Distributional Effects of Optimal Commodity Taxes with Minimum Income Programs: micro-simulations for Brazil

Transcription

1 Dstrbutonal Effects of Optmal Commodty Taxes wth Mnmum Income Programs: mcro-smulatons for Brazl Ana Luza N.H. Barbosa*, Eduardo P.S. Fuza**, Marcel Scharth**, Sek Asano*** Keywords: Almost Ideal Demand System, Equvalent Income, Optmal commodty taxaton, Socal Welfare Functon, Lump sum subsdy. JEL: H21, H23, H31, D12, D31, D63, C33. ÁREA 6 Economa do Trabalho, Economa Socal e Demografa Abstract Commodty taxes play an mportant role n Brazl and rase around 60% of the total tax revenue. Ths heavy relance renders commodty taxaton one of the man tools avalable to the government for rasng revenue and securng redstrbuton. In fact, Brazlan ncome nequty s one of the hghest n the world: the wealthest 1% of populaton, equvalent to 1.6 mllon people, earn together as much as the 50% poorest, around 80 mllon. The purpose of ths paper s a partal equlbrum numercal smulaton of the dstrbutonal effects of optmal commodty taxaton combned wth mnmum ncome transfers made by the government to households. The approach used to measure households welfare s a money metrc ndrect utlty or equvalent ncome` [Kng (1983)], that underles the Almost Ideal Demand System parameters. We plug t nto the equvalent varaton measure to evaluate the equty effects specfed n terms of the equvalent ncome. The data source s a 1995/1996 natonal household expendture survey, though estmated parameters come from a sample comprsng a 1987/88 wave as well. We fnd that our proposed mnmum ncome programs combned wth selectveness n commodty tax structure would be useful as redstrbuton ncome nstrument among households n Brazl. These results can provde some valuable contrbuton n the context of the ncreasng dscusson about mnmum ncome programs n Brazl assocated wth demographc characterstcs such as educaton and famly sze. * Dretora de Estudos Macroeconômcos, Insttuto de Pesqusa Econômca Aplcada - IPEA, Avenda Pres. Antôno Carlos, 51 / 15 º andar, Ro de Janero - RJ , Brazl. ** Dretora de Estudos Macroeconômcos, Insttuto de Pesqusa Econômca Aplcada IPEA and Pontfíca Unversdade Católca PUC, Ro de Janero, Brazl. *** Department of Economcs, Tokyo Metropoltan Unversty, 1-1 Mnam-Osawa, Hachoj, Tokyo, Japan e-mal: Barbosa: aluza@pea.gov.br, Fuza: fuza@pea.gov.br, Scharth: marcelsf@pea.gov.br, Asano: asano-sek@c.metro-u.ac.jp.

2 1. Introducton The Brazlan tax system as a whole s extremely complex and generates producton and consumpton dstortons. Commodty taxes play an mportant role n the country and rase around 60% of the total tax revenue. Ths heavy relance renders commodty taxaton a subject of consderable polcy mportance and one of the man tools avalable to the government for rasng revenue and securng redstrbuton. In fact, Brazlan ncome nequty s one of the hghest n the world: 1% of the wealthest, equvalent to 1.6 mllon people, earn together as much as the 50% poorest, around 80 mllon. Moreover, absolute poverty reaches around one thrd of the populaton wth a standard of lvng under basc necesstes. Therefore, an analyss of optmal commodty taxes combned wth ncome transfers n Brazl and ts mpact n ncome dstrbuton s of utmost mportance. There has been an ncreasng debate on the use of mnmum ncome programs as a consstent socal polcy measure to cope wth ths problem. In partcular, the Bolsa-Escola s a good example of a cash transfer program targeted to low-ncome famles wth chldren. Ths program, where elgblty s condtoned to school frequency by the dependents at school age, has shown good results both n terms of focus and coverage n the ctes of Brasla, Campnas, Belem and Belo Horzonte. As from Aprl of 2001, the federal government has been mplementng Bolsa-Escola, coverng nearly 98% of the 5,561 muncpaltes n the country. It should be noted that, besdes Bolsa-Escola, there are other cash transfer programs n Brazl that vares from the elgblty crtera to the target populaton. Some of them are the followng: ) Pensons, for whch enttlement s based on contrbutons made to the socal securty system; ) Unemployment Beneft, as long as the worker has not resgned nor been lad off for far reasons; ) a famly allowance pad for all chldren less than 14 years old or dsabled of any age to employees and temporary workers who earn less than near two mnmum wages; v) Bolsa Almentação whch s a transfer targeted to pregnant women and to chldren aged from 6 months to 6 years. 1 Optmal commodty tax systems elct the conflct between effcency and equty concerns n the desgn of commodty taxes that rase requred revenue and obtan gven dstrbuton objectves, at the lowest cost n terms of effcency. Optmal tax models can be solved under alternatve assumptons regardng: the government s concern wth nequalty; household preferences; requred government revenue level; and constrants on ts ablty to tax. The purpose of ths paper s to examne the dstrbutonal effects of optmal commodty taxaton combned wth mnmum ncome programs n Brazl. Our analyss s restrcted to a partal equlbrum treatment and focuses on the effects of alternatve optmal commodty tax systems, combned wth ncome transfers, on the economc welfare of dfferent ncome classes of household. We extend our prevous work (see Asano, Barbosa and Fuza [forthcomng]) by allowng for an ncome transfer to be made by the government accordng to households demographc characterstcs such as the number of chldren at school age n lower ncome households (followng Bolsa-Escola features). Our benchmark s a more generalzed program n whch the government combnes an optmal commodty taxaton system and a unform lump sum ncome transfer made only to the lower ncome households. The approach used to measure household s welfare s a money metrc ndrect utlty, n some contexts also referred to as equvalent ncome and ntroduced by Kng (1983). Welfare 1 Most of these programs stll reman after the new Federal government was establshed n offce n January Recent announcements released by the press ndcate that there wll be some changes towards unfyng the varous cash transfer programs, but the debate has not settled yet. 2

3 effects smulatons are based on the equvalent varaton concept specfed n terms of the equvalent ncome that underles the Almost Ideal Demand System parameters (Almost Ideal Demand System, by Deaton and Muellbauer, 1980). The data source s a 1995/1996 natonal household expendture survey, though estmated parameters come from a sample comprsng a 1987/88 wave as well. The analyss of the extent to whch dstrbutonal goals can be reached n Brazl through commodty taxes, wthn an optmal taxaton framework, has already been consdered n Squera (1997). Her model employed the equvalent measure of consumer surplus to estmate the effects of alternatve tax structures on the welfare of households. The tax structures vared from a system wth two rates of value added tax (VAT), n addton to zero rate on food, combned wth excse dutes on alcoholc beverages and tobacco to a tax system wth a proportonal value added tax on all goods. She also consdered the case that n addton to an optmal commodty tax structure the government made a unform lump sum payment to all households. Squera s results ndcated that a tax system based on two or three rates of VAT, plus some food subsdes and/or drect ncome support for certan household groups, and supplemented by excse on luxury goods, could effectvely mprove socal welfare and advance the objectve of greater equty. The household demand pattern n Squera s model was a Lnear Expendture System (LES) and the data source was ENDEF (Estudo Naconal de Despesas Famlares), a comprehensve survey undertaken from August 1974 to August 1975 n all metropoltan and urban areas, and rural areas n the Southern, Southeastern and Northeastern regons. Sah (1983) consdered the use of commodty taxaton and subsdes n order to mprove the welfare of the worst-off ndvdual, based on U.K. data. He obtaned an upper lmt n terms of the maxmum budget share of the worst-off as a rato of the mnmum average share n the economy. The results pont out to nadequacy of commodty taxaton as a redstrbutve nstrument. Nevertheless, n the Indan case, Majumder (1988), contrary to Ray (1986), found out that the possblty of commodty taxes actng as a major source of redstrbuton cannot be ruled out. Creedy (1999) presented an emprcal analyss of the welfare effects of several ndrect tax reforms n Australa. Emphass was placed on the mplcatons for selected household types. When lookng at all households combned, the results suggested that redstrbuton through possble tax reforms s small. Creedy argues that these results conceal dfferences between dfferent household types. Comparsons among several types of households show that the largest welfare losses are experenced by low total expendture couples wth one chld whle the smallest losses accrue to hgh total expendture couples wth two chldren and low total expendture sngle person retred households. 2 The paper s organzed as follows: n the next secton we ntroduce the approach to measure the welfare mpacts of the alternatve tax systems. We also present the demand specfcaton of the model wth the emprcal results regardng mean shares and elastctes from the AIDS estmaton. Secton III presents optmal commodty tax structure when the only tax polcy nstrument avalable to the government s commodty taxaton. It also shows optmal commodty tax structures combned wth dfferent ncome transfer schemes and presents our methodology to smulate the welfare mpacts on households under these programs. Secton IV descrbes the data used for the smulatons. The results regardng the mpact on household welfare are dscussed n secton V and we sum up wth the concludng remarks of secton VI. 2 Although some of these studes do use an optmal commodty tax framework, none of them appled the same methodology as ours. 3

4 2. The Model 2.1 Welfare Changes Measure The method of measurng the welfare and dstrbutonal effects on dfferent households of alternatve commodty tax reforms uses the Hcksan concept of the equvalent varaton. Ths s defned usng the expendture functon, E(p,U), whch represents the mnmum amount of ncome requred to acheve a gven utlty level U at prces p. The equvalent varaton, EV, s: EV = E(p 0, U 1 ) - E(p 0,U 0 ) (2.1) We assume that there are H households n the economy ndexed by h. Savngs s not ncluded n the model, so ncome and total expendture are treated as synonymous. In the prereform poston household h faces a vector of prces p 0. The term E(p 0,U 0 ) s the total expendture before the prce change, denoted by y 0. After the reform the household faces a new prce vector p 1. Suppose that the prce vector changes from p 0 to p 1, and that U 1 represents the post-change utlty. The equvalent varaton s, therefore, the amount the household would be wllng to pay, n the new stuaton, to avod the prce change. Followng Kng (1983), we use a money metrc measure known as equvalent ncome. Ths s defned as the value of ncome, y e, that at some reference set of prces, p r, gves the same utlty functon as the actual ncome level. In terms of the ndrect utlty functon, y e s therefore defned by V(p r, y e ) = V(p, y). If we use the expendture functon we have y e = E(p r, V(p, y)). When pre-change prces, p 0, are used as reference prces p r the equvalent ncome s gven by E(p 0,U 1 ). 2.2 Demand Specfcaton We assume that the preferences and household demand patterns are based on the Almost Ideal Demand System AIDS, proposed by Deaton & Muellbauer (1980). Ths system has desrable propertes and provdes a flexble approxmaton to the consumer preference structure. The AIDS expendture functon s gven by: β 0 log E(U, p) = log a (p) + U p β (2.2) where U s the utlty ndex, and: ( ) p p p 1 * log a p = α 0+ α log + 2 γ j log.log j j (2.3) The lnear homogenety of the expendture functon wth respect to the prce vector requres the followng constrants: α = 1, γ = γ = β = 0 (2.4) * * j j j By applyng Shephard s lemma to (2.2), we obtan the share equatons: 4

5 y = 1,.., n. (2.5) h h w = α + γjlog p j + β log ( /P) j where w h s the expendture share of good for ndvdual h; p j s the prce of good j (j=1,..,n); y h s total expendture. Hereafter, we drop superscrpt h, for the sake of smplcty n exposton. The prce ndex s a non-lnear prce functon represented by P: log P= log a( p) (2.6) Under (2.4), addng up constrants and homogenety of the demand functons, correspondng to (2.5), are all satsfed. The expendture functon n (2.2) represents the mnmal amount of ncome necessary to acheve a gven level of utlty U at prces p. The parameter α 0 can be nterpreted as the subsstence expendture when all prces are normalzed to one. The model defned by equatons (2.5) and (2.6) s the Almost Ideal Demand System - AIDS (Deaton and Muellbauer [1980]). Also, the AIDS expendture elastctes are gven by: η = 1 + β / w (2.7) It follows that f β s negatve the th group s a necessty, and f β s postve t s a luxury. Invertng the expendture functon, we obtan the AIDS ndrect utlty functon: log( Y) log a( p) v( p, Y) = β 0 p β (2.8) The value of the AIDS ndrect utlty functon les between 0 and 1, and ts monotonc transformaton can be used as a welfare measure. The AIDS parameters, used n our ndrect utlty functon, were estmated by Asano and Fuza (2001). The model ncorporates demographc varables nto the share equatons, n the followng form: = 1,..,n. (2.9) w = α + γ log p + β log( Y / P ) + ω Z h h h j j j k k k where Z k s (k = 1,..,K) are demographc varables, such as famly sze, educaton of the household heads, etc. Underlyng to ths extenson s an adaptaton of the subsstence level to ncorporate demographc varables: h 1 h log a ( p) = α0 + α log p + γ log p log p ω Z log p + j j k 2 j j jk k j (2.10) 5

6 The AIDS form of equvalent ncome s gven by: 1 log y = α + α log p + γ log p log p + ω Z log p 2 h e 0 r j j r rj j k jk k r j β pr 1 h + β0 log y α0 α log p + γj log p log pj + ωjkzk log p j j k j p Shares and Elastctes (2.11) The analyss of the mpact of dfferent tax systems on household s welfare requres consderable nformaton about the preferences and demand patterns of households. Ths secton provdes some useful nformaton about the mean budget shares and the (expendture) elastctes for sample used n the AIDS estmaton. 3 Table I presents the mean expendture and own prce elastctes for Total expendture elastctes (a proxy for ncome elastctes) ndcate that food and housng are the only necesstes n Brazlan utlty functons, whereas furnshngs, clothng, transportaton, health care and personal expenses are found to be luxures. The own-prce elastctes are found to be sgnfcantly negatve. Among them, those for food, housng and furnshngs are sgnfcantly less than one (own-prce nelastc), whle clothng, transportaton and communcaton, health care and personal expenses dsplay own-prce elastctes around 1. 3 See secton 4 for more nformaton about the data sources used n AIDS estmaton and n welfare smulatons. 6

7 Table I Mean Expendture and Prce Elastctes, 1996 Expd. Food Hous Furn Clth Tran Hlth Pers Exp Shares Elastcty (s.e.) ( 0.030) ( 0.043) ( 0.104) ( 0.122) ( 0.069) ( 0.068) ( 0.055) Prce Elastctes Food Hous Furn Clth Tran Hlth Pers Exp Food (t-value) ( ) ( 0.723) ( 0.019) ( 0.322) Hous (t-value) ( ) ( 0.877) Furn (t-value) ( 0.726) ( ) ( 0.294) ( 0.335) ( 0.783) Clth (t-value) ( 0.019) ( 0.293) ( ) ( ) Tran (t-value) ( 0.336) ( ) Hlth (t-value) ( 0.343) ( 0.830) ( ) ( ) Pers Exp (t-value) ( 0.872) ( ) Source: Asano e Fuza (2001). 3. Optmal Commodty Tax Systems and Income Transfers Optmal tax models feature the maxmzaton of a socal welfare functon, subject to a balanced government budget requrement. The trade-off between equty and effcency s taken nto account by ntroducng the government s averson to nequalty nto the socal welfare functon. Ths secton presents alternatve tax structures based on optmal commodty taxes and unform ncome transfers calculated n Asano, Barbosa and Fuza [forthcomng]. We present an extenson of these optmal commodty tax structures n whch we allow for a per capta payment to be made by the government only to the lower ncome households. The case whch optmal commodty taxaton s combned wth a mnmum ncome program drected to lowncome households wth chldren of age from 6 to 15 enrolled n publc elementary schools ( Bolsa-Escola ) s also presented. 3.1 Optmal Tax Rates and Unform Transfers Table II presents three cases of optmal commodty tax structures for seven groups of commodtes. They are: 1.FOOD; 2.HOUS (Housng); 3.FURN (furnshngs); 4.CLOTH (clothng); 7

8 5.TRANS (transportaton and communcaton); 6.HLTH (health and personal care) and 7.PERS_EXP (personal expenses, educaton and readng). The tax rates and lump sum subsdes are presented for two dfferent levels of nequalty averson (ε = 0.25 and ε = 2.00). Ths approach explctly allows for the ntroducton of alternatve value judgments from the government. Needless to say that for hgh levels of ε (n our case when ε = 2.00), the government has a stronger commtment to equty. We assume that the government revenue corresponds to 10% of the consumer s total expendture. Table II OPTIMAL TAX RATES (%) AND UNIFORM LUMP SUM TRANSFERS COMMODITY CASE I CASE II CASE III GROUP ε = 0.25 ε = 2.00 ε = 0.25 ε = 2.00 ε = 0.25 ε = FOOD 12.2% -29% % 301,77% 21.12% % 2. HOUSING 11.5% -10% % 215,85% 18.48% 1.43% 3. FURN 8.1% 27.3% 29.07% 20.62% 10.22% 22.04% 4. CLOTH 10.2% 35.9% 50.47% 73.47% 13.53% 34.29% 5. TRANS 11.4% 53.1% 57.83% 95.29% 15.16% 50.67% 6. HLTH 10.8% 36.8% 60.68% 96.74% 14.76% 36.71% 7.PERS_EXP 10.9% 53% 50.09% 78.99% 14.28% 50.14% LUMP SUM TRANSFER In R$ Sep 1996 Note: n Case III, lump sum transfers are constraned wth a bndng celng equvalent to 50% of the mnmum observed ncome n 1996 (source: Asano, Barbosa and Fuza [forthcomng ]). Case I presents tax rates based on the assumpton that the only tax polcy nstrument avalable to the government s consumpton goods and servces taxaton. When ε = 0.25, the optmal commodty tax structure shows a movement towards unformty. 4 However, for ε = 2.00, when redstrbuton objectves are hgher, there s a selectveness n tax rates. In partcular, food and housng groups are subsdzed and the tax rates for the other groups ncrease sgnfcantly. Case II reports tax rates for the stuaton when the government, n addton to commodty taxes, sets an optmal unform per-capta lump sum payment to all households. Ths payment works as a unversal beneft unform to all households. The man results for ths case are the strkngly hgh levels of commodty tax rates and optmal lump sum subsdes for all levels of nequalty averson parameters. In Case III we show optmal tax rates combned wth a constrant n the optmal lump sum transfer (obtaned n the last case). The constrant s equal to 50% of the mnmum observed ncome (we call t as bonus rate = 0.5). The results show hgher tax rates (and lower subsdes) than Case I and lower values of lumps sum transfers (than case II), whch remans stable for both values of nequalty averson ε = 0.25 and ε = When ε s near zero, there s no concern for nequalty. In ths stuaton, a unform rate of tax on all goods s equvalent to a tax on labor alone. Ths corresponds to the conventonal prescrpton f there s a completely nelastc factor, ths should bear all the tax (Atknson and Stgltz, 1972). Therefore, as Asano, Barbosa and Fuza [forthcomng] assumed that labor supply s completely nelastc, the optmal commodty tax rate structure s unform. 8

9 3.2 Transfers to the Lower Income Households We also extend our analyss by allowng for a per capta payment to be made by the government only to the lower ncome households. Ths ncome transfer s also a unversal beneft but, dfferent from case III, t s restrcted only to the poorer households. We assume that ths mnmum ncome polcy guarantees that each household gets 50% of the mnmum wage per capta. So f the household already earns that ncome or more, he gets no transfer. Ths exercse was based on the mnmum wage value of September 1996, whch was R$ per month. Table III OPTIMAL TAX RATES (%) AND TRANSFERS TO LOWER INCOME HOUSEHOLDS (CASE IV) COMMODITY ε = 0.25 ε = 2.00 GROUP 1. FOOD 11.01% % 2. HOUSING 10.57% -4.08% 3. FURN 7.73% 25.69% 4. CLOTH 9.65% 30.47% 5. TRANS 10.84% 35.40% 6. HLTH 10.25% 30.78% 7.PERS_EXP 10.39% 38.76% TRANSFERS (PER-CAPITA)* R$ R$ In R$ Sep 1996 *half of the mnmum wage per year. The commodty tax structure shown n Table III s qute smlar to Cases I and III presented n Table II. For low value of nequalty averson parameter (ε = 0.25), the commodty groups that are more prce nelastc, food and housng, hgh tax rates n comparson to the stuaton when there s a stronger commtment to redstrbuton objectves (ε = 2.00). In ths case, both tems are subsdzed. As regards the other commodtes, there s a sgnfcant ncrease n ther tax rates when the nequalty averson ncreases. One mportant feature to stress n the results presented n table III s that even when the mnmum ncome polcy s ntroduced, a subsdy for food and housng s needed to mprove welfare. Therefore, we conclude that ntroducton of transfers to lower households may stll gves some room to commodty taxaton as a redstrbuton nstrument. Ths subject s analyzed n secton Optmal Commodty Tax Rates combned wth Bolsa-Escola The Bolsa-Escola federal program was ntroduced n 2001 by the federal government. It came out as an unfoldng of a prevous federal mnmum ncome program Programa de Garanta de Renda Mínma (PGRM), whch lasted two years (1998/2000 perod). The target populaton of Bolsa-Escola Federal low-ncome households wth chldren of age from 6 to 15 enrolled n publc elementary schools. The elgblty crtera of Bolsa-Escola federal s the students mnmum school attendance of 85% and that households must have a per-capta ncome of maxmum 50% of mnmum wage. The ncome benefts are R$ per chld, wth a constrant to R$45.00 per household. 9

10 Followng the characterstcs of Bolsa-Escola federal, we also extend our analyss by allowng for a per capta payment to be made by the government only to the lower ncome households wth chldren of age from 6 to 15. The ncome transfer s equal to R$15.00 per chld and t s constraned to R$45.00 per household. Ths program s a more specfc mnmum ncome polcy than the one presented n secton 3.2 (case IV), n the sense that the latter has a larger target populaton than ths one. Optmal commodty tax structures n ths case are qute smlar to the ones obtaned n case IV presented n table III, for both values of nequalty averson parameter (ε = 0.25 and ε = 2.00). Table IV dsplays the optmal tax rates for the case n whch ncome benefts are assocated to Bolsa-Escola program (case V). Table IV OPTIMAL TAX RATES (%) AND BOLSA-ESCOLA (CASE V) COMMODITY ε = 0.25 ε = 2.00 GROUP 1. FOOD 12.26% % 2. HOUSING 11.56% -9.80% 3. FURN 8.06% 26.76% 4. CLOTH 10.15% 35.47% 5. TRANS 11.40% 52.65% 6. HLTH 10.83% 36.47% 7.PERS_EXP 10.90% 52.53% TRANSFERS (PER-CHILD)* R$ R$ In R$ Sep 1996 *ths value, R$145.20, represents the ncome beneft per year n 2001, R$ , adjusted for nflaton to September,1996. We choose to ncrease the beneft value of Bolsa-Escola n order to have a hgher dstrbutonal mpact. Therefore, we ncrease the ncome benefts regardng Bolsa-Escola Federal to R$ 22,50 and R$30.00 per chld, n R$ 2001 terms (1.5 and 2 of the federal value beneft, respectvely). Ths ncrease n the ncome transfer generated an optmal commodty tax rates structure qute smlar to the prevous Bolsa-Escola case presented n table IV. Optmal commodty tax rates for the hgher benefts values are presented n table V. Table V OPTIMAL TAX RATES (%) AND CONSTRAINED BOLSA-ESCOLA (CASE VI) COMMODITY ε = 0.25 ε = 2.00 ε = 0.25 ε = 2.00 GROUP 1.5 X FEDERAL VALUE 2 X FEDERAL VALUE 1. FOOD 12.30% % 12.32% % 2. HOUSING 11.58% -9.68% 11.59% -9.35% 3. FURN 8.05% 26.49% 8.44% 25.75% 4. CLOTH 10.15% 35.29% 10.14% 34.77% 5. TRANS 11.39% 52.39% 11.39% 51.64% 6. HLTH 10.83% 36.32% 10.82% 35.91% 7.PERS_EXP 10.89% 52.28% 10.88% 51.57% TRANSFERS (PER-CHILD)* R$ R$ R$ R$ In R$ Sep 1996 *all values represent the ncome beneft per year n 2001, adjusted for nflaton to September,

11 4.Data The AIDS parameters are obtaned from Asano and Fuza (2001). The estmaton of the demand system was based on famly-level expendture data for seven consumpton categores and ther correspondng prce ndexes: food; housng; furnshngs; clothng; transportaton and communcaton; health and personal care and personal expenses, educaton and readng. The data sources for expendtures are two waves of natonal expendture surveys conducted n 1987/88 and 1995/96, and sources for prce ndexes are the monthly natonal survey consumer prces. Correspondng prce ndexes were constructed n a way to allow a comparson of prces both across perods and regons. The regons surveyed are the metropoltan areas of São Paulo, Ro de Janero, Porto Alegre, Belo Horzonte, Recfe, Belém, Fortaleza, Salvador and Curtba, besdes the ctes of Brasla-DF and Goâna. In the present study we assume that the tax structure s common to all ndvduals. We also restrct our ntal analyss to Sao Paulo households. The welfare analyss assocated wth the alternatve optmal commodty tax structures presented n prevous sectons s smulated for 587 household observatons sampled from a populaton of more than 8,200,000 ndvduals. 5 The household sample s dsaggregated n 40 per-capta expendture classes (2.5 percent quantles). Therefore, equvalent varatons are calculated for each of these 40 total expendture classes. 5. Results Gven the estmates of AIDS parameters and the equvalent varaton calculaton we present the results regardng welfare effects on consumers for the specfed optmal tax structures combned wth the dfferent ncome transfers schemes It s mportant to pont out that postve values of EV (or EV/y 0 ) mean a gan n household s welfare from the tax change. Fgures I, II, III, IV and V report the welfare effects for all the fve cases regardng the alternatve tax systems, presentng the EV/y 0 rato, for ε = 0.25 and ε = One mportant feature presented n all fgures s that, except for case II, the rato EV/y 0 remans qute stable along household expendture classes (quantles), for ε = Ths s an expected result because as t s shown n tables II and III of secton 3 and table IV of secton 4 there s a movement towards unformty n commodty tax rates for low values of nequalty averson parameter (specfcally, n cases I, IV and V). Therefore an equal proportonal ncrease n all prces has no redstrbuton effect snce there s an equal proportonal reducton n real ncomes for all households. As our man concern s on the dstrbutonal mpact on households welfare we focus on the equvalent varatons ratos for the hgher value of nequalty averson, ε = It can be seen n all fgures that the rato EV/y 0 s postve for the lowest expendture quantles, a result that presents a welfare mprovement through the optmal tax systems for these expendture classes. Fgure I presents the equvalent varaton rates based on an optmal tax structure wthout ncome transfers to households (case I). We can see n ths fgure that the welfare gan for the lowest expendture quantles (less than 5% of orgnal expendture) s the lowest among all cases. Therefore, the ncluson of a mnmum ncome polcy assocated wth ncome transfers s extremely mportant for the welfare mprovement of lower ncome households. Except for fgure I, all fgures represents a combnaton of optmal commodty taxaton wth some knd of mnmum program polcy. 5 Ths populaton correspond to households earnng from 1 to 40 mnmum wages (total household ncome). 11

12 The results n Fgure II, regardng case II optmal tax rates and unconstraned optmal lump sum, show a hghest welfare gan for the lower ncome households among all the alternatve taxbeneft systems. The hghest EV/y 0 rato s near 400% of orgnal expendture. However, the hgher ncome households ncur the hghest losses (a negatve EV/y 0 rato) n comparson to the other fgures. As t s shown n table I (secton 3), the optmal lump sum transfers obtaned n ths case are extremely hgh. Although commodty tax rates are also hgh and present a regressve nature n ts structure, the strkngly hgh levels of optmal transfers can be the man explanaton for such a remarkably welfare gan n comparson to the others tax systems. Fgure III presents the EV/y 0 ratos regardng case III optmal tax rates combned wth a constrant n the optmal lump sum transfer. The results shown n fgure III are smlar n structure to those presented n fgure II. However, as we are constranng ncome transfers to all households, welfare gans are lower than those shown n fgure II. The hghest gan accrung to low ncome households corresponds to a fgure slghtly above 40% of the orgnal expendture. Fgure IV show the results based on the mnmum ncome polcy that guarantees a per-capta ncome for the lower households. The hghest welfare gan for ths program s close to the one obtaned n fgure III (more than 40% of orgnal expendture). However, the very low expendture quantles present a welfare gan less than ths rate. The results for the school stpend program ( Bolsa-Escola ) are presented n Fgure V. We note that there are fewer households who effectvely beneft from ths program as compared to the other mnmum ncome programs we have shown. Ths s explaned by the fact that the populaton elgble for the Bolsa-Escola s smaller than the other three mnmum ncome programs. 6 Ths fact appears also n fgure IV, the mnmum ncome polcy that guarantees 50% of mnmum wage to the lower ncome households. In the last case, the mnmum program also restrcts the target populaton. The results shown n Fgure VI are qute smlar to the ones n Fgure I, where no transfer s made. As regards the household welfare mpact wth the ncrease n the ncome beneft (we double t), we found an ncrease n the EV/y 0 ratos, whch ndcates an mprovement n welfare gan for the lower ncome households. Ths result s presented n Fgure VI. 7. Concludng Remarks The purpose of ths paper s to examne the dstrbutonal effects of optmal commodty taxaton combned wth cash transfer programs n Brazl. Our analyss s restrcted to a partal equlbrum treatment and focuses on the effects of alternatve optmal commodty tax-beneft systems on the economc welfare of households n dfferent ncome strata. Preferences and household demand patterns used n our welfare mpact smulatons are based on a complete demand system estmated wth a flexble functonal form, the Almost Ideal Demand System (by Deaton and Muellbauer [1980]). Preference parameters estmates are consstent wth mcroeconomc demand theory and allow for an accurate evaluaton of the smulated gan and losses of households welfare. As stressed by Creedy (1999, p.56): the fundamental requrement of any detaled analyss of welfare changes s a sutable consumer demand model, along wth emprcal estmates of the requred demand functons. 6 It s worth remndng that the man targets of ths knd of program may be outsde our restrcted populaton, snce we have performed a lower bound ncome truncaton for the sake of the demand estmaton and we are smulatng results only for São Paulo, the wealthest State of Brazl, so the coverage of the program n the real world s expected to be much hgher. On the other hand, extendng the smulaton to the households wth total ncome below one mnmum wage wll not take nto account the dfference on consumpton behavor (self-consumpton, donatons, etc), whch are heavly present n these segments. 12

13 Our results show that under the proposed optmal commodty tax systems combned wth mnmum ncome programs the welfare gans on low ncome households are hgher n comparson to the stuaton n whch the government gves no ncome transfers to households. These results ndcate that mnmum ncome programs combned wth selectveness n commodty tax structure (wth subsdes n food and housng) would be useful as redstrbuton nstrument among households. More specfcally, the Bolsa-Escola smulatons can be of help n the debate on the need of a general expanson of educaton to reduce poverty and nequalty n Brazl. Barros, Henrques and Mendonça (2000) found out that 40% of overall nequalty n the Brazlan personal dstrbuton of ncome could be ascrbed to educaton. In a recent study, Ferrera and Lete (2002) present some results of a mcro-smulaton exercse for the State of Ceará. They suggest that broad-based polces amed at ncreasng educatonal attanment would have substantal mpacts on poverty reducton, but muted effects on nequalty. Some fnal remarks should be stressed. Our results focus on detaled comparsons of the equvalent varatons resultng from alternatve optmal tax-beneft systems. No attempt was made to produce an overall summary measure. Socal evaluatons of the tax-benefts systems are certanly a useful extenson and can be made usng a specfed socal welfare functon, expressed n terms of the dstrbuton of equvalent ncomes. The ncluson of labor supply n our model would be also desrable to examne the labor supply responses of ndvduals, as well as the mpact of the alternatve reforms on the households welfare. Nevertheless, we beleve that our emprcal fndngs provde a valuable contrbuton n the context of the ncreasng dscusson about mnmum ncome programs and the current tax polcy debate n Brazl, where dstrbutve goals have a great mportance n the agenda. 13

14 Fgures Fgure I 10 Optmal tax No Lump Sum Transfer 5 (% Intal Income) ε =0.25 ε = Orgnal Total Expendture Quantles Fgure II Optmal tax Optmal Lump Sum Transfer (% of Intal Expendture) ε =2.00 ε = Orgnal Total Expendture Quantles 14

15 Fgure III Optmal tax Constraned Lump Sum Bonus Rate= 0.5 (% of Intal Expendture) ε = ε = Orgnal Total Expendture Quantles Fgure IV 50 Optmal tax Mnmum Income Poll Transfer 40 (% of Intal of Expendture) ε =2.00 ε = Orgnal Total Expendture Quantles 15

16 Fgure V Optmal tax Constraned Bolsa-Escola (1x Federal value) (% Intal Income) ε =2.00 ε = Orgnal Total Expendture Quantles Fgure VI (% Intal Income) Optmal tax Constraned Bolsa-Escola (2 x Federal value) 1 40 ε =0.25 ε =2.00 Orgnal Total Expendture Quantles 16

17 References ASANO, S. & FIUZA, E.P.S. (2001). Estmaton of the Brazlan Consumer Demand System. Ro de Janero, IPEA. (Texto para Dscussão No. 793). & BARBOSA, A.N.H. (forthcomng). Optmal commmodty taxes n Brazl wth AIDS preferences. Revsta Braslera de Economa. ATKINSON, A. & STIGLITZ, J.E. (1972). The Structure of Indrect Taxaton and Economc Effcency. Journal of Publc Economcs 1: BARROS, R., HENRIQUES,R. & MENDONÇA,R. (2000). Pelo Fm das Décadas Perddas: Educação e Desenvolvmento Sustentado no Brasl, n Henrques (ed.) Desgualdade e Pobreza no Brasl. Ro de Janero, IPEA. CREEDY, J. (1999). Modellng Indrect Taxes and Tax Reform. Cheltenham (UK) e Northampton (MA, USA): Edward Elgar. DEATON, A. & MUELLBAUER J.(1980). An Almost Ideal System. Amercan Economc Revew, 70 (3): FERREIRA, F. & LEITE, P. (2002). Educatonal Expanson and Income Dstrbuton. A Mcro-Smulaton for Ceará. Ro de Janero, PUC.(Texto para Dscussão No.456). KING, M. A. (1983). Welfare analyss of tax reforms usng household data. Journal of Publc Economcs, 21, LAVINAS, L & VARSANO, R. (1997). Programas de Garanta de Renda Mínma e Ação Coordenada de Combate à Pobreza. Ro de Janero, IPEA. (Texto para Dscussão No.534). MAJUMDER, A. (1988). A note on optmal commodty taxaton n Inda. Economc Letters, 27, RAY, R. (1986). Senstvty of Optmal commodty tax rates to alternatve demand functonal forms. An econometrc case study of Inda. Journal of Publc Economcs, 31, ROCHA, S. (2000). Applyng Mnmum Income Programs n Brazl. Two Case Studes: Belém and Belo Horzonte. Ro de Janero, IPEA. (Texto para Dscussão, N0.746) SIQUEIRA, R. (1997). Redstrbutve effects of alternatve ndrect tax reforms for Brazl. Economa Aplcada 1 (3):