UNIVERSITÀ DEGLI STUDI DI ROMA LA SAPIENZA DIPARTIMENTO DI ECONOMIA PUBBLICA Workng Paper No. 40 Paolo Lberat DID VAT CHANGES REDISTRIBUTE PURCHASING POWER IN ITALY? Roma, novembre 2000
Abstract Ts paper assesses te effects of te canges of te Value Added Tax (VAT) rates n Italy for te perod 1988-1997, questonng ter overall redstrbutve mpact compared wt an equal-revenue proportonal tax rate. By usng te teory of margnal tax reforms and te AWARETAX mcrosmulaton model, t s sown tat most real relatve prce canges due to te selectve use of VAT rates ave caused a welfare loss wc s ncreasngly negatve n te degree of nequalty averson. It means tat many VAT canges mplemented durng te last decade ave been welfare nferor to a dstrbutonally neutral proportonal cange of all VAT rates provdng te same revenue. JEL Classfcaton: D12, D63, H23. Keywords: Margnal tax reforms, VAT, Redstrbuton, Socal Welfare, Italy.
DID VAT CHANGES REDISTRIBUTE PURCHASING POWER IN ITALY? PAOLO LIBERATI ( * ) Introducton Most European countres apply te Value Added Tax (VAT) wt dfferent tax rates for redstrbutve reasons. Italy s not an excepton to ts model of taxaton; snce 1985, VAT was appled wt four dfferent tax rates (fve from md 1991 to end 1992) and only n 1997, te number of rates was reduced to tree. Durng te same perod, owever, te levels of VAT rates were often canged and, more mportant, many goods were reallocated among tem. A natural queston to ask s ten ow muc redstrbuton as been aceved, f any, by mplementng tese canges. In partcular, ow large was te redstrbutonal power of eac adjustment? And wc was ter total redstrbutve mpact? Was te polcy-maker dstrbutonally concerned n usng multple-rate VAT? Was a redstrbutve logc pursued n te overall perod? * ADDRESS FOR CORRESPONDENCE: Unverstà d Roma La Sapenza, Facoltà d Economa, Dpartmento d Economa Pubblca, Va del Castro Laurenzano, 9 00161 Roma (Italy). E_MAIL: lberat@dep.eco.unroma1.t. I ws to tank two anonymous referees for ter useful comments on a prevous verson of ts paper. Te usual dsclamer apples.
Emprcal works on te redstrbutve effects of ndrect taxes n Italy are not uncommon, but tey ave been generally based eter on a sngleyear perspectve or on te dfferental effects of proposed tax reforms. 1 In ts paper, we take a longer perspectve, by studyng te pattern of VAT redstrbuton as a sequence of yearly canges from 1988 to 1997 wt te elp of a mcrosmulaton model and repeated cross-sectons of Houseolds Expendture Surveys (HES) for te same perod. Te perod analysed s of partcular relevance because t almost concdes wt bot te convergence process of Italan economy to te Maastrct parameters and te armonsaton process of VAT structures suggested by te European Unon, two factors tat may ave lmted te redstrbutonal power of te VAT. To evaluate te welfare mpact of te adjustments, ts paper uses te metodology developed by Newbery (1995) and already appled by Gastald and Lberat (1998) and Lberat (1999) to analyse te consequences of ntroducng a two-rate Vat n Italy. Te approac reles on te teory of margnal tax reforms and t focuses on te redstrbutve effect of real relatve prce canges. Accordng to ts, eac VAT cange may be nterpreted as te combnaton of a proportonal cange of all exstng rates (to neutralse te revenue effect) and a selectve adjustment of tax rates at constant revenue. Ten, te mplct baselne of te calculated welfare effects s tat arsng from a proportonal adjustment of all tax rates provdng te same revenue. Accordng to ts nterpretaton, te results sowed n ts paper ave te meanng of dfferental welfare effects aceved by te polcy-maker by beng selectve rater tan adjustng all tax 1 An excepton s Patrz and Ross (1991), wo analyse VAT canges from 1970 to 1987. 2
rates by te same percentage. Te structure of te paper s as follows. Secton 1 provdes basc nformaton on VAT adjustments n Italy durng te perod analysed. Secton 2 sets te framework of analyss, llustratng bot te standard teory of te margnal tax reforms and te real relatve prce approac. Secton 3 dscusses te results, frstly nvestgatng te pattern of te dstrbutonal caracterstcs of goods over te perod analysed; ten sowng te pattern of welfare canges under dfferent assumptons n order to test te robustness of results. Secton 4 concludes. 1. VAT adjustments n Italy: 1987-1997. In te decade from 1987 to 1997, n Italy, te legal structure of VAT rates as been canged egt tmes (Table 1). In 1987, VAT was caractersed by te exstence of two reduced rates (2 and 9 per cent), one ordnary rate (18 per cent) and one rate on luxury goods (38 per cent). For four monts of te same perod, a surtax of 4 per cent on selected goods was also appled. One year later, te ordnary rate was ncreased to 19 per cent, wle two years later, te frst reduced rate was ncreased to 4 per cent. In May 1991, an ntermedate rate (12 per cent) was added to te prevous structure so tat a fve-rate structure prevaled untl te end of 1992, wen te g rate on luxury goods was abolsed. One year later, te ntermedate rate was ncreased by one pont (to 13 per cent) and agan ncreased to 16 per cent n 1995, were also te second reduced rate was ncreased by one pont (10 per cent). Te end pont of ts pat s te tree-rate VAT of 1997, were te ntermedate rate as been abolsed and te ordnary rate ncreased to 20 per cent. In te same perod, some canges of te legal structure were 3
frequently accompaned by reallocatons of goods among rates. Te overall result of tese two knd of canges (legal structure and reallocaton) s sowed n Table 2, were te sgn of te adjustment of eac VAT cange s reported for eac good used n te paper. A blank space ndcates nvarance, wle te sgns + and - ndcate te drecton of te adjustment. Lookng at te table by columns, t s easly seen tat tere s a majorty of + sgns n eac adjustment (for 1989 tere are only + sgns), and ts may be nterpreted as an evdence of te large use of VAT as a revenue-rasng nstrument n te last decade. However, te ncrease of tax rates as been generally dfferentated among commodtes; ten, relatve prces ave canged n eac year and some redstrbuton of purcasng power s terefore always expected. Lookng at te same table by rows, nstead, gves a pcture of te pattern of VAT rates appled to eac good. Most of tem ave borne repeated ncreases n te overall perod, and only for about one trd of te goods ts trend as been broken by negatve adjustments. 2 2. Te teory of margnal tax reforms 2.1. Te standard approac Te framework of analyss s based on te teory of margnal tax reforms developed by Feldsten (1972), Amad and Stern (1984) and summarsed by 2 Te analyss of VAT canges n te perod wll be conducted condtonal to te canges tat consumpton patterns ave dsplayed n te perod. In oter words, t s not consdered ere tat te cange of some VAT rates may ave condtoned or t may ave been encouraged by te evoluton of ouseolds consumpton patterns (e.g. some luxury goods becomes normal, etc.). 4
Newbery and Stern (1987). 3 Consder an ndvdualstc Socal Welfare 1 H Functon (SWF) W = W ( v,..., v ) were v ( c,p) v = s te ndrect utlty functon of ouseold defned over total expendtures c and te prce vector p. 4 Te mpact on socal welfare of a sngle prce cange may ten be wrtten as: W p Ten, by Roy's dentty: W p = W v = β x v p [1] [2] were W v β = s te socal margnal utlty of money for ouseold v c and x s te consumpton of good for te same ouseold. For a margnal cange of te prce of good, te welfare cange s terefore approxmated by te sum over ouseolds of te socal margnal utlty of money wegted by te actual consumpton of good. Assumng fxed producers prces, wc mples dp =dt (were t s te tax on te -t good), 3 Ts approac s teoretcally consstent only for small polcy canges; yet, t provdes a smple framework to nfer welfare drectons of complex tax reforms. In partcular, t mgt ave greater nformatonal effcency wen appled on dsaggregated data, compared wt welfare analyss based on te estmaton of demand systems. In ts latter case, te potental power of dsaggregated data s often reduced by te use of a smaller set of aggregate commodtes. For a toroug treatment of demand systems and mcrodata see Blundell et al. (1993). For dfferent opnons on te proper way to estmate welfare canges see, for example, Banks, Blundell and Lewbel (1996) and Slesnck (1998). 4 We coose to defne te ndrect utlty functon n terms of total expendtures, rater tan ncome, as below n te text socal wegts wll be assumed to be a functon of expendtures per equvalent adult. About ts coce see, for example, Slesnck (1998). Te unt of analyss s te ouseold ; ten, we do not address te ssue of wtn-ouseold allocaton of resources. For some consequences of ts coce see, for example, Haddad and Kanbur, 1990 and Apps and Rees, 1996. 5
expresson [2] yelds te effect of a margnal cange of tax rates on socal welfare under te assumpton of full forward sftng of taxes on prces. 5 Beng condtonal on ts assumpton, results obtaned from [2] sould be consdered wt some cauton. Accordng to Feldsten (1972), te dstrbutonal caracterstc of a commodty can be defned as: were d β βx x X = s te total consumpton of good and β s te average x socal wegt. 6 Solvng [3] for te numerator of te r..s. and replacng t nto [2] yelds: [3] W p = βd X [4] Snce equaton [4] contans bot te average socal wegt and te dstrbutonal caracterstcs, wc n turn depend on ndvdual socal wegts, an explct functonal form for te socal welfare functon s needed to make expresson [4] an operatonal tool. To ts purpose, we can make recourse to te soelastc socal welfare functon by Atknson (1970) 5 Te assumpton of fxed producer prces s a smplfed way to andle redstrbutonal ssues. From a teoretcal pont of vew, t olds only n te case of constant returns to scale. In te case of mperfectly compettve markets, oversftng s also a possblty. Emprcal work on ts topc s dffcult and tus uncommon. For emprcal evdence of oversftng, see Besley and Rosen (1998). For examples of studes assumng fxed producer prces, see Pecman and Okner (1974), Kaser and Span (1989), Metcalf (1994), Ncèle and Robn (1995). Gven te lack of emprcal evdence on approprate sftng assumptons, te results must be nterpreted condtonal to te coce made. 6 Te dstrbutonal caracterstc s a measure of ow concentrated s consumpton of a gven commodty on te most deservng part of te populaton (.e. among tose wt g β s). Indeed, f all ouseolds would consume te good n te same absolute amount, d 6
1 ( v ) e W = 1 e (were e s te degree of nequalty averson) 7. Also assume expendture per equvalent adult (K) as a reasonable welfare ndcator for eac ouseold. 8 Ten, socal wegts,.e. te frst dervatve of te socal welfare functon wt respect to te welfare ndcator, are gven by ( K ) e β =. 9 10 2.2. Te real relatve prce effect Te man concluson of te prevous secton s tat expresson [4] can be used to calculate a local approxmaton of te welfare cange as a consequence of a tax cange on a gven commodty. For our purposes, tat expresson can be modfed to solate te dfferental effect of usng selectve tax canges combned wt a proportonal ncrease of all tax rates provdng te same revenue. By ts way, welfare canges wll not be affected by te dfferent revenue-rasng power of dfferent tax canges; rater, tey wll solate te convenence of beng selectve compared wt a would be equal to 1; d=1 also f tere s no dstrbutve concern,.e. f all socal wegts are equal. 7 For e=1, W = ln( v ). Te use of an explct functonal form for te SWF may ndeed be a sortcomng, because results are condtonal on te specfc form cosen. 8 See, for example, Slesnck (1998; p.2147) and Deaton and Muellbauer (1980; p.226). 9 Applcatons of ts metodology s wdespread. Only to quote some examples see Amad and Stern, 1984; Newbery and Stern, 1987; Kaser and Span, 1989; Amad and Stern, 1991; Newbery and Révész, 1995; Newbery, 1995; Madden, 1995a; Madden, 1995b. 10 As easly seen, welfare wegts are ndependent of bot oter agents' utltes and prces, consstent wt te margnal caracter of te approac tat dentfes drectons of reforms but not ter amount. In te words of Amad and Stern (1984; p.279), β sould be seen as a convenent local approxmaton gven current prces and ncomes rater tan an exact expresson oldng for all prces and ncomes. Te dependence of socal wegts on prces 7
unform percentage ncrease of all tax rates. Te mportance of glgtng ts effect arses from te observaton tat te man reason of cangng VAT rates n a selectve way sould be redstrbutve,.e. tryng to elp poor people to gan purcasng power. Newbery (1995) developed a smple way to analyse te redstrbutve power of tese canges, by calculatng real relatve prce canges. 11 By te omogenety of degree zero n prces and nomnal ncome of te ndrect utlty functon, we can rewrte t as ( C v, π), were C = c / P s now real expendture, π = P s te real relatve prce of good at te base p / perod, P ω p s te general prce ndex at te base perod and ω s = te sare of good n total expendtures. After a tax cange of good, te new real relatve prce becomes * * * π = p /P because bot te orgnal p and te general prce ndex P would be affected by te cange. For margnal canges, P * ω = * p s ten te new general prce ndex, wt fxed wegts ω referrng to te base perod. By te defntons of real relatve prces before and after te tax cange, we get π = π * π. By a normalsaton of all consumers prces of as been nvestgated by Roberts (1980). An emprcal applcaton of prce-dependent welfare wegts can be found n Murty and Ray (1989). 11 In ts framework, ouseolds money expendtures s assumed to vary n te same proporton as te general prce ndex. By ts way, real expendtures, relatve to te general prce ndex, are kept constant. In oter words, we are dsregardng te socal welfare cange due to real ncome canges between polcy at tme t and polcy at tme t+1 for a gven set of relatve prces. We are nstead consderng te socal welfare cange due to te canged set of relatve prces between t and t+1 for a constant level of real ncome. 8
* te base perod to one ( P = 1 and π = 1 ), π = π 1. 12 Now, suppose tat all tax rates are ncreased proportonally; ten, all * p s and * P would grow at te same rate, wc mples π * = 1 and π = 0. Ten, by usng te transformed ndrect utlty functon (expressed n terms of real varables) and generalsng expresson [4] to multple tax canges, te local approxmaton of socal welfare canges s: W β d X π [5] Expresson [5], mplctly, provdes te welfare cange caused by a selectve cange of VAT rates aganst te bencmark of a zero cange for a proportonal ncrease of all ndrect tax rates provdng as muc revenue as te actual adjustment. 13 By ts way, we can dsregard te dfferental revenue-rasng power of eac tax cange n te perod 1987-1996, by mplctly assumng tat eac selectve cange ad as a natural alternatve a unform ncrease of all tax rates provdng te same revenue. In ts latter case, t s mmedate to verfy tat π =0 and W = 0. 14 Terefore, eac non-zero value of [5] can be nterpreted as eter te beneft (f postve) or 12 Furter, te way n wc te general prce ndex s constructed mples ω π = 0 * * ω Ts s easly sown by ω ( π 1 ) = ω π ω = p 1 = 0 *. P * = p * * as P ω. 13 For wat s of nterest ere, labour supply s assumed to be fxed and labour ncome treated as lump-sum. Ten, a proportonal ncrease of all ndrect tax rates s ere equvalent to a tax on all lump-sum ncome (fxed labour ncome plus true lump-sum ncome) 14 It s wort notng tat W = 0 also n te case were all d = 1,.e. eter wen all goods are consumed n te same amount by everybody or wen te socal welfare s dstrbutonally nsenstve (β s are equal). 9
te cost (f negatve) of mplementng a gven selectve cange rater tan ncreasng all tax rates by te same percentage. To normalse expresson [5], te ntal level of welfare W can be used. Wt Atknson s socal welfare functons, te proportonal cange n welfare can be expressed by: W 1 W W 0 0 = 1 e 1 ( s ( K1 ) )/( 1 e) s ( K 0 ) 1 e ( s ( K 0 ) )/( 1 e) e ( )/( 1 e) were s s te number of equvalent adult and K s agan expendture per equvalent adult. Usng te defnton of socal wegts, te prevous W expresson can be smplfed as W = β X β X. By te defnton of te dstrbutonal caracterstcs and by notng tat te measure of consumpton cange, n our case, s X = x π, we can obtan a fnal expresson: W W d ω π [6] were bot numerator and denomnator ave been dvded by total expendture X. d ω 3. Results 3.1. Data and smulaton tools Data used n te smulatons are from te Houseold Expendture Survey (HES) conducted by te Italan Natonal Insttute of Statstcs (ISTAT). 15 15 Istat, Indagne su consum delle famgle, 1987-1996, Rome. Istat does not bear any responsblty for te analyss and te nterpretaton of te data reported ere. Extensve 10
For our purposes HES from 1987 to 1996 are used. Orgnal data are reclassfed to obtan a vector of 68 goods and servces to wc te correspondng vectors of tax rates and canges of VAT rates (one for eac year) are assocated. In some cases, te groupng of goods provded by ISTAT ncludes more tan one commodty. Tere are also cases were commodtes of te same group bear dfferent tax rates; n ts case, a wegted average tax rate as been calculated usng te wegts of te Retal Prce Index (RPI) for te correspondng years. Ts metod mples tat te sare of tose commodtes on te aggregate group s te same for all ouseolds. Welfare effects and oter results sowed n ts paper are obtaned usng AWARETAX (Assessng Welfare And Revenue Effects of TAXes), a mcrosmulaton model for te analyss of te redstrbutve effects of drect and ndrect taxes, as well as of te man cas benefts for ouseolds. 16 3.2. Te dstrbutonal caracterstcs of goods Dstrbutonal caracterstcs of te 68 goods ave been calculated accordng to equaton [3] and reported n Fgure 1 (ranked by decreasng level at e=1) for te tree levels of nequalty averson cosen (e=0.5, e=1 and e=2) and for te two extreme years of te perod, 1987 (lnes) and 1996 (symbols). Te grap s plotted aganst te cumulatve budget sare of te treatment of HES data can be found n Brandoln (1993). See also Brandoln (1999). Attenton n te text wll be confned to te basc features of HES data and on te assumptons used trougout te analyss. 16 AWARETAX as been bult by F. Gastald and P. Lberat. Basc nformaton on te mcrosmulaton model can be found at ttp://dep.eco.unroma1.t/lberat/awaretax.tm. For some features of te model, see Gastald and Lberat (1998) and Gastald and Lberat (1999). Furter nformaton s also avalable from te autors upon request. 11
same goods wen e=1. Te lnes and te symbols n te mddle, referrng to e=1, are monotoncally non-ncreasng as tey are used to rank goods. At e=0.5 and at e=2 volatons of te monotoncty may occur f goods are ranked dfferently under dfferent degrees of nequalty averson. For bot 1987 and 1996, te grap sows tat te rankng of goods s rater nsenstve to te degree of nequalty averson. However, some rerankng occurs n bot cases at e=2, more eavly for 1996 especally n te mddle part of te grap. Dfferentaton among commodtes, on te oter and, appears substantal n bot cases, reflectng, to some extent, te specalsed pattern of consumpton of poorer and rcer ouseolds n market economes. 17 Comparng 1996 wt 1987, owever, te dfferentaton among commodtes tends to reduce; te dstrbutonal caracterstcs n 1996 are always ger tan tose of 1987 for all goods, but symbols n te grap appears slgtly more equally dstrbuted n 1996. Indeed, ts olds true at e=0.5 and at e=1 for wc te standard devaton n 1996 s actually lower tan tat of 1987; t does not old at e=2, were te standard devaton n 1996 slgtly ncreases. It s wort notng tat, n bot years, at e=2 te gest dstrbutonal caracterstc s for electrcty. Indeed, ts good becomes ncreasngly mportant wt te ncrease of wegts to te lowest part of te ncome dstrbuton; ten, any tax on tat good as te greater potental regressve mpact f evaluated aganst eter total consumpton or total ncome. 18 17 See Newbery (1995) for a comparson between UK and Hungary. 18 A g value of te dstrbutonal caracterstc means tat consumpton does not ncrease sgnfcantly wt ncome; ten, any tax on consumpton of ncome nelastc goods s lkely to be regressve wen measured aganst total ncome. Note also tat electrcty ranks at te frst place n 1987 for bot e=1 and e=2. Ts feature s partcularly mportant 12
Te dfference between 1987 and 1996 n terms of sngle commodtes s captured n Table 3, tat sows te rank of eac good by decreasng value of te dstrbutonal caracterstc at e=1. Te frst ten places are for te same goods n bot years, even toug ranks are dfferent. For example, bread, tat ranked frst n 1987, ranks fourt n 1996; wle sugar, frst n 1996, was only fft n 1987. Wt te excepton of te last tree goods (otels, transport and communcatons, and furnture) almost eac oter good as a dfferent poston n 1996 tan t ad n 1987. Of partcular mportance s te rerankng of gas for domestc uses (oter tan eatng) wose consumpton appears less concentrated on poor ouseolds n 1996 tan t was n 1987; te same lne of reasonng olds for gasolne, kerosene, wne and publc transport. For oter goods, te reverse s true: n partcular for telepone, blankets and seets, bcycles and motorcycles, reflectng, to some extent, cangng consumpton patterns durng te last decade. Consumpton of durable goods are also sgnfcantly reranked. Furter nformaton arses from Fgure 2, were te dstrbutonal caracterstcs of all goods for eac year, agan at e=1, are reported aganst te budget sare of 1996. Te degree of dfferentaton n some years s more substantal tan n 1987 and 1996 (e.g. n 1989). Indeed, te dsperson of te symbols around te empty boxes of 1996, n te grap, s a sgnal of te ntensty of bot rerankng of goods among dfferent years and ter canged wegt n total consumpton, as well as of vared consumpton nequalty. From 1987 to 1996, not only te poston of specfc for tax polcy, as electrcty, n Italy, s taxed wt bot VAT and excse taxes at varous 13
commodtes mgt ave canged, but also te sape of te expendture dstrbuton. Ten, a natural queston to ask s wy dstrbutonal caracterstcs n 1996 are ger tan tose of many oter years reported n Fgure 2. Is total consumpton less unequally dstrbuted n 1996 or as consumpton of more goods become more wdespread among te poor (.e. decreasng expendture elastctes across years)? In oter words, s te result drven by a general factor or by commodty-specfc factors? Followng Newbery (1995), a smple way to analyse ts ssue s to assume tat total expendture s lognormally dstrbuted wt mean µ and varance σ 2, and tat eac good as a constant expendture elastcty λ. 19 From equaton [3], te value of te dstrbutonal caracterstc can be defned n terms of expected values as follows: E(βx )/EβEx. From te assumpton of constant elastcty, te demand for good can be wrtten as x =α (K ) λ, were K s agan expendture per equvalent adult and α s te constant. Ten, te value of d s: d = λ e Eα K e EK Eα K λ were te numerator comes from te defnton of te socal wegt β. From te propertes of te lognormal dstrbuton we know tat EK z =exp(zµ+0.5z 2 σ 2 ) for any varable z; ten, t s easy to sow tat: levels of government (central government and muncpaltes). 19 A constant expendture elastcty at all levels of expendture s a smplfyng assumpton, but t elps to get a broad mpresson of wc factors are more relevant n determnng te level of te dstrbutonal caracterstcs. Applcatons of te lognormal dstrbuton n ncome/consumpton nequalty s wdespread. See, for example, Newbery and Stgltz (1981), especally capter sx, Newbery (1995), Creedy (1996). 14
2 2 2 2 2 exp( eµ + 0. 5e σ )exp( λµ + 0. 5λ σ )exp( λeσ ) 2 d = = exp( λeσ ) [7] 2 2 2 2 exp( eµ + 0. 5e σ )exp( λµ + 0. 5λ σ ) Ts expresson allows us to compare dfferent values of te dstrbutonal caracterstc of te same good n two dfferent years and to understand f te cange depends on te commodty specfc-factor λ (te expendture elastcty) or on te general factor σ 2 (te varance of te log of expendture per equvalent adult). Table 4, n te frst column, reports te dfference between te dstrbutonal caracterstcs of 1996 and 1987 for e=1. As also evdent from Fgure 1, tere are only few goods wt negatve sgns (.e. decreasng value of te dstrbutonal caracterstc: gas for oter domestc uses, bread, pasta and wne). Te explanaton of tese canges accordng to equaton [7] s reported n te oter two columns. Te second column measures te dfference between a ypotetcal dstrbutonal caracterstc defned by te same expendture elastcty as n 1987 and te varance of 1996 expendture, and te dstrbutonal caracterstc of 1987. By ts way, we solate te effect of te canged sape of te expendture dstrbuton,.e. te nfluence of te general factor. Te trd column, nstead, measures te dfference between te dstrbutonal caracterstc of 1996 and a ypotetcal dstrbutonal caracterstc defned by te expendture elastcty of 1987 and te varance of 1996. d 2 In symbols, for e=1, one can defne 96 exp( λ 96σ96 ) 2 ( λ ). Ten: 2 2 d 96 d 87 = exp( λ 96σ96 ) exp( λ87σ87 ) = 2 2 2 2 [ exp( λ σ ) exp( λ σ )] + [ exp( λ σ ) exp( λ σ )] 87 exp 87σ87 96 96 87 96 d and Te frst term n square brackets s te commodty-specfc factor (te trd 87 96 87 87 15
column of Table 4), wle te second term s te general factor (te second column n Table 4). By [7], we also know tat te value of te dstrbutonal caracterstc ncreases f bot σ 2 and λ decreases. Snce, n 1996, 64 values of te dstrbutonal caracterstcs out of 68 ave ncreased, compared wt 1987, t s of some nterest to capture te reasons of ts pattern. Frst of all, one can observe tat te varance of expendture per equvalent adult s less n 1996 tan n 1987; ten, te general effect s always postve,.e. t causes te value of te dstrbutonal caracterstcs to ncrease. In te trd column t s sown tat n 48 cases out of 68, te commodty specfc factor (te elastcty) s also postve, renforcng te general effect of te varance of te expendture dstrbuton. Te combnaton of ts result explans, for 48 cases, te growt of te value of te dstrbutonal caracterstcs. But wc effect s stronger? For nondurable goods, tere are only sx cases out of 48 were te elastcty effect s ger tan te varance effect (bcycles, blankets and seets, wasng powders, medcnes, publc telepone and sugar). In all oter cases, te varance as te predomnant role. For durable goods (te bottom egt goods n te table), nstead, tere are fve cases were te elastcty effect s ger. Tese results gves partal support to te ypotess tat te dstrbutonal caracterstcs of 1996 are ger because expendtures are less unequally dstrbuted rater tan te goods ave sgnfcant dfferent expendture elastctes n te two perods. At least, te frst factor always explans more of te cange tan te second does. In te remanng twenty cases were te commodty-specfc effect s negatve, tere are only four cases were ts negatve effect s predomnant (tose for wc te cange 16
of te dstrbutonal caracterstc s negatve). Tere are also cases were te two effects almost compensate (gas for domestc uses oter tan eatng, edble ol, bread), gvng mportant addtonal nformaton compared wt a smple comparson between te values of te dstrbutonal caracterstcs between te two perods. For gas for domestc uses (oter tan eatng), for example, d 87 =0.823, wle d 96 =0.816; but t would be wrong to nfer tat notng relevant for redstrbuton as appened, because te decomposton sows tat expendture elastcty as ncreased, sgnallng a potental lower concentraton of ts knd of expendture among te poor. But t also sows tat ts reducng effect on te dstrbutonal caracterstc as been compensated by a low varance of total expendture. Te understandng of te magntude of tese two factors s terefore crucal to set proper tax polces on consumpton. 3.3. Welfare canges Te applcaton of expresson [6] s reported n Table 5 for te tree degrees of nequalty averson cosen, wt greater numbers reflectng greater averson. In te frst panel, welfare canges are calculated by usng, for eac cange, te survey of eac year as a base year. It means tat dstrbutonal caracterstcs and budget sares are tose of te year before te cange was made. In oter terms, any cange n a gven year s mplctly assessed assumng te prevous year as a reference year. Te table sows tat, on balance, sx out of nne adjustments are dstrbutonally adverse. From 1988 to 1993 and wt te excepton of 1990, any cange of VAT rates s assocated wt a negatve cange n welfare for 17
any degree of nequalty averson. Furter, most of tese negatve canges are ncreasng n te degree of nequalty averson, wc means a partcular adverse redstrbutve effect for poorer ouseolds. To generate ts effect, t must be tat te reductons of real relatve prces were not correlated wt te level of te dstrbutonal caracterstcs. For example, n te adjustment of 1988, te bggest reducton n real relatve prces was tat of books, an tem wc rank partcularly low n te dstrbutonal caracterstcs rankng at e=2; and te second bggest reducton n real relatve prces was for some durable goods wc agan rank very low n te same rankng. In 1989, tere were two reductons n real relatve prces muc better targeted, tose for electrcty and telepone, but tey were accompaned by reductons of te relatve prce of goods rankng low on te dstrbutonal ground and, even worse, by te ncrease of real relatve prces of many foodstuffs (e.g., bread, edble ol, ceese, butter) wt g dstrbutonal caracterstcs and budget sares. Ts latter effect as more tan compensated, on te welfare sde, te postve mpact of te reductons of electrcty and telepone. In bot 1991 and 1992, te only relevant reducton n real relatve prces s tat of gas for domestc uses oter tan eatng, wc s owever compensated, n bot years, by some ncreases n relatve prces of foods (especally fs, wose budget sare s tree tmes as muc tat of gas at comparable levels of dstrbutonal caracterstcs). In 1993, tere were a few reductons of real relatve prces, especally for transport and communcatons and motorcycles wc, agan, are not partcularly valuable on redstrbutonal grounds. Fnally, n 1995, te bggest reducton n real relatve prces s for medcnes, wc s md-ranked at e=2. Ts 18
effect, owever, s manly counterbalanced by te ncrease of te real relatve prce of electrcty (wc s top-ranked) and gas for domestc uses oter tan eatng. Partal evdence of ts rater marked ndependence of real relatve prce canges and dstrbutonal caracterstcs can be found by lookng at te covarance between te two varables. A negatve covarance means tat g postve values of te dstrbutonal caracterstcs are assocated wt negatve values of te real relatve prce canges, wc s expected f tere s dstrbutonal concern n tose canges. For te adjustments n te perod 1988-90, te covarance s postve and of very small magntude, wle for 1991 and 1992 t s almost zero. Relatvely g covarances can be found for 1993 (toug postve) and for 1997 (negatve), were a postve welfare cange occurs. Ts latter adjustment appears of partcular relevance and somewat surprsng, because te 1997 cange as reduced te number of rates from four to tree and te reducton of VAT rates s generally tougt to be dstrbutonally adverse to te extent tat te prevous targetng s dstrbutonally grounded. Te bulk of ts postve adjustment s drven by a sgnfcant reducton of tax rates on beef meat and sugar, wc explans about one alf of te overall welfare cange for all degrees of nequalty averson, as bot commodtes ave g dstrbutonal caracterstcs. Ts casts some doubts on te redstrbutve logc of te four-rate structure used untl 1996; on te oter and t may contrbute to 19
reduce te resstance to swtc to a two-rate VAT structure as suggested by te European Unon. 20 Snce eac welfare cange s assessed aganst a dfferent base year, te sum of te canges does not provde te total welfare cange of te perod. In order to get ts result and to test te senstvty of welfare canges to dfferent coces of te base year, we performed te same exercse usng te 1996 HES survey as a reference year for all adjustments. Results are sowed n te second panel of Table 5. Te stuaton appears even worse tan tat sown by te frst panel: now, seven adjustments out of nne sow a negatve sgn. Indeed, te sgn of te welfare cange s now negatve also for te 1990 adjustment and for two cases n te 1994 adjustment. On te oter and, te 1995 adjustment provdes now a postve welfare cange. If evaluated on te overall perod, owever, te balance s strctly negatve for all degrees of nequalty averson. Te same concluson can be nferred by lookng at te trd panel of Table 5, tat sows te results of canng te canges of varous years on two dfferent panels, 1996 and 1992. Results are agan negatve and ncreasng n te degree of nequalty averson. 21 All tese results provde support to te concluson tat VAT canges adversely contrbute to redstrbute purcasng power n te last decade. 20 On te welfare canges and VAT ncdence of suc a swtc n Italy see Gastald and Lberat (1998) and Lberat (1999). 21 Te exercse as also been repeated wt oter years, not reported n te table, and te only evdence of postve (but not statstcally sgnfcant) welfare canges are from 1987 HES (and 1988 HES for e=0.5). 20
Ten, te adjustments made durng tat perod cannot fnd ter justfcaton on te redstrbutve ground. 22 However, gven te uncertanty of some results, t proves useful to test te statstcal sgnfcance of te numbers sown n Table 5. A test for te statstcal sgnfcance of te welfare canges can be performed by regressng te dstrbutonal caracterstc of eac commodty on te product between te budget sare and te real relatve prce canges. 23 To see wy, one can nterpret te numerator of equaton [6] as te covarance between te dstrbutonal caracterstc and ω π. Te covarance * between te two varables s Cov ( d, ω π ) = ( d d )( ω π ω π) were d * s te expected value of te dstrbutonal caracterstc and te expected value of ω π s ω π ( ω π I, were I s te number of / goods). Snce te numerator of ts latter term s by constructon equal to zero, ( d ω π ) = d ω π Cov,. It s known tat, by te relaton between te correlaton coeffcent ρ, te covarance can be expressed by were te two σ s are te standard devaton of d and ρσ d σ ω, π ω π respectvely and ρ s te correlaton coeffcent between te two varables. Ten, te values of Table 5 wll be statstcally sgnfcant from zero f te correlaton coeffcent s statstcally sgnfcant from zero. Table 5 reports statstcally sgnfcant welfare canges (at 5 per cent level) n bold caracters, wle talcs sows tose statstcally sgnfcant at 22 In prevous papers (Gastald and Lberat, 1998 and Lberat, 1999), t as been sown tat also a two-rate VAT mgt ave a postve mpact on welfare, renforcng te doubts about te dstrbutonal superorty of a four-rate VAT used untl 1997. 23 See, for example, Newbery (1995). 21
10 per cent level. It s of some nterest to note tat te only sgnfcant postve welfare canges are tose of te 1997 adjustment. All oters postve welfare canges (tose of 1990 and 1994) are not statstcally dfferent from zero. On te oter and, tere s strong evdence tat te negatve welfare canges of 1988, 1989 and 1993 are statstcally sgnfcant. Ts fndng s rater mportant as t casts many doubts on bot te redstrbutve concern and te logc of te polcy-maker n usng VAT n te past ten years. Recalculatng te total welfare cange of te perod by consderng only tose results tat are statstcally sgnfcant eter at 5 or 10 per cent level, gves te followng negatve entres for te overall perod: -0.07 for e=0.5; -0.1098 for e=1; -0.1315 for e=2, wt a total welfare loss wc s ncreasng n te degree of nequalty averson. Conclusons Te man concluson emergng from ts paper s tat VAT as been an neffcent means of addressng redstrbutonal ssues n Italy n te last decade. Te total redstrbutve mpact as been found negatve and ncreasng n te degree of nequalty averson. Moreover, negatvty appears suffcently robust to dfferent coces of te base year and dfferent ways of calculatng real relatve prce canges. Terefore, we can conclude for te lack of any clear evdence of bot polcy-maker s redstrbutonal concern and ntentonal redstrbutve logc, wt te possble excepton of te reducton of te number of tax rates occurred n 1997. Prevous results on te dstrbutonal superorty of a two-rate VAT n Italy agan support te ypotess tat te redstrbutve targetng of te multple-rate VAT n te 22
last decade was very poor. Ten, te conclusons of Sa (1983; p.101) tat anyone serously nterested n redstrbuton as reasons to be qute skeptcal of ndrect taxes fnds, n Italy, emprcal evdence. Gven te now wdespread avalablty of repeated cross-secton surveys on ouseolds consumpton n oter European countres, t wll be of some nterest to verfy weter ts effect s country-specfc, or weter t can be generalsed to oter VAT experences. References Amad, Etsam, and Ncolas Stern. 1984. Te Teory of Reform and Indan Indrect Taxes, Journal of Publc Economcs, 25, pp. 259-298. Amad, Etsam, and Ncolas Stern. 1991. Te Teory and Practce of Tax Reform n Developng Countres. Cambrdge: Cambrdge Unversty Press. Apps, Patrca F., and R. Rees. 1996. Labour Supply, Houseold Producton and Intra-Famly Welfare Dstrbuton, Journal of Publc Economcs, 60, pp. 199-219. Atknson, Antony. 1970. On te Measurement of Inequalty, Journal of Economc Teory, 2, pp. 244-263. Banks, James, Rcard Blundell and Artur Lewbel. 1995. Quadratc Engel Curves, Indrect Tax Reform and Welfare Measurement. Insttute for Fscal Studes. Besley, Tmoty J, and Harvey S. Rosen. 1998. Sales Taxes and Prces: An Emprcal Analyss. NBER Workng Paper n.6667. Blundell, Rcard, Panos Pasardes and Guglelmo Weber. 1993. Wat Do We Learn About Consumer Demand Patterns from Mcro Data?, Amercan Economc Revew, 83, pp. 570-597. Brandoln, Andrea. 1993. A Descrpton and an Assessment of te Sample Surveys on te Personal Dstrbuton of Incomes n Italy. Mcrosmulaton Unt Workng Papers, Unversty of Cambrdge, n.9303. 23
Brandoln, Andrea. 1999. Te Dstrbuton of Personal Income n Post-War Italy: Source Descrpton, Data Qualty and te Tme Pattern of Income Inequalty. Banca d Itala, Researc Department. Creedy, Jon. 1996. Fscal polcy and socal welfare. Celtenam: Edward Elgar. Deaton, Angus and Jon Muellbauer. 1980. Economcs and consumer beavour. Cambrdge: Cambrdge Unversty Press. Feldsten, Martn S. 1972. Dstrbutonal Equty and te Optmal Structure of Publc Prces, Amercan Economc Revew, 62, pp. 32-36. Gastald, Francesca and Paolo Lberat. 1998. Towards a Two-Rate VAT n Italy: Dstrbutonal and Welfare Effects. Mcrosmulaton Unt Workng Papers, MU9802, Unversty of Cambrdge. Gastald, Francesca and Paolo Lberat. 1999. Imposte e redstrbuzone n Itala, fortcomng n Garofalo, Guseppe and Antono Pedone (eds.). Dstrbuzone, Redstrbuzone e Crescta - Gl effett delle dseguaglanze dstrbutve. Mlano: F.Angel. Haddad, Lawrence and Rav Kanbur. 1990. How Serous Is te Neglect of Intra-Houseold Inequalty Economc Journal, 100, pp. 866-881. Kaser, H. and Paul B. Span. 1989. On te Effcency and Dstrbutve Justce of Consumpton Taxes: a Study on VAT n West Germany, Journal of Economcs, 49, pp. 199-218. Lberat, Paolo. 1999. Te Dstrbutonal Effects of Indrect Tax Canges n Italy, fortcomng n Internatonal Tax and Publc Fnance. Madden, Davd. 1995a. An Analyss of Indrect Tax Reform n Ireland n te 1980s, Fscal Studes, 16, pp. 18-37. Madden, Davd. 1995b. Labour Supply, Commodty Demand and Margnal Tax Reform, Economc Journal, 105, pp. 485-497. Metcalf, Glbert. 1994. Te Lfetme Incdence of State and Local Taxes: Measurng Canges Durng te 1980s, n Slemrod, Joel (Ed.). Tax Progressvty and Income Inequalty. New York: Cambrdge Unversty Press. Murty, M.N and Ranjan Ray. 1989. A Computatonal Procedure for Calculatng Optmal Commodty Taxes Wt Illustratve Evdence from Indan Budget Data, Scandnavan Journal of Economcs, 91, pp. 655-670. Newbery, Davd M.G. 1995. Te Dstrbutonal Impact of Prce Canges n Hungary and n te Unted Kngdom, Economc Journal, 105, pp. 847-63. 24
Newbery, Davd M.G. and Tamas Révész. 1995. Te Burden of te Hungaran Personal Tax System, Unversty of Cambrdge, Department of Appled Economcs. Newbery, Davd M.G. and Ncolas Stern (eds.). 1987. Te Teory of Taxaton for Developng Countres. Oxford: Clarendon Press. Newbery, Davd M.G. and Josep Stgltz. 1981. Te teory of commodty prce stablzaton. Oxford: Clarendon Press. Ncèle, Veronque and Jean-Marc Robn. 1995. Smulaton of Indrect Tax Reforms Usng Pooled Mcro and Macro Frenc Data, Journal of Publc Economcs, 56, pp. 225-244. Patrz, Vncenzo and Ncola Ross. 1991. Preferenze, prezz relatv e redstrbuzone. Bologna: Il Mulno. Pecman, Josep A. and Benjamn A. Okner. 1974. Wo Bears te Tax Burden? Wasngton DC: Brookngs Insttutons. Roberts, Kevn. 1980. Prce-Independent Welfare Prescrptons, Journal of Publc Economcs, 13, pp. 277-297. Sa, Raaj K. 1983. How Muc Redstrbuton s Possble Troug Commodty Taxes?, Journal of Publc Economcs, 20, pp. 89-101. Slesnck, Danel T. 1998. Emprcal Approaces to te Measurement of Welfare, Journal of Economc Lterature, 36, pp. 2108-2165. Stern, Ncolas. 1990. Unformty versus selectvty n ndrect taxaton, Economcs and Poltcs, 2, pp. 83-108. 25
Table 1 - VAT Rates n Italy, 1987-1997 R1 R2 I N H 1987 2 9 18 38 snce 30_7_88 2 9 19 38 snce 1_1_89 4 9 19 38 snce 13_5_91 4 9 12 19 38 snce 1_1_93 4 9 12 19 snce 1_1_94 4 9 13 19 snce 24_2_95 4 10 16 19 snce 1_1_97 4 10 20 R1= Frst reduced rate; R2 = Second reduced rate; I = Intermedate rate; N = Normal rate; H = Hg rate 26
Table 2 - Canges of VAT Rates n Italy, 1988-1997 88 89 90 91 92 93 94 95 96 97 88 89 90 91 92 93 94 95 96 97 Mneral Water + + + Laundry + + + Hotels and motels + + - + + Fres vegetables + - Oter foodstuffs ++ ++ +- - Books - +- + + Oter drnks + + + + - + + Medcne + - + Oter meat + + - - Furnture - + + Oter ome + + + + Edble ol + Oter cereals + + Bread + Butter and margarne + Hardresser + + + Oter furnture + + + Pasta + + - + Oter clotes + + + + + + + Meals out + Clotes + + + + + Cutlery and pots + + + Personal ygene ++ +++ + Fs + +++++ Sport artcles + + + + Poultry + Bus and taxs + + - - + + Oter goods + + + + + + + Gasolne + + + Recreatonal goods + + + + + + + Bkes + + + Oter recreatonal + + + + Leater artcles + + + Cured pork meat + + - - Coffee and tea + + + Healt + + - Statonery artcles + + + Soes + + - - + + + + TV lcence fee + Tobacco + + + Coal + + + Telepone + + Meat + + - - Publc telepone + Blankets and seets + + + + + Transport, comm. - + - - - + + Wasng powder + + + Eggs + Electrcty + Wne + + + + + + Flowers + + + + + - Sugar + - Ceese + Ktcens - + + Frut + + Cameras + + + Gas dom. uses - - + + Refrgerator - + + Gas for eatng + + + Jewels + + + Toys + + + Dswaser - + + Newspapers - + + + Wasng macne - + + Kerosene ++ - - + + + Motocycles +++ - + Mlk + + - + Rado-TV sets + + + 27
Table 3 - Rankngs by dstrbutonal caracterstcs, e =1 1996 1987 1996 1987 Sugar 1 5 Statonery artcles 35 37 Eggs 2 3 Refrgerator 36 45 Electrcty 3 2 Publc telepone 37 43 Bread 4 1 Ktcens 38 52 Mlk 5 6 Gas for eatng 39 35 Poultry 6 7 Newspapers, magazne 40 34 Oter meat 7 9 Oter clotes 41 44 Coffee and tea 8 8 Jewels 42 60 Pasta 9 4 Cameras and oter 43 50 Fs 10 10 Hardresser 44 36 Wasng powder 11 15 Bkes 45 61 Meat 12 13 Cutlery and pots 46 41 Telepone 13 20 Flowers 47 40 Ceese 14 16 Toys 48 42 Fres vegetables 15 14 Laundry 49 39 Edble ol 16 11 Bus and taxs 50 38 Frut 17 18 Rado-TV sets 51 46 Oter cereals 18 19 Motocycles 52 62 Butter and margarne 19 23 Clotes 53 49 Gas dom. uses 20 12 Dswaser 54 55 Cured pork meat 21 22 Meals out 55 47 Tobacco 22 21 Books 56 54 Mneral Water 23 25 Oter recreatonal 57 51 Medcne 24 27 Oter furnture 58 58 Oter foodstuffs 25 24 Healt 59 64 Wasng macne 26 30 Recreatonal goods 60 53 Blankets and seets 27 48 Oter goods 61 56 Wne 28 17 Leater artcles 62 63 Soes 29 31 Oter ome 63 65 TV lcence fee 30 33 Kerosene 64 57 Personal ygene 31 29 Sport artcles 65 59 Oter drnks 32 28 Hotels and motels 66 66 Gasolne 33 26 Transport, comm. 67 67 Coal 34 32 Furnture 68 68 Source: AWARETAX 28
Table 4 - Decomposton of te cange of te dstrbutonal caracterstcs (dc), 1987-1996, e=1 dc σ2 λ dc σ2 λ Mneral Water 0.0927 0.0474 0.0453 Laundry 0.0542 0.0602-0.0059 Hotels and motels 0.0275 0.0761-0.0486 Fres vegetables 0.0372 0.0340 0.0033 Oter foodstuffs 0.0434 0.0445-0.0011 Books 0.0791 0.0695 0.0097 Oter drnks 0.0741 0.0514 0.0227 Medcne 0.1074 0.0507 0.0567 Oter meat 0.0490 0.0276 0.0214 Furnture 0.0539 0.0729-0.0190 Oter ome 0.1227 0.0745 0.0482 Edble ol 0.0044 0.0294-0.0250 Oter cereals 0.0392 0.0368 0.0024 Bread -0.0074 0.0142-0.0216 Butter and margarne 0.0582 0.0402 0.0180 Hardresser 0.0507 0.0567-0.0059 Oter furnture 0.0855 0.0706 0.0149 Pasta -0.0330 0.0155-0.0485 Oter clotes 0.1277 0.0636 0.0641 Meals out 0.0544 0.0657-0.0113 Clotes 0.0893 0.0666 0.0227 Cutlery and pots 0.0860 0.0619 0.0241 Personal ygene 0.0825 0.0521 0.0304 Fs 0.0484 0.0291 0.0193 Sport artcles 0.0150 0.0710-0.0560 Poultry 0.0458 0.0250 0.0208 Bus and taxs 0.0303 0.0587-0.0284 Oter goods 0.0485 0.0703-0.0218 Gasolne 0.0627 0.0501 0.0126 Recreatonal goods 0.0537 0.0691-0.0154 Bkes 0.2006 0.0714 0.1293 Oter recreatonal 0.0578 0.0679-0.0102 Leater artcles 0.0707 0.0718-0.0011 Cured pork meat 0.0467 0.0394 0.0073 Coffee and tea 0.0435 0.0270 0.0165 Healt 0.1222 0.0733 0.0489 Statonery artcles 0.0975 0.0571 0.0404 Soes 0.0960 0.0533 0.0427 TV lcence fee 0.1040 0.0544 0.0496 Tobacco 0.0434 0.0391 0.0043 Coal 0.0820 0.0542 0.0278 Telepone 0.0765 0.0390 0.0376 Meat 0.0406 0.0308 0.0098 Publc telepone 0.1317 0.0624 0.0693 Blankets and seets 0.2245 0.0657 0.1589 Transport, comm. 0.0448 0.0761-0.0313 Wasng powder 0.0769 0.0341 0.0428 Eggs 0.0362 0.0152 0.0210 Electrcty 0.0287 0.0142 0.0145 Wne -0.0281 0.0359-0.0639 Flowers 0.0820 0.0618 0.0202 Sugar 0.0640 0.0185 0.0455 Ceese 0.0505 0.0352 0.0153 Ktcens 0.1869 0.0680 0.1189 Frut 0.0498 0.0366 0.0133 Cameras and oter 0.1662 0.0677 0.0985 Gas dom. uses -0.0072 0.0306-0.0378 Refrgerator 0.1631 0.0643 0.0987 Gas for eatng 0.0646 0.0559 0.0086 Jewels 0.2130 0.0710 0.1420 Toys 0.0795 0.0620 0.0175 Dswaser 0.1158 0.0698 0.0459 Newspapers, magaz. 0.0523 0.0550-0.0027 Wasng macne 0.1054 0.0528 0.0526 Kerosene 0.0389 0.0703-0.0314 Motocycles 0.1652 0.0715 0.0937 Mlk 0.0361 0.0233 0.0128 Rado-TV sets 0.0850 0.0647 0.0203 Source: AWARETAX 29
Table 5 - Welfare canges VAT adjustment HES e=0.5 e=1 e=2 1988/87 1987-0.0164-0.0278-0.0387 1989/88 1988-0.0345-0.0674-0.1319 1990/89 1989 0.0172 0.0281 0.0402 1991/90 1990-0.0150-0.0228-0.0234 1992/91 1991-0.0055-0.0074-0.0033 1993/92 1992-0.0997-0.1752-0.2736 1994/93 1993 0.0121 0.0225 0.0415 1995/94 1994-0.0134-0.0177-0.0118 1996/95 1995 0.0000 0.0000 0.0000 1997/96 1996 0.0505 0.1036 0.2101 1988/87 1996-0.0088-0.0149-0.0212 1989/88 1996-0.0240-0.0466-0.0879 1990/89 1996-0.0001-0.0013-0.0058 1991/90 1996-0.0286-0.0470-0.0629 1992/91 1996-0.0086-0.0131-0.0134 1993/92 1996-0.0593-0.1050-0.1696 1994/93 1996-0.0023-0.0024 0.0024 1995/94 1996 0.0118 0.0253 0.0528 1996/95 1996 0.0000 0.0000 0.0000 1997/96 1996 0.0505 0.1036 0.2101 Total -0.0694-0.1012-0.0955 Caned 1996-0.0879-0.1331-0.1432 Caned 1992-0.0979-0.1476-0.1502 Note: Bold = Sgnfcant at 5 per cent level; Italc = Sgnfcant at 10 per cent level Source: AWARETAX 30
Fgure 1 - Dstrbutonal caracterstcs n Italy, 1987 and 1996 Dstrbutonal caracterstcs 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.0 0.2 0.4 0.6 0.8 1.0 Cumulatve budget sare (ranked at e=1) e=05_87 e=1_87 e=2_87 e=05_96 e=1_96 e=2_96 Fgure 2 - Dstrbutonal caracterstcs n Italy, 1987-1996 Dstrbutonal caracterstcs 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.0 0.2 0.4 0.6 0.8 1.0 Budget sare 1996 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 31