop- World Bank, 1991, for a discussion).2 ing country settings. W. Max Corden There are very special models in which (1987) also makes the interesting obser- extra domestic borrowing has the same, vation that extra borrowing (foreign or or "equivalent" effect on the economy domestic) may lead to capital flight in as extra taxation (see Robert J. Barro the anticipation of future tax burdens. 1974, and B. Douglas Bernheim 1987 for One could add that "no avoidance by em- a review). Strong conditions are required igration" should be included in the nec- for "Ricardian equivalence" to hold, in- essary conditions for "equivalence." cluding: Given that debt is not neutral in the (i) perfect capital markets and ra- sense of equivalence, we may ask what tional expectations, constrains its use. Essentially there are (ii) private agents being able to bor- two related problems, inflation and row or lend on the same terms as crowding out of private investment. the government, Debt can be inflationary because a gov- (iii) non-distortionary taxes. ernment will have an incentive to renege on part of the debt by a burst of unex- While the conditions look implausible pected inflation. Further if debt is used as a description of less-developed coun- excessively, there will come a time when tries it is nevertheless an empirical mat- it will no longer be accepted and the gov- ter as to whether the "equivalence" prop- ernment will have to resort to monetizing osition holds good. Nadeem U. Haque the debt. If sufficiently anticipated this and Peter Montiel (1987) attempt to test can lead to inflation now (see Thomas J. the proposition for sixteen developing Sargent and Neil Wallace 1981).4 2There has hardlybeen an adjustmentprogram 3The evidence from developed countries where in recentyears that has not seen the reformof the empiricalresearch has been more intensive is incon- taxsystem and the containmentof publicexpenditure clusive and difficultto interpret (see Bernheim 1987 as centralobjectives (World Bank 1991). Borrowing for a discussion). andmoney creation are no longerseen as sustainable 'On the relationshipbetween debt and inflation or reliablesources of governmentfinance. in Latin America in the 1980s, see Jeffrey D. Sachs

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 768 Journal of Economic Literature, Vol. XXXI (June 1993)

Closely associated with these inflation- in taxation, which may be difficult or ary aspects of debt financing are the costly to implement (Richard Goode problems of crowding out. Given the 1984, p. 197). As with other schemes, non-neutrality of debt, the authorities' however, the surpluses in "randomly decision making directly affects the pri- good" years may not be generated to fi- vate allocation of resources. For exam- nance the deficits in the "randomly bad." ple, substitution by the government of Simple rules of thumb in relation to do- borrowing for current taxes on labor in- mestic borrowing have occasionally been come may increase private consumption. used to guide policy (see Goode 1984, In an economy with full utilization of re- pp. 197-98). From the perspective of sources this may lead to private invest- standard economic analysis they gener- ment being "crowded out," or may result ally have serious shortcomings, although in an increase in the deficit on the cur- they can have advantages in constraining rent account of the balance of payments governments from pursuing undesirable (see Willem H. Buiter 1989). policies (also see Section 4.5 below). For The role of domestic debt finance will example, one view is that it is appropriate depend in part on the degree of moneti- to use borrowing to finance investment, zation in the economy and the sophistica- but not consumption, outlays on the tion of the financial markets. The experi- grounds that returns to finance the debt ence of low-income countries in South service and repayment will be generated. Asia has shown that relatively complex This rule fails to distinguish between the financial markets can develop within a relative costs and benefits that accrue short period. This can have an impact from financing "the project" from alter- on the composition and level of savings. native sources. Borrowing will involve For example, in India between the early greater taxation, or reduced expenditure, 1950s and the early 1970s the proportion in the future and the government will of household savings going into financial have to compare the costs of doing this assets rose from around 4 percent to over with taxing now. 50 percent (Government of India, Raj Foreign borrowing differs from domes- Committee 1982). Given that tax reve- tic in that it enables a country to com- nue involves collection costs and gener- mand current goods and services in addi- ates "consumer resistance," which may tion to those arising from domestic be substantial in developing countries, resources. Furthermore, the displace- and that the purchase of government se- ment of private investment may be less curities is more or less voluntary, govern- severe, although public foreign indebt- ments may well be tempted to make edness may make finance more difficult more use of the latter if financial markets for private borrowers. Over time debt and institutions permit. servicing and repayment entail a reverse Governments may seek to use domes- transfer of resources abroad. For foreign tic borrowing when they face an unex- borrowing to be attractive the govern- pected shortfall in revenues, or to cover ment should believe that the productiv- emergencies such as floods or defense ity of the loans along with the taxes gen- requirements. Such borrowing can sub- erated will provide sufficient resources stitute for sharp and substantial increases for the loan to be serviced and repaid in the future. Excessive optimism can lead to serious problems as we have seen (1989)and on the problemsand circumstancesassoci- ated with reneging on debt, see Jonathan Eaton, with the debt crisis of the 1980s (Sachs MarkGersovitz, and Joseph Stiglitz (1986). 1989).

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 769

Governments have at times resorted The evidence would suggest that the rev- to financing expenditure by printing enue potential of noninflationary money money. In this way governments can finance is limited. Seignorage typically claim real resources from the population. accounts for only about 0.5 percent of This claim is known as seignorage. A GDP in stable, low inflation countries. measure of seignorage can be taken as Even moderate money financing of the the product of the money growth rate order of 1-2 percent of GDP has been (a) and money balances (M). Thus we associated with inflation in excess of 50 have, in real terms, the seignorage, R., percent in many developing countries in given, where P is the price level, by the 1980-85 period (World Bank 1988). High inflation rates can be associated Rs = a MIP. (1) with claims on resources that are significant. For example, in Bolivia and Argen- Because demand for monetary assets tina, seignorage accounted for 4.0 and tends to rise with economic growth the 6.2 percent of GDP respectively in 1980- government can, to a limited extent, fi- 85 (World Bank 1988).6 nance itself by expanding the monetary High inflation, however, has many dis- base without causing inflation. However, ruptive and distortionary effects on eco- when the rate of new money creation ex- nomic functioning (Fischer 1986; Sachs ceeds the increase in the demand for 1989; and Buiter 1989). In addition, due money, inflation may result. Inflation to collection lags and the structure of then imposes a tax on the population by taxation, inflation will have a negative reducing the value of real money bal- impact on real tax revenues in less- ances (see Martin J. Bailey 1956). The developed countries (see Section 5.4 and cost of this tax (Re)may be written as: Vito Tanzi 1991a). Further, the inflation RRi=,TrMIP (2) tax may hit the poor, particularly as they tend to hold a large proportion of their where aTis the rate of inflation. This is wealth as currency (see e.g., Francisco not a revenue directly received by the Gil Diaz 1987). Taken together these fac- government in the particular period at tors mitigate against consideration of issue but is a capital loss sustained by seignorage or the inflation tax as attrac- holders of money. The terms R. and Ri tive methods of public finance. are sometimes used interchangeably Two central messages have emerged though the two quantities are identical from the experience of less-developed only when a = Tr,for example in steady countries during the last two decades. states (see Olivier J. Blanchard and Stan- First, there is no viable, long-term, and ley Fischer 1989, p. 188). Out of steady substantial alternative to taxation as a state there will be an intertemporal link means of financing government expendi- between the resources captured by the tures. Second, successful adjustment and government through seignorage and the macroeconomic stability rely on pro- capital losses suffered by the population. 5 found and permanent fiscal corrections being made in the short and medium run (see Sweder van Wijnbergen 1989; Mario 5To keep things simple we will be looking only at currency held by the public. However, it should Blejer and Ke-young Chu 1989; Tanzi be kept in mind that inflation will tend to erode the value of domestic government debt taken as a whole which includes currency, interest and nonin- 6 Note that total tax shares for these countries in terest bank deposits, and government securities (Gil 1987 accounted for 5.9 and 16.9 percent of GDP Diaz 1987). respectively (see Table 2).

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 770 journal of Economic Literature, Vol. XXXI (June 1993)

1990; Ronald I. McKinnon 1991; World taxation and on average the share of non- Bank 1991; Pedro Aspe Armella 1992). tax revenue in total revenue is much At the heart of these fiscal corrections larger than for industrial countries (see are structural changes involving tax re- Section 3.2). form. Taxation is thus a central concern Much of the data on the existing tax which cannot be relegated to the medium systems of developing countries pre- or long term. sented here come from IMF statistics (see IMF 1989).8 Our eighty-two country 3. Sources of GovernmentRevenue sample includes most less-developed countries, all having per capita In all countries government revenue incomes less than $6000. Given the difficulties is raised from a number of sources and of collecting comparable the size through a variety of mechanisms. In this data, of the IMF sample is section we provide a description of those impressively large. Data on twenty-one industrial sources, examine the mechanisms and at- countries, all with per capita incomes greater than tempt to explain their relative impor- $6000, are provided for comparison. tance in different countries. The main In Table 2, (see appendix) we set out source of government revenue is taxa- the basic data on total tax tax tion, which contributes close to 80 per- revenue, and revenue by type of tax, as a cent of total government revenue for less- percentage of GDP for each of these countries. Table developed countries as a whole (IMF 3 summarizes this 1989). In Section 3.1 we describe the information by averag- ing over five income level and structure of taxation in develop- classes while Table 4 shows regional averages (as ing countries, contrasting these with the percentages of GDP). The situation in industrial countries, and in percentage breakdown of total tax revenue on a regional basis is Section 3.2 provide a brief investigation provided in Table 5.9 of non-tax sources of revenue of which The GNP per capita (1987 dollars) figures are calculated in the there is a rich variety in developing countries. conventional manner without making ad- justments for purchasing power parity 3.1 The Level and Structure of Taxation (see, e. g., Robert Summers and Alan in Developing Countries7 Heston 1988, and Stern 1989, on the effects of such an adjustment). Presentation The overall proportion of taxation in of the data, GDP is higher in industrial than develop- essentially using a static, cross-sectional framework based on GNP ing countries and there is a greater share or GDP of direct taxation in that total. Tax struc- per capita as a basis, is somewhat crude. The data and method, however, ture varies strongly between countries should be sufficient and across regions. For most countries to give an impression of some of the broad and taxation constitutes the main source of regularities differences. government revenue, but some developing countries obtain a substantial share 8 Though the data are the best availableon a cross- of total revenue from sources other than countrybasis there are problems including:(i) unreli- ability of data due to factors such as poor and biased accounting,(ii) hidden variabilityand diversity across 7 This section builds on the work of Tanzi (1987) tax systems, (iii) underrepresentation or complete and on data prepared by Christine Wu of the Fiscal absence from the data of some taxes (e.g., local AffairsDepartment of the IMF. We are very grateful taxes). to them for their advice and assistance. Note that 9 Breakdownsare presented as averages over the Table 2 follows the text as an appendix. The terms three years closest to 1987 for which data are avail- "industrial"and "developing"follow IMF classifica- able-some year-by-year fluctuations are smoothed tions. in this way.

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions to r 0 11 o I atOK e 0 II OOb.] 9~~~~~~ to to o Go Zj:

C;ID ;]C C C; CC; C;, C

0'-~~~~~~~~~~~~~~2

H~~~~~~~~( cc ltr* :

c~i H: -: 0~f00 0~ 0 g.

| 2o | E - | Ot- t- c t- | 8

N 98 E = 4 S 0iC) G 6 C; C6 o6 9

0.00 C) t- In C,) t

Z: } | > | t t"0 |,) C,eU) e CD | C4 0 Y ?1 tC

_C,c) ? i

E c; | 8 2 x t1 3 4 11 ,, E

o _ t 7 - '~~~~~~~C~~-1 C,, It -4

> ec Go>>c 1a :38a5C l 0 C4 k4 - 0- '4 4 ~~'~~~~00~ lll~ elt~~~~~~~,crto|V 0~~~~~~~~~ 9 : c 8e;2

C.0~~~~~~~~~~~~~~~.

~~~~~~~~~~~~~zC5~~~~~~~~~~~~~~C

4 o6 ci o - 'Z4H~c

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions C.-)O O ?C

nt84 t 1 X _ O O e t- 1ti 8 e

D o gD ct6 g i6X -

;~~~~~~~~~~~~~~~~~~~~~~~

~~C 3 i g t X > > t 0 e 4 o a @ o~~~~~~~~CAl|: w=>|eN ~~~~~ ao m ?O oq U 2Nt Q~~~~~~~~~~~~~~~Nt t~~

~~| : | ||C0 - 6 6 x c E m c .R l~~

~~r oo o '"So t-z vv~~

~~40 16 _Oi _ oi C,> oi Z tO8 o2Wo~~

~~I;| E : gguu~~

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions oo0ea 0) m O oica_ 0 ces ^~~

~~3 cl ci14( :4~~

~~Co 0c to *C) X 0 X 0~~

~~0 8o6 c6 -i '>t-.8 v~~

~~| tog o NO~~~~~~8t C,O e S~~

~~E o .R g ffi ~~~~~~~~~~~008 Ot~~

~~0~~~~~~~~~~~~~~v Eg |> it at sm 8 <||5 o w 0tc e S.~~

~~Z 2 ~~~~~~~~~~~~~Ooi c6eg 6 t6 .~~

~~0So~~~~~~~~~~~~~~~~~~tHX0 ? F b Z~~~~~~~~~~~~~~~~~~~~.R .c in JO@~~

~~> E l | a | ' mQ NcF s X t V)e > 3 || * 3 ^~~

~~Z ~ OC m -- -- C,, ~~~~~~~~~~~~~~~~~o~~~~~~~~~~~~~~~~~~lC'soCa 11 4008GH 16 4.:~~

~~<>L)O a eEW||XXQ~~

~~I~~~~~~~~~~~~~, Z o6 c'6 o i C'E||7; =B 4-JQ** *>13E~~

~~0 ct, )~~

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 774 Journal of Economic Literature, Vol. XXXI (June 1993)~~

3.1.1 Level of Taxation. Among the The data may be described in terms developing countries the lowest propor- of a regression of a tax ratio on log GNP tion of total tax revenue in GDP is found per capita for the whole sample of 82 in Sierra Leone at 5.5 percent and the countries (Equation 3). The coefficients highest in Guyana at 40.5 percent (Table in parentheses are t values. 10 For 82 de- 2). The (unweighted) average ratio of to- veloping countries: tal tax revenue to GDP, for the eighty- Total tax revenue two developing countries, is 18.1 percent compared with the figure of 31.2 percent GDP for industrial countries. The ratio is 14.0 = 5.78 + 1.84 log(GNP per capita). percent for the twenty countries with per (1.02) (2.02) capita income less than $360, however, R2 = 0.04 (3) the 21 countries and 19.8 percent for The finding then is of a weak but signif- with per capita income of $1620-$6000. icant (at the 5 percent level) relationship the 41 countries in between, it is For between tax share (T/GDP) and income about 19 percent (Table 3). At the re- for the whole sample of 82 developing gional level, the Middle East and Asia countries (Equation 3). The very low R2 tax ratios of between 14 and exhibit low indicates a wide scatter of points around 15 percent, the Western Hemisphere is the linear relationship between T/GDP higher at about 18 percent and slightly and log GNP and any commentary should Africa the highest with almost 20 percent bear this firmly in mind. The positive (Table 4). Among industrial countries 76 coefficient in (3) indicates that an increase ratios above 25 percent percent have tax in GNP per capita is associated with an compared to 16 percent of developing increase in the tax share. This finding is countries, and not a single industrial consistent with the idea that the ability country has a tax ratio below 15 percent to tax grows faster than income. And it of GDP (Table 2). is in accordance with Musgrave's (1959, The patterns shown in Tables 2-4 sug- 1969) idea that the "handles" to which gest that though higher income exerts the revenue system may be attached be- some influence on overall tax share, there come more diverse with economic devel- are many other influences involved; his- opment resulting in a widening in the torical, social, political, and economic tax base (see also Harley H. Hinrichs factors all contribute to the explanation 1966). Not only are new sources of reve- of differences in both the level and struc- nue tapped but there may be extensions ture of taxation. Studies such as Richard of the coverage of different types of tax. A. Musgrave (1969), Raja J. Chelliah, Ability to tax is closely associated with Hassel J. Baas, and Margaret R. Kelly administrative capability and this is likely (1975), Alan A. Tait, Wilfrid L. M. Gratz, and Barry J. Eichengreen (1979), Guido Tabellini (1985), and Tanzi (1981, 1987 0 The point of the regression analysis here is sim- and 1991b), indicate that other factors, ply to describe a mass of data and to discern broad for example, the share of non-tax reve- patterns and regularities. Not too much importance nue, import and export ratios, literacy should be attached to them as models of processes. Other regression forms are also in common usage. rates, urbanization, level of indebted- The coefficient on income in a time series log reve- ness, the share of agriculture, and the nue-log income regression gives a measure of "tax monetization and openness of the econ- buoyancy" which tells us whether a given tax revenue is growing faster than income. These estimates are, omy are all correlated with the total share however, unsatisfactory in that they conflate policy of taxes in GDP. and economic change.

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 775 to improve with economic development duties, sales taxes, and excises). The thus reducing the gap between the statu- need to service rising foreign debt often tory tax system and the real or effective generates tax increases.12 A high share tax system (see Section 5). As tax bases of agriculture in GDP proxies the diffi- broaden the temptation to impose high culty of imposing taxation (see Section rates on constricted bases lessens and this 4.3 and Section 5.3). may have a positive effect on enforce- 3.1.2 Structure of Taxation in Devel- ment and hence revenue (Malcolm Gillis oping Countries. In this subsection we 1989a). discuss the relative importance of various If the sample is split into two (ranking types of taxes in developing countries. by income per capita) around the median The situation in industrial countries is income per capita we find that the rela- presented as a comparison. Examining tionship is stronger for the poorer subset. the top two rows of Tables 4 and 5 we In the richer developing countries there see that there are marked differences be- is often a substantial revenue contribu- tween developing and industrial nations. tion from such sources as state-owned en- Developing countries obtain the bulk of terprises and mineral resources which is their revenue from (i) domestic taxes on not captured by tax statistics. For the goods and services (5 percent of GDP richer subset of 41 countries, the rela- and 30 percent of tax revenue-partly tionship of the type embodied in Equa- from taxes on sales and partly from ex- tion (3) is insignificant. In industrial cises), (ii) foreign trade taxes (5 percent countries we also find an insignificant re- of GDP-mainly import duties) and (iii) lationship. There are fewer constraints income taxes (6 percent of GDP-mainly on raising revenue in rich countries and on corporations). In contrast, the three the level of taxation is a reflection more big sources of government revenue in in- of the orientation of government policy dustrial countries are (i) income taxes (11 rather than the level of development." percent of GDP and 36 percent of tax A central finding then is that per capita revenue- mainly on individuals), (ii) do- income is not particularly good at "ex- mestic taxes on goods and services (9 per- plaining" the level of taxation. Tanzi cent of GDP and 29 percent of tax reve- (1991b), who has tracked a similar sample nue mainly on sales) and (iii) social of 88 less-developed countries between security contributions (9 percent of 1978 and 1988, identified the share of GDP). imports in GDP and the share of foreign The different makeup of taxes in devel- debt in GDP as having a substantial posi- oping and industrial economies may be tive impact on tax levels, and the share illustrated using the triangle diagram of of agriculture as having a negative im- Figure 1. The points A, B, and C in the pact. Together these three variables ac- triangle represent 100 percent of tax rev- count for almost half of the variation in enue from income and social security levels between countries. Per capita in- taxes, 100 percent from trade and other come had a smaller impact, which de- taxes, and 100 percent from domestic in- clined over the decade. There are several direct taxes, respectively. A point on the reasons for the importance of these fac- line BC corresponds to a zero level of tors. Imports constitute a significant tax income and social security taxes, and sim- base for various types of taxes (import ilarly the lines AC and AB represent the

11Compare, for example, tax levels in Canadaand 12 See Tanzi (1988, 1991a) on the effects of macro- the USA to those in many of the European states economic policies on the level of taxation in less- (see Table 2). developed countries.

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 776 Journal of Economic Literature, Vol. XXXI (June 1993)~~

Domestic Indirect Taxes Income taxes Income taxes from all sources account for 5.5 percent of GDP and 28.9 percent //f / // of total tax revenue for the 82 country sample of developing countries. This may /+ + \ be compared to industrial countries /+ O+ +/ 000 ++ where income taxes comprise 11.0 percent of GDP and 35.8 percent of tax revenue. Personal income taxes are roughly X0 / + ++ \+ constant (and low) over all income ranges

/ +/ +/+ + + + + + (except for the lowest, <$360) while corporate income taxes exhibit a bell-shaped + + /+++ + ++ + relationship with income. For rich coun- A Z t + + B Income Taxes and SS Trade Taxes and Others tries (>$6000) personal income taxes, to a substantial extent, replace taxes on cor- Legend +:Industrial Countries porations (Table 3 and see below). o :Developing Countries Individual income taxes account for 2.1 Figure 1. Breakdownof TaxRevenue by Type: (8.5) percent of GDP and 10.6 (27.7) per- Industrialand Developing Countries cent of total tax revenue for the 82 devel- Sources:IMF Government Finance StatisticsYearbook oping (21 developed) countries exam- (1989) and Table 3. ined. There is no significant relationship Note:We are grateful to Tony Atldnsonand Stephen Howes with GDP for developing countries. for guidance and help with this diagram.SS= social security The constraints on raising revenue contributions. through personal income taxation in developing countries are many and include other axes. It can be seen that the indus- problems of income measurement, ad- trializedcountries cluster close to the AC ministrative capability, low literacy, and axis and towards the direct tax corner poor accounting. An economic structure (A), whereas the developing countries dominated by agriculture and small-scale are spread out with an average close to (often unregistered) enterprises makes it the center of gravity of the triangle (i.e., difficult to trace, and hence tax, incomes where one-third of revenue comes from (see Musgrave 1969; Goode 1984; Rich- M. each category)+ 13 ard Bird and Oliver Oldman 1990). We examine more closely in the analy- As a result, revenue from this source in sis below the relative importance of in- developing countries tends to accrue come taxes, domestic taxes, trade taxes, largely from taxes on the wages of em- social security contributions, and wealth ployees in public-sector enterprises and and property5).3 taxes (Tables 2 to foreign corporations to whom tax laws can 13 Over 95 percent of industrialcountries have less be more easily applied. Collection from than a 15 percent share of trade taxes in their tax such a narrow base, often at high rates, totalswhile more than 86 percent of developing coun- creates resistance which is apparent in tries have a share above 15 percent (see dotted line). While many developing countries have a sizable di- the experience of a number of countries rect tax component this mainly consists of corporate income taxation,the individualincome tax and social security contributionsbeing relatively small. revenue for the Western Hemisphere), to taxes on 14 One might suggest from Tables 4 and 5 a rough individual incomes and social security contributions progression (with income) from import duties (e.g., (e.g., 56 percent of tax revenue for industrialcoun- 28 percent of tax revenue for Africa), to domestic tries). This is by no means a fixed progression;how- taxes on goods and services (e.g., 37 percent of tax ever, a broad pattern is discernible.

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 777~~

(see Gillis 1989a; Tanzi 1991a). In indus- This share of tax revenue is almost identi- trial countries, wage and salary employ- cal to that in industrial countries (29.3 ment is more widespread and the capa- percent), though the share of GDP is bility to tax those outside this net (e.g., lower. The revenue raised through do- the self-employed) greater, thus partly mestic taxes on goods and services by explaining the greater importance of in- developing countries may be seen as an dividual income taxation in revenue. impressive achievement and its share in Corporate income taxes are more im- total tax revenue has been on the rise portant than taxes on the incomes of indi- (John F. Due 1988; Tanzi 1991b). The viduals in developing countries, account- share of domestic indirect taxes in GDP ing for about 3.3 percent of GDP and shows no significant correlation with 17.8 percent of total tax revenue, while GNP per capita and these taxes are im- the reverse is true in developed coun- portant for all regions (less so for the Mid- tries.15 Large corporations represent an dle East at 14.7 percent of tax revenue), attractive target for taxation. Their profits comprising between 26 and 37 percent are often substantial and they are re- of total tax revenue (Table 5). quired to comply with statutory account- The allocation of revenue between doing requirements from which the major- mestic sales taxes and excises is roughly ity of small farmers and traders are equal for developing countries while in exempt. Taxation at source avoids some industrial countries the contribution of of the monitoring problems associated sales taxes outweighs that of excises (see with personal income taxation and the Tables 4 and 5). This partly reflects the difficulties in taxing a diffuse set of trans- administrative ease for developing coun- actions. Corporations, however, consti- tries of collecting revenue via excise sys- tute a narrow base and high marginal tems. Excises are often based on quanti- rates can act both as a disincentive to ties as opposed to values of goods and investment and as an incentive to adjust tend to be restricted to a limited subset internal pricing and costing structures to of identifiable items (e.g., tobacco, alco- minimize tax liabilities. hol, petroleum). We discuss sales taxes and excises in turn. 16 Domestic taxes on goods and services Sales taxes average 2.5 percent of GDP Whereas corporate income taxation has and 13.8 percent of tax revenue in the limited scope and individual income developing country sample. Their impor- taxes require sophisticated administra- tance thus slightly exceeds that of indi- tion, domestic taxes on goods and vidual income taxes (2.1 percent of GDP services can be applied, or at least and 10.6 percent of tax revenue). Sales attempted, wherever markets or produc- taxes are not homogeneous. Aside from tion exist. In the developing country VAT, which began to be adopted in de- sample as a whole, domestic taxes on veloping countries in the 1960s (e.g., by goods and services average 5.2 percent Brazil and Colombia), there are a num- of GDP and 30.4 percent of tax revenue. ber of other forms of sales tax. Though in some respects easier to administer than VAT, these other taxes are often 1 In four countries (Botswana, Congo, Gabon, Venezuela) the share of these taxes in GDP exceeds cascading, in that tax is added to tax as 10 percent and in thirteen it exceeds 5 percent. These -a product moves from production to final are mainly countries that depend heavily on oil and other mineralexports and, on a regionalbasis, corporate income taxes are most importantfor the Middle 16 See John Kay (1987)for a note on the distinctions East (see Table 5). between different types of indirect taxation.

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 778 Journal of Economic Literature, Vol. XXXI (June 1993)~~

sale. With cascading, tax liabilities cumu- no tax incentive to buy imported rather late through the system and the effect than domestically produced products and on final prices is very hard to determine. vice versa. They provide artificial incentives for ver- In most countries the VAT extends tical integration and it is difficult to assess through the retail stage (Cnossen 1991). the tax content of exports or to decide Nearly all categories of goods can be the amount of tax to be imposed on taxed using VAT, though basic foodstuffs imports'7 (see Ehtisham Ahmad and and associated inputs are often exempt Stern 1991; Tait 1988). The coverage of or taxed at a lower rate. "Luxuries" are domestic indirect taxation is generally often taxed at a higher rate. Services, weak. Tanzi (1987) estimates that at most and this is of considerable importance, 20 percent of the domestic value added can be taxed using the VAT, though this is reached by domestic indirect taxation aspect may not have been given sufficient in most developing countries. There is emphasis in developing countries (Cnos- a common tendency to concentrate on sen 1991). taxing the sales of imports which repre- VAT systems have been introduced in sent an easier target. over 30 developing countries and further The introduction and successful opera- introductions, especially in Asia, Africa, tion of value-added tax (VAT) systems and Eastern Europe are planned. Its im- have been heralded as "the most signifi- portance in Latin America has been in- cant event in the evolution of tax struc- creasing over time, and it now accounts ture in the latter half of this century" for between 1 and 5 percent of GDP for (Sijbren Cnossen 1991, p. 72). Because Uruguay, Peru, Mexico, Guatemala, Co- a VAT is confined to the value added lombia, and Argentina, and for about 9 (value of sales minus purchases), tax neu- percent of GDP in Chile. In nearly all trality is ensured in the sense that pro- developing countries where it has been ducers and users of a good in a produc- introduced its contribution to tax reve- tion process face the same price for it. nue has been increasing (see Tait 1988). Thus some inefficiencies associated with It is also notable that VAT has been in- most other indirect tax systems are creasing in importance in industrial avoided: the marginal rates of transfor- countries despite the falling importance mation between two goods (in competi- of domestic taxes on goods and services tive markets) will be the same wherever as a whole (again, see Tait 1988). Increas- the pair of goods enter into a firm's activi- ing recognition of its advantages in terms ties. In practice the common price is usu- of revenue buoyancy, a broad base con- ally achieved by giving registered firms sisting of most goods and services, neu- credit for tax paid on purchases against trality as concerns both domestic and taxes paid on sales. This has the added international trade and difficulties of eva- benefit of creating a good audit trail. Tax sion would all seem to imply that the neutrality in international trade is also ascent of VAT will continue. ensured by applying a zero rate to ex- The redistributive scope of indirect ports (and allowing reclaim of tax paid taxes is limited and extensive rate differ- on inputs) and by taxing imports equally entiation can be problematic administra- with domestic products, thus providing tively. However a system with some modest 17 progressivity can be designed as For example, in Korea, recognition of these follows: an in VAT problems led the government to replace eight indi- exemption the for ba- rect taxes with a VATsystem and a "specialconsump- sic foods plus a higher rate of VAT, or tion tax," i.e., an excise system (see Tait 1988). excises, on some "luxury"items (Ahmad

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 779~~

and Stern 1991; Tait 1988; and Section voices, understated business turnover or 4 below).18 spurious claims for exemption which are Excises are almost as important as sales common for other forms of indirect taxa- taxes in developing countries, contribut- tion. In addition, externalities may pro- ing 2.1 percent of GDP and 12.6 percent vide additional reasons for singling out of tax revenue. Their share in tax revenue certain goods.19 is thus larger than in industrial countries, In general, excises should be restricted as is their contribution within domestic to a small set of goods with the bulk of indirect taxes as a whole. Though excises goods being taxed under a coherent and are important in all developing regions unified system of indirect taxation (e.g., they are of special significance in the VAT), otherwise the system is likely to Western Hemisphere and Asia where become unwieldy. In India, for example, they account for 17.9 and 16.2 percent the central excise system resembles a of tax revenue respectively (Table 5). In form of -general production taxation with developing countries, the rise in the im- highly differentiated rates. Rationaliza- portance of excises is associated with a tion, through the imposition of a VAT decline in customs revenue as domestic system to cover most goods, would seem production of excisable goods grows to be desirable (Ahmad and Stern 1987 (1988). Administrative simplicity and and Cnossen 1991). revenue buoyancy form a major part of Foreign trade taxes the attraction of excises in developing countries (see Cnossen 1977). Foreign trade taxes account for 5.1 Excise goods typically exhibit large percent of GDP and for 29.4 percent of sales volume, few producers, relatively total tax revenue in developing coun- inelastic demand, and easy observability tries. They are thus similar in importance (Cnossen 1991). Three products-alco- to income taxes- (5.5 percent and 28.9 hol, tobacco, and petroleum-account percent) and domestic taxes on goods and for most excise revenue though motor services (5.2 percent and 30.4 percent). vehicles, consumer durables, coffee, tea, In industrial countries their role is very cocoa, soft drinks, sugar, and confection- limited, accounting for only 0.7 percent ary are also often targets for excise taxa- of GDP and 2.8 percent of total tax reve- tion. The fact that there are few produc- nue. As we shall see in Section 4 the ers (e.g., on account of license control) main justification for using trade taxes implies that close physical monitoring is administrative feasibility. They are can be exercised over these goods. Thus generally unattractive on efficiency excises may be effectively levied on the grounds. basis of quantities (e.g., packets of to- As a percentage of GDP, trade taxes bacco, litres of alcohol) either leaving the in developing countries are highest for factory or at the import stage. This sim- countries with incomes between $360 plifies measurement and collection, al- and $750 and lowest for countries with lows extensive coverage (aside from illicit incomes between $1620 and $6000 (Table production and smuggling) and limits (al- 3). Africa has the highest share of trade though does not eliminate) evasion through channels such as fraudulent in- 19In the case of alcohol and tobacco, one might want to protect the user (and society) from the negative effects of smokingand drinking.Taxes on motor 8 The strength of the argumentfor differentiation fuels may be seen as charges for externalities such of indirect taxes is weakened if income support and as pollution, congestion and road maintenancecosts social systems are in place (see Section 4). (David M. G. Newbery et al. 1988).

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 780 Journal of Economic Literature, Vol. XXXI (June 1993) taxes in GDP (6.8 percent) and five coun- cient interest to warrant a more detailed tries (Zaire, Lesotho, Mauritius, Swazi- investigation. To do this we regress share land, The Gambia) exhibit shares above of import duties in GDP (ID/GDP) 10 percent compared to only four coun- against per capita income (Y), the share tries in all other regions (Belize, Mal- of imports in GDP (IM/GDP), and the dives, Solomon Islands, Western Sa- share of domestic taxes on goods and ser- moa)-see Table 2. Trade taxes are also vices in GDP (DOM/GDP).20 Numbers important for Asia (5.5 percent of GDP) in parentheses represent t values. For and the Middle East (4.2 percent) but the 82 developing countries in our sam- less so in the Western Hemisphere (3.5 ple we have: percent) or developing Europe (2.8 per- ID/GDP = 2.28 -0.0007 Y cent-see Table 4). Given this pattern (3.64) (3.24) it is not surprising that we observe (for + - the developing countries' sample) a nega- 0.145 (IM/GDP) 0.311 DOM/GDP. tive correlation between the proportion (12.22) (4.27) of total trade taxes in GDP and log GNP R2 = 0.68 (4) per capita (-0.22, significant at the 5 per- The level of import duties in GDP cent level). This may be attributed partly would appear to be (i) positively related to the fact that in very poor countries to the degree of openness of the econ- with less developed markets and limited omy, (ii) negatively related to per capita administrative capability (e.g., sub-Sa- income and (iii) negatively related to a haran Africa),trade taxes represent a con- country's raising of revenue from domes- venient handle for the raising of revenue. tic taxes on goods and services. The relative decline in the importance Export duties are a much less impor- of foreign trade taxes and the increase tant form of taxation than import duties, in importance of domestic indirect taxes contributing only 0.6 percent of GDP for with economic development is a well- developing countries and 4.1 percent of established finding (see Hinrichs 1966; total tax revenue (Table 5). They are most Musgrave 1969; Norman Gemmell 1990; important for countries with GNP per David Greenaway 1984, 1985; Due 1988; capita less than $360 (1.1 percent of Theo Hitiris 1990; and Tanzi 1991b). GDP) and least important in countries Countries that make heavy use of im- with incomes between $1620 and $6000 port duties have poorly developed sys- (0.4 percent-see Table 3). A negative tems of domestic taxation of goods and correlation coefficient between their services, while those countries which share in GDP and log GNP per capita have developed extensive systems of do- of 0.29 (significant at the 1 percent level) mestic taxation make little use of import indicates that their importance falls with duties (see Table 6). It is notable that rising income. When we list the top eight this variation occurs within a total share users of export duties,2' we see that they of indirect taxes which is roughly similar are predominantly poor exporters of pri- for both sets of countries. Those which use import duties heavily are predomi- 20 See Tanzi (1987) for similar calculations. The nantly small and poor open economies right hand variables cannot be thought of as exoge- nous since importswill be a functionof import duties, with less developed domestic markets. and domestic taxes will be chosen at the same time The domestic-tax users, on the other as import duties, rather than determining them. The hand, tend to be larger, richer countries regressionshould be interpretedonly in a descriptive way. with diversified economies. 21Uganda, Zaire, Zambia, Ghana, Solomon Is- The relationship seems to be of suffi- lands, El Salvador, Mauritius, Cote d'Ivoire.

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 781~~

~~TABLE 6 INDIRECT TAXES (percent of GDP)~~

~~GNP per Domestic Capita Taxes on Total (1987 Import Goods and Indirect Country dollars) Duties Services Taxes Imports~~

Lesotho $ 372 27.99 5.73 33.77 136.2 Swaziland $ 701 16.87 2.16 19.07 96.1 Western Samoa $ 560 13.83 4.10 19.06 na Th-e Gambia $ 190 11.40 1.25 13.56 51.7 Solomon Islands $ 421 10.25 0.55 13.75 79.5 Belize $1324 9.79 2.58 13.23 68.4 Maldives $ 344 9.52 3.87 14.36 68.3 Mauritius $1500 9.03 4.12 15.55 59.8 Average $ 682 13.59 3.05 17.79 80.0

Nicaragua $ 828 2.21 18.05 21.62 24.0 Djibouti $1008 1.62 15.39 17.04 56.0 Jamaica $ 942 1.37 14.71 16.08 53.7 Guyana $ 390 1.74 13.96 18.58 65.8 Greece $4015 0.34 12.73 13.07 30.9 Portugal $2874 0.81 11.81 12.62 38.7 Chile $1358 2.62 11.81 14.63 26.1 Mexico $1771 0.76 9.85 10.65 12.0 Average $1648 1.43 13.54 15.54 38.4

Sources: Same as Table 2; except imports, which are from the IMF International Finance Statistics (Dec. 1986, Dec. 1989, Apr. 1990, Aug. 1990). Notes: Averages are unweighted. Total indirect taxes are the sum of total foreign trade taxes and total domestic taxes on goods and services. The countries in the first block are the eight developing countries with the highest share of import duties in GDP and those in the second block represent the highest eight users of domestic taxes on goods and services as a percentage of GDP. mary agricultural commodities mainly in small, accounting for 1.3 percent of GDP Africa. As will be discussed in Section and 6.2 percent of tax revenue. This is 4.3 this may be connected to the difficul- in sharp contrast to industrial countries ties of taxing agricultural production and where they constitute one of the major incomes at source, although some coun- sources of government revenue, con- tries raise considerable revenue via mon- tributing 8.9 percent of GDP and 28.4 opsonistic marketing boards (see Section percent of tax revenue. While this may 3.2 below). The convenience of export partly reflect a low priority being at- taxes must, however, be balanced against tached to social security in developing their deleterious effect on the competi- countries one must recognize that there tiveness of goods in world markets (see are considerable difficulties connected Christopher Bliss 1987). with operating formal social security systems in developing countries. The prob- Social security contributions lems of extracting social security contri- The importance of social security con- butions are similar to those for individual tributions in developing countries is income taxation (see Burgess and Stern

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 782 Journal of Economic Literature, Vol. XXXI (June 1993)~~

1991). For this reason, coverage is often ures (Tanzi and Casanegra de Jantscher restricted to employees in the formal sec- 1989). tor leaving out most of those in the infor- Land and buildings are immovable, mal sector and in rural areas, while in visible assets usually with legal titles of industrial countries coverage is often ownership. Land, being easily observ- close to universal (see Ahmad et al. 1991; able, in inelastic supply (at least "unim- and Anthony B. Atkinson 1989 for a dis- proved" land) and with an unequal distri- cussion). Given that the base of social bution, would seem to be a natural base security taxes in developing countries is for taxation from the perspectives of ad- largely wages, and that the share of wages ministration, efficiency, and equity (see in national income rises with per capita e.g., Ahmad and Stern 1991; and Nicho- income, it is not surprising to find a las Kaldor 1964). While historically there strong positive correlation between the are times and places where the land tax share of these taxes in GDP and log GNP has been important, the recent record per capita (0.48-significant at the 1 per- has been poor (see Jonathan Skinner cent level). 1991, and Section 4.3 below). Wealth and property taxes 3.2 Non-Tax Revenue Taken together, these taxes constitute In most countries the bulk of govern- the least important form of taxation in ment revenue is raised through taxation. developing countries, contributing a Governments do, however, obtain sub- mere 0.5 percent of GDP and 2.6 percent stantial revenue from a number of other of total tax revenue. Their importance sources. On the whole this non-tax reve- in industrial countries is also negligible. nue is more important for developing As wealth and ownership of property is than opposed to industrial countries, commonly concentrated in the hands of comprising about 21 percent compared a small section of the population this re- to 10 percent of total revenue (IMF sult is striking, especially as these taxes 1989). have exhibited considerable importance Regional variation in the importance in the past (Hinrichs 1966, and Haskell of non-tax revenue is marked. Non-tax P. Wald 1959). The poor performance revenue is by far the highest in the Mid- of these taxes may be traced to two fac- dle East where it accounts for over half tors (i) difficulties of observation, mea- (53.3 percent) of total revenue due surement and administration, (ii) political largely to the importance of direct reve- and social resistance. nue from oil. Shares in total revenue are Wealth is inherently difficult to ob- also high for developing Europe (22.4 serve and measure. It may be concealed, percent) and Asia (21.8 percent) which exported (e.g., capital flight) or shifted is partially explained by state involve- into assets which are exempt from taxa- ment in industry. Shares for Africa (15.8 tion. Due to lack of other criteria, pre- percent) and the Western Hemisphere sumptive criteria often have to be used (15.4 percent) are somewhat lower al- to estimate wealth (see Tanzi and Milko though a number of countries in these Casanegra de Jantscher 1989). Poor regions do exhibit high dependence on administration and stiff and effective non-tax sources.22 resistance on the part of the poten- 22 tial taxpayers have meant that at- The IMF statistics (IMF 1989) show eight countries where revenue from non-tax sources exceeds tempts to introduce taxes such as the that from taxation. These for the main part are oil net-wealth tax have often been costly fail- exporters-United Arab Emirates (100 percent of to-

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 783~~

A notable feature of many non-tax in- revenue by consistently setting farm-gate terventions is that revenue can accrue prices below world prices. While there directly to the government. In some is a distinction between prices and taxes cases, people may not be fully aware that it should be noted that the principles of they are being taxed, thereby easing col- taxation (see Section 4) also apply to pub- lection. On the other hand, non-tax lic-sector pricing (see e.g., Atkinson and sources of revenue can exhibit a substan- Stiglitz 1980, and Newbery and Stern tial degree of variability. In countries 1987). where there is a strong dependence on (i) Nationalized Industries.25 Develop- revenue from mineral and agricultural ing countries devote a much higher pro- bases, fluctuations in world prices lead portion of expenditure to the supply of to large swings in government revenue "economic services" than industrial and these "shocks" have many adverse countries. This often reflects a view of effects (see Newbery and Joseph E. the role of the state as the driving force Stiglitz 1981). behind growth and development through Sources and mechanisms for raising ownership and control of natural re- non-tax revenue are diverse. Our intent sources, public utilities, transport and here is not to attempt comprehensive communications, construction and manu- coverage but rather to present and com- facturing, heavy industry, and so on. ment briefly on a few prevalent non-tax Revenue from public-sector enterprises mechanisms of revenue generation.23 In constitutes the main component of non- (i) we discuss revenue from state-owned tax revenue and is important in some industry. Revenue from mineral monop- countries (as are losses of public-sector olies, of particular importance, is exam- enterprises). In cases where a surplus is ined in (ii) and in (iii) we look at agri- generated, this form of implicit taxation cultural marketing boards. 24 Though may carry benefits in terms of low costs diverse, these different sources of non- of administration and difficulties of eva- tax revenue all involve government com- sion (see (ii) and (iii) for examples).26 mand over resources. In the case, for ex- The failure of public-sector enterprises ample, of a nationalized industry, gov- to generate investable surpluses in a wide ernment revenue is extracted through variety of settings has focused recent dis- pricing as opposed to taxation. Similarly, cussion on their efficiency relative to agricultural marketing boards may raise their private counterparts and of the gains from privatization (Paul Cook and tal revenue), Kuwait(97.6 percent), Oman (76.2 percent), Bahrain(71.2 percent), Nigeria (62.9 percent). 25 In discussing revenues from state enterprises A few others have heavy concentrationsof revenue one should distinguishbetween net and gross figures. from state-owned industry (Romania-71.2 percent; Public enterprises often receive investment funds Brazil-53.5 percent) and the final one, Singapore from the center. The size of revenues channelled (50.6 percent) is a city state. None of the industrial backto the center will then be overstatedif expressed countries exhibits high shares of non-tax revenue, in gross terms. In the case of a planned economy, the highest share is recorded by Norway (17.6 per- the government and enterprises have, in essence, a cent). unified budget. 2 Also it should be realized that the purpose and 26 In a decentralizedeconomy the governmentmay effect of non-taxinterventions are often not primarily not, however, find it easy to induce enterprises to to raise revenue for the state. transferfunds to the exchequer and there are a host 24 Revenue from these sources often does not make of accounting and other procedures which hinder its way into the government accounts and as a result this. For a full discussionof the substantialeconomic, their importancemay be underestimated.Ideally the financial, and social costs of administering implicit discussion and statistical analysis in Section 3.1 taxes (and subsidies) through state-owned enter- should have focused on total revenue (tax and non- prises, see Cook and Kirkpatrick(1988) and Gillis tax-see Tabellini 1985). (1989b).

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 784 Journal of Economic Literature, Vol. XXXI (June 1993)~~

Colin Kirkpatrick1988; Robert Millward in 1978 total nationalization of the indus- et al. 1983; IMF 1986; World Bank 1988; try was implemented (see Edwin Lieu- John A. Vickers and George Yarrow1988; wen 1985). Alternatively there may be Dieter Bos 1991). From the public fi- a mixture of instruments. In Indonesia, nance perspective, a number of points where operations are jointly managed emerge. First, changing ownership per and financed by private companies se may achieve little savings in resources (mostly foreign) and the state company in the absence of effective competition. Pertamina, there is a requirement that Second, revenue from privatization sales 85 percent of the output be handed over cannot constitute a sustainable depend- to Pertamina. When taken together with able source and even in the short term other mineral taxes, Garnaut and Clunies its revenue potential is not large. Third, Ross (1983) find that this arrangement from the perspective of the public fi- is equivalent to an 89 percent corporation nances the critical question is whether tax. From a revenue perspective one can- a greater quantum of discounted reve- not assume the superiority of one form nues can be obtained from a privatized of rent extraction over another. Both the- industry via the receipts of privatization ory and empirical evidence suggest that and taxation than from state-owned en- the effect on revenue of a move to mo- terprises via pricing and profits. nopoly control may be insignificant if (ii) Revenue from Mineral Sources. profit or corporate income taxes are al- The mineral sector with its high visibility ready extracting the entire rent element and large rents is often seen as a conve- in profits.27 nient revenue handle. Governments em- (iii) Agricultural Marketing Boards. ploy a variety of means to extract mineral In many developing countries, agricul- rents which include royalties, property/ ture is the largest sector. As we saw in wealth (or related) taxes, corporate in- Section 3. 1 (see also Section 4.3), taxation come taxes, and direct monopoly control. of agriculture is problematic. In some All these mechanisms may be effective countries, there has been an attempt to in raising significant revenue. The raise revenue through the establishment method or combination of methods em- of agricultural marketing boards.28 The ployed is determined by factors such as government acts as a monopolistic buyer the stage of development of the domestic and where farm-gate prices are set sys- mineral industry, administrative con- tematically below world or consumer cerns, judgments concerning the relative prices, with the margin designed to ex- efficiency of public versus private min- ceed transport, storage, and distribution eral companies, and considerations of na- costs, government revenue is generated. tional security and bargaining power (see This system resembles a trade tax. It is Ross Garnaut and Anthony Clunies Ross 1983). 27 Lieuwen(1985), finds that the effectsof the oil As concerns oil, we see that as the do- industry nationalizationin Venezuela in 1976 on the profit split between government and companies was mestic industry develops there tends to insignificant.In Kuwaitthe revenue boost which fol- be a shift from output and profit sharing, lowed legislationpassed in 1973 to allow the national enforced through taxation, to govern- oil copmany KNOC to purchase40 percent of equity in private companies is attributed more to the oil ment direct monopoly control. Vene- boom of 1974 than to any other factor, and there zuela, for example pioneered the 50-50 was in fact a drop in revenue followingtotal national- profit split in 1948. From 1969 to 1978, ization in 1975 (Garnautand Clunies Ross 1983). 28 Other objectives include the stabilization of the government appropriated 85 percent prices or incomes and support of particulartypes of of profits as corporation income tax, and production(see John C. Abbott 1987).

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 785 often presented as a device for maintain- limited availability of tax tools and their ing price stability to farmers. Volatility restricted coverage, lack of resources and will, however, be translated into govern- poor administration, different institu- ment revenue. In most cases only a few tional and sectoral structures, and a vari- goods are taxed in this way. The disincen- ety of types of market functioning. This tive aspects of imposing lower prices can makes, in many ways, the analysis of the lead to lowered domestic production of impact of taxation in developing coun- these goods and illegal informal market- tries more challenging and more interesting and smuggling. ing. In this section we try to summarize The extent to which revenue is raised and assess what theory has to say about varies. Abbott (1987) reviews a number tax policy in developing countries. of modern examples and finds only a few While this section is not very technical, cases where significant revenue was it is based largely on formal models and raised. Peter T. Bauer (1954), in his now results in the relevant existing literature classic study of West African trade, ar- on the economic theory of taxation. For- gued that the agricultural marketing mal models must necessarily be simple boards set up in this region in 1947 for and omit a great deal if they are to be such commodities as cocoa, palm oil, tractable. In the economic theory of taxa- groundnuts, and cotton, regularly main- tion this has generally meant that many tained prices paid to producers at be- of the problems of administration and po- tween 32 percent and 58 percent of litical pressures have been left out of for- commercial values, and thus raised sub- mal policy models. Those involved in tax stantial revenue. Angus Deaton and analysis, whether theoretical or empiri- Dwayne Benjamin (1988) found that the cal, do not, however, regard these issues Cote d'Ivoire Coffee and Cocoa Market- as unimportant. For instance, they are ing Boards, responsible for all purchasing often taken into account by concentrating and distributing of the country's two most the analysis on those tax tools which are important crops, had in some years gen- seen as implementable in a given con- erated as much as 40 percent of govern- text. The problem is that it is difficult ment revenue. Real price series for the to get very far in the analysis of policy period 1963-1987 for both crops exhibit models that try to incorporate adminis- a flat administered price trend consis- trative and political problems explicitly. tently below a highly variable world price Our approach in this paper, in the way (the mean ratio of the two prices being of theoretical insights, is to distil theoret- 47 percent for cocoa and 37 percent for ical insights arising from models and then coffee). The search for revenues can, set them beside a less formal analysis, however, be carried too far and the role provided in Section 5, of the problems of the Cocoa Marketing Board in the of administration, political pressures, and demise of the cocoa sector in Ghana is the like. Any final judgment about the well documented (see Gillis 1989b, Mi- workings of taxation, and on policy, chael Roemer 1984). would then come from taking the more formal and less formal approaches to- 4. Theory and Principles gether. In our view that is the way tax analysis is likely to proceed for some There are a number of features of de- time, and wisely so given what theory veloping countries which are not present is likely to be able to manage. in the standard tax models and which are The positive theory of taxation forms critical to tax analysis. These include the the subject matter of Section 4.1. In Sec-

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 786 Journal of Economic Literature, Vol. XXXI (June 1993) tion 4.2 we present the relevant insights are importantfor developing countries, from theories of normative taxation in they have not been centerpieces for re- terms of basic principles which can guide search, partly because the kind of data- the design of tax structures. By far the bases used in their empiricalanalysis are largest sector in most developing coun- rarelyavailable for developingcountries. tries is agriculture. This creates special Rather, for developing countries, there problems for taxation and the structure has been greater emphasis among theo- and tax treatment of this sector can exert rists on the effects of various indirect particular influence on the economy and taxes and government pricing decisions the effectiveness of other taxes. Agricul- in models which incorporateparticular tural taxation is the subject matter of Sec- features of relevance for developing tion 4.3. We comment briefly in Section countries, with a focus on restricted 4.4 on some aspects of the political econ- tools, special economic structures, or omy of taxation. By this we mean how particularproblems in markets, such as one takes into account the ways in which that for labor(Newbery and Stern 1987). agents in the economy may try to influ- The differencein emphasisaccords with ence tax policy to suit their own interests. the different balance in the sources of Much of the analysis of taxation is based revenue (see Section 3) with, typically, on static models, and we discuss in Sec- in developing countries around two- tion 4.5 some of the problems involved thirds of tax revenue coming from indi- in extending the analysis to dynamic con- rect taxes, and in developed countries texts. two-thirds from direct taxes (including social security contributions)for which 4.1 The Positive Theory of Taxation questions of labor supply, risk, and sav- Analysis of taxation for developing ings have been more importantin the countries diverges from that for devel- literature (Atkinson and Stiglitz 1980). oped countries in several important re- A number of authorshave contributed spects, in particular concerning the tools lists of special features of developing and bases which may be used and the countries of particularrelevance for tax structure of the models analyzed. In this analysis (see, for example, Newbery section of the paper we will illustrate the 1987a, p. 167; Raaj K. Sah and Stiglitz differences which arise from these two 1987, pp. 428-29; AnwarShah and John perspectives. At the same time, how- Whalley 1991, Table 11.4). These in- ever, we shall want to emphasize that clude: (i) the importanceof the primary many of the principles, methods, and ele- sector;(ii) dualism-economic and social ments of models will be common to the organizationin traditionalactivities may analysis of taxation for both developed be differentfrom that in moderncapital- and developing countries. Therefore ist enterprises;(iii) segmentationin the many lessons learned in one sphere will labor market; (iv) the fragmentationof be relevant to the other. capitalmarkets; (v) largenumbers of peo- Subjects which are prominent in the ple living in very poor conditions plus positive analysis of taxation for developed small sections of the populationwho are countries include the effects of taxes on extremelyrich; (vi) many small-scaleen- labor supply, on risk taking, and on sav- terprises;(vii) poor educationlevels; (viii) ing, and the incidence of taxes on the the prevalence of trade distortions,par- corporate sector (see, for example, Atkin- ticularlyquotas; (ix) a majorrole for plan- son and Stiglitz 1980; and Kay and Mer- ning, includingextensive use of permits, vyn A. King 1986). While these issues licenses, and rations; (x) a large public

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 787 sector; (xi) extensive foreign ownership; fluence of other taxes in the system. (xii) weak administrative capabilities; While incidence is a word that can some- (xiii) pervasive corruption; and (xiv) sub- times have useful intuitive appeal, it is stantial evasion. Limited information hazardous in formal analysis, and that is plays an important role in a number of why we would recommend presenting the special features just described. Ex- the problem in terms of the analysis of tensive use of physical controls in trade gainers and losers from reform. and in planning, together with a small A useful table is provided by Shah and industrial base, may be associated with Whalley (1991), the entries of which oligopolies, rents, and rent-seeking. We show the implications of different model- do not have the space to follow through ing assumptions for the effects of different the theoretical implications of the kinds taxes. For example, an increase in a tariff of difficulties implied by the features de- on a good with a binding import quota, scribed, however, in addition to the ref- but competitive world supply conditions, erences cited, see Newbery and Stern would be expected to reduce the rents (1987), Javad Khalizadeh-Shirazi and from the quota, but would not affect (at Shah (1991) and Ahmad and Stern (1991). least in the first round) the domestic An important goal of tax analysis is to price of the good. This is in contrast with characterize who gains and who loses the case without a quota where an import from tax reform or tax changes, and by tariff would simply push up the price of how much. It has been common in the the good subject to tariff. To take a sec- literature to phrase this type of concern ond example, if a company is foreign- in terms of questions such as Who bears owned and taxes incurred in the devel- the tax burden? and What is the inci- oping country are creditable for tax dence of a particular tax? (For an excel- purposes by the foreign company in its lent discussion see Atkinson 1990.) There country of origin, then an increase in cor- are two problems with the expressions poration tax may have no effect on prices of the question in terms of "burden" and in the developing country or on net of "incidence." The first is that the implied tax profits for the firm worldwide, but contrast is apparently between the state might simply imply a reduction of tax of affairs with -the tax and that with zero revenue for the finance ministry in the taxation, which is generally not a sensible origin country.29 Third, in models (such or feasible option. Replacement of zero as that of John R. Harris and Michael taxation as the standard of comparison P. Todaro 1970) where labor markets do by some kind of proportional tax struc- not clear and there is movement from a ture can often be rather artificial because rural to an urban sector with urban the system that is supposed to yield this search unemployment, then taxes that af- proportionality is often difficult to specify fect the level of income in either sector in a plausible manner. Second, there is will have an influence on the level of a temptation to think that the "incidence" equilibrium unemployment. Further ex- can be analyzed by looking at one tax at amples of the influence of developing a time and the particular assumptions country characteristics on tax analysis are surrounding that tax. Such an approach provided in the section on agriculture can be misleading because predictions of (Section 4.3). the'effects on the economy of a tax change should take into account the whole struc- 29 Under these conditions the provisionof tax con- ture of the model, including behavior of cessions to the foreigncompany entail an unnecessary agents, structure of markets, and the in- loss of revenue (see World Bank 1991).

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 788 Journal of Economic Literature, Vol. XXXI (June 1993)~~

There have been, broadly speaking, example, Shah and Whalley 1991, for a three approaches to questions of deriving recent survey, and Kehoe and Serra- the impact of tax changes for developing Puche 1983, for an early example). countries in general equilibrium models. We shall confine ourselves here to Some theoretical analyses (see, for exam- some brief comments concerning the use ple, Sah and Stiglitz 1987; Christopher of CGE models in policy discussion. Heady and Pradeep Mitra 1987; Avinash First, they require a very large number K. Dixit and Stern 1974) have used sim- of parameters, many or most of which ple theoretical general equilibrium mod- are essentially imposed exogenously. els with two or three sectors and a trans- Second, the scope for sensitivity analysis parent analytical structure. One can try is rather narrow. Thus one can vary an to bring out the effects of tax changes elasticity of substitution fairly easily but in terms, for example, of the contribu- it would generally require substantial tions of different elements in an appropri- work to change in any serious way the ate equation or formula. Second, there institutional structure of a market or sec- have been attempts to calibrate this kind tor. Third, and related to the first two of model using explicit functional forms points, it is not easy to make an intuitive (see, for example, Avishay Braverman, assessment of the role of crucial assump- Jeffrey Hammer, and Choong Y. Ahn tions in determining the answers. 1987, or Newbery 1987b). Third, we Fourth, the detail they provide on the have computable general equilibrium consumption side is generally rather less (CGE) models. than would be required in coming to a There is now a substantial selection of judgment about the identification of dif- examples of CGE models. It is unneces- ferent types of gainers and losers-typi- sary to review these in detail because cally there may be 20 or so household there are excellent surveys (see, for ex- groups compared with a household sur- ample, John B. Shoven 1983; Shoven and vey of five or ten thousand household Whalley 1984; and Sherman Robinson groups (see, for example, Atkinson and 1989). For applications see, for example, Holly Sutherland 1988). On the more Kemal Dervis, Jaime de Melo, and Sher- positive side the models are explicit and man Robinson (1982), or for a country they do allow some flexibility. And the case-study (Mexico) see Timothy J. Ke- models allow, in many contexts, estima- hoe and Jaime Serra-Puche (1983). Typi- tion of changes in factor prices, an essen- cally, production functions exhibit a tially general equilibrium phenomenon. constant elasticity of substitution (CES), 4.2 Normative Tax factor markets are perfect, and prefer- Theory ences are of a fairly standard type (often Normative tax theory uses a combi- also CES). The free parameters in the nation of positive theory on the effects model are chosen to be compatible with of tax changes, together with ethical the national accounts structure for a par- criteria for the evaluation of these effects, ticular base-year. Policy variables are to appraise changes in the tax system. then changed and the new equilibrium Most normative arguments involve a bal- is computed. Household utilities can be ancing of the economic criteria of effi- compared before and after the tax change ciency, equity, and revenue. The explic- to come to a judgment as to whether the itly economic analysis would itself have change is beneficial. More recently there to be balanced against administrative, has been a focus on introducing price ri- political, constitutional, or any other rel- gidities and non-market clearing (see, for evant constraints or objectives (see Sec-

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 789 tion 5). In this section our main focus is be nonlinear (see Newbery and Stern on the economic criteria. 1987, for example, for an introduction). Optimal tax theory combines economic Of particular interest are models where criteria into a single objective and finds optimal commodity taxation and optimal the best tax system subject to the various income taxation have been combined. constraints. The objective is usually ex- We shall not go into the detailed analy- pressed using a social welfare function, sis of optimal tax models. We shall, how- which itself depends on the utility or real ever, present and discuss briefly the ba- income of households. If one household sic principles for tax design and tax gains and no others lose then the social- analysis which emerge from these models welfare function shows an improvement. and which have importance for the analy- It therefore embodies efficiency in its basis of taxation in developing countries. sic sense. Moreover, the social welfare They are all based on specific results and function incorporates, in a way which is theories in the literature (see Ahmad and central to the whole analysis, a trade-off Stern 1989, 1991, and Newbery and between efficiency and equity. To be ex- Stern 1987, for references). plicit the function usually takes the form (i) Where possible, lump-sum taxes W(u1, u2,. ,uh . . ,uH) where uh is and transfers, or close approximations, the utility of household h. A transfer of should be used to raise revenue and income Ay from household 1 to 2 (if there transfer resources. Examples are land are no further effects) would therefore30 taxes (although disincentives to the im- - change welfare by (p32 P')Ay where provement of land must be considered) ph iS the social marginal utility of income and transfers, or subsidized rations, of household h. It is often helpful to think linked to the demographic structure of of distributive value judgments in terms households (although incentives concern- of specification of the ,Bs(see Stern 1987, ing family size and saving would have or Ahmad and Stern 1984, 1991). The to be considered). It is not easy to find theory is not, of course, about how to other examples where lump-sum taxation impose the ,Bs but shows how to work can be appropriately linked to relevant out policies for an arbitrary set of ,Bs or criteria (particularly wealth or poverty) social welfare function. It provides, without the tax or transfer ceasing to be therefore, the basic grammar of norma- lump sum. Head or poll taxes are further tive tax analysis. examples but are unattractive on distri- Given that one is optimizing over a butional grounds. range of possible tax measures, the tax (ii) It can be very misleading to look tools themselves and the model of the at one set of tax tools in isolation from economy will usually have to be kept what is happening elsewhere in the tax fairly simple to keep the analysis tract- system. For example, we should not allo- able. The best-developed optimal tax cate redistribution to the income tax and models are those of optimal commodity revenue raising to indirect taxes. Both taxation, where there is a simple linear taxes affect distribution, affect resource tax rate for each commodity, and of opti- allocation, and raise revenue. Further, mal income taxation where, usually, the presence and role of the one set of there is a single source of income, but taxes strongly influences the appraisal of where the tax rate on that income may the other (see, for example, principle (vi) below). 3 Where we also assume that there are no external (iii) The focus of indirect taxation effects. should be final consumption. This means

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 790 Journal of Economic Literature, Vol. XXXI (June 1993)~~

that intermediate goods should not be of shadow prices. Similarly a reform rule taxed unless there are special distribu- based on the other indicators, such as tional reasons or there is difficulty in tax- adjusting tariffsto move towards uniform ing final goods. This applies also to tariffs, protection, is incorrect. which should be rebated on intermediate (vi) The cost of a policy measure, such goods and linked to other taxes on final as an expanded education or pension goods. They should be used for protec- scheme, should be seen in terms of tion only when the case for supporting shadow tax revenue rather than market a particular domestic industry (and pe- tax revenue. Shadow revenue is that as- nalizing its users) is strong and where sociated with shadow taxes defined as the other means of stimulating the industry difference between market prices and are less satisfactory. Tariffs are generally shadow prices (Dreze and Stern 1987 and superior to quotas in providing protec- 1990). For example, if producer prices tion. However, because they act like a are equal to shadow prices then shadow production subsidy financed by a con- producer taxes are zero, shadow consumer tax on the same good, there is sumer taxes are equal to actual commod- no presumption that they are the best ity taxes and shadow tax revenue and fiscal means of aiding an industry through market tax revenue coincide. a learning period. It must be recognized (vii) Indirect taxes should be guided that the elimination of tariffs is a long- by a trade-off between efficiency and term goal which, for revenue reasons, equity, and, in the absence of well- could not be achieved in the short or functioning schemes for income redistri- medium term in countries with very few bution, there is no theoretical presump- tax handles. But it should be pursued tion in favor of uniformity of indirect in the sense that tariffsshould be reduced taxation. Specifically one can show (see as and when the revenue from final goods Deaton and Stern 1986) that proportional taxation can be built up (e.g., through commodity taxes are optimal if Engel VAT systems). Again, in the short term, curves are linear and differences across it is generally preferable to replace quo- households in their demographic struc- tas by tariffs so that the rent from the ture shift the position but not the slope quota flows directly to the government of their Engel curves, and if there is a rather than to those agents who allocate universal and optimal set of demogrants or receive the quota. (i.e., transfers linked to household struc- (iv) Public-sector prices should be set ture). If any of these assumptions fail, according to the same principles as indi- then generally indirect taxes should de- rect taxes: price equal to marginal social part from uniformity to take account of cost for intermediate goods (except for whether goods are relatively more im- the cases noted in (iii) above) and mar- portant in the consumption of the worse ginal social cost plus an element for taxa- off (mitigating in favor of lower taxes on tion for final goods. those goods). (v) The appropriate microeconomic cri- (viii) There are important examples of terion for the expansion of industries is externalities as a basis for taxation. These profitability at shadow prices (see below) include road usage, energy consumption, of the incremental output. Other indica- tobacco, and alcohol. As with other taxes tors (such as effective protection rates or in analyses where "first-best" assump- domestic resource costs) are reliable only tions do not apply, income distribution where policies arising from them coin- and revenue effects elsewhere (and not cide with those associated with the use simply marginal externality costs) should

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 791~~

enter into the judgment of the appropri- approaches are closely related since, by ate tax. definition, the optimum is a state of af- (ix) Theoretical arguments for a corpo- fairs from which no beneficial reform is rate income tax as such are weak. This possible. One can show (see Dreze and tax should be seen as a means for taxing Stern 1987 and 1990) that an improve- personal incomes and an analysis of it ment in a government instrument (be it should be closely linked to the personal a ration, tax, subsidy, or a quota) may income tax. be characterized by first looking at the These results are based on basic propo- direct welfare effect on households of the sitions in the theoretical literature. Al- change and then subtracting the cost at most all of them also have some simple shadow prices of the induced changes in intuitive appeal. It is nevertheless the net demands. For example, if we are con- case that they run counter to many poli- sidering an increase in an old-age pension cies currently implemented and to then we can compute the net social bene- "common sense" arguments that one of- fit as the social value of the income going ten hears advanced. We often hear uni- to the pensioners less the cost at shadow formity of proportional indirect taxes ad- prices of the increases in pensioners' de- vocated as a requirement for efficiency- mands. If that calculation yields a posi- see principles (ii) and (vii). That argu- tive number, then an increase in the old- ment is inadequate because there are age pension is a social improvement. The generally many sets of taxes which are optimality rule for the old-age pension efficient subject to revenue and other rel- is that this calculation should yield zero evant constraints (e.g., the absence of (i.e., there should be no further room lump-sum taxes). If we add equity as a for improvements in social welfare). Ana- criterion in addition to efficiency, we find lytically this approach is identical to that that uniform taxes become unattractive involving shadow tax revenue embodied unless the income redistribution system in principle (vi). through taxation and transfers is very One concept we have not used in for- powerful. The less effective the transfer mulating our set of principles of guide- and tax system, the weaker the argument lines is that of the "shadow price" of pub- for uniformity of commodity taxes be- lic funds.3' In our judgment it has limited comes. usefulness. Crudely speaking, the con- We shall not provide further detailed cept is intended to encapsulate the extra discussions of the derivation or applica- dead-weight losses associated with rais- tion of the basic principles beyond noting ing extra revenue-it is argued that the briefly some of the analytical discussion losses would imply that the cost of a dol- around principles (v) and (vi). These are lar of public revenue is in excess of a a little different from some of the other dollar. There are a number of difficulties. principles in two ways: they introduce First, the shadow price or Lagrange mul- an important analytical concept, that of tiplier on the government revenue con- social opportunity cost, or shadow price; straint will depend on how the problem and they concern reform rather than opti- is formalized, rather than on the basic mality. The distinction here is that the economic structure of the model (see optimality analysis involves a search for Dreze and Stern 1987). For example, in the best, whereas the analysis of reform is simply concerned with finding im- "For a recent and useful discussion which provements from some given starting provides a number of references, see Charles L. Bal- point. Analytically it is clear that the two lard and Don Fullerton (forthcoming).

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 792 Journal of Economic Literature, Vol. XXXI (June 1993)~~

many general equilibrium models the ond, there are strong limitations on the government revenue constraint will not tax tools available to the government to appear at all (it will follow from Walras' tax agriculture. In particular it is often Law). Second, even in models where the impossible to tax transactions between shadow price can be appropriately de- producers and consumers, the difficulty fined as the Lagrange multiplier on this arising both when the "transaction" is constraint, its level and moreover its rela- within the household and when sales are tionship to private marginal utilities of between households or in informal mar- income will depend on the choice of kets. Third, the rural labor market, numeraire (see Atkinson and Stern 1974) which is dominated by agriculture, inter- so that whether the value is above or acts directly and indirectly with labor below unity requires careful interpreta- markets throughout the economy. tion. Third, if taxes are not optimally set, Fourth, the government is often the main "marginal dead-weight losses" appropri- or only supplier of vital inputs such as ately defined, can be negative. For exam- water, fertilizer, and electricity so that ple, a small increase in an indirect tax its pricing policy must be integrated into might generate substantial revenue if it the taxation of production. Fifth, food- led to substitution in demand towards its availability, distribution, and price- heavily taxed goods. The loss in welfare is of such importance to welfare that all required to generate a unit of revenue governments need to take some responsi- may therefore be small, indeed it can bility for its price, quality, and security. be smaller than that associated with rais- These characteristics make agriculture an ing extra revenue through an increase arena where the application of normative in lump-sum taxation. Fourth, the (high) tax theory of the kind described in Sec- shadow price of public funds is often used tion 4.2 is both important and requiring as an argument for high social discount of careful attention to the special features rates. This argument requires careful described. With agriculture playing such definition of social discount rates, of the an important role in output and labor structures of projects being analyzed and markets, it is clear that one has to take of the movement over time in the shadow a general equilibrium view and, there- price of public funds (see Dreze and fore, there will be many influences on Stern 1987). Such care is rarely exer- the incidence of taxes. We would empha- cised. For some examples of the many size four of these: the difficulty of taxing costs of funds see Ahmad and Stern food transactions within the country (so (1991). that subsidies to consumers act like taxes on farmers and vice versa); the elasticities 4.3 The Taxation of Agriculture of supplies and demands for agricultural products; effects operating through the There are many reasons why the tax- labor market (for example, changes in ru- ation of agriculture deserves special ral incomes may influence urban wages), study in developing countries and cannot and the nonuniform distribution of farm- be treated as just another example of a ers' access to technology (so that taxing production activity in the standard com- or subsidising certain inputs may have petitive model. First, it is of central im- strong distributional effects). portance in both employment and out- An obvious and important example of put, the contributions often being in the a possible method of agricultural taxation region of one-half to three-fourths and is the land tax. The distribution of land one-fourth to one-half respectively. Sec- in developing countries is often very un-

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 793 equal and it can be argued that land is sidized in this way. An attempt at such in inelastic supply. Land is visible, im- a system would therefore distort incen- movable, and serves as a good indicator tives towards purchased inputs. The ex- of wealth. Effective land taxation re- ample does show, however, that one quires careful land records. This does not must examine carefully the effects of in principle raise difficulties which are combinations of taxes. excessive when compared to measuring Other possibilities for taxing agricul- the base for other taxes-landowners ture include an agricultural income tax have a strong incentive to establish legal and export duties. The first of these raises titles to their lands. From the viewpoint severe administrative difficulties in mea- of equity, and efficiency and administra- suring income. The second has been tion, land would seem the natural base quite popular (for example, for cocoa in for agricultural taxation and has been Ghana and cotton in Pakistan) but can seen as such by economists from David create considerable distortions given that Ricardo to Henry George. elasticities of supply of particular crops There is a problem with land quality. may be quite high (see Newbery 1987a One would want to measure this by look- and 1987b). Note that if supply is inelas- ing at potential income. A number of tic, an export tax is like a land tax. If countries (for the example of Pakistan, the exportable is produced by richer see Ahmad and Stern 1991) do have farmers, this need not have an adverse records which are linked to quality. This distributional impact. raises, however, the further question of The different possible methods we whether improved land is in inelastic have indicated suggest that a careful supply and the answer would often be study of the potential of the reform of negative. To this extent the attraction of various combinations of the taxation of the efficiency argument for taxing land outputs, the pricing for publicly supplied is diminished. Another disadvantage of inputs, and the subsidization and taxation the land tax is that it involves no co- of purchased inputs may well suggest insurance between government and possibilities for substantial improve- farmer, as compared with, say, an output ments for revenue, efficiency, and distri- tax (see Karla Hoff 1991). bution. It is an area where it can be very These drawbacks are small relative to misleading to look at one agricultural tax the overall attractiveness of land taxes. in isolation and for which a general equi- The small contribution of land taxes to librium framework will be important overall tax revenue (see Section 3.1) and (see, e.g., Braverman, Hammer, and its relative decline over time are thus Ahn 1987; and Newbery 1987b). mainly explained by fierce and effective resistance to this form of taxation (see 4.4 Political Economy Skinner 1991, and Section 5.3). It is interesting to ask how far taxes Recently there has been particular on inputs and outputs can substitute for attention in the study of public policy a land tax. Clearly if the prices of all out- in developing countries to "rent-seek- puts and inputs are reduced in the same ing." Government policy, such as tariffs proportion then this is equivalent to a or quotas on particular goods, raises the proportional tax on land. This would in- incomes of certain individuals or groups. volve an output tax and an input subsidy. Those individuals or groups seek the Such a combination is clearly impossible, rents which might be conferred by such hQwever, because labor could not be sub- policies by spending resources on trying

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 794 Journal of Economic Literature, Vol. XXXI (June 1993) to get them implemented (see also Sec- portance for democratic decision making tion 5.3). Those who decide or administer of clarity in tax-prices (defined as the ex- the policies can thereby gain substan- tra tax due from individuals or groups tially and favor systems which allow arising from an extra unit of public spend- themselves to benefit from the exercise ing). He therefore argues the virtues of of their discretion. In this kind of theory, well-defined and understandable rules policy has an endogenous element in that governing the allocation of costs and ben- it is determined in part by agents acting efits of public transactions." A related in their own self-interest. line of argument emphasizes the impor- "Rent-seeking" theories provide one tance of constitutions both in protecting example of a type of approach often called some groups from others (e.g., a minority "political economy." Further, important from the majority), but also in the restric- and interesting examples concern voting tion of "Leviathan's fiscal appetites" models where public choices are deter- (Geoffrey Brennan and Buchanan 1980 mined by votes of individuals, models and 1985). Constitutional limitations on of bureaucracy where those who adminis- revenue, bases, rates and types of tax ter the organs of state exert strong influ- might then be seen as controlling the size ence on outcomes, and models where in- of government and the direction of its terest groups compete against others for tentacles. In addition the theory offers group gains and for control. A leading insights into why certain taxes encounter figure in this literature has been James resistance while others do not (see Sec- M. Buchanan and many of these positive tion 5.3) and thus can help explain ob- theories are grouped under the heading served tax structures. "public choice" (see e.g., Buchanan and All sensible tax models, however, Robert D. Tollison 1972; Agnar Sandmo whether or not policy is determined 1990; and Buchanan 1991). In these mod- within the system, share the require- els bargains are struck and transfers may ment that the effects of policy have to be made. An optimistic view of such a be calculated explicitly. This is the first, process would see it as leading to Pareto and often most difficult, part of any analy- improvements (Wicksell writing in 1896 sis of policy in terms of its consequences. provides an early discussion).32 More Thus, for example, the normative ap- pessimistic pictures would be of special proach to a change in income tax should interest groups gaining power or a levia- calculate who gains and who loses in a than bureaucracy dominating. It must be similar manner to a model where changes recognized, however, that a systematic take place according to whether particu- analysis of policies in terms of their lar interest groups would benefit. For consequences is necessary for this ap- further discussion of "rent-seeking" and proach also. For a recent discussion "neoclassical political economy" see of some of the issues and of relations Krueger (1974), Buchanan, Tollison, and with normative theories see Atkinson Gordon Tullock (1980), David Colander (1987). (1984), T. N. Srinivasan (1985), Jagdish Some analysts have embodied the con- Bhagwati (1987), Bos (1989), and William siderations just raised into guidance for Ascher (1989). the construction of tax systems. For example Buchanan (1966) stressed the im- 33The importance of both transparency and of reaching a consensus between the various players (i.e., government, workers, private sector institu- 32 This articlemay be found reprintedin translation tions)has been stressed in the recent tax reformliter- in Musgraveand Alan T. Peacock (1958). ature (see Section 6).

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 795~~

4.5 Dynamics suggest that inter-firm dynamic transactions should not be taxed and indirect The models and approaches dis- taxes should be between producers and cussed in the preceding subsections have final consumers. Where the household been largely static. Many of the problems and firm overlap, a situation more com- of developing countries involve growth mon for developing than developed and change. It is natural to ask, there- countries, this suggests lower taxation on fore, whether the theories of growth and household savings than might otherwise of taxation can be combined to provide be the case. useful insights into the role of policy in Third, we have taxation or subsidiza- growth (for a discussion see Newbery and tion for externalities, for example Stern 1987 and Stern 1992). Only a little through learning from others, or to cor- progress has been made in this direction. rect for distortions introduced elsewhere This has been clearly not because the in the tax system. Fourth, the govern- writers on one topic have been ignorant ment may wish to take greater responsi- of the other as the number (and the dis- bility for future generations than current tinction) of authors who have contributed generations are prepared to take. In- to both is striking (including Wicksell, deed, the government may wish to take Ramsey, Hotelling, and Samuelson). greater responsibility for individuals' Rather the reasons include the follow- own future than they are prepared to take ing: we know much less about the dy- themselves-a version of the merit-good namic behavior of individuals; there are argument (Musgrave 1959). These last important problems which arise for two sets of results may be of relevance dynamic models; and dynamic optimiz- for developing countries and point away ing models are less tractable than static from the taxation of saving and invest- ones. ment. The static theory of optimal taxation To the problems of applying and ex- allows a dynamic interpretation if we tending static models to a dynamic frame- view goods produced or consumed at dif- work must be added those issues which ferent times as different goods. From this are intrinsically new when we pass to the perspective there are four groups of re- dynamic from the static. These include: sults of particular relevance. First, we incorrect expectations; the development have results on the circumstances under of knowledge; revision of policy and cred- which indirect taxation is optimally uni- ibility; the influence of future genera- form. These focus on cross-elasticities tions; pensions and intergenerational with leisure and on whether a well- transfers; whether the tax base should functioning direct tax system is in place. be consumption, income, or wealth;34 Uniformity, if translated into a dynamic and the process of adjustment. Each of context, is a characteristic of expenditure these could be of great significance for taxation in the sense that consumption tax policy. Together they imply that in different periods is made subject to while the simple intuition developed the same rates of tax. Given the strength in static theory has some usefulness it of the assumptions and the weakness of may have only a limited role to play direct tax systems in developing coun- in the understanding of dynamic tax pol- tries, this kind of result can furnish little icy. support for expenditure taxation. Sec- ond, we have the intertemporal ana- 3 In a one-period model there is no distinction logues of the efficiency theorems which between consumption, income, and wealth.

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 796 Journal of Economic Literature, Vol. XXXI (June 1993)~~

Growth theory has provided us with tax policies. The rate of growth is rightly valuable tutoring on the logic of the de- at the center of the stage in development terminants of growth-a simple but cru- economics, and in most theories from cial example being the Harrod-Domar Roy F. Harrod (1939) to W. Arthur Lewis growth rate slv for capital stock (where (1954), to more modern growth theories s is the savings rate and v the capital- (Romer 1986; Robert E. Lucas, Jr. 1988; output ratio) and slv for output, where and Maurice F. Scott 1989, for example), v is the incremental capital-output ratio), investment is seen as a crucial engine and equal to slv if v is constant. Growth of growth. It might seem therefore that theory has pointed also, however, to tax policy should be oriented towards the what we are bad at explaining, particu- encouragement of investment and sav- larly technical progress. There have been ings. However, the justification of special some efforts in recent years to improve incentives in a market economy requires our understanding here many of these arguments to tell us why market incen- have been grouped under the heading tives to invest are inadequate. On the "endogenous growth theories." In these savings side we remain unsure not only models technological advance comes about the interest elasticity of savings, from the creation of ideas, either indi- and how it varies within the population, rectly through the act of capital invest- but also about the appropriate models ment or directly through investment in within which the elasticity should be esti- research and development. The public- mated. We do not know, then, whether ness of ideas, however, creates problems tax policy is very important in its effects for the efficient functioning of markets. on savings rates. We would be sceptical, If the externalities from investment do for example, whether it has a lot to do influence the long-run growth rate then with the rise in the savings rate in India the importance of institutions, and of in the decades following independence government activities and tax-subsidy or in the fall in the rate in economies in policies, which encourage them could, sub-Saharan Africa (for further discussion in principle, be substantial. see, e.g., Stern 1989 and 1991b). A number of the models have an externality from capital investment operating 5. Constraints and Pressures through learning which increases labor productivity (e.g., Kenneth Arrow 1962; We saw in Section 3 that the differ- Paul M. Romer 1986 and 1989). A partic- ences in levels and patterns of taxation ular form of the production function and across countries reflect a diversity of eco- externality which has been popular has nomic structures and administrative ca- the effect of making output proportional pabilities as well as widely varying politi- to capital so that the long-run rate of cal, cultural, and institutional histories growth becomes slv. Hence any policy and environments. The generation of that increases s or reduces v increases revenue via taxation in developing coun- the long-run rate of growth (see, e.g., tries faces a number of constraints and Robert G. King and Sergio Rebelo 1990). pressures. These need to be brought into For further discussion of this "new" sharper focus and considered alongside growth literature, see Robert J. Barro the principles of taxation derived from and Xavier Sala-i-Martin(1990) and Stern theory (Section 4) in order to understand (1991b). the design of tax reform (Section 6). Prob- The uneasy state of theory leaves us lems of administration, implementation, in a difficult position as regards dynamic and political feasibility represent the

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 797 greatest challenge and in Sections 5.1- The process of taxation may be divided 5.3 we dissect the different constituent into four distinct phases: design and en- elements. At the most basic level, infor- actmhentof the tax codes; information col- mation, resources, and managerial ability lection and identification of taxpayers; tend to be inadequate and, in Section the assessment of liabilities, tax litiga- 5.1, under administration, we discuss at- tion, and enforcement; and tax collec- tempts to increase the efficiency with tion. Weaknesses in the tax administra- which taxes are assessed, processed, and tion can exist at each of these stages and collected. Administrative difficulties are this is the subject of Section 5.1. Evasion also related to the propensity and ability and noncompliance compounds these dif- to evade which is discussed in Section ficulties and these factors are discussed 5.2. Tax analysis must also take into ac- in Section 5.2. count political responses and feasibility, Administrative constraints may be the subject of Section 5.3. These impinge eased via changes to tax laws and the strongly on administration and are overall tax structure. As argued in Sec- closely related to evasion. In Section 5.4 tion 3.1 different taxes make very differ- we examine the difficulties that inflation ent demands on administrative capacity. imposes on taxation and finally, in Sec- The use of physical barriers for taxation tion 5.5, we discuss the international con- (for example, excises and customs), constraints on the design of taxation and the centration on sectors with greater reve- role of harmonization. nue potential (for example, the modern corporate sector), the withholding of 5.1 Administrative Constraints taxes (for example, PAYE), the use of presumptive taxation for many areas Much of the practical work on tax of difficult verification, the restriction of reform over the last three decades or so sales taxes to easily taxable goods and has concentrated on finding taxes and sensible exemption limits to exclude very procedures that can be implemented small firms, traders, and individuals on with the administrative capabilities found low incomes can all lessen administrative in developing countries (see Bird and demands. Self-assessment of income or Oldman 1964, 1990; Gillis 1989a; Mi- VAT can help to shift the burden of ap- chael J. Boskin and Charles E. McClure, praisal from the public authorities to the Jr. 1990; Khalizadeh-Shirazi and Shah private sector where it may be dealt with 1991; World Bank 1991; and Section 6 more efficiently. Self-assessment also re- below). More effective enforcement of duces scope for collusion between au- taxes has been another central theme of thorities and taxpayers. The removal of tax reform. While the focus on adminis- exemptions, loopholes, and concessions trative effectiveness and simplification is can simplify administration and reduce to be welcomed, indeed is indispensable, evasion. These are often central elements one should not lose sight of the objectives in tax reform (Section 6). Taking a sys- of economic efficiency, equity, and mac- tematic view of the tax system, rationa- roeconomic stability, which are central lization, simplification, and the removal concerns of tax and public finance theory of anomalies should have the effect of (see Section 4 and Newbery and Stern reducing the administrative costs of iden- 1987). Administrative and other prob- tification, assessment, auditing, and en- lems should not be set headlong against forcement. The administrative simplicity theory. The task of the tax analyst is to of "taxhandles" however, while influenc- combine them constructively. ing tax policy, should not be allowed to

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 798 Journal of Economic Literature, Vol. XXXI (June 1993) dictate it. Concentration on just a few provide valuable weapons for improving handles can lead to a highly distortionary administration. Computerization can structure. help update and improve records and fa- There are a number of shortcomings cilitate cross-checking between, for ex- common to most tax agencies in develop- ample, direct and indirect tax authorities ing countries, albeit in varying degrees, (see Frangois Corfmat 1990). which affect abilities both to use informa- Factors such as improving the gather- tion to assess taxes, and to collect the ing of information, raising the pay of tax assessed taxes. These include insufficient collectors, linking pay and indeed em- staff with the appropriate skills, lack of ployment to performance, and improving sufficient and up-to-date equipment and the training of these officials can lead to facilities, and complex legal structures significant improvements in terms of which are often ill-defined, or in a state both efficiency and revenue collected of flux. The enforcement of penalties for (Nicholas Kaldor 1956; Shankar N. Ac- evasion is generally poor (World Bank harya 1985; Goode 1984, 1990; and Bird 1991; Richard K. Gordon 1990). The 1989). shortage and inadequacy of staff and Improvements in identification and as- equipment, and the inadequate legal sessment need to be coupled to penalities structure may be compounded by the ab- for evasion which are both clearly sig- sence of incentives to collect taxation. naled and enforced.36 Amnesties have Wages of public-sector employees typi- been tried occasionally in an attempt to cally do not depend on performance, jobs encourage revelation. To be effective, are secure and there is little supervision penalization for not declaring hidden in- (Newbery 1987a). Legal complications come must represent a likely event. Reg- can be extreme: it is often difficult in ular and widespread auditing of accounts developing countries to find a document can help in this respect. or documents setting out what the rules The "administrative constraint" should and rates for a particular tax really are not be seen as an unchanging reality for (see, for example, Ahmad and Stern developing countries. Significant im- 1991). Overly complex tax codes make provement may be achieved even in the it difficult to assess and collect taxes. This short term (see Section 6). Strengthening vagueness and complexity often gener- tax administration is critical to the suc- ates great scope for evasion.35 cessful implementation of tax reform (see Administration may be improved by World Bank 1991). Investing to improve direct reform of the tax administration. one's tax agencies is an infrastructural Records are often poor or inconsistent project which may show considerable re- and under the control of different tax au- turns, both economic and otherwise. thorities. The introduction of a system of numbers that uniquely identifies each 36Auditing, prosecution, and enforcement in Mex- individual taxpayer and the creation of ico, for example, have been used to increase the a number of taxpayersfrom 1.7 million to 2.8 million master tax file for use by the tax authori- between 1988 and 1991. An amnesty was provided ties can greatly aid both identification whereby taxpayers who declared themselves were and enforcement (see World Bank 1991). liable only for the past year's taxes, while tax evaders would be fully liable. Fiscal enforcement was made Cross-checking across authorities can credible by the widespread auditing of both firms and individuals, and by the prosecution of evaders. Between 1989 and 1991 there were more than 200 3 Clarity regarding both liabilities and penalties recorded offences for fiscal evasion while between for evasion is often a central objective of tax reform 1929 and 1988 there were only two (see Aspe Armella (see Section 6). 1992).

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 799~~

5.2 Evasion37 Problems of information and measurement imply that individual income taxa- Tax evasion in developing countries tion is particularlyvulnerable to noncom- is rife. We discuss in turn cultural factors, pliance (see Gordon 1990). For some, the evasion of direct and indirect taxes, evasion may be relatively passive in that corruption, and rent-seeking. The effect there is little attempt by the government of evasion activities can be to drive a to impose the tax. This is especially the wedge between the statutory and effec- case for the self-employed (e.g., farmers, tive tax systems to the point that the rev- informal sector workers, or profession- enue that is legally due bears little rela- als), as the gathering of information on tion to that collected. the incomes of such individuals is difficult Information on incomes, production, and costly.39 For similar reasons many transactions, property records, and in- small enterprises remain invisible to the heritances is notoriously difficult to ob- tax authorities.40While under- or nonre- tain in developing countries. As Alex porting of income may be easier in the Radian (1980) argues, information prob- self-employed and small-business sectors lems do not originate solely from the defi- than elsewhere, it is by no means re- ciency of collecting agencies but also stricted to them. Large enterprises. also from the fabric of economic, social, and often underreport taxable income. Taken cultural relations which exist in a given as a whole, the effects of evasion are im- country. There may be no conventions pressive. Acharya (1985) estimated that, of issuing receipts, recording transac- for India, of the total income assessable tions, reporting the existence of enter- for tax, the actual percentage declared prises to the government, complying was 53.3 percent in 1975-76 and 41.9 with accounting and bookkeeping stan- percent in 1980-81. Frederico J. Her- dards, defining farm boundaries, and so schel (1978) reports that, for Argentina on. Taxation is a strange, unwelcome, in 1959, only 21.5 percent of total taxable and sometimes incomprehensible con- income was reported for tax purposes. cept to many people in developing coun- Tax evasion is not limited to income tries. Differences in the tradition of com- taxes. Sales taxes and excises are evaded pliance probably explain as much of the in many ways. A popular method is un- worldwide pattern of taxation as do un- derinvoicing. The problem seems to be der-resourced or poorly organized tax ad- particularly severe in the service sector, ministrations. The success or failure of where clients are often presented with systems of taxation such as VAT depend an option: a higher fee if tax is to be on the level of voluntary compliance as declared, a lower one if the transaction well as on enforcement. Very few like is to go unreported. Elsewhere greater paying taxes but the hostitility to taxation and the propensity to evade depend on cultures as well as economic incentives.38 ments give the impressionof immutabilityin tax hab- its, however, and rest uncomfortablywith the rapid progress Argentina has made in tax collection over the last few years. 37 This section draws on James J. Thomas (1992). 39 As a result, many individualswho are in princi- We are grateful to him for his advice and making a ple liable may not become registered as taxpayers. draft available to us. For an excellent theoretical 40 Rather than depend on declaration of income treatment of tax evasion see FrankA. Cowell (1990). by small taxpayersand businesses, Musgrave(1990) 3 For example, one hears it arguedthat the superi- and others suggest that presumptive-incomeand esti- ority of Chile over Argentinawith respect to tax col- mated-income approaches, where incomes and re- lection comes in part from the Germanic tradition turns on capital are calculatedindependently, might in the formerand the Italianin the latter. Such argu- be used instead.

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 800 Journal of Economic Literature, Vol. XXXI (June 1993) checks are in principle available, for ex- oping countries. The procedures used ample, through the verification of stock vary, and do not always require avoid- purchases, but nevertheless evasion is ance of customs officials. B. P. Alano, extensive. The introduction of VAT sys- Jr. (1984) reports various ways, for the tems, with their built-in incentives to Philippines, whereby duties are evaded seek correct invoicing of one's purchases even though goods are physically (which are another person's sales), seems brought through customs: goods can be to contribute to a reduction of noncom- misdeclared; goods that are correctly de- pliance in domestic indirect taxation (see scribed may have their values under- Tait 1988 and Goode 1990). The revenue stated; fake delivery papers may be used security advantages of VAT over simple to remove the goods from the customs sales and business taxes are twofold: (i) area before duties are paid, and so on.4' only buyers at the final stage have an The importance of these activities can be incentive to undervalue purchases; (ii) gauged by the estimate that the value nonpayment of tax at one stage can be of smuggled goods into the country could reversed by payment at a later stage. It have been some 29 percent of the total should be remembered, however, that value of imports into the Philippines be- in developing countries many or most do- tween 1965 and 1978 (Alano 1984).42 Tax- mestic goods and services do not come ation is thus constrained in the sense that under any type of taxation (Tanzi 1987). if customs duties or excises are raised Evasion is thus unnecessary for the ma- smuggling may increase. The response jority of transactions. may be sufficient to result in a net de- The evasion of taxes gives rise to the crease in government revenue from this phenomenon of transactions in black source.43 money. For example, the purchase of a Finally, we mention briefly some as- house in India is usually partly in "white" pects of corruption in public life and rent- and partly in "black." The white part of seeking via the currying of favor with offi- the transaction is recorded and the black cials. Corruption, bribes, kickbacks, and is not. This not only keeps down taxes leakages from public expenditure in gen- on property and transactions but also eral are included here (see Thomas 1992). conceals from the authorities the posses- Reduction in rent-seeking may be sion of large sums gained through un- achieved by reducing discretionary con- taxed activities the origins of which trols available to officials. One can for would otherwise have to be explained. example, keep down the number of li- Occasionally attempts are made to allow cences and permissions needed to do official laundering of black into white money, for example, through the is- 41 Underinvoicingof imports can be used to evade suance of government "bearer bonds" high importduties. For Tanzania,T. L. Maliyamkono which can be bought "no questions and M. S. D. Bagachwa(1990), by comparingTanza- asked" with black money. In this way nia and U.K. records, found that in 1985 imports from Britain were underinvoiced by 18.7 percent. governments can extend their borrowing 42This is not an isolated example. Mark M. Pitt capabilities. Further, tax revenue may (1981)estimates that between 25 percent and 30 per- increase subsequently as this part of cent of all rubber exported from Indonesia in the period from 1959 to 1965 was smuggled. wealth can be spent more openly outside 43 Gil Diaz (1987)notes that in Mexico in the early the black economy. 1970s a 30 percent excise tax on mink coats caused Evasion of import or export duties- tax collections to fall to one-third of former levels when the tax had been 10 percent, as many rich we shall use the term "smuggling"-re- Mexicans now purchased the coats in U.S. border mains a serious problem in many devel- towns.

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 801 business. Similarly import quotas are to systems (for models of this type see Sec- be eschewed on these grounds (in addition 4.4). Individuals who have spent tion to their inferiority, from the perspec- considerable resources devising means of tive of revenue, to tariffs). Rent-seeking evading the outgoing set of taxes are un- involves real losses of resources in the likely to welcome new measures de- efforts spent to secure favors and is not signed to overcome the porousness of the just a transfer payment to corrupt offi- existing system. Those who may benefit cials. from tax reform may be less aware of the Once corruption gets into tax authori- significance of proposals, and less able ties it is not easily eliminated. Those who to organize and ensure implementation are supposed to root it out can them- of new tax measures. The net result is selves be corrupted or threatened. In that, at the public level, tax reforms tend principle, raising penalties on officials to be greeted with antagonism and con- and increasing the probability of detec- siderable resources may be spent lobby- tion can improve performance but that ing against tax measures (Gillis 1989a). is easier said than done. Raising wages This in turn restricts the feasible set of and salaries in the tax authorities at the tax instruments and influences the ob- same time as increasing the likelihood served structure of taxation. of dismissal can lead to improvement, as Political power is often closely allied the experience of Ghana in the 1980s ap- to groups which have large asset or pears to have shown. Alternatively, one wealth holdings. These groups are in can accept that corruption is ingrained many cases attentive in blocking the in- in the public sector and resort to private troduction of unfavorable measures, for sector firms to collect taxes on a payment example, progressive income taxes, by results basis. Indonesia, for example, wealth and property taxes, and land taxes in 1985 turned customs assessments over (Robert H. Bates 1989; and Ascher 1989). to a foreign firm; in Mexico commercial It is likely that the relatively small contri- banks play a central role in the collection bution of personal income taxes, land, of domestic taxes and the customs ware- property, and wealth taxes in developing houses are now operated by private cus- countries (see Section 3.1) reflects in part tom agents (see Gillis 1989c; World Bank the use of political power to suppress 1991; Aspe Armella 1992). The introduc- such measures or to evade them when tion of new management and organiza- implemented.44 Michael H. Best (1976) tional structures are often central aspects studied the incidence of taxation in sev- of tax reform (see Section 6). eral Central American republics in the 1960s and and 5.3 Political Pressures early 1970s, suggests that the observed tax structure closely resem- Problems in introducing and main- bled the preferences of the most power- taining effective systems of taxation relate ful group, and that this mapping became not only to administration and evasion, closer over time. but also to political pressures. Taxation involves an attempt by the state to exer- 'The dramatic drop in the importance of land cise control over the resources which revenue, for example, in India from 50 percent of total revenue in the mid-19th century to a mere 0.9 would otherwise be controlled by private percent of total tax revenue in 1984-85, and the con- agents. Political coalitions, or representa- comitantrise in the importanceof indirect taxes may tives thereof, will attempt to avoid such partly reflect the successful efforts of landowners in shifting the burden of taxationto the general popula- claims on resources by influencing both tion by politicaland other means (see DharmaKumar the development and functioning of tax 1982; and Ahmad and Stern 1991).

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 802 Journal of Economic Literature, Vol. XXXI (June 1993) Political blocking power is illustrated (Bird 1989; Gordon 1990; Bos 1989; and in the failure of a number of attempted Goode 1990). As more individuals (for reforms. In Colombia, landowners, act- example, not just public-sector employ- ing through the Ministry of Agriculture, ees) or goods (for example, not just luxury managed to suspend a tax law creating items) are taxed, the perceived "discrimi- a presumptive income tax on agricultural natory" nature of some forms of taxation land in Colombia in 1973, despite the may be removed or lessened. Indirect will of Congress to the contrary (Ascher taxes, being less visible, tend to encoun- 1989). In 1986, in Argentina, as part of ter less resistance than direct taxes. a sectoral adjustment loan, the World A central message here is that reforms Bank proposed that agricultural export facing strong and active opposition can- taxes be replaced by a less distortionary not be imposed upon countries. For gov- land tax. A tax bill to this effect did not ernment commitment to tax reform to become law because of opposition both be credible, the likely behavior of poten- from the farm lobby and from govern- tial gainers and losers needs to be taken ment members who argued that the tax into account and a broad consensus ar- was administratively infeasible. rived at. Gil Diaz (1987), for example, Despite its theoretical attractiveness attributes much of the success of Mexican (Section 4.3) and the importance of agri- tax reform (see Section 6) to the ability culture, resistance to land taxation is of the Secretary of the Treasury to align fierce and effective. It is interesting to different interest groups and views ob- speculate why this is so. First, given that taining approval of the groups as "the land ownership is related to wealth, the most difficult part of the reform" (p. 355). the success of the effectiveness reflects 5.4 The Effects of Inflation resistance of the rich and powerful to measures which harm their interests. The balance among taxation, borrow- Second, land taxation is very visible. ing, and money creation was discussed Third, it is not easy to organize a deduc- in Section 2.2. In this section we focus tion at source in a way which is possible on some of the effects of inflation on the for other direct taxes. It seems that part- functioning of tax systems. 45 In less- ing with money is more painful than not developed countries, these effects are, receiving it. Fourth, for distribution and on the whole, disruptive and negative. administrative reasons some land taxes The distortionary effect on the tax base, have had an exemption limit (see, e.g., especially as regards changes in the value Ahmad and Stern 1991, on Pakistan). of assets, can be important and lags in This allows landowners to divide up land collection can greatly reduce the real between relatives for tax purposes (or to value of tax revenue to the government. create fictitious holders) to reduce liabil- Tanzi (1977, 1978, 1980) has argued ity. that the effects of inflation on the tax sys- Potential resistance to taxation can be tem depend on two factors: the so-called countered by reducing the incentives for "price elasticity of tax revenue"; and the opposition. From this perspective high length of the collection lag, described as rates targeted on a narrow base are likely the time elapsed between the moment to prove unproductive. Widening the when the amounts of tax due are deter- coverage, broadening the tax base, and lowering tax rates may improve political 45 For a more complete treatment of the effect of acceptability and reduce conflict be- macroeconomicpolicies and inflation on tax issues, tween different political interest groups see Tanzi (1988, 1991a).

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 803 mined and the moment when they are 1991a). Specific (as opposed to ad valo- actually transferred to the government. rem) taxes may be adjusted only with a If the price elasticity is equal to one and lag so that real tax revenue falls.48Similar the lag is negligible, real tax revenue col- effects arise with import duties (even if lected is not affected by inflation. If, how- they are ad valorem) if inflation leads to ever, the collection lag is not negligible periods of exchange rate overvaluation or the price elasticity is less than one, (see Jesus Seade 1990). Where interest inflation will tend to lead to reductions payments are deductible, inflation may in the real value of tax revenue collected. also reduce corporate income tax reve- The "price elasticity of tax revenue" nue and create incentives against equity measures the response of total tax reve- finance.49 nue to a general rise in prices and in- 5.5 International Constraints comes and depends on the composition of revenue. Inflation can increase the We have so far in this section focused yield from a system of progressive taxa- on domestic pressures and constraints as- tion when tax brackets are not index- sociated with the raising of tax revenue linked. Brackets may be adjusted for in- in developing countries. But, albeit to flation but this is usually done with a varying degrees, virtually all of these lag.46 Another route by which inflation countries are open economies. The inter- may, in principle, raise tax revenue from national interaction of tax systems places individuals is through taxes on capital further restrictions on the raising of tax gains if these are inadequately indexed. revenue (see Mark Gersovitz 1987; Bird Inflation will, however, decrease the and Oldman 1990; Tanzi 1990; Boskin yield of indirect or direct taxes which are and McLure 1990; and Khalizadeh-Shi- set at specific rates and which do not ad- razi and Shah 1991). Many of the factors just quickly enough to match value ero- subject to taxation are moveable and may sion by inflation. be relocated internationally in search of In industrial countries, with low to more favorable tax treatment. Interna- moderate inflation, where a large part tional mobility of factors may thus con- of revenue comes from income taxes col- strain the effectiveness of domestic tax lected with progressive rates and with- systems and limit the scope for reform. held at source (i.e., negligible lag), in- Because capital is the most internation- flation can increase tax revenue. 47 In ally mobile of factors, capital income taxa- less-developed countries where the share tion (be it on interest, corporate income, of these taxes is small, where taxes with or assets) is perhaps the most subject of specific rates (e.g., import duties, ex- all forms of taxation to international con- cises) have a significant role, where col- straints. lection lags are long and where inflation Regulatory controls on capital outflows is often high, the reduction in the real cannot be fully effective. Attempts to tax value of tax revenue collected can be sub- capital income heavily can result in capi- stantial (see Tanzi, 1977, 1978, 1980, tal flight (see Donald R. Lessard and John

4 Bracket creep can undermine the progressivity of individual income taxation. Inflation adjustment 'The buoyancy of VAT systems under inflation to prevent this involves linking the nominal base to is thus an importantconsideration. some indicatorof inflationsuch as the consumerprice 49 In Mexico the combinationof inflation,long col- index. lection lags, and the deductibilityof nominalinterest 47 This indeed was the experience of many paymentsled to a fall in corporateincome tax revenue industrial countries during the 1970s (see Tanzi from 2.6 percent of GDP in 1982 to an average of 1980). 1.6 percent during 1983-87 (World Bank 1991).

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 804 Journal of Economic Literature, Vol. XXXI (June 1993)~~

Williamson 1987). Gil Diaz (1987), for rate tax policy is thus effectively con- example, emphasizes the constraints on strained. Transnational corporations that interest income taxation which arise be- supply a large industrial market face a cause so many Mexicans have tax-free wide choice of locations and the more bank accounts in the USA-the net re- similar the other factors (e.g., cost of la- turn in Mexico cannot, therefore, fall sig- bor) the more attention will be paid to nificantly below interest rates in the corporate tax structure. Many develop- USA. ing countries are highly dependent on Manual Pastor (1990) lists interest rate foreign direct investment and the desires differentials, expected devaluations, for revenue and for investment must be availability of liquid capital and accelerat- balanced. Tax free zones and tax holidays ing inflation as the main causes of capital are often designed to attract foreign in- flight from Latin American countries. vestment. Revenue losses may, however, Stocks of foreign assets held by nationals be unnecessary if the decision to invest of this region are very large. Argentine does not hinge on tax costs or the capital and Mexican assets abroad, for example, exporter can credit foreign income taxes represent over 70 percent of each coun- made on overseas investment. Competi- try's respective debt burden. Large hold- tion over tax incentives may be avoided ings of foreign assets represent severe by harmonization of tax policies across erosion of the wealth and earnings tax countries through multilateral agree- base. Pastor (1990) estimates that if all ments. By ensuring that similar rates ap- interest earnings on the capital stocks ply across different countries and that held abroad by Latin Americans could there is some degree of constancy and be taxed the taxable income base would credibility in tax policy, harmonization increase by around 5 percent in both Ar- can provide a more stable environment gentina and Mexico. This would entail for investment and business. an improvement in the budget deficit in these countries of the order of one per- 6. Tax Reform in Practice cent of GDP. Aside from the effects on investment and growth, tax base erosion The 1980s saw many examples of major through capital flight is particularly seri- tax reforms in developing countries. For ous at a time when foreign borrowing, a review of the literature see Newbery an important means of financing budgets and Stern (1987), World Bank (1988, in the 1970s, is increasingly unavailable 1991), Gillis (1989a), Bird and Oldman (see Sachs 1989). The greater use of (1964, 1990), Tanzi (1990 and 1991a), broad-based taxes (e. g., VAT)whose bur- Boskin and McLure (1990), Ahmad and den is spread more evenly over the popu- Stern (1991), and Khalizadeh-Shiraziand lation, and a lowering and harmonization Shah (1991), and the references cited in of tax rates can both help reduce capital Table 7. A number of related pressures flight. came together which led to several gov- Corporate income taxation is also sub- ernments grasping the nettle. First, a se- ject to international constraints (see, ries of ad hoc tax measures in response e.g., Shah and Joel Slemrod 1991). A firm to increasing expenditure and changing planning to build a plant in the industrial economic structure had accumulated, belt along the Mexico-USA border will over a number of decades, into systems naturally consider tax conditions in both which were cumbersome, riddled with countries along with other site factors. complications and anomalies, and subject Mexico's leeway in setting its own corpo- to widespread evasion. Second, views on

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burg.ess and Stern: Taxation and Develovment 805 trade and industrial policy were changing were chosen for four reasons: (i) they are with an increasing emphasis on reducing reasonably well documented; (ii) they in- trade barriers and protection. A tax sys- volve significant shifts in tax structure; tem heavily focused on trade taxation, a (iii) they are diverse and cover a wide common position as we saw in Sec- range of circumstances; and (iv) they il- tion 3, is inconsistent with these trade- lustrate some broad lessons which will oriented policies. Third, the difficulties be brought out in Section 6.2. of the international recession of the early In Table 7 the countries form the rows 1980s, following the increase in oil prices and we have selected information for the of the late 1970s, led to severe macroeco- columns which gives a picture of prere- nomic problems in many countries. The form conditions and pressures, central resulting short-term revenue difficulties elements of the reform, and outcomes. accentuated those of the long term which We discuss the table initially in terms had already been mounting. Finally, of the columns focusing respectively on these pressures together generated the prereform conditions, direct taxes, indi- perceived necessity for "stabilizationand rect taxes, and administrative aspects of structural adjustment" in which public the reforms. We then consider two exam- finances would have to play a key role ples of tax reform in detail (Indonesia, (see World Bank 1988, 1991; Blejer and Mexico) because, as we have empha- Chu 1989; Tanzi 1990 and 1991a). sized, it is important to see the tax system In this section we discuss eight exam- in its entirety. The examples of Bangla- ples of major tax reforms in developing desh, India, and Pakistan, where reform countries (namely Korea, Mexico, Co- has yet to take place, are provided as a lombia, Turkey, Jamaica, Indonesia, Sri contrast. Lanka, and Malawi; see Table 7 for refer- Prereform Conditions. While coun- ences) taken mostly from the 1980s, to- tries vary greatly in the accumulation of gether with, briefly, three examples from problems and pressures which lead to the South Asia (India, Pakistan, and Bangla- decision to embark on tax reform, there desh) where pressures for tax reform are are a number of characteristics of tax sys- severe but where substantial reform has tems which often appear to be relevant. not yet occurred.50 Brief histories of the Their joint effect is to make tax systems selected examples of reform are pre- unwieldy and ineffective, thus encourag- sented in Section 6.1 and general lessons ing reform. A number of the problems are drawn from them, in a way that links are related and include: (i) perceived with our earlier theory, in Section 6.2. overreliance on particular sources such We must emphasize that tax reforms as the oil sector or customs duties; (ii) have been taking place for centuries. Our narrow bases, particularlyfor income tax- choice of examples must necessarily be ation, as a result of, inter alia, loopholes, selective and we have chosen only some lack of clarity of laws, and evasion; (iii) of those from recent experience. complications in indirect taxation often from a cumulation of ad hoc measures; 6.1 of Tax Reform Experience and (iv) weak administration. These con- Space constraints prevent a detailed ditions do indeed provide strong argu- discussion of the selected examples and ments in favor of reform but it is interest- instead we set them out analytically in ing that they can be present for extended summary form in Table 7. The examples periods of time without reform actually taking place. Indeed several of them are 5 The remarkrefers to the time of writing (late 1991). likely to apply to any developing country

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions >0 ~~ 00 i 0~~

~~r- _ c s 0 0 *~ 0~~

~~_c _ i Co d~~

~~C~C~~

~~E o E~~~~~~~~~i o~~

~~C~~~~~~~~~~~~~~~~~~~~~~

~~0 . D 0~~

~~400 0 ~ ~ ~ 0- ( W~~

~~Q) cb1 cb~~

~~~~~~() z Y d~~

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 0 0 F."I5 0~~~~~~~~~~~~~~~~~0r 0 4 -~~

~~c~~~~~~~~~~~~~~~~~~~~~~~1~ ~ ~ ~ ~ ~ ~ ~ ~ 1~~

~~$.. ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ .~~

~~cli V) ~ ~ ~ ~ ~ ~ ~ ~ +~~

~~" E4 0' 0 q.) 15C ~ ~ ~ 0 ~ ~ ~ C.22 0 C=~~~~~~~~ 0 ~ -=~~

~~Z ~~~Q ~~~~~~~~~~~~mZ~~

~~C, cl #)ci~~

~~t 00bi0~~~~~~~~~0 05~~

~~Z~~~~~~~~0r >~~~

~~t w 0~~~~~~ Q0~~~~~~~ Q ~~~~~~~~~~~~~~>t~~

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 4-1 1 4-1~~

towo U) C.) 0 4-i tt " - = .- (L) VI 4.1 C) (U 4-J tn > C= M > > rA 4-J to, 0 0 4-J Co 4-J 4-1 z. llc (:z . CZ 4-j _V bJD 4.1 GO 4-J CZ bio 4-J tt0) .2 (n e. 4-J 4-i bC M t4 r. 41 4-J tt , Q) 0) : o r" - $.. E > Q) v, m 4-1 41 -v 4-1 > IM 0 4-J CU M > r" s.. 0 > 4-1 >, 0) = = o 4--J cn 4-i9 > m

~~C9 C9~~

4-1 4-J 4-1 (Z (U w E (U u r. , M 36. 0 -0 cd C) CZ 4.1 CO bJD -0 7:1 (U CZ 4-J bJD C,4,) q) m 4-1 Z 4-j s.. $.. U 4-1 ,j v, r" .= 4-1 0 i34 o s.. cq 4- 0 CO i=4 CZ 4-1 > 1-4 C,) 41 4-J -4 4- 4-1 > 1-4 4., 4-1 4-, 4.1 Q) C) 4-1 b-C r.) 4..o ..C, 4., 4.5 41 r, 4-, r. z 4-1 M = = JD -- >< 4-J s.. 43 > -b" -C;

~~cq CO cq M~~

~~4- 4-J M 0~~

~~0 4-1 u 40 = $.. w 4.1 0 Co (A 79 0~~

~~V Q) 4 4-J (=) 't Co Co) 4-1 0 4.1 (U W a)~~

~~4.1 > bt bJD CK) 4-1 IM A-4 4-1 C. 4~~

~~cq M cq~~

rA w rA 4-J >1 bJD 4-i 0 0 cd tn 'n s.. U M 4-1 4-J W 4., (U .0 cd :t4 g (L) = tn = - .- 6. "I t') 4-1 0 0 >, biC 4- rl 4-1 t) CA - 4-i (C) 4 4-i (1) $" 4-1 lt $.. Cd > (1) iZg biD 4-1 b.0 41 r, ct s.. -14 0 -b-b (Z) 14. $.. % E ,a _= s.. (1) b-C C) cl 0 >, c) tt 4- Ct 12 t I : 9 -t 4- > r, a blO Cl 41 rA ct 4- 4- 4- 4-1 M 4-J 4-J M Cl > 0 > > > 4-1 1,1 > 4-J 4-J c) C, 4-J

~~cq CO CO~~

~~4-J~~

~~cli 0 M~~

~~C.0~~

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~>~~~

~~0 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0.t _~~~~~~~~~~~~~~~~~~~~~~~~~~~~b C.) _~~

~~U 0) It $-- 4-, - 0..~~

~~0 'C (1)~~~~~bi C~~

~~o~~~~~~~~~~~~~~~2 4- a) 4 a5E~~

~~I7~~~~~~~~~~~~~~~~~~~-~ ~ ~ ~ - X 3 b | j | S 4 ';~~~~~~~~~~~~~~~~~~~~~~~~~c> a)|2~=b~~

~~0 I ~~~~~~~~~~~~~~~~~~~~~~~~41c q ~~~~~~~~~~~~~~~~~c -0 qc~~

~~> > cc00 > 00 00~~~~~~~~~~~~~~~~~~~~~~i 0) 00~~~~~~~~~ bD.~~

10~~~~~~~~~~~~~~~~~~~~~~~~~10 Ez. 41z -2 m0 c) in S. zt-=~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~INe C 0 v~~~~~~~~~~~~~~~~C r" _t bi 0AC AC o~ 0-s ~ ~ ~ ~ ~~~~~~Q p~~~~~~~~~~~~~~~~~~--~~~~~~~~~~~~~~~~~~~~~~~ i _H _ 05 p 00~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~: 5 U~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~c

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions I M~~

~~Cd bJD 0 bo .- (U V) = 4.1 > 10 m tl : III z > 0 $-- bJD 4-1~~

~~TO 4 cd u Q.) bO > cd~~

~~N) 0 O 0 - Cd ci t 4-1 bt 4-1 11Z O V, r. '4-1 bo 4-0 O. c) 0 4-1 (U~~

~~4-1 0 (u Cd 4-i 4-1~~

~~00 4-0 (u O 04 4.j~~

~~to 00m b-C O 4-j~~

~~4-1 ;S 4.1 C= cd bJD CZ bID x bJD < VI = E - r. Cd Cd r- bb 0 (1) 4-1~~

~~0 I.. cd"~~

~~00 4-1 w (L) cd biO~~

~~c) Q) O Cl 0 -14N-my, 0 w $.. r. (u > > 0 4) C, >~~

~~W (U 0 .Cd 4.j~~

~~.4..' Cd 4-j~~

~~Cq (1)~~

~~IM ND~~

~~4.1 M~~

~~41 bt C:~~

~~0.)~~

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 811 selected from, for example, Table 2. in Table 7 were in relatively strong posi- There does not appear to be a combinations domestically. Finally, in four out tion of conditions which will inevitably of the eight cases tax reform was a logical lead to reform. complement to revisions in trade and in- In addition to these difficulties of the dustrial policy. In these cases, desire for tax system, Table 7 points to the impor- a more open trade policy was associated tance of macro problems and the desire with moves to reduce trade taxation and for changes in trade and industrial policy often coupled with reductions in corpo- as pressures contributing to reform. A rate taxation to attract investment. At macro crisis is, however, neither neces- given levels of expenditure these reduc- sary nor sufficient for the occurrence of tions require, in the long run, an alterna- reform. Indeed there are arguments tive source of revenue.52 which suggest it may hinder reform. Cri- Direct Taxation. Reform of direct taxa- sis sometimes acts as a catalyst, in partic- tion has been an integral part of the re- ular when the IMF or World Bank are form experience of all countries in Ta- called upon. On the negative side, gov- ble 7. The effect of these measures has ernments facing macroeconomic crises not, on the whole, been to increase reve- may be dominated by short-term con- nue from direct taxes, and the overall cerns and may postpone tax reform, rise of indirect tax revenue (especially which might be seen as a medium- to VAT) has meant that their relative role long-term venture.51 It must also be kept has generally been reduced. Instead the in mind that both measures to deal with measures were intended to promote the short-term crises and those associated taxation of incomes in a more rational with tax reform may generate political (less distortionary, anomalous, or dis- unpopularity. Layering tax reform on top criminatory) way, requiring less effort, of austerity may pose political difficulties using broader bases with fewer loopholes even for secure governments. Finally re- and providing a more even distribution forms may be "tainted" by association of liabilities. A common concern in the with international lending organizations prereform period was that income tax or with particular domestic political par- systems were, and were seen as, ad hoc, ties. New administrations may have the irrational, and inequitable and that this advantage of political goodwill in imple- contributed to widespread evasion and menting tax reform (see e.g., Turkey, nonpayment. There have been attempts Mexico). to reach personal incomes by curtailing It would appear in general that govern- opportunities for evasion, especially in ments which are more secure and are the higher income brackets. Different prepared to take a longer-term view are types of realized income (e.g., capital more likely to embark on tax reform than gains, rents, dividends, etc.) have in those which are not. At least half, but some countries been included, along not all, of the governments represented with labor income, in the taxable base (for example, Mexico and Indonesia). 51There is, however, a growing consensus, both Deductions, exemptions, and loopholes among governments and internationalorganizations, which had represented significant routes that macroeconomic stability is impossible without a profound and permanent correction of the public finances. Tax reform is thus increasingly becoming 52 It is possible in some cases that reductions in an essential, short-termconsideration as opposed to tax rates could increase revenues. This is more likely a secondary, exclusively long-term concern (see e.g., to occur for a specific element, such as a tax on a World Bank 1991; Aspe Armella 1992; McKinnon particularitem, than an across-the-boardreduction 1991). for a tax category (e.g., all tariffs).

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 812 Journal of Economic Literature, Vol. XXXI (June 1993) for evasion have been cut back (e. g., Co- forms (Mexico, Turkey, Colombia, In- lombia, Mexico) and effective rates have donesia, and Malawi) in Table 7 involved been reduced in order to improve incen- the introduction of a VAT and Korea had tives for payment and to adjust for the already adopted one in 1977 (see Section effects of inflation (e.g., Turkey, Jamaica, 3.1). The introduction was in most cases and Mexico). a considerable success in terms of cover- Another major element of the reform age, revenue, and simplicity relative to of direct taxation has been the integration the existing system. Table 7 shows that of the systems of personal income and this success covers countries with a range corporate income taxation. Corporate of GNPs per capita and economic struc- rates have often been lowered to be com- tures. In all cases compliance with indi- patible with personal rates (see, e.g., rect taxation was felt to be improved and Mexico, Colombia, Jamaica, and Indone- administration simplified. The compli- sia). This has been achieved, for example, ance effects, both through cross-checking by changing the corporate tax rate struc- within the VAT itself and with the linking ture into a single tax rate equivalent to of information for the corporate and per- the highest personal tax rate. Integration sonal income taxes, do appear to have has been carried out to encourage invest- been significant (for further discussion ment, prevent tax arbitrage, and remove see, for example, Tait 1988). The revenue distortions against, for example, equity effects of VAT introduction can be sub- financing. Greater integration has also stantial, even over a short period, with, had, to some degree, the effect of im- for example, VAT revenues rising from proving compliance by standardizing lia- 1 percent to 4 percent of GDP between bilities, pooling information, coordinat- 1984 and 1989 in Indonesia. ing administrative and enforcement Administration. In most of the coun- procedures, closing loopholes, and re- tries, notwithstanding substantial tax col- ducing incentives for nonpayment. lections, administrative weaknesses and Indirect Taxes. The emphasis of tax re- evasion remain as severe problems. In form for the eight countries shown in Ta- most of the countries seen in Table 7, ble 7 has been on indirect taxes. Indirect the provision of a clear and effective legal tax reform has been characterized by two structure has been seen as critical for the prominent and related features. First, in smooth functioning of administration. line with a wider economic policy of trade Constancy, transparency, and honesty as liberalization, there has been a general concerns government policy have been reduction in the role of trade taxes. Sec- needed to maintain popular support for ond, particular emphasis has been given the reforms. The inclusion of different to the development of effective systems tax types under a coordinated administra- of taxation of domestic goods and ser- tive system has helped with information vices, especially through the introduc- gathering and enforcement. Self-assess' tion of VAT systems.53 ment for income tax purposes has been The rise of the VAT has undoubtedly used effectively in a number of countries constituted the single most important el- (for example, Indonesia), and VAT sys- ement of tax reform in developing coun- tems create a good audit trail for the au- tries over the past decade (see Tait 1988; thorities. Of course, shifting costs from Cnossen 1991). Five out of the eight re- the public to the private sector does not make these costs go away. Small businesses can find the VAT, for example, a 5 Imports and exports can be treated within VAT systems in a nondistortionarymanner thus further very frustrating and irksome burden. reducing the role of foreign trade taxes. Indonesia and Mexico. We discuss

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 813 briefly two examples of tax reforms taking coordinated sales and excise taxes. As re- each as a whole, Indonesia and Mexico. gards direct taxation, personal and corpo- They are both large countries, one low rate tax rates have been reduced both income and the other middle income, to improve compliance and to make them both are oil exporters and both carried competitive with those of the USA. out their reforms over similar periods These reductions have been most (see Tables 2 and 7). Indonesia faced marked in the 1989-91 period. The per- most of the problems listed in our discus- sonal income tax base has been broad- sion of the prereform conditions. The re- ened to include such sources as capital liance on oil sector taxes was particularly gains, and auditing and penalties for eva- striking (contributing over 60 percent of sion have increased the number of taxrevenue). It did not however face severe payers. Corporate and personal tax sys- macroeconomic pressures and was not at- tems have been brought into closer tempting, simultaneously with the tax re- alignment to prevent, for example, the form, to restructure trade and industrial double taxation of income, and both sys- policies and restore macroeconomic sta- tems have been inflation-proofed with bility. Indonesia's tax reform seems to some success. have seen some successes. The income The VAT as an ad valorem tax has pro- tax system has been improved through vided inflation protection for government reduction of evasion, clarification of the revenue in a way that the specific excises, law, self-assessment, closing loopholes, which it, inter alia, replaced, did not. and broadening the base and is now re- More than 30 federal excise taxes and garded as more equitable. VAT, which 300 state taxes were eliminated when the has been the major success of the reform, VAT was introduced in 1980, thus simpli- was introduced and now raises 4 percent fying tax administration and leading to of GDP and has also assisted in improv- improved compliance with indirect taxa- ing the effectiveness of the income tax. tion. It is also argued for Mexico that As a result of the reforms, oil tax revenue the zero-rating of agricultural exports via now provides only one-third of govern- the VAT provided more efficient incen- ment revenue. The overall revenue ef- tives than the preceding system. Com- fects have been broadly neutral, with to- pared to the previous amalgamation of tal government revenue as a percentage taxes which were proportional the VAT of GDP falling slightly from 19 percent is slightly progressive. It now raises 2.5- to 17 percent between 1983 and 1989 3 percent of GDP. (Mukul Asher 1990). However, revenue Sustained commitment to tax reform is collected in a more coherent and effi- and fiscal correction have led to reduced cient manner which is less dependent on inflation and debt, and the macroeco- the vagaries of the world oil market. nomic stability which prevails in the Mexico's tax reform was introduced in country today is in large part attributable the late 1970s and continues today, a pe- to the success of the reforms (see Aspe riod covering, although it was initiated Armella 1992). The overall effect of ra- prior to, the debt crisis. Rising inflation tionalization has also been to make taxes and fiscal deficits were important moti- more progressive. vating forces. Measures have included a Tax reform in Mexico has been influ- broadening of the income tax base and enced by its federal structure and its the removal of anomalies and distortions, proximity to the USA. Whereas the fed- many of which were associated with infla- eral turnover tax and various excise taxes tion, excessively high marginal rates, and which the VAT replaced had been admin- the operation of a large number of un- istered by the states, indirect tax admin-

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 814 Journal of Economic Literature, Vol. XXXI (June 1993) istration is now the preserve of the fed- less distortionary manner. In both coun- eral government. This led to the need tries dramatic reform of tax administra- to provide incentives to collect VAT reve- tion with the imposition of new manage- nue and relinquish previous revenue ment and organizational structures has rights. A carrot took the form of a 12 led to significant improvements in tax percent share of total federal revenues collection. On the whole these measures to be distributed to the states the year have helped reduce dependence on oil that VAT was introduced. The closeness revenue and have contributed to macro- of the large USA economy also meant economic stability. that individual income tax rates and post- Pressure Without Reform. Finally in tax returns on liquid assets, import taxes, our discussion of the experience of tax excises on luxury items, and corporate reform we point briefly to three large tax rates had to be consistent with those countries where, notwithstanding severe in the USA to prevent, respectively, eva- problems with the existing system of the sion and capital flight, smuggling, and kind associated with the countries dis- diversion of investment. cussed in Table 7, reform has not oc- Broadly speaking, the tax reforms of curred. These are India, Pakistan, and Indonesia and Mexico have shown im- Bangladesh. The three countries share portant elements of success both on the a legal structure for taxation inherited indirect and direct side. Rationalization from the British colonial period in undi- of the direct tax system through rate low- vided India. In all three countries this ering, base-broadening, simplification, has remained largely intact, although and administrative reform has been effec- they have had different experiences as tive in both countries. However, in In- regards total revenue. India collects donesia the personal income tax raises around 17 percent of GDP in taxes, Paki- less than 1 percent of GDP and in Mexico stan around 13 percent, and Bangladesh the figure is less than 2 percent; the over- 8 percent (figures for the mid-1980s-see all tax to GDP ratio in Mexico is only Ahmad and Stern 1987, and 1991).56 14 percent and in Indonesia 15 percent The Indian constitution divides taxes (Table 2). The scope for sustained reve- into a central list and a state list. The nue gains through personal income taxes constitution is taken seriously in India. or social security contributions is limited States guard their rights jealously and the and there is no doubt that in both coun- Center does not want to give up its reve- tries low coverage and evasion remain nue sources. The constitution thus places as major problems. considerable constraints on reform. Do- The main success has been with the mestic indirect taxes raise around 11 per- increased role of domestic indirect taxa- cent of GDP, split roughly in half be- tion, in particular VAT. Though for the tween the central excises (mostly on main part revenue effects have been broadly neutral,55 it is clear that the changes have meant that revenue is now 56In comparing these figures with those of Table 2 it should be emphasized that only central taxes raised in a more rational, equitable, and have been included in that table and revenues collected at lower administrative levels (for example, state for India and province for Pakistan)are thus 5 The share of corporateincome taxationis larger unrepresented in that table (see Section 5.5). For due mainly to the importanceof the mineral sector Pakistanand Bangladeshthe collections at this lower in both countries. level are negligible but for India another 5-6 percent 55Taxrevenues, however, have recently increased of GDP is added, largely from state sales taxes and as a share of GDP in Mexico in the 1988-91 period excises on alcohol, which are under state control (see (Aspe Armella 1992). RajaJ. Chelliah 1989).

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 815 large-scale production) and the state with the strong dependence on trade tax- taxes. An attractive reform would be to ation is making reform in that country integrate indirect taxes into a single co- an urgent priority. herent system such as the VAT, but con- Bangladesh's revenue problems have stitutional constraints make this difficult been greatly eased by foreign aid. Ban- and the states would not wish to give gladesh has managed to raise similar up their revenue rights. These rights quantities in external assistance (around were relinquished with some difficulty 8 percent of GDP) to those raised inter- in the case of VAT introduction in Mex- nally through the tax system. With an ico. This provides an important lesson- essentially free external source of finance a tight constitutional specification of per- of such importance, the incentive to em- mitted taxes can inhibit innovation and bark on reform of internal revenue rais- reform. India introduced in the mid- ing, with its attendant political dangers, 1980s a modification of the central excise is considerably weakened. system called MODVAT which allows for 6.2 Lessons from Tax Reform in some rebating and partially mitigates the Developing Countries cascading effect of this system but its scope is fairly narrow. Notwithstanding We have discussed briefly the expe- the constraints on reform and widespread rience of eight countries where tax re- evasion, the proportion of GDP raised form has taken place (see Table 7). They in tax revenue by India has not been out- come from Central and South America, standingly low by the standards of devel- Africa, and Asia. Some are very large oping countries (see Table 2), and has countries, others small islands, some are increased significantly over recent de- diversified, others have a strong depen- cades. An important question, however, dence on particular sectors, some are is whether government revenue is being poor and some are richer. Their experi- raised in an efficient manner and there ences have been very different, but we the answer is clearly negative. can nevertheless draw out some common Constraints on reform in Pakistan and themes. Bangladesh have not come from respect Direct Taxes. Principles of equity and for the constitution. On the other hand efficiency suggest that the base for in- the administrations appear to be weaker come taxation should be the sum of in- than in India and the political systems comes an individual or household re- less stable. Government may have rea- ceives from all sources. This task, for soned that the administration is too weak understandable reasons, seems to be be- and the dangers of political unpopularity yond most developing countries. Re- too great to risk embarking on a major forms have been largely pragmatic, reform. On the administrative side this aimed mainly at reducing evasion and in- conclusion would not seem to be justi- volving rationalization of the method of fied, at least for Pakistan. A number of assessment and collection through, inter poor countries with weak administrations alia, administrative reform, stronger en- have carried out reform with some suc- forcement, base-broadening, reduced cess (take for example the case of Malawi differentiation and reductions in marginal in Table 7). Yet Pakistan has consistently rates. In part, one might see the expan- ducked the problems of reform, it always sion of VAT, which is a tax on wages and appearing attractive to postpone that par- profits, as part of this pragmatic re- ticular problem. However, the magni- sponse. Income taxes can be collected tude of the deficits in Pakistan together more successfully from large corpora-

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 816 Journal of Economic Literature, Vol. XXXI (June 1993) tions, which is where many developing lems, including the introduction of VAT countries have concentrated their efforts. first at the import and manufacturing This can be supplemented with reform stage, and then progressing to the whole- of the administration. We do not, how- sale and retail, and the use of presump- ever, find the same degree of linkage be- tive methods. Experience has taught that tween economic principles and public fi- administrative abilities in developing nance practice that we find for indirect countries should not be underestimated. taxes. While it is foolhardy to suggest that Indirect Taxes. As we saw in Section everything is possible, one must at the 3, there has been a long-run movement same time avoid the corresponding mis- over time away from import duties and take of assuming that current administra- towards domestic taxes. This is in the tion is so poor that nothing is possible direction indicated by theory as being except a simplification of the existing sys- more efficient, although such movement tem. is dictated in large part by the growth While we have seen that broad-based of administrative capabilities and the do- indirect domestic taxation is not only de- mestic production base. Within domestic sirable but also possible in many coun- taxes we have seen in the last three de- tries, we cannot assume that it is a realis- cades a remarkable expansion of the tic option everywhere. Countries under value-added tax as the basis for indirect heavy revenue pressure may have to con- taxation. It has been seen to have many tinue to rely on import duties while the advantages including: a broad base; built- domestic indirect tax system is devel- in administrative checks; allowance of a oped although, as we have seen, the tran- neutral treatment of imports and domes- sition need not be of great length. In tic supplies; the rebating of exports; this case, too, experience and theory avoidance of the taxation of inputs thus come together in pointing the way. Quo- encouraging efficiency in domestic pro- tas are inferior to import duties and if duction activities; and encouragement to quotas are not auctioned, an opportunity keep records and thus help improve man- for revenue for the government is agement and organization. The VAT has missed. Very high import duties run into been a major part of the tax reform in problems of loss of revenue through sub- six of the eight countries considered, and stitution away from the imported com- all of the larger ones. modities and smuggling. Hence the tax With the advantages we have de- reformer will be well advised to begin a scribed, we may naturally ask why the study of commercial policy by reviewing, VAT has not been adopted elsewhere. and recommending abolition of all the The argument usually advanced against quotas, and by examining import duties it is its involvement of very many agents to see whether revenue could not be in- throughout the productive chain, many creased by reducing the highest ones. of whom may not be in the habit of keep- This is not to say that import duties ing proper records. There is no doubt should be uniform or that they should that this is indeed a major problem, but be adjusted so that effective protection experience shows that it can be exagger- is uniform; that is not the lesson from ated. It is impressive how successful theory (see Section 4). But it is sensible many countries which have introduced to review those which are outstandingly VAT have been in the collection of reve- high. A number of countries have found nues. There are a number of ways of cut- that reducing these very high import du- ting through the administrative prob- ties has led to an increase in revenue

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and DeveloDment 817 from those goods. Even if one rational- sonably stable over time. Tax reform izes import duties, however, the prob- measures should be publicized in ad- lems of the inefficiencies they create can- vance of implementation and, to the ex- not be eliminated. It is difficult to tent possible, agreed by those who will organize import duties so that they fall be affected (e.g., private sector institu- on consumer goods only. If they fall on tions, unions, farmers, etc.). Penalties for inputs then we have the problems of cas- evasion should be transparent and en- cading and inefficiency. Further, it is not forced from the start. The successful im- easy to organize domestic taxation so that plementation of fiscal policy depends domestically produced and imported critically on government credibility. goods are subject to the same tax, unless Haphazardimplementation of apparently one has a system like the VAT which is arbitrary taxes, rates, and codes breeds organized to do exactly that. Theory resistance. would recommend a move in this direc- Self-assessment does throw burdens on tion. One contribution that theory and individuals and firms, but in many cases empirical study can make to the analysis those can be burdens which have some of a complex system is the calculation of benefit to them in the long run, in terms the cumulative effect of taxes on prices, of their own management and organiza- which will clarify some of the conse- tion. It provides substantial relief to the quences of government action (see Ah- administrative authority, which can rely mad and Stern 1986 and 1987). on random checks and presumptive Administration. We have already ar- methods when carrying out those checks. gued that one must be careful to take a Increased dependence on self-assess- balanced view of administration-one ment can be important in the reform of can make the mistakes of assuming both direct taxes but is also an integral part that too much and too little is possible. of the VAT method. We will review here some lessons from It is important to go for simple, well- tax reform for administration. defined bases. We saw that the measure- It is much easier to stop people from ment of total personal income from all receiving money than to collect it from sources was difficult and that it may be them afterwards. Withholding may be an simpler to proceed component by com- important canon for the taxation of in- ponent rather than attempting to aggre- comes generated by companies and the gate them together. One can use the han- financial sector. 57 One can withhold taxes dles of interception, i.e., control of the on wages, interest, and dividends at passage of a good, as in the case of excises source. This may mean some distortions and customs. For countries with very in the sense that individuals may be pay- weak administration it may be that this ing different marginal rates on different must remain the major source of revenue sources of income, or there may be some for some time, though one should not violations of equity, but these are proba- underestimate scope for greater integra- bly minor compared to the revenue and tion with domestic taxation. We should administrative gains. emphasize the importance both of pre- The tax laws to which individuals are sumptive methods for taxing hard-to- subject should be transparent and rea- reach groups and of keeping the differentiation of rates within reasonable bounds. We should also stress the quality of, and 57Withholding also minimizes the collection lag which can significantlyreduce the real value of tax incentives facing, administrators. The revenue in high inflationcountries (see Section 5.4). period of reform may well be a time when

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 818 Journal of Economic Literature, Vol. XXXI (June 1993) administration can be improved. There the political party in government, but have sometimes been dramatic advances also high variability in liabilities over in revenue when incentives (carrots and time, as the tax burden is shifted from sticks) for tax collectors to collect tax have one group to another during political been improved together with an im- transition. Tax instability creates uncer- provement in the quality of personnel tainty and disrupts business, commerce, and the resources at their disposal.58 Fi- and the management of personal finally, one must establish and implement nances. stiff penalties for evasion. There is, unfortunately, often a real Politics. A tax reform will not be suc- tension between economically efficient cessful unless there is the political will taxes and taxes which are attractive to to carry it through and some degree of politicians. For example, protective im- acceptance and support by the popula- port duties can benefit small sections of tion. The acceptance may well be related the population which may in turn pass to the perceived honesty and efficiency that benefit to politicians. The losers are of government expenditure and a judg- dispersed throughout the economy in the ment as to the equity of the tax system form of those who have to pay higher itself. Where corruption and self-interest prices for what they consume. They may pervade the expenditure side, not only be difficult to organize and they may not are they likely to be present in tax admin- perceive the accumulation of small losses istration as well, but will also seriously very clearly. The taxation of agriculture undermine public willingness to be is often unsatisfactory as direct methods taxed. Tax reform may well overcome (e.g., a land tax) encounter stiff resis- some resistance if the newer taxes are tance. Generally indirect taxes are less seen to be more equitable by the popu- visible to consumers than direct taxes lace, but this cannot by itself overcome since they do not involve a direct pay- public perception of misuse of funds. It ment (at least by households) to the gov- may be that expenditure reform and ad- ernment and the tax is simply part of ministrative reform have to be coupled the total price paid by the purchaser. with tax reform (see Section 2.1). It is From the economic point of view, the partly for this reason that tax reform is best taxes are the most direct and the often associated with incoming govern- least avoidable. These are precisely the ments, where, at least in the initial taxes which generate the most opposi- stages, acceptance of and enthusiasm for tion. It is privately rational for an individ- reform is heightened. Once imple- ual to campaign very strongly against an mented, continuity is an attractive fea- unavoidable redistributive tax that falls ture of tax systems and different types on him in the hope that it may be re- of taxes should not be too closely associ- placed by an avoidable tax which falls ated with one political regime or another. on someone other than him. The land Not only would such association create tax is a clear example where this problem opposition by those who do not support arises.

5 For example, in the latter half of the 1980s Gha- 7. na's tax administrationunderwent radical reform. Conclusions This involved bonuses for meeting revenue targets, the recruitmentof qualifiedaccountants, and greater Our review of taxation and develop- stringency in dismissing errant staff. It is claimed ment began with the suggestion that the that largely as a result of these changes tax revenue objectives of taxation should be similar rose by 7 or 8 percentagepoints of GDP over a three- year period (personal communicationfrom Finance for developed and developing countries: MinistryOfficials, Nov. 1990). to raise resources for government expen-

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 819 diture in a way that is administratively We saw in Section 3 that developing feasible, equitable, and efficient. The countries are already collecting substan- main differences between developed and tial amounts in government revenue, developing countries then, lie not in the amounting to 20 percent or 25 percent objectives of government but in the con- of GDP, with 17 percent or 18 percent straints facing government. The weak- coming from tax revenue, and 5 per- ness of administration, limited experi- cent or so of GDP from non-tax revenue. ence with taxation, poor accounting, the These are, however, aggregate figures, low level of monetization in the econ- and there is very substantial variation omy, the high share of agriculture, and both across and within continents. Gen- the fact that "tax handles" are generally erally speaking, total government reve- few in number relative to developed nue in developing countries is about 10 countries, together imply that some as- percentage points of GDP below indus- pects of tax analysis for developing coun- trial countries where government reve- tries will have a different structure and nue averages between 30 percent and 35 emphasis. Broadly speaking, however, percent of GNP, with only a small we argued that similar techniques and amount of this coming from non-tax methods could be used for the analysis sources. If only taxation is considered, of taxation in both developed and devel- levels in less-developed countries are a oping countries, though, where models little over a half those in industrial coun- are used, their special features might tries (18.05 percent compared to 31.21 look rather different, reflecting the above percent of GDP). Within developing differences. countries we found that there was a weak In Section 2 the close relationship be- but significant relationship between the tween taxation and expenditure was em- tax ratio (tax revenue to GDP) and GNP phasized. We argued that it was a mis- per capita but no significant relationship take to see the tax problem as simply for industrial countries. This suggests one of financing a given aggregate expen- that economic structure may constrain diture. On the expenditure side, our re- taxation expansion in developing more view of arguments concerning the role severely than in developed countries, of the state pointed to substantial and and this view is supported by, for exam- important areas of government expendi- ple, the existence of a strong relationship ture, implying that the problem of raising between the tax ratio and the shares of resources will be a major one in less-de- agriculture and of imports. veloped countries. While expenditure The structure of taxation in developing must be controlled, this challenge cannot countries looks radically different from be tackled simply by reducing expendi- that of developed countries. For devel- ture to negligible levels. Attention also oping countries we have roughly two- needs to be paid to the composition of thirds of tax revenue coming from indi- expenditures. Finally in Section 2 we re- rect taxes, split roughly equally between viewed the different possible sources of domestic taxes and trade taxes. The re- finance available to the government, maining one-third consists largely of the namely raising government revenue, incorporate income tax. For developed ter-naland external borrowing, and print- countries, however, roughly two-thirds ing money. Debt and money finance of- comes from direct taxes, with personal ten prove to be unsustainable sources of income tax and social security payments revenue and in the long run there is no constituting the bulk of this. The struc- real substitute for taxation, broadly de- ture we have described for developing fined, for raising revenue. countries constitutes an average and

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 820 Journal of Economic Literature, Vol. XXXI (June 1993) there is enormous variation between suggested were useful for both devel- countries in their sources of revenue. As oped and developing countries. As well far as movement over time is concerned, as providing such guiding principles, the there seems to be a gradual move away normative analysis also tells us where to from trade taxes to domestic sales taxes look for key issues in judging policy and as domestic economies develop and, cor- helps us to structure the detail of our respondingly, tax tools can be fashioned applied investigations. to tax domestic production and transac- We also discussed in Section 4 theories tions. It seems, however, that the per- of political economy which attempted to sonal income tax has developed only model and explain the way in which gov- slowly and in some countries has stag- ernments behave, rather than the ways nated or declined as a fraction of GDP in which governments should behave. and tax revenue. We emphasized that such theories re- We examined formal tax theory and quired essentially the same type of em- models in Section 4 and argued that the pirical investigation as an approach which objective of positive tax analysis should was explicitly normative. From the per- be the design of feasible reforms and the spective of public choice or political tracing of the effects of tax changes economy we will want to know who will through to changes in not only revenue be likely to support or oppose a change but also household real income (and, and, from the normative, we have to more generally, living standards)i.e., the compare the gains to gainers with the identification of who gains and who loses, losses to losers. and by how much. For developing coun- In Section 5 we discussed constraints tries the critical assumptions in that anal- and pressures on tax systems in develop- ysis would concern the availability and ing countries. Weaknesses in the basic coverage of tax tools, the institutional and administrative functions of identification, sectoral structure of the economy, and assessment, enforcement, and collection the way in which domestic markets, par- often undermine the effectiveness of tax ticularly the labor market, function. systems in developing countries. We also While a calculation of the detailed effects emphasized the extensive scope of non- on final consumers (now and in the fu- compliance and evasion in less-devel- ture) might often prove overambitious, oped as opposed to developed countries. given the availability of data and prior Irrespective of how well a tax system is work on the economy, it should be kept designed on paper, tax reform cannot be firmly in mind as an ultimate goal. Posi- expected to be successful unless it ad- tive tax analysis for developing countries dresses these problems. We found that has concentrated largely on indirect the ability of powerful groups to block taxes, public-sector pricing and the cor- reforms greatly aggravates problems of poration tax, reflecting the main sources administration and evasion by constrain- of revenue, rather than on labor supply, ing the tax tools available to the govern- risk-taking, and saving, which have been ment. Macroeconomic laxity, manifest- the prominent topics in developed coun- ing itself in inflation, causes severe tries where direct taxation dominates. problems for the tax system. It also in In our normative discussion in Section large part arises from the weaknesses of 4 we showed how theories of optimal tax- public finance and expenditure control. ation could be analyzed to derive basic The openness of many developing coun- principles and we set out nine such guid- tries creates further stringent constraints ing principles for tax policy which we on tax options. For example, the kinds

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 821 of taxes that Mexico can expect to admin- we set out. We would argue, particularly ister are going to be closely related to for indirect taxes, that there are available income-earning opportunities in the reasonable compromises between what USA. is theoretically attractive, administra- Finally, in Section 6 we examined the tively reliable, and politically possible. experiences of a number of countries Consider an indirect tax system based which have attempted tax reform in the on the following elements: (i) a VAT with recent past. We saw that tax reform was one or two rates and some exemptions; most likely to come about when some (ii) excises on alcohol, tobacco, petro- or all of a number of preconditions ex- leum products, and some luxury goods; isted: (i) a desire to diversify away from and (iii) direct support for certain groups, high rates and a narrow base for revenue; possibly through subsidized rations. It is (ii) widespread evasion; (iii) increasingly a system which is broadly consistent with irrational and severe complications; (iv) the economic principles set out in Sec- extreme revenue pressure; and (v) a tion 4 and is capable of being adminis- change in economic strategy involving, tered by most less-developed countries. for example, trade liberalization, or mac- It can raise the necessary revenue, pro- roeconomic stabilization. vide the appropriate industrial incen- We saw that, generally speaking, the tives, cope with externalities, and protect reform of indirect taxation had been the poor. It may be supplemented by rather more successful than that of direct temporary tariffs to maintain revenue or taxation. As administrative abilities grow where infant industry arguments have and as the economy becomes more mon- genuine empirical support (and other etized with an increasing domestic base more direct means of support are unavail- for taxation, domestic indirect taxation able). In general, however, as adminis- will replace the taxation of foreign trade. trative capabilities permit, there should Within domestic indirect taxation, a be a move away from reliance on foreign movement to a system like the VAT, trade taxes to greater use of domestic in- which is noncascading and does not inter- direct taxes. fere with production efficiency, has be- It is much less easy to prescribe a di- come increasingly attractive and has rect tax system which lives happily with indeed been implemented in many coun- both the theory and what is administra- tries. As far as direct taxes are con- tively possible. An exception would be cerned, rationalization of the method of a land tax but this is often ruled out for collection and closer alignment of the in- political reasons. Corporate taxation is dividual and corporate bases can mean likely to remain a mainstay of taxation that, relative to the cumbersome systems in less-developed countries. But the that exist in many countries, revenue can measurement of personal and small-busi- be collected in a less distortive manner ness income from any one source, let with less effort. A tax system that is ra- alone from several, is extremely difficult, tionalized is easier to enforce and as a except for individuals whose occupations number of countries have shown, returns are entirely as employees in the more to more effective enforcement can be formal sectors. Even there, fringe bene- very substantial. fits and evasion can cause severe prob- Our analysis has shown that there are lems. feasible revenue-raising tools for the gov- To summarize, although there exists ernments of developing countries which no unflawed and detailed tax package fit perform fairly well relative to the criteria for universal prescription, it is evident

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 822 Journal of Economic Literature, Vol. XXXI (June 1993) that the combination of theoretical, ad- ARROW, KENNETH J. "The Economic Implicationsof Learningby Doing," Rev. Econ. Stud., June 1962, ministrative, and political considerations 29, pp. 155-73. can, when supported by careful empirical ASCHER, WILLIAM. "Risk, Politics and Tax Reform: analysis, provide reasoned assessment Lessons from Some Latin American Experiences" and informed advice. We ignore any one in GILLIS, ed. 1989a. ASHER, MUKUL. "Reforming the Tax System in In- of these elements at our peril. donesia." Paper presented at the World Bank Con- Work on taxation and development has ference on Tax Policy in Developing Countries, shown great advances in the last twenty 1990. ASPE ARMELLA, PEDRO. Stabilisation and structural or thirty years, both from the perspective change: The Mexican experience. Lionel Robbins of what governments have achieved and Lectures, LSE, London. MIT Press, forthcoming. in terms of the application of analytical ATKINSON, ANTHONY B. "JamesM. Buchanan'sCon- tributions to Economics," Scand. J. Econ., 1987, economics. The reform of the tax system 89(1), pp. 5-15. and the control of expenditure are at the . Poverty and social security. London: Har- heart of most structural adjustment prob- vester Wheatsheaf, 1989. . "The Distribution of the Tax Burden: 30 lems. Indeed, without fiscal stabilization, YearsAfter the Theory of Public Finance." Mimeo, structural adjustment will surely fail. The Welfare State Programme, STICERD, London countries of Eastern Europe, of the for- School of Economics, 1990. ATKINSON, ANTHONY B. AND STERN, NICHOLAS H. "Pi- mer Soviet Union, and those elsewhere gou, Taxation and Public Goods," Rev. Econ. relaxing or dismantling tight control of Stud., Jan. 1974, 41, pp. 119-28. the economy must find, and are finding, ATKINSON, ANTHONY B. AND STIGLITz, JOSEPH E. Lec- tures on public economics. NY: McGraw-Hill, that new tax systems are an urgent re- 1980. quirement of transition. The lessons of ATKINSON, ANTHONY B. AND SUTHERLAND, HOLLY. "TaxBenefit Models." STICERD OccasionalPaper theory and the experience of developing 10. London: STICERD, 1988. countries have much to offer these coun- BAHL, Roy. "The Political Economy of the Jamaican tries too. Tax Reform,"in GILLIS, ed. 1989a. . The Jamaican tax reform. Cambridge, MA: REFERENCES Lincoln Institute of Land Policy, 1991. BAILEY, MARTIN J. "TheWelfare Costs of Inflationary ABBOTr, JOHN C. Agriculturalmarketing enterprises Finance,"J. Polit. Econ., Apr. 1956, 64, pp. 93- for the developing world. Cambridge:Cambridge 100. U. Press, 1987. BALASSA, BELA. "OutwardOrientation," in CHENERY ACHARYA, SHANKAR N. Aspects of the black economy AND SRIVINASAN, eds. 1989. in India. New Delhi: Governmentof India, 1985. BALLARD, CHARLES L. AND FULLERTON, DON. "Dis- AHMAD, S. EfITISlIAM ET AL., eds. Social security in tortionary Taxes and the Provision of Public developing countries. Oxford: Clarendon Press, Goods,"J. Econ. Perspectives, forthcoming. 1991. BARRO, ROBERT J. "Are Government Bonds Net AHMAD, S. EHTISHAM AND STERN, NICHOLAS H. "Ef- Wealth?"J. Polit. Econ., Nov.-Dec. 1974, 82(6), fective Taxes and Tax Reformin India." Develop- pp. 1095-1117. ment Economics Research Centre Discussion Pa- BARRO, ROBERTJ. AND SALA-I-MARTIN, XAVIER. "Pub- per No. 25, U. of Warwick, 1983. lic Finance in Models of Economic Growth." . "The Theory of Reform and Indian Indirect Working Paper no. 3362, National Bureau of Eco- Taxes,"J. Pub. Econ., Dec. 1984, 25(3), pp. 259- nomic Research, 1990. 98. BATES, ROBERT H. "A Political Scientist Looks at Tax . "Tax Reform for Pakistan: Overview and Ef- Reform," in GILLIS, ed., 1989. fective Taxesfor 1975-76," PakistanDevelop. Rev., BAUER, PETER T. West African trade. Cambridge: Spring 1986,25, pp. 43-72. Cambridge U. Press, 1954. . "Alternative Sources of Government Reve- BERNHEIM, B. DOUGLAS. "RicardianEquivalence: An nue: Illustrationsfrom India, 1979-S0," in NEW- Evaluationof Theory and Evidence," NBER Mac- BERY AND STERN, eds. 1987. " roeconomic Annual, 2, Ed.: STANLEY FISCHER. "Taxation for Developing Countries," in Cambridge, MA: MIT Press, 1987, pp. 263-304. CHENERY AND SRINIVASAN, eds. 1989. BEST, MICHAEL H. "Political Power and Tax Reve- . The theory and practice of tax reform in nues in Central America," J. Develop. Econ., Mar. developing countries. Cambridge: Cambridge U. 1976, 3(1), pp. 49-82. Press, 1991. BHAGWATI, JAGDISH. "Directly-Unproductive-Profit- ALANO, B. P. JR. "Import Smuggling in the Philip- Seeking Activities" in The new palgrave. Eds.: J. pines: An Economic Analysis,"J. Philippine De- EATWELL, M. MILGATE, AND P. NEWMAN. London: velop., 1984, 11, pp. 157-90. Macmillan, 1987.

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions RIfss and Stern: Taxation and Develonment 823~~

BIRD, RICHARDM. "The Administrative Dimension cial Security in Developing Countries: What, Why, of Tax Reform in Developing Countries," in GILLIS, Who and How?" in AHMAD ET AL., eds. 1991. ed. 1989. CASS, DAVID. "Optimum Growth in a Aggregative BIRD, RICHARD M. AND OLDMAN, OLIVER, eds. Read- Model of Capital Accumulation," Rev. Econ. Stud., ings on taxation in developing countries. Balti- July 1965, 32, pp. 233-40. more: Johns Hopkins U. Press, 1964. CHELLIAH RAJA J. "Changes in Tax Revenue Struc- . Taxationin developingcountries. Baltimore: ture: A Case Study of India," in Changes in revenue Johns Hopkins U. Press, 1990. structure. Eds.: A. CHIANCONE AND K. MESSERE. BLANCHARD, OLIVIER J. AND FISCHER, STANLEY. Lec- Detroit: Wayne State U. Press, 1989. tures on macroeconomics, Cambridge MA.: MIT CHELLIAH, RAJAJ.; BAAS, HASSEL J. AND KELLY, MAR- Press, 1989. GARET R. "Tax Ratio and Tax Effort in Developing BLEJER, MARIO AND CHU, KE-YOUNG, eds. Fiscal pol- Countries, 1969-71, IMF Staff Papers, 1975, 22, icy, stabilisation, and growth in developing coun- pp. 187-205. tries. Washington, DC: International Monetary CHENERY, HOLLIS AND SRINIVASAN, T. N., eds. Hand- Fund, 1989. book of development economics. Amsterdam: BLISS, CHRISTOPHER. "Taxation, Cost-Benefit Analy- North Holland, 1989. sis and Effective Protection" in NEWBERY AND CHOI, KWANG. "Tax Policy and Tax Reforms in Ko- STERN, eds. 1987. rea." Paper presented at The World Bank Con- Bos, DIETER. Public enterprise economics: Theory ference on Tax Policy in Developing Countries, and application. Amsterdam: North Holland, 1990. 1986. CNOSSEN, SIJBREN. Excise systems: A global study . The political economy of progressive taxa- of the selective taxationof goods and services. Bal- tion. Berlin: Springer Verlag, 1989. timore: Johns Hopkins U. Press, 1977. . Privatisation: A theoretical treatment. Ox- . "Design of the Value Added Tax: Lessons ford: Clarendon Press, 1991. from Experience," in KHALIZADEH-SHIRAZI AND BOSKIN, MICHAEL J. AND MCLURE, JR., CHARLES E. SHAH, eds. 1991. World tax refornt. San Francisco: ICS Press, 1990. COLANDER, DAVID C. ed. Neoclassicalpolitical econ- BRAVERMAN, AVISHAY; HAMMER, JEFFREY AND AHN, omy. Cambridge, MA: Ballinger Press, 1984. CHOONG Y. "Multi-Market Analysis of Agricultural CONRAD, ROBERT F. AND HOOL, R. BRYCE. Taxation Pricing Policies in Korea," in NEWBERY AND STERN, of mineral resources. Lexington, MA: Lexington eds. 1987. Books, 1980. BRENNAN, GEOFFREY AND BUCHANAN, JAMES M. The COOK, PAUL AND KIRKPATRICK, COLIN, eds. Privatisa- power to tax. NY: Cambridge U. Press, 1980. tion in less developed countries. Brighton: . The reason of rules. Cambridge: Cambridge Wheatsheaf Books, 1988. U. Press, 1985. CORDEN, W. MAX. "The Relevance for Developing BROWNING, EDGAR K. AND BROWNING, JACQUELINE Countries of Recent Developments in Macroeco- M. Publicfinanceand the price system. Macmillan, nomic Theory," World Bank Res. Observer, July 1979. 1987, 2(2), pp. 171-88. BRUTON, HENRY. "Import Substitution" in CHENERY CORFMAT, FRANCOIS. "Computerising Tax Adminis- AND SRINIVASAN, eds. 1989. tration," in BIRD AND OLDMAN, eds. 1990, pp. 467- BUCHANAN, JAMES M. Public finance in the demo- 74. cratic process. Chapel Hill: U. of North Carolina COWELL, FRANK A. Cheating the government: The Press, 1966. economics of evasion. Cambridge, MA: MIT Press, . The economics and the ethics of constitu- 1990. tional order. Ann Arbor: University of Michigan DEATON, ANGUS S. AND BENJAMIN, DWAYNE. "The Press, 1991. Living Standards Survey and Price Policy Reform, BUCHANAN, JAMES M. ANDTOLLISON, ROBERTD. The- A Study of Coffee and Cocoa Production in Cote ory of public choice. Ann Arbor: U. of Michigan D'Ivoire." Living Standards Measurement Study Press, 1972. Working Paper no. 44. Washington, DC: The BUCHIANAN, JAMES M.; TOLLISON, ROBERT D. AND World Bank, 1988. TULLOCK, GORDON, eds. Toward a theory of the DEATON, ANGUS S. AND STERN, NICHOLAS H. "Opti- rent-seeking society. College Station, TX: Texas mally Uniform Commodity Taxes, Taste Differ- A&M U. Press, 1980. ences and Lump-Sum Grants," Econ. Letters, BUITER, WILLEM H. Principles of budgetary andfi- 1986, 20, pp. 263-66. nancial policy. London: Harvester Wheatsheaf, DERVIS, KEMAL; DE MELO, JAIME AND ROBINSON, 1989. SHERMAN. Generalequilibrium modelsfor develop- BUITER, WILLENI H. AND PATEL, URJIT. "Debt, Defi- mentpolicy. London: Cambridge U. Press for the cits and Inflation: An Application to the Public Fi- World Bank, 1982. nances of India,"J. PublicEcon., Mar. 1992, 47(2), DIXIT, AVINASH K. AND STERN, NICHOLAS H. "Deter- pp. 171-205. minants of Shadow Prices in Open Dual Econo- BULUTOGLU, KENAN AND THIRSK, WAYNE. "Tax Re- mies," Oxford Econ. Pap., Mar. 1974, 26(1), pp. form in Turkey." Paper presented at The World 42-53. Bank Conference on Tax Policy in Developing DORNBUSCH, RUDIGER. "Expectations and Exchange Countries, 1990. Rate Dynamics," J. Polit. Econ., Dec. 1976, 84(6), BURGESS, ROBIN S. L. AND STERN, NICHOLAS H. "So- pp. 1161-76.

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 824 Journal of Economic Literature, Vol. XXXI (June 1993)~~

DREZE, JEAN P. AND SEN, AMARTYA K. Hunger and tion, Unemployment and Development: A Two- public action. Oxford: Clarendon Press, 1989. Sector Analysis," Amer. Econ. Rev., Mar. 1970, DREZE, JEAN P. AND STERN, NICHOLAS H. "The The- 60(1), pp. 126-42. ory of Cost-Benefit Analysis," in Handbook of pub- HARROD, RoY F. "An Essay in Dynamic Theory," lic economics. Eds.: ALAN AUERBACH AND MARTIN Econ. J., Mar. 1939, 49, pp. 14-33. FELDSTEIN. Amsterdam: North-Holland, 1987, pp. HEADY, CHRISTOPHER AND MITRA, PRADEEP. "Opti- 909-89. mal Taxation and Shadow Pricing in a Developing . "Policy Reform, Shadow Prices and Market Economy," in NEWBERY AND STERN, eds. 1987, ch. Prices," J. Public Econ., June 1990, 42, pp. 1- 15. 45. HERSCHEL, FEDEIUCO J. "Tax Evasion and Its Mea- DUE, JOHN F. Indirect taxation in developing coun- surement in Developing Countries," Pub. Finance, tries. Baltimore: Johns Hopkins U. Press, 1988. 1978, 33(3), pp. 232-68. EATON, JONATHAN; GERSOVITZ, MARK AND STIGLITZ, HINIUCHS, HARLEY H. "A General Theory of Tax JOSEPH. "The Pure Theory of Country Risk," Structure. Change During Economic Develop- Europ. Econ. Rev., June 1986, 30(3), pp. 481- ment." Harvard Law School International Tax Pro- 514. gram, 1966. FISCHER, STANLEY. Indexing,inflation, and economic HITIRIS, THEO. "Tax Structure, Trade Taxes, and Eco- policy. Cambridge, MA: MIT Press, 1986. nomic Development: An Empirical Investigation," GARNAUT, RosS AND Ross, ANTHONY CLUNIES. Taxa- in TANZI, ed. 1990, pp. 29-41. tion of mineral rents. Oxford: Clarendon Press, HOFF, KARLA. "Land Taxes, Output Taxes, and 1983. Sharecropping: Was Henry George Right?" World GEMMELL, NORMAN. "Fiscal Dependence on Trade Bank Econ. Rev., Jan. 1991, 5(1), pp. 93-111. Taxes and Development: Some New Evidence." IMF. World economic outlook. Washington, DC: CREDIT Research Paper, Nottingham U., Dec. IMF, 1986. 1990. . Governmentfinance statistics yearbook, Vol. GERSOVITZ, MARK. "The Effects of Domestic Taxes XIII. Washington, DC: IMF, 1989. on Foreign Private Investment," in NEWBERY AND . Internationalfinance statistics. Washington, STERN, eds. 1987. DC: IMF, Dec 1986, Dec 1989, Ap 1990, Aug GIL Diaz, Francisco. "Some Lessons from Mexico's 1990. Tax Reform," in NEWBERY AND STERN, eds. 1987. JENKINS, GLENN P. "Tax Changes Before Tax Policies: GILLIS, MALCOLM. Tax reform in developing coun- Sri Lanka 1977-88," in GILLIS, ed. 1989a. tries. Durham: Duke U. Press, 1989a. KALDOR, NICHOLAS. Indiantax reform:Report of a . "Tacit Taxes and Sub Rosa Subsidies through survey. New Delhi: Ministry of Finance, 1956. State-Owned Enterprises." Mimeo, Duke U., . "Will Underdeveloped Countries Learn to 1989b. Tax?" in BIRD AND OLDMAN, eds. 1964, pp. 255- . "Comprehensive Tax Reform: The Indone- 65. sian Experience 1981-1988," (1989c) in MALCOLM KAY, JOHN. "Indirect Taxes" in The new Palgrave GILLIS, ed. 1989a. dictionary of economics. Eds.: JOHN EATWELL, GOODE, RICHARD. Governmentfinancein developing MURRAY MILGATE, AND PETER NEWMAN. London: countries. Washington, DC: Brookings Institution, Macmillan, 1987. 1984. KAY, JOHN A. AND KING, MERVYN A. The British tax . "The Economics of Tax Administration" in system. 4th ed. London: Oxford U. Press, 1986. BIRD AND OLDMAN 1990, pp. 475-91. KEHOE, TIMOTHY J. AND SERRA-PUCHE, JAIME. "A GORDON, RICHARD K., JR. "Income Tax Compliance Computational General Equilibrium Model with and Sanctions in Developing Countries," in BIRD Endogenous Unemployment: An Analysis of the AND OLDMAN 1990, pp. 455-65. 1980 Fiscal Reform in Mexico,"J. Pub. Econ., Oct. GOVERNMENT OF INDIA, MINISTRY OF PLANNING, RAJ 1983, 22(1), pp. 1-26. COMMITTEE. "Capital Formation and Saving in In- KHALIZADEH-SHIRAZI, JAVAD AND SHAH, ANWAR, eds. dia, 1950-51 to 1979-80." Bombay: Reserve Bank, Tax policy in developing countries. Washington, 1982. DC: World Bank, 1991. GREENAWAY, DAVID. "A Statistical Analysis of Fiscal KING, ROBERT G. AND REBELO, SERGIO. "PublicPol- Dependence on Trade Taxes and Economic Devel- icy and Economic Growth: Developing Neoclassi- opment," Public Finance, 1984, 39(1), pp. 70- cal Implications." National Bureau of Economic 89. Research, Working Paper No. 3338, Apr. 1990. . "Trade Taxes as a Source of Government KRUEGER, ANNE 0. "The Political Economy of the Revenue: A Comment on the Bedrossian-Hitiris Rent-Seeking Society," Amer. Econ. Rev., June Re-estimation," ScottishJ. Polit. Econ., June 1985, 1974, 64(3), pp. 291-303. 32; pp. 205-08. . "Government Failures in Development," J. HAQUE, NADEEM U. AND MONTIEL, PETER. Ricar- Econ. Perspectives, Summer 1990, 4(3), pp. 9-23. dian Equivalence, Liquidity Constraints and the KUMAR, DHARMA. "The Fiscal System," in The Cam- Yaari-Blanchard Effect: Tests for Developing bridge economic history of India. Vol. II. Eds.: Countries." Development Research Department, DHARMA KUMAR. Hyderabad: Orient Longman, World Bank, Washington, DC, 1987. 1982. HARRIS, JOHN R. AND TODARO, MICHAEL P. "Migra- LESSARD, DONALD R. AND WILLIAMSON, JOHN. Capi-

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions Burgess and Stern: Taxation and Development 825~~

tal flight and third world debt. Washington, DC: ROEMER, MICHAEL. "Ghana: Missed Opportunities," Institute for International Economics, 1987. in World economic growth. Ed.: ARNOLD HARBER- LEWIS, W. ARTHUR. "Economic Development with GER. San Francisco: Institute for Contemporary Unlimited Supplies of Labour," Manchester Studies, 1984, pp. 201-25. School, May 1954, 22, pp. 139-91. ROMER, PAUL M. "Increasing Returns and Long- LIEUWEN, EDWIN. "The Politics of Energy in Vene- Run Growth," J. Polit. Econ., 1986, 94, pp. 1002- zuela," in Latin American oil companies and the 1037. politics of energy. Ed.: JOHN WIRTH. Lincoln: U. . "Capital Accumulation in the Theory of of Nebraska Press, 1985. Long-Run Growth," in Modern macroeconomics. LUCAS, ROBERT E., JR. "On the Mechanics of Eco- Ed.: ROBERT BARRO. Cambridge, MA: Harvard U. nomic Development," J. Monet. Econ., July 1988, Press, 1989. 22(1), pp. 3-42. SACHS, JEFFREY D., ed. Developingcountry debt and MALIYAMKONO, T. L. AND BAGACHWA, M. S. D. The economic performance. Vols. I-III. Chicago: U. second economyin Tanzania. London:James Cur- of Chicago Press, 1989. rey, 1990. SAH, RAAj K. AND STIGLITZ, JOSEPH E. "The Taxation MCKINNON, RONALD I. Theorder of economicliberal- and Pricing of Agricultural and Industrial Goods ization:Financial control in the transitionto a mar- in Developing Countries," in NEWBERY AND STERN, ket economy. Baltimore, Johns Hopkins U. Press, eds. 1987, ch. 16. 1991. SANDMO, AGNAR. "Buchanan on Political Economy: MCLURE, CHARLES E. "Analysis and Reform of A Review Article," J. Econ. Lit., Mar. 1990, 28(1), the Colombian Tax System," in GILLIS, ed. pp. 50-65. 1989a. SARGENT, THOMAS J. AND WALLACE, NEIL. "Some MCLURE, CHARLES E. AND ZODROW, GEORGE. "Tax Unpleasant Monetarist Arithmetic," Fed. Res. Reform in Colombia: Process and Results," in Bank Minn. Quart. Rev., Fall 1981, 5(3), pp. 1- KHALIZADEH-SHIRAZI AND SHAH, eds. 1991. 17. MILLWARD, ROBERT ET AL. Publicsector economics. SCOTT, MAURICE F. A new view of economicgrowth. Harlow: Longman, 1983. Oxford: Clarendon Press, 1989. MUSGRAVE, RICHARD A. Thetheory of publicfinance. SEADE, JESUS. "Tax Revenue Implications of Ex- NY: McGraw-Hill, 1959. change Rate Adjustment," in TANZI, 1990, pp. 54- . Fiscal systems. New Haven, CT: Yale U. 65. Press, 1969. SHAH, ANWAR AND SLEMROD, JOEL. "Taxation and . "Income Taxation of the Hard-to-Tax Foreign Direct Investment," in KHALIZADEH-SHI- Groups," in BIRD AND OLDMAN, eds. 1990, pp. RAZI AND SHAH, eds. 1991. 299-312. SHAH, ANWAR AND WHALLEY, JOHN. "The Redistribu- MUSGRAVE, RICHARD A. AND PEACOCK, ALAN T. Clas- tive Impact of Taxation in Developing Countries" sics in the theory of publicfinance. London: Mac- in KHALIZADEH-SHIRAZI AND SHAH, eds. 1991. millan, 1958. SHALIZI, ZMARAK AND THIRSK, WAYNE. "Tax Reform NEWBERY, DAVID M. G. "Taxation and Develop- in Malawi" in KHALIZADEH-SHIRAZIAND SHAH, eds. ment, (1987a) in NEWBERY AND STERN, eds. 1987, 1991. ch. 7. SHOVEN, JOHN B. "Applied General Equilibrium Tax . "Identifying Desirable Directions of Agricul- Modelling,"IMF Staff Papers, 1983, 30, pp. 350- tural Price Reform in Korea," (1987b) in NEWBERY 93. AND STERN, eds. 1987, ch. 18. SHOVEN, JOHN B. AND WHALLEY, JOHN. "Applied NEWBERY, DAVID M. G. ET AL. "Road Transport Taxa- General-Equilibrium Models of Taxation and In- tion in Developing Countries," World Bank Dis- ternational Trade,"J. Econ. Lit., Sept. 1984, 22(3), cussion Paper No. 26. Washington, DC: World pp. 1007-51. Bank, 1988. SKINNER, JONATHAN. "Prospects for Agricultural Land NEWBERY, DAVID M. G. AND STERN, NICHOLAS H. Taxation in Developing Countries" in KHALIZADEH- eds. The theory of taxationfor developing coun- SHIRAZI AND SHAH, eds. 1991. tries. NY: Oxford U. Press, 1987. SRINIVASAN, T. N. "Neo-classical Political Economy, NEWBERY, DAVID M. G. AND STIGLITz, JOSEPH E. the State and Economic Development." Lecture The theory of commoditystabilization. Oxford:Ox- to Asian Development Bank, 1985. ford U. Press, 1981. STERN, NICHOLAS H. "The Theory of Optimal Com- NOZICK, ROBERT. Anarchy, state and utopia. NY: Ba- modity and Income Taxation: An Introduction," sic Books, 1974. in NEWBERY AND STERN, eds. 1987, ch. 2. PASTOR, MANUEL. "Capital Flight from Latin Amer- . "The Economics of Development: A Sur- ica," in World Development, 1990, 18(1). vey," Econ. J., Sept. 1989, 99(397), pp. 597- PITr, MARK M. "Smuggling and Price Disparity," J. 685. Int. Econ., Nov. 1981, 11(4), pp. 447-58. . "Public Policy and the Economics of Devel- RADIAN, ALEX. Resource mobilizationin poor coun- opment," Europ. Econ. Rev., Apr. 1991a, 35(2- tries. New Brunswick, NJ: Transaction Books, 3), pp. 241-71. 1980. . "The Determinants of Growth," Econ. J., ROBINSON, SHERMAN. "Multisectoral Models, in Jan. 1991b, 101(404), pp. 122-32. CHENERY AND SRINIVASAN, eds. 1989. . "From the Static to the Dynamic: Some

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions 826 Journal of Economic Literature, Vol. XXXI (June 1993)~~

Problems in the Theory of Taxation,"J. Public on the Level of Taxation(and on the Fiscal Balance) Econ., Mar. 1992, 47(2), pp. 273-97. in Developing Countries." IMF Working Paper SUMMERS, ROBERT AND HESTON, ALAN. "A New Set WP/88/95, Washington, DC, 1988. of International Comparisons of Real Product and . Fiscal policy in open developing economies. Price Levels Estimates for 130 Countries, 1950- Washington, DC: IMF, 1990. 1985," Rev. Income Wealth, Mar. 1988, 34(1), pp. . Public finance in developing countries. Al- 1-25. dershot: Edward Elgar, 1991a. TABELLINI,GUIDO. "InternationalTax Comparisons . "Structural Factors and Tax Revenue in De- Reconsidered." Mimeo. Washington, DC: IMF, veloping Countries: A Decade of Evidence." 1985. Mimeo, International Monetary Fund, Washing- TAIT, ALANA. Value added tax: International prac- ton, DC, 1991b. tice and problems. Washington, DC: IMF, TANZI, VITO AND CASANEGRA DE JANTSCHER, MILKO. 1988. "Presumptive Income Taxation: Administrative, TAIT,ALAN A.; GRATrZ,WILFRID L. M. AND EICHEN- Efficiency and Equity Aspects." Fiscal Affairs De- GREEN, BARRY J. "International Comparisons of partment, International Monetary Fund, Washing- Taxation for Selected Developing Countries, 1972- ton, DC, 1989. 76," Int. Monet. Fund StaffPap., Mar. 1979, 26(1), THOMAS,JAMES J. Informal economic activity. Hert- pp. 123-56. fordshire: Harvester Wheatsheaf, 1992. TANZI,VITO. "Inflation, Lags in Collection, and the VICKERS, JOHN A. AND YARROW, GEORGE. Privatiza- Real Value of Tax Revenue," Int. Monet. Fund tion: An economic analysis. Cambridge: MIT Staff Pap., Mar. 1977, 24(1), pp. 154-67. Press, 1988. . "Inflation, Real Tax Revenue, and the Case WALD, HASKELL P. Taxationof agricultural land in for Inflationary Finance: Theory with Application underdevelopedeconomies. Cambridge, MA: Har- to Argentina,"Int. Monet. Fund Staff Pap., 1978, vard U. Press, 1959. 25(3), pp. 417-51. VAN WIJNBERGEN, SWEDER. "External Debt, Inflation . Inflation and the personal income tax. Cam- and the Public Sector: Towards Fiscal Policy for bridge: Cambridge U. Press, 1980. Sustainable Growth," World Bank Econ. Rev., . "Taxation in Sub-Saharan Africa: A Statistical Sept. 1989, 3(3), pp. 297-320. Evaluation," IMF Occasional Paper, 1981, 8, pp. WORLD BANK. World developmentreport 1988. NY: 43-73. Oxford U. Press, 1988. . "Quantitative Characteristics of the Tax Sys- . World development report 1989. NY: Oxford tems of Developing Countries," in NEWBERY AND U. Press, 1989. STERN,eds. 1987. . Lessons of tax reform. Washington, DC: . "The Impact of Macroeconomic Variables World Bank, 1991.

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions =~~~~~~~~~~8. 8.?OOd?gg?gOcD ? g _ O Pg~~

~~? v Q Sc ~~~~~~~~~o0 -- x N N - oo~~

~~o ~ ~ ~ ~ ~ ~ ~ ~ c-o~~

~~8 - t 8 cos X X g t-t <> coO e0 t S b t X N X g x O t <- t =; .?~~~~~~~0 o eq o o q t- 0 "i 0 o 0 o to 0 0~~

~~A X; = Q 80 80~~~~~~~~iq 80 ?O0 ?O 80 880 80 80 80 ?O 80 80 0 8 0 0 8080~~

~~.t E o = $ o~~~~~~~~10e t "i N CDt N o 8"i o CD oi CD t- N x e~~

~~_d CD - m co m O CD cO CD co 10 -- 0 0"q c~~

~~0~~~~~~~~~ co 10O C.0 O csb C.0O O co Ol O CO CO t- o ~~~~0 ~O ~ O~O~ d~ O IC O~Oi O O~O~ O OOO~~

~~0 Cd q, -m > cod d c 00 d t- oq 0 d0c 00 Xo t- t- O o co 00~~

~~X~~~~~~~~~~~~~~~Cuz0X0eCadC C DG F~~~~~~~~~~~~~~~~~~~~~s0 *i oo.eO ci cqd6 d. d6 cs csi csi d4 5t6 ood6 co~~

= . ^ ' ^ n Q ~~~~~~~~~~~~~~~~~~~O CsO t- O0 CO n OOn W E~~~~~~~~~~~~~~~~~~~~C00 0 co 10 t- -4 O Ol 10 CO m t- cl 10 C nOX O~~~~~~~~~~~~~~~~~~~~~~O4 O4 4 16 6 Oic- C4 N_; c; -4 c;sb C) C4 oi X A~~~~~~~~~~~~~0

Q ,. _ N e t t N O X s o O b O Q b s e o n n z~~~~~~~~~~~~~~~~~~~~~~~~~~.= 11 O 0 C) oM 0 0s t-s CO (M c 4 o. (M (M C) Cn o X~~~~~~~~~~~~~~~~~~~~~~~~~~r-0 c-iF o6 c- ; 4 i6 t- c4 6 -, 4 i6 6 o- 4 ) m C4 4

~~<~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~OQ -- C) t-1 l liC1- 0 0 C~~

~~Q~~~~~~~~~~~~~~1lZ 10 CnO CO t- 0 t- s bln o o6 t-- t-n t-- o6 1 Co F~~~~~~~~~~~~~~~~~~~~oU o ct c c s CO CO CO CO~~

~~=?~~~~~~~~~~~~~~~~~~~~CCO CO -- CbO o1 1 co CO o o _~~~~~~~~~~~~~~~~~~~~0t- t- 00 00 t- 0 00 t- 0 t- t- t- t- t-o O~~

~~Y o 0 CO 10 l lZ lZ 10 cZ Z 10 cZ 10 10 10 lZ c~~

~~odt- o o o o o co o on o on p 0 0 0 0 C) t- 4t oo - 0- -0 O O _m v )C n t- c s to ut 05 n Co~~

~~C' 00 CD b 00 0b0 0b0 0 CZb~~

~~v Y b O~~~~~~~~-O 0 00 Oq 00 00 0 0000 r.~~

~~04 E z u z~~

This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions X 8 _ t o 000o x N > 8 'C O t 'O 0 0 00 000 0 O 0 0g 00 00 0 ~ COC0t 00o o 0 o co co o 0 0o 0 Co C o oo- oC0 00"-0000 000

~~0 6~~~~~~~~~&6 6~~~~0"tI 6 0 t 0)= COO 00 O00 C 000 X oo C- s CO o00 -0 - o ? s - t~~

~~O) 0000 CO-O00 COCO OCO OO O O OOO -O O-0 000LO 0 O0 0 0 00 0 0CO O~~

~~XQ; z o ? ~~~otCO ?mIC- TO"It an2 8 8 8 8 8 m,? 8 t- 48 8 o t Q t =*~o C o 6 6 c- -c o o) ) co oi 6 6 6 c o o C5 o o) o o o o 6 o 6 o o o 6 CZ 2) O 8CO o88 O 80t 0 8CO 00o t0 8 8 0 00--o 8 8 "I88t'~~

~~4O o4 o o sooooooo o o o o6 o o o o o~~

~~.t c;~~~~~~~~C 8, o- C)8 108 8. 8 8. 8 8 8? -tt to - D ? oN ms m N t- N t- o- o oc - 0~~

~~10 o o o o o o- o c =~~~~~~~~~~~~~~~~L o oC 0 0 0 0 0 0 x t-d O 00 N 0 -o0 C0N0 C - 4 0 o C900~~

~~oo e0 C0 0 0 " 00 0 x 0 _N > 00m m_ o0 _ Qm - xO m 10 " 1C _O j M~~

~~0 tvt~~~ I- 0 C) a: o 00 8 0 0 m 0CoCC 0 oCt- C-o N00 C - 0 t- C c?t- O xCO t-~~

~~O >~~

~~CO OOCOO v c I - (3 80-ox C- - X - t N m 00 m C O CO "- 0 0 0 0 - OCO C. C_ o o- cq v O0 _ c 00 O0 O t cC,) _ q Cos v sr _ v C)o0 o to v _ o= -_ -4_~~

~~s~ ~ ~~~~~~o4 ~~~~~~~~ 8 C5C6 4 C5-4 c- 16 c0t> 4 8- -4 C4-. 4 C4c- ?,0 8C' 4 V C8-- (:: C5 C- t -4 ? 0) H o oC C O C O0C - CID00 o 0 10 _~~

t- oc 0 000 C- 0 LO oov"C v o00 N 00? I 0C-O) m1- - 0 In00 N '0 oo0 - Co 0 "0 C t- v 0s COX0 0C,o v * 0C v _ 00 v 0 000 0 v 00 COt_C t- N In v 0O t- 0 C, t _, ~ 000nnto000000006obN_ oooo&66n_ oi_oo Xo 8 8 m~~~~~~~~~~~~~~~ t. 4 No~ t 6 N t. o~as v 4 6 s 86 --q3 C' 4 t0 o - 3lo t? 6 N tgYg888sXto o o4

H 0C-00 00CO000-4* " 00 ~00 0 - C * COO00 CO 000 COC-c _ '" O0 000* 4 0 0 0 > 04 x ?f x c et88'c cq v >cm >0sc nc 0 8:4. 888g 0 0 0 - o '0" 00 0 t 0 C - 00 0 0 '-4-40 H> I- to v CID -Cc v C I-' to v-00-0

0) 10 V O I- --4 ~~ 0) OCO~~~~~~~~000 0M CO000 0-oC0 500t C0 00 0 C-00M C O00001mt ts C0 Cb 1,1 uCt OX OX t0 -I:bC? _0 t- o oM 0 Cs -4 -4 --4 ~ ~ ~ ~ 00 C 0 - - C > Z 0 0 CO X, . 00Z0 00 CO CO 0"CO 0 0 v000 v 0 0000t 0 0 cO > W I oq u: ? 0 ut O G ~ W)O N0 09 v C_00 CO "C 0ci O0 0 0 0v

H ~ ~ ~ ~ ~ ~ ~ ~~~~00~~VC- C) C- 0CM ")t VInO0- CO )O 0COC-0 --0 OC- C,,C-v( C-0 -t-0 0 CO 0) CO C0000"005-VO 0 - 6 -- 6 6 q<:SOi1500C-~0"-4"-4 000 o"- -() 5cqcil 00 00 0 C~ o0 CID00 0 i CID000 0. 0-4 CO O OCOOO 4 -4 0000 -6- 4- . 4-4 - q C)-4

~~0 0l On 0 l 0 0 Ci 0! M 0~ C, Vu 0~ 0i CI 0a m0c~~

~~co -0 0 00 00 0000 00 0000000 CO 00 CO 00 c 0 d0 0)O) " 4-00xxx 0 C-I-- 0~~

~~0 C-CO000CO00M C)VC-00 --4000CO00NC-1O C0000C 000 v0000 00 CC xo(C CO0 " - 0 0 -4 - "C 0 c -000-700to H - - ~~~~~~~~~~~~~~~~-4- - - 4 - 4 - 4 4 4------~~

~~CO~~~~~~~~~~C O 0 CO 0 L00 )0 0 0 000>00 0 0 0 00 ~ 0 O O O OooUoO O O ooo 00 00 000 00 00~~

~~0 ^ 00 * '0"000C) " - 00000 CZ = D0 0 E >C0 COOO) = 0 000000._ 0 (U~~

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions '.14 -14 O O cl) cq O 01 - 0 ci C5 6 cl c; -4 c c; c 6 6 6 6 6 ci~~

~~in. cl)~~

~~4.1~~

~~ll Cli~~

~~co co m 00 co cq 0 4.1~~

~~t- co t- 00 t- t- ol M t- CO M~~

~~004 cccq cocco cqc~~

~~fc 00 10 00 t- t- co t- c 00 10 cq 0 13 o! O Oi 4.1 0 co cq 10 c co 10 01 c co 114 t- 0 44 10 0 0 10~~

~~t- -4 10 t- t- 01 m 10 m c 10 M 00 Ic -4 Cd 10 t- co m 10 O cq m cq co t- 10 m 10 - cq 10 cq co co 4 C6 t- 10 6 C6 16 cq Ic 4 4 r. _4 16 -4 co 6~~

~~t- __4 00 t- 00 00 co cq 0 m m 01 co o co Ic t- m cq cq t- co 0 co co 0 co co co -4~~

~~t- 00 0 00 co -4 0 c Ic 10 t- 00 co oo m 01 t- 00 00 co fc cq m cq -4 co m -4 oi 0 C6 -,v c C6 co -4 cq cq 0 C5 c~~

~~C.0 cq -4 00 co cq cq 00 cq 10 0 co __4co 00 cocq 8 m I'll 0 0 10 Cd c cq cq 0 co C4 0 -q lo 0 C5 0 to co 0 0 oi C5 16~~

~~00 t- cq 1114 .,v 00 Cd t- fc t- co 00 co 0 fc 0 c co co 10 10 Ic Ic oo t- t- t- ,,:V M 4.1 co 0 cq 10 m to cq z cl) 00 t- C) M (:: lp lr (:: ll O O O ci O O O 0 0 O O~~

~~4.1 00 O c 00 00 10 Cd rl r17 lp lr ci Cl! cl) cl) O 00~~

~~0 cd c O cl) O 00 00 cli 10 O m 00 = m m t- m cl) 10 ci c (:: rl lp _! 4 75! c c; 16 C6 cq 6 c; t- 01 c) 10 0 10 O >~~

~~-4 00 00 t- Ic __4 Ic cli 00 ol C) t- Cd cli cl) 10 __l "'r O N 10 lp N X 0 O __l lll __l cli cl) cli cl) m 10 cl) 01 cl) 0 t- m t- cli~~

~~C's CD t- in m CD CD N N t- N m (M in (M 00 O CD 00 C-1 O (M in 0 C.0 C.0 C.0oo in oo in C'l -4 -4 -4 co 01 01 01 01 01 01 -4 -4 -4 C'I~~

~~O 00 00 t- C.0 C.0 t- C.0 t- t- t- 00 Co C.0 "!:v c CD 00 00 00~~

~~00 C.0 in C.0 "!:v in "!:v in in in C.0 ml "!:v C-1 "!:v C.0 C.0 t- (M~~

~~CZ t- cl) in 0 .6.1 "!:v 00 cl) 01 C.0 C-1 in O 00 in t- 00 c) oo - m 00 __4 -4 00 in "!:v C-1 O t-oo C-1 (m in00 in cl) 00 00 00 t- C.0 C.0 in in in in in "!:v "!:v "!:v "!:v CZ VD- U~~

~~CZ 0 Q 0 CZ CZ CZ > C) C'S V) (1) C's CZ z (3) CZ> = .$. _Q-m m m 'Q O - = 0 CZ K cuta C's.5 (1) CZ CZ U - -0 CZ U -M - (3) 0 > U) 1 (3) CZS- C, z z O a (3) (3) 0 0 0 = Q C13 O' E E = E =O -0 Oc'!~~

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions ;T4~~

~~X 0 0 in v N O 0 C> 0 0 0 N Cl) 0 00 Cl) 0 0 v~~

~~C-) C.') -4 m 1-4 m v m to 1-4 00 m N co 00 N -0 (O 00 01 (D C-1 0 C) C-1 (D -4 1-4 --4 t- - m N 0 to N v blo (1) o C%~~

~~NV (OD OM OMO CNI tN- 8 (IC-01 -C1, It z 6 C5 C; 6 6 6 6 6 6 1-4 6 6 6 6 6 6 6 6 _; 6 6 6 6 6~~

~~0 co c 0 t- C-1 t(=-> (1100 1110, C-"> "MO -V-l (18 231 C5 ci o o o O 0 o 0 o 6 6 ci o o C5 O c; C5 O o C; ci~~

~~C-1 00 00 10 -4 C,) v 00V C,) 10 -4 m -4 o 1-4 10 v 00 10 0110 0- C)0C) v 0 to - C cq- C.0 C--4 (M C-1 00 ce)-co 00 in, C6 6 ci66 6 6 c;6 c;6c; C.0 _; ce5 o o -4 C50000C's 14~~

CD v (M 9 00 00 eq (M C,) in mcli C,)vm ---4 OC) (7) vOC) cli C-1 C-1 o C tl-: 10 t- C,) to--4 OC) (M 10 0c t- 00 -4 C,) t- 00 ml to OC) (M 0 A) C-1 t- t- -4 cli 0) C,) 10 C,5 6 C C6 C6 -4 _; ei -4 -v C,5 oi c C 01 >

t- C,) V - M N c I- C.0 t- C lp (M t- 00Vt-M Mt- CD NN--4 00 N C.0MMt-Xr- -4 t- MM-4N0C) 10 c,5 o6 oi o6 16 C4 -4 4 C13 C 16 Ci Cli C64 cli 1-4 -4 -4 0 MV t- -4 01 t- -4 V00 ,C VD (C> C-1 C) (M 10VCII t- t- C.0 tO (O -4 -4 108V00 t- -4 -4 --4 04 -4 C.0 10001 IV (=> (O CII C-1 'Aw 666C66 0 666 666 6 6 66666666

M NNt- t- 00 MMM-4 M 100010 t- t- 10 10 v(M CD CD CD vvoo0 C,) oo 00 .c-010 0 C4 Co66ci 4 lo <6 6C,5 oi <6 0-40<6 _; <6 c-i 0 0) co ol xN mm t- 00 v CD 'IV 00 mN t-m CZ 10 -4 ;T. 0 00 t- 100 -4 04- (=> 00 (M 10 oo 0 > > 0 6ci ci 164 6C,5 6616 oi w CL4

~~-4 '.4M C') -i tl,7 (01 oq aq (M t- - V o! (lq NIq CiNtIl: C 10V,I! M-4M0VM1- U-i 16 Oi NM00V MV-4NNDMV10 cq 4 ct~~

~~z > M --I (10 (10M -4 10 CII CII Cj C 0-4 cli -4 10 -4 C-110 C5 C5 0C56ci0C; i gi0(O0 0c; c; c; C5 c; <6 (o c) C5 (M C) 00~~

~~-4 (M v10m t- 0 0) 10 -4 10vvt-xt-m00m C? C-1 C') Cl t- C'I t- 10 I.. C C- 0-.4 OC) p C')0 (3 oq -4 (M --4 -4 C.0 C) t.: 6 o-44 N0 6646C'34 WCZ~~

~~00 ci In 0) ci ci 'i C112 6C15 C56 C4 C5 64 -40-4 CiNC5 C) > C)~~

-4 MMMM M C.0 X V VVCD MM-4 V'"4 t- -4V00 110 C,) vv10 v (M (M 0 v r- r- C,) r- 00 in IcoSCO) 00v0) oi4-44C.6 616 Ci 6C 6 644Cli Ci4t-Z6 6 4 cd > bC CZ cd C")V-4 t- t-Mt- MVMN-40NM 0MM M tl,7 :: C Vt- 14 o66 6 i6 t--.: vi 4M t-.: C4 6 ci t---: t--: r-: ci t-.: -C$ 00 ,,i gi 6 16 -4 -4 J--4 -4 -4M --4 -4 -4 -4 -4 -4 -4 -4 --4 -4 CII C) m

bo 'n co t- t- t- zt- 00 t- 00 t- (1) oo 00 00 00 00 OC)00 00 00 E 05 -4 -4 -4 -4 -4 1-4 1-4 -4 -4 -4 -4 -4 --I -4 -4 -4 -4 -4 -4 -4 --4 -4 -4 -4 > Cd t4-4 C) inm-41 ":r inm910 "!:v to 't in v(M c LI) C,) "iq ":vVc to o in In 4., (L) 00 00 00 00 00 00 00 OC) 00 &O oc 00 1- &O 00 OC) 00 oc 00 OC) OC) 00 00 00> (M (M (M (M (M 0) m (M (M (M (M 0) 0) 0) 0) 0) 0) 0) (M cm (Mm (3) 4-4 -4 -4 -4 -4 -4 -4 -4 -4 -4 -4 -4 -4 --( -4 -4 --I -4 -4 -4 -4 -4 -4 -4 0-4 C14 CZ -4 00 10 10 -4MNV -4 t- -4 CIIO A4 oo :$ C -4 -4 11-0 ICOOtc--l I- 10 C,) ce) -4 -4 -4 z 4 co C) co C,)C,) CO) ce) C,) C,) C-1 C-1 Cj

~~0 C'sX 4- mCZ CZ CZ C's > CZ 'm CZ rW 0 Cd 4-i CZ 0 Cd Cl 4-4 zO a) bf,, 0CZ z Z75 D =N Uco Zo NM cn~~

~~This content downloaded from 158.143.41.7 on Mon, 9 Sep 2013 09:19:07 AM All use subject to JSTOR Terms and Conditions~~