indirect taxation in greece: evaluation and possible reform

International Tax and Public Finance, 10, 511–533, 2003c© 2003 Kluwer Academic Publishers. Printed in the Netherlands.

Indirect Taxation in Greece:Evaluation and Possible Reform

GEORGIA KAPLANOGLOU [email protected] of Cambridge, England; Economic Research Department, Bank of Greece, Athens, GR-102 50, Greece

DAVID MICHAEL NEWBERY [email protected] of Applied Economics, University of Cambridge, Cambridge, CB3 9DE, UK

Abstract

The paper assesses the distributional and efficiency/disincentive aspects of the Greek indirect tax system, whichprovides 60% of total tax revenue. The marginal welfare costs of broad commodity groups were computed toidentify welfare-improving directions of reform. The disincentive effects were estimated from marginal indi-rect tax rates using Household Expenditure Survey data. The indirect tax structure is shown to be unnecessar-ily complicated and inefficient, without achieving any redistributive goals. The UK indirect tax structure wasshown to be simpler, more equitable and more efficient to implement and administer when simulated on Greekconsumers.

Keywords: indirect tax reform, inequality, tax efficiency, disincentive effects, tax simulations

JEL Code: H21, H23, H31

1. Introduction

Greece is unusual among EU countries in the high share of indirect taxes in total tax revenue(60% compared to an average of around 40% for the OECD countries or the EU-15, OECD2001), and the considerable variation in individual indirect tax rates. This variability in taxrates results from the large number of different taxes levied often on the same good, and toa large extent survives despite joining the European Union and adopting the EU system ofValue Added Taxes (VAT). Perhaps more surprising, both the heavy emphasis on indirecttaxes and their variability have survived substantial tax reforms that have raised Greece’stax share in GDP from 24 to 37.4% from 1980 to 1999 (Ministry of Finance and Economy,2001), though the share of indirect taxes in total revenue has fallen from 70 to 60% over thatperiod. The introduction of VAT in the late 1980s simplified indirect taxes somewhat, andsince then there have been further but modest simplifications. Greece is now contemplatinga major reform of its entire tax system, and thus needs to consider whether these particularfeatures of the current tax system are the result of historical accident, political expediency,or a considered view of the social objectives of the Greek people that should be defendedand retained.

512 KAPLANOGLOU AND NEWBERY

The balance between indirect and direct taxation is not the direct concern of this paper,and will depend on the efficacy of direct and indirect tax collection (Newbery, 1997), thecosts of tax collection (Slemrod, 1990), the extent to which indirect taxes are considered anefficient way of taxing tourists (compared, say, to hotel taxes and airport charges), and theextent to which indirect taxes are less visible than direct taxes, and hence less objectionableto swing (middle class) voters. This paper addresses the second feature of the indirect taxsystem, and the extent to which the high degree of variability in tax rates can be defendedon social welfare grounds. Not only is this timely, given the current interest in major taxreform in Greece, but, despite the overwhelming importance of indirect taxes in Greece,few studies are available and those are either fragmentary (Grevenitis and Sapounas, 1988;Georgakopoulos, 1989), or outdated (Karageorgas, 1973; Provopoulos, 1979).

As with all such studies, we note that the component taxes within a properly designedtax system can only be assessed within the context of the whole system, including theexpenditure side of the budget. If direct taxation is reasonably effective, indirect taxes canbe left to make minor improvements to distributional and efficiency goals (e.g. addressingexternalities). A well-designed system of social expenditure (on health, education, andwelfare) and transfers (to the poor, as child support, etc.) is a better method of addressingdistributional concerns than most taxes. It follows that uniform VAT combined with well-designed expenditures can be part of an efficient yet egalitarian tax and expenditure system.Criticising indirect taxes because they are not adequately progressive or redistributive isunreasonable unless it can be shown that such distributional goals cannot be better achievedby direct taxes, transfers, or public expenditures.

To see how important this is for the reform of indirect taxation, Section 2 derives formulaefor the marginal social cost of raising revenue from particular indirect taxes. If these are notequal across goods, then the same revenue can be raised at lower social cost by raising taxeson the good with the lower marginal social cost and lowering the tax on one with a highersocial marginal cost. If the only possible reform consists of revenue-neutral adjustmentsof indirect tax rates, then this approach (of marginal tax reform analysis) has some merit,though it is not without its own problems, as we shall see. If distributional considerations arenot too important, then taxing inelastically demanded goods more heavily than elasticallydemanded goods appears defensible (but conversely if distributional considerations are veryimportant, and are reflected sufficiently in different consumption patterns of the rich andpoor).

In order to estimate these elasticities, moderately strong restrictions have to be placedon the form of consumer demands, and these, together with the choice of tax instruments,can largely predetermine the outcome. Thus weak separability between goods and leisurecombined with Engel curves which are linear and have the same intercept, the ability totax all goods, and the ability to redistribute tax revenue in the form of lump-sum grants,implies that indirect taxes should be made uniform (Atkinson, 1977; Deaton, 1979, 1981).Even if the Engel intercepts vary with observable household characteristics, uniformity isstill optimal provided that the government can apply an optimal system of family grants(Deaton and Stern, 1986). If non-linear income taxes can also be optimally set, then all weneed for the optimality of uniform indirect taxes is weak separability of goods and labourand individuals differing only in the wage rate (Atkinson and Stiglitz, 1976; Deaton, 1981;

INDIRECT TAXATION IN GREECE 513

Stern, 1987). Uniform commodity taxes then deliver the consumption bundle to workers atleast cost (by creating minimum relative price distortion), encouraging them to work harderby confining distortions to the supply of labour (effort).

The case for non-uniform commodity taxation then rests on the extra leverage suchdifferentiation gives over labour supply—there is then a case for heavier taxes on goodsthat are complementary to leisure (golf clubs?), and lighter taxes on goods complementaryto work (commuting?). As it is hard to think of such taxes that do not create unreasonableextra distortionary costs, and whose elasticities can be estimated sufficiently reliably, thereis a presumption for uniformity.

It is therefore likely to be difficult for Greece to defend its very non-uniform indirect taxstructure. One possible defence is that there are very few other instruments that can be usedto improve the overall tax and expenditure structure. For example, it is clearly perceivedto be difficult to achieve all the desired redistribution from personal income taxes (whichcollect remarkably little revenue from some segments of the population that are clearlyrelatively well off, such as the professional classes and the self-employed). We observethat Greece finally balanced its budget and met the Maastricht criteria exclusively throughsubstantial increases in the share of GDP taken in taxes, so that the expenditure side of thebudget is also severely constrained.

Can indirect taxes offer any serious prospect of relaxing these constraints, and contribut-ing anything to distributional equity without compromising either revenue or efficiency(i.e. the total amount of after-tax income available)? If not, then the current variability oftaxes reflects a failure, despite repeated opportunities, to further simplify indirect taxes andmake them more uniform (which should help compliance, reduce collection costs and movetowards the preferred EU pattern of taxation).

This paper therefore examines the efficiency as well as distributional properties of thecurrent tax system, and inquires how it may be improved.1 This is done in three ways. Thestandard theory of marginal tax reform (Feldstein, 1972; Ahmad and Stern, 1984; Stern,1987) can, given sufficient information about aggregate demand elasticities, indicate whichindirect taxes should be reduced and which increased (assuming, as noted above, no othertaxes can be changed). We find little support for the current pattern of tax variability evengranted the restrictive assumptions needed to justify this approach.

Marginal tax reform analysis has the attraction of combining both efficiency and distri-butional considerations in a theoretically consistent utilitarian framework (which it shareswith standard welfare economics). It has, however, a number of limitations, in addition tothose noted above. The most serious, and the one that bedevils all arguments for differen-tial indirect tax rates, is that it relies on good estimates of own and cross-price elasticities(Deaton, 1987). This restricts the analysis to fairly high level of aggregation, which is likelyto conceal important detail. Further, the analysis does not indicate by how much to changetaxes, nor even which changes would deliver the best result.

We address this by examining whether the aggregate marginal indirect tax rate variesacross the income (or expenditure) distribution in a logical and defensible way, followingthe approach of Newbery and Revesz (2000). We find that once we correct for householdtype, there is essentially no variation in the marginal tax rate across the income distribution,and hence the individual variability in indirect tax rates is at best only redistributing income


between different household types at each income level. It has no advantage over moreuniform indirect taxes for taxing the rich more heavily than the poor. This provides strongevidence that the guiding principle of uniformity of VAT rates within a complete tax andbenefit system holds even within the more restrictive exercise conducted here.

Finally, we criticise the current structure of indirect taxes by examining the impact ofa revenue-neutral tax reform of the entire system. Section 4 investigates the distributionaland efficiency effects of simulating the indirect tax system of the United Kingdom onGreek consumers. This simulation exercise is interesting both in its own right as a thoughtexperiment and as an attempt to use as a benchmark a tax system which is already inoperation elsewhere in the EU and seems to be politically sustainable. Furthermore, it canbe shown that the UK indirect tax structure can largely be rationalised in terms of optimalcommodity taxation principles (Kaplanoglou, 1999). These advantages of reasonablenessand political relevance in using existing tax systems as comparison benchmarks have alreadybeen recognised. Thus Gale (1997) and Keen (1997) discuss the relevance of aspects ofBritish tax policy for the US. This is, however, the first attempt to put the idea into effectwithin a microsimulation analytical framework.

2. Marginal Tax Reform Analysis

Optimal tax theory starts with a description of the economy and the objectives of thegovernment, and characterises the optimal level of each of the available tax instruments,given the information assumed available to the government. Solving for the set of optimaltaxes requires the ability to predict how the economy behaves not just around its presentequilibrium but at all other possible equilibria, in order to identify the optimum. Marginaltax reform analysis also requires an adequate description of the economy and governmentobjectives (summarised in a social welfare function), but considers the more modest andinformationally less demanding task of identifying the direction in which taxes can bechanged to increase social welfare. That requires only local information about economicresponses to tax changes, in contrast to the global information required for determiningoptimal taxes (Stern, 1987).

2.1. The Theory of Marginal Tax Reform

Suppose that the government ranks distributional outcomes according to a Utilitarian orBenthamite social welfare function W (V 1, . . . , V h, . . . , V H ),2 where agent h enjoys utilityV h = V h(yh + g, q) that depends on net income yh , plus any transfers, g, and the vectorof consumer prices, q = p + t, where p is the vector of producer prices and t is the vectorof indirect taxes. All we need to assume about the description of the economy is that theproducer prices do not vary with indirect taxes (see note 3 below). Consider the impact onsocial welfare of a change in consumer price qi , caused by a change in the tax, ti :

∂W

∂qi=

∑h

∂W

∂V h· ∂V h

∂qi= −

∑h

βh xhi , βh ≡ ∂W

∂V h· ∂V h

∂g, (1)


where βh is the social marginal utility of transferring 1 Euro to agent h, xhi is consumption

of good i by agent h, and the last equality in equation (1) makes use of Roy’s identity.The impact of the tax change will thus depend on both the level of consumption and itsdistribution amongst the population. It is convenient to isolate these two effects by definingdi , Feldstein’s (1972) distributional characteristic of good i :

di ≡∑

h βh xhi

β Xi, Xi ≡

∑h

xhi , β ≡ 1

H

∑h

βh, (2)

where Xi is aggregate consumption of i , β is the average over the H agents of βh, so thatdi is a measure of how concentrated the consumption of good i is on the socially deserving(those with high social marginal values of consumption, βh). The social welfare impact ofa price change is then

∂W

∂qi= −βdi Xi . (3)

The standard approach to determining desirable directions of tax reform is to computethe marginal social cost of raising one Euro of revenue by increasing the indirect tax on thei th good. On the standard assumption that the tax is 100% shifted forward to the final price(this is the force of assuming that producer prices do not vary with taxes),3 the impact onsocial welfare is ∂W/∂ti , while the extra revenue collected is ∂ R/∂ti , where R = �ti Xi istotal tax revenue. The extra revenue collected is then

∂ R

∂ti= Xi +

∑k

tk∂ Xk

∂ti= Xi

(1 +

∑k

qk Xk

qi Xi· tk

qk· qi∂ Xk

Xk∂qi

)

= Xi

(1 +

∑k

ωkτkεki

ωi

), (4)

where ωi is the budget share of good i , qi Xi/�qi Xi , τi = ti/qi is the ratio of the indirecttax to the tax-inclusive price, and εki is the cross-price elasticity of good k with respect tothe price of i (or the tax ti ).

In order to collect 1 Euro more revenue, the tax rate will have to be increased by 1/∂ R/∂ti ,so the marginal social cost of raising a Euro of tax is

λi = −∂W/∂ti∂ R/∂ti

= di Xi

Xi(1 + ∑

kωkτkεki

ωi

) . (5)

This expression is rather inelegant, and its inverse can be more readily examined. Definethe marginal social tax productivity of tax i as the extra revenue collected for a unit increasein social pain, −∂W/∂ti , θi :

θi ≡ − ∂ R/∂ti∂W/∂ti

= 1

λi= fi

(1 +

∑k

τkωkεki

ωi

)= fi

(1 + τiεi i +

∑k �=i

τkωkεki

ωi

), (6)

where fi =1/di is the inverse of the distributional characteristic of good i , and may betermed the tax appeal of good i , the extent to which it is well targeted on the consumptionof the better off.


Equation (6) can now be interpreted. The marginal social productivity of taxing good iincreases with the tax appeal of the good, is reduced the higher is the product of the taxrate on the good with the (absolute) value of the own-price elasticity, εi i , and is furtherinfluenced by other taxes and cross-price elasticities, εki . High own-price elasticities signalthat the distortionary cost of raising revenue by tax increases as the tax rate rises, but thismay be counterbalanced by the equity considerations represented by the good’s tax appeal,fi .

Unfortunately, the direction of reform depends to a considerable extent on the valuesof the elasticities, εki , that are hard to estimate. Deaton (1987) argues that their values areoften largely predetermined by the choice of demand system in any econometric estimation.For example, the Linear Expenditure System (LES) has zero (uncompensated) cross-priceelasticities, so that (6) then reduces to

θi = fi (1 + τiεi i ). (6a)

If we ignore distributional considerations, this is the familiar Ramsey rule that argues thatindirect taxes should be inversely proportional to the (absolute) value of the price elasticity.Goods that are price inelastic should be more heavily taxed. If instead of assuming additiveseparability (of which the LES is a special case), we assume indirect additivity (Houthakker,1960; Deaton, 1987), then (6) reduces to

θi = fi (A − τi (εi − ξ )), A ≡ 1 − (1 + ξ )τ +∑

j

ω jε jτ j , (6b)

and where εi > ξ is the expenditure elasticity of good i , and ξ is a parameter to be estimated.In this case, ignoring distributional considerations, indirect taxes should be inversely pro-portional to the adjusted expenditure elasticity, (εi − ξ ), and income inelastic goods shouldbe more heavily taxed. There is a close relation between these two formula, as Deatonand Muellbauer (1980a, p. 139) point out, for additivity implies Pigou’s Law, namely thatprice elasticities are likely to be roughly proportional to income elasticities in the case ofadditivity (and a reasonable degree of commodity disaggregation). It also reminds us thatgoods that on efficiency grounds are attractive for higher indirect taxes (low price or incomeelasticities of demand) are likely to be unattractive on distributional grounds (low incomeelasticities, i.e. necessities consumed by the poor). These two opposing factors are capturedby fi and the various elasticities, as in (6a) and (6b).

The conflict between equity and efficiency can be demonstrated for the special case ofconstant expenditure elasticities εi , and a log-normal distribution of expenditure (whichis a reasonable approximation to the actual expenditure distribution in the case of mostcountries). The tax appeal of good i can be written as EβExi/{E[βxi ]), where Exi is theexpected or average value of xi over the whole population. The trade-off between equity andwelfare requires a method to calculate the welfare weights βh that allows us to experimentwith differing value judgements about the importance of distributional considerations. Thesimplest and most easily parameterised approach is given by the additive iso-elastic socialwelfare function, W = �uh , where uh = (ch)1−ν/(1−ν), (ν �= 1), uh =log ch , (ν = 1), ch

is real consumption per equivalent adult, and ν is Atkinson’s (1970) coefficient of inequalityaversion.


For this social welfare function, βh = (ch)−ν . The higher is the value of ν, the moreconcerned the government is with inequality. Thus, if ν = 1, transferring one Euro tosomeone at double the living standard of another has a social value of only one-half that ofthe reference person. A value of zero indicates no inequality aversion with βh the same forall households, while at the other extreme, a value of 5 approaches the Rawlsian “maxi-min”principle in which only the impact on the poorest counts (where 1 Euro is worth 25 = 32times the value of 1 Euro to someone with twice that income). Then if the log of expenditureis normally distributed with a mean µ and variance σ ,

xhi = ai (c

h)εi , βh = (ch)−v, Eca = exp

(αµ + 1

2α2σ 2

).

The last equality comes from the properties of the lognormal distribution (Newbery andStiglitz, 1981, p. 89). It follows that

fi = (Ec−vEai cεi )/(Ec−vai c

εi ) = exp(−εiνσ 2)

and the values of fi are a simple function of, and will be correlated with, the expenditureelasticities. A high value of εi will give a high value of fi , but from (6b) on efficiencygrounds should have a low tax.

2.2. Estimation

In order to apply this theoretical framework to the Greek case, we have to combine in-formation from different sources. Household expenditure data are readily available fromcross-section microdata of the 1987/8 Household Expenditure Survey (HES) (NSSG, 1994).The sample consists of 6,489 households, is representative of the population and providesinformation on household expenditure on 293 categories of goods and services.

Demand derivatives and elasticities can be obtained from estimates of aggregate demandsystems. Since there is effectively no price variation in the HES data, we draw on two recenttime series estimates of complete demand systems for Greece.4 The first, by Andrikopoulos,Brox and Georgakopoulos (1992), contains an estimation of the Almost Ideal DemandSystem (AIDS) for Greece for thirteen commodity groups based on National Accounts timeseries data for the period 1958–1986. The second study was conducted by Alogoskoufis,Misoulis and Karavitis (1996) and involves again an estimation of the AIDS for Greece fornine commodity groups using annual time series data for the period 1970–1990.

The marginal social tax productivities, θi , are calculated for both sets of elasticity es-timates from the two studies, in order to test for the sensitivity of results to the differentdemand derivative estimates. The number of commodity groups handled in this frameworkis therefore restricted to the level of commodity aggregation of these two studies. The ma-trices of price elasticities have been adjusted to satisfy the 1987/8 HES vector of shares ofconsumer expenditure. This is done as in Ahmad and Stern (1984), where adjusted elas-ticities ε∗ are chosen to minimise the distance, S, between the original and the adjusted


elasticity matrices, where S ≡ ε∗ − ε and we minimise:

Min|S| =√∑

i j

(ε∗i j − εi j )2 s.t.

∑i

wiε∗i j = −w j (7)

where w is the vector of the HES shares of consumer expenditure.Table 1 presents the values of θi , as well as their components as presented in equation

(6), of different commodity groups and the corresponding ranks, for five different degreesof inequality aversion (ν = 0, 0.1, 0.5, 2 or 5). The interpretation of ranking followedhere is that goods ranked lower (say 12) are preferred to goods ranked higher (say 1) ascandidates for additional taxation—i.e. low numbers mean lower taxes. In the same tablewe present the tax appeal values fi (the inverse of the distributional characteristics, di ) ofthe commodity groups and the corresponding ranks for different values of the inequalityaversion parameter. Goods with high tax appeal are candidates for higher tax rates. If we donot care about equity (ν = 0), they are equal to one. The fi s indicate directions of reformfor a government that is willing to sacrifice efficiency for equity.

From Table 1 is apparent that the existing tax structure is not optimal for any of the levelsof ν chosen, since the θi s differ across commodities. Directions of tax reform are moderatelysensitive to inequality aversion. If government cares little or nothing about redistribution(ν = 0 to 0.5), one would propose raising taxes on communication, health, education, andpossibly furniture and food, and decreasing taxes on recreation (including durables relatedto entertainment such as TV sets, hi-fi stereo systems), transport, alcohol and housing,using the elasticities by Andrikopoulos, Brox and Georgakopoulos (1992). The results arenot very different using the elasticities from Alogoskoufis, Misoulis and Karavitis (1996),taking into account the difference in the number of commodity groups. For the latter setof elasticities, for example, taxes on communication, furniture and possibly food are to beraised, while taxes on alcohol, tobacco, housing and transport should be lowered.

Inequality aversion affects the choice of tax reforms. Distributionally sensitive com-modities like food, communication and heating move from moderately high number ranks(indicating that taxes should perhaps be raised) for low values of inequality aversion to lowernumber ranks (lower taxes). Commodity groups like transport, clothing and alcohol thathave low ranks on efficiency considerations (ν = 0), become more attractive candidates foradditional taxation when the government strongly cares about redistribution. Some goods,such as tobacco and housing (which have low ranks), and education (high rank), hardlychange their position regardless of how averse to inequality the government is. The resultsseem robust to the set of elasticities used, taking into account the differences in commoditygroups.

As the level of inequality aversion increases, the ranking of the θi ’s tends to approximatethe ranking of tax appeal (see Table 1). Closer inspection of the table, however, suggests thatin general the rankings of the θi s are dominated by the efficiency element. The introductionof distributional considerations at low levels of ν does not significantly affect the ranking ofthe θi ’s and it is only at high values of ν that significant changes in the ranking occur. Thisresult supports the finding by Decoster and Schokkaert (1989), who demonstrate a similarresult for Belgium. The authors conclude that this indicates that consumption patterns arenot sufficiently differentiated to allow an important redistributive role to indirect taxes. Part

INDIRECT TAXATION IN GREECE 519Ta

ble

1.E

ffici

ency

and

equi

typa

ram

eter

sfo

rva

ryin

gva

lues

ofin

equa

lity

aver

sion

.

f i(a

):Ta

xap

peal

for

vari

ous

θ i(a

)fo

rva

riou

sin

equa

lity

aver

sion

(v)

ineq

ualit

yav

ersi

on(v

)(r

ank

inita

lics)

(ran

kin

italic

s)(6

)(7

)=

(3)∗ (

6)

Goo

d

Bud

get

shar

e(1

)

Ran

kef

ficie

ncy

(2)

θ i(f

orv

=0)

(3)=

1+

(4)+

(5)

t ie i

i

(4)

Sum

of(t

kw

ke k

i/w

i)fo

ri�=

k(5

)v

=0.

10.

52

5v

=0.

10.

52

5

And

riko

poul

os,B

rox

and

Geo

rgak

opou

los

(199

2)R

ecre

atio

n/en

tert

ainm

ent

3.6%

10.

352

−0.2

99−0

.349

1.01

101.

0510

1.18

91.

4811

0.43

10.

441

0.50

10.

637

Alc

ohol

0.7%

20.

429

−0.0

61−0

.510

1.01

81.

038

1.05

60.

907

0.52

20.

532

0.54

30.

475

Tra

nspo

rt10

.3%

30.

574

−0.2

16−0

.210

1.01

111.

0611

1.27

111.

4510

0.69

30.

735

0.87

81.

019

Hou

sing

5.9%

40.

616

−0.0

34−0

.350

0.99

40.

944

0.74

40.

424

0.73

40.

703

0.55

50.

312

Toba

cco

2.6%

50.

669

−0.2

22−0

.110

0.98

20.

912

0.66

20.

392

0.78

50.

734

0.53

20.

313

Hea

ting/

light

ing

4.3%

60.

755

−0.0

12−0

.232

0.98

10.

901

0.60

10.

161

0.88

60.

816

0.54

40.

151

Clo

thin

g14

.3%

70.

866

−0.1

590.

025

1.01

91.

059

1.21

101.

9712

1.05

71.

099

1.26

92.

0712

Oth

ergo

ods

and

serv

ices

7.4%

80.

872

−0.0

55−0

.072

1.02

131.

0813

1.32

132.

0213

1.06

91.

1210

1.37

112.

1313

Food

29.9

%9

0.89

7−0

.005

−0.0

980.

993

0.94

30.

743

0.40

31.

068

1.01

70.

806

0.43

4C

omm

unic

atio

n1.

4%10

0.92

3−0

.053

−0.0

240.

995

0.95

50.

785

0.56

51.

0910

1.05

80.

877

0.62

6Fu

rnitu

re9.

1%11

1.01

4−0

.011

0.02

51.

017

1.03

71.

087

0.70

61.

2211

1.24

111.

3210

0.86

8H

ealth

7.3%

121.

070

−0.0

790.

148

1.02

121.

0812

1.28

121.

208

1.30

121.

3813

1.64

131.

5510

Edu

catio

n3.

1%13

1.11

1−0

.010

0.12

11.

006

1.03

61.

128

1.23

91.

3413

1.36

121.

4912

1.66

11

Wei

ghte

dav

erag

e10

0.0%

0.83

6−0

.078

−0.0

861.

001.

001.

001.

001.

001.

001.

001.

00

Alo

gosk

oufis

,Mis

ouli

san

dK

arav

itis

(199

6)A

lcoh

ol0.

7%1

0.36

3−0

.269

−0.3

681.

017

1.03

71.

056

0.93

70.

451

0.46

10.

462

0.41

4To

bacc

o2.

6%2

0.55

6−0

.317

−0.1

280.

982

0.91

20.

662

0.39

20.

672

0.61

20.

451

0.27

2T

rans

port

10.3

%3

0.69

0−0

.239

−0.0

711.

018

1.06

81.

279

1.49

80.

853

0.89

51.

077

1.24

8H

ousi

ng5.

9%4

0.76

7−0

.017

−0.2

160.

994

0.94

40.

754

0.43

40.

935

0.88

40.

694

0.40

3H

eatin

g/lig

htin

g4.

3%5

0.76

9−0

.102

−0.1

290.

981

0.90

10.

601

0.17

10.

924

0.84

30.

563

0.16

1Fo

od29

.9%

60.

827

−0.0

11−0

.162

0.99

30.

943

0.75

30.

413

1.00

60.

946

0.75

50.

415

Com

mun

icat

ion

1.4%

70.

840

−0.0

68−0

.092

0.99

50.

955

0.78

50.

575

1.02

70.

987

0.80

60.

586

Oth

ergo

ods

and

serv

ices

35.9

%8

0.86

0−0

.048

−0.0

921.

019

1.06

91.

238

1.68

91.

068

1.11

81.

299

1.75

9Fu

rnitu

re9.

1%9

0.94

2−0

.079

0.02

11.

016

1.03

61.

097

0.72

61.

169

1.18

91.

248

0.82

7

Wei

ghte

dav

erag

e10

0.0%

0.81

9−0

.069

−0.1

121.

001.

001.

001.

001.

001.

001.

001.

00

Not

e:R

anke

dby

effic

ienc

y.L

owes

tsco

rem

erits

low

erta

x.(a

)D

ivid

edby

the

wei

ghte

dav

erag

eof

the

resp

ectiv

eco

lum

n.


of the problem is the high degree of aggregation forced by the need to estimate cross-priceelasticities, as this tends to obscure variations in consumption patterns (and also variationsin tax rates within the commodity groups). However, the conclusion that the indirect taxsystem is not particularly well-suited for redistribution is well taken, and was argued above.Equity considerations are better addressed using direct taxes and even more by transfers(via the expenditure side of the budget).

The marginal tax results presented in Table 1 should be further qualified before they can beused to guide policy. In addition to the theoretical limitations of the approach listed above, itis worth applying some elementary common sense to the prescriptions that appear to emergefrom the analysis. Thus tobacco, which on efficiency and equity grounds appears to requirelower taxes, is heavily taxed in most countries to discourage use of an addictive and harmfulproduct. Similarly alcohol is taxed to discourage consumption, though here the externalitiesare more obvious (road accidents, battered wives and children, etc.). Similarly, taxing educa-tion may appear similarly attractive, but ignores the merit good case of beneficial externali-ties (to children, in reduced crime, etc.) Transport fuels are frequently heavily taxed as a usercharge for the public road system, and the pure tax element (above this user charge element)is considerably smaller than shown in the table (Newbery, 1995a). Replacing the nominaltax rate with the amount above the user charge element would greatly weaken the argumentfor lowering that tax. Similarly, the apparently low tax on food which on efficiency groundsis a reason for some increase, ignores the heavy implicit tax on consumers caused by theCommon Agricultural Policy. What this suggests is that a simpler indirect tax system woulddistinguish between corrective or user charge excises (on alcohol, tobacco, fuel), and a smallnumber of VAT rates to deal with other goods (the lower or zero rate being retained for food,education and other goods which are deemed for various reasons to require lower taxes).

3. Disincentive Effects of Indirect Taxation

The inefficiency of a tax depends on its marginal rate (and on various elasticities), as thisdetermines its disincentive effect. That concept is familiar in the case of personal incometaxes, but applies just as much to indirect taxes as a whole. Here the appropriate conceptis the marginal indirect tax rate, defined as the extra indirect tax paid on additional goodspurchased when income (or expenditure) increases by one unit. If this marginal rate isconstant over the income or expenditure distribution, then there is no distributional case fordepartures from uniformity. In this section we explore the cross-sectional characteristics ofthe marginal indirect tax rate.

3.1. Two Models of Marginal Indirect Tax Rates

The marginal indirect tax rate is the weighted sum of indirect taxes on different goods,weighted by the marginal expenditure shares of the goods (for an application see Newbery,1995b; Newbery and Revesz, 2000). The marginal expenditure shares can be estimatedgiven some estimable form of individual preferences (for example, the linear expendituresystem or the AIDS). This is readily done by estimating the marginal indirect tax rates asa direct function of expenditure, given a suitable functional specification. Two alternative


formulations of the relationship are adopted, one corresponding to the Linear ExpenditureSystem, or LES (Stone, 1954) and the other to the Almost Ideal Demand System, or AIDS(Deaton and Muellbauer, 1980b). More precisely, according to the LES, Engel curves takethe form:

qi xhi = γi + αi

(yh − yh

0

)(8)

where, using the same notation as before, qi xhi is equivalent expenditure on good i , yh

is income (or total expenditure) of household h (by equivalent adult), yh0 is subsistence

expenditure, and �αi = 1. The formula for total tax collected from household h is:

T h =∑

i

τi qi xhi (9)

where T h is total taxes paid by household h (per equivalent adult) and τ i is the tax rate ongood i (as a fraction of the tax inclusive price). Substituting (8) into (9):

T h =∑

i

τiγi +∑

i

τiαi(yh − yh

0

)(10)

Differentiating (10) with respect to expenditure, yh, gives the marginal indirect tax rate,Bh = �τiαi , a constant that does not vary with the level of expenditure. If prices do notvary, AIDS gives the non-linear Working-Leser Engel curve (Working, 1943; Leser, 1963):

whi = qi x

hi

yh= αi + δi log yh ⇔ qi x

hi = αi yh + δi yh log yh (11)

where whi is the expenditure share of good i , and �αi = 1, �δi = 0. The tax equation is:

T h =∑

i

τi qi xhi =

∑i

τi (αi yh + δi yh log yh) = yh∑

i

τiαi + yh log yh∑

i

δiτi

(12)

Differentiating (12) by expenditure, yh, gives the marginal indirect tax rate as T h/yh +�δiτi , so that the value of �δiτi describes departures from the average.5

Marginal indirect tax rates were estimated on the basis of equations (10) and (12) acrossthe expenditure distribution, in a first stage not controlling, and in a second stage con-trolling, for demographics.6 More precisely, in the first stage the household expendituredistribution was split in quintiles of equivalent non-durable expenditure and equations (10)and (12) were estimated for each group. In the second stage, the equations were estimatedacross the total sample controlling for demographic and other status characteristics. Dueto multicollinearity problems, the latter were allowed to influence the intercept of the taxequation, but not its slope. The estimated equations appear in Table A1 in Appendix Aand allow the calculation of marginal indirect tax rates across the expenditure distributionholding demographic composition and other socio-economic characteristics constant. Inthe present context, marginal indirect tax rates were calculated on the basis of the estimatedequations for the average values of non-durable equivalent expenditure for each quintilefor two “representative” household types. Household type A is a married couple with twochildren living in Athens in own accommodation, where the head belongs to the 26–50 agegroup, is an employee in the private sector and has high school education. Household type B


is a childless couple living in own accommodation in Athens, the head is a pensioner over65 years old and has basic or no education. Note that the Working-Leser model produces amore realistic picture of the relation between paid taxes and expenditure, since it capturesnon-linearities in this relation. The LES, by forcing a linear relationship between taxes andexpenditure, in general produces higher and less reliable estimates of marginal tax ratesthan the Working-Leser model.

Figure 1 shows the variation in marginal tax rates across quintiles of non-durable ex-penditure for the total sample. This variation is non-negligible and statistically significant,in terms of the deviations from the values of the middle quintile. If, however, the socio-demographic structure is controlled for, marginal indirect tax rates become surprisinglyuniform (Figure 2). Most of the variation in average and marginal indirect tax rates seemsto be captured by demographic and socio-economic characteristics, rather than the expen-diture level itself—note in Table A1 that the estimated expenditure coefficient is positive

0%

5%

10%

15%

20%

25%

1 2 3 4 5

Quintiles of equivalent non-durable expenditure

Mar

gin

al t

ax r

ate

W -L (grossed-up) LES (grossed-up)

Figure 1. Marginal indirect tax rates—all households.

0%

3%

6%

9%

12%

15%

1stquintile(W-L)

2ndquintile(W-L)

3rdquintile(W-L)

4thquintile(W-L)

5thquintile(W-L)

LES

mar

gin

al t

ax r

ates

households type A households type B

Figure 2. Marginal indirect tax rates—households type A and B.


and significant, but very small in magnitude for the LES, and negative and not statisticallysignificant for the W-L model.

The main conclusion is that when the demographic and other status characteristics of thehouseholds are controlled for, the marginal indirect tax rates are fairly constant across theexpenditure distribution. Thus the actual tax system produces the same marginal indirecttax rates as a uniform VAT, but with more distortion because different taxes are charged atdifferent rates. These distortions have no obvious logic, providing a strong argument foruniformity.

4. Simulating the UK Indirect Tax System on Greek Consumers

4.1. The Simulation Methodology

The UK indirect tax structure is considerably simpler than the Greek one. It involves astandard VAT rate applicable to most expenditure items, a lower rate (mainly for domesticenergy), with certain goods zero-rated (for example food, children’s clothing, medicines,books, newspapers and transport) or exempt, coupled with special excises on tobacco,alcohol, petrol and diesel.7 Table 2 gives the tax rates for this system that give the currentGreek tax revenue.

Table 2. Implications of the tax reform for the structure of indirect tax revenue.

UK system UK systemGreek system (constant expenditure) (constant quantity)

Average Tax shares Average Tax shares Average Tax sharestax rate (%) (%) tax rate (%) (%) tax rate (%) (%)

Food 6.0 11.9 5.0 9.9 5.0 9.8

Alcohol 33.6 1.5 53.2 2.4 50.1 3.0

Tobacco 67.6 11.8 78.9 13.8 75.9 17.9

Clothing/footwear 13.8 13.3 14.6 14.1 13.4 12.7

Housing 10.8 6.3 8.9 5.2 9.4 5.1

Central heating 44.0 4.4 0.0 0.0 0.0 0.0

Household goods 13.5 8.2 18.2 11.0 16.1 10.0

Medical 1.0 0.3 0.0 0.0 0.0 0.0

Personal care 12.1 1.7 18.2 2.6 16.1 2.4

Education 0.5 0.1 1.6 0.2 1.6 0.2

Recreation 15.6 5.2 14.3 4.7 12.6 4.1

Transport 44.0 30.3 36.9 26.1 40.3 23.2

Communication 13.5 1.2 17.9 1.6 15.8 1.5

Other 7.6 3.8 16.7 8.4 18.2 10.1

Total 14.9 100 14.9 100 15.1 100


In the absence of own- and cross-price elasticities estimated for a sufficiently detailedlevel of commodity disaggregation, two extremes of behavioural response to the resultingchange in the retail price of commodities can be modelled. The first assumes that purchasedquantities remain constant, corresponding to zero own-price elasticities. Final expendi-tures may change and hence the household budget constraint might be violated (ignoringborrowing), but this describes the short-run impact of the tax reform. The alternative sce-nario assumes constant expenditures, corresponding to an own-price elasticity of −1 as inRedmond (1995). In both cases no cross-price effects are modelled. These extremes providea ‘confidence interval’ for the results.

4.2. Distributional and Disincentive Effects

The results are presented in Table 2. As expected, revenues from food, housing, heatingfuel and transport have substantially declined. This fall is counterbalanced by an increasein revenues from alcohol, tobacco, household durables and ‘other goods’. The reform hasbeneficial distributional effects, since the distribution of tax gains and losses across decilesof the population is progressive regardless of the assumptions on behavioural responses.Losers are concentrated in the top deciles, with small proportions of large losers in thebottom deciles, though there are some exceptions to the norm in each decile of expenditure(see Figure 3). The progressive impact can be explained by the zero or low rating of food,medicines, domestic energy and domestic fuel, and higher rates on eating out, householddurables and personal care. Their progressive effect is not offset by the regressive effect ofincreasing tax rates on tobacco and the large decrease in taxes on cars and their use.

0%

5%

10%

15%

20%

25%

30%

35%

40%

Per

cen

tag

e

>+

3

+1

to +

3

+0.

4 to

+1

-0.4

to +

0.4

-1 to

-0.

4

-3 to

-1

<-3

Dec

ile 1

Dec

ile 2

Dec

ile 3

Dec

ile 4

Dec

ile 5

Dec

ile 6

Dec

ile 7

Dec

ile 8

Dec

ile 9

Dec

ile 1

0

Gains and losses

Figure 3. Proportions of each decile gaining/losing certain amounts, in percentages of total initial expenditure(constant expenditure assumption).


0,00

0,10

0,20

0,30

0,40

0,50

0,60

0,70

0,80

0,90

1,00

0,00 0,20 0,40 0,60 0,80 1,00

0,000

0,005

0,010

0,015

Total expend.after UK taxes

Total expend.after Greektaxes

Difference(UK-Greek)RHS scale

Figure 4. Lorenz curves of expenditure distributions under Greek and UK taxes.

The positive equity impact on the expenditure distribution is demonstrated by comparingthe relevant Lorenz curves in Figure 4. The Lorenz curve post-reform lies closer to the45◦ line of equality and actually dominates the initial curve of after-Greek tax expenditure,indicating clearly that the UK indirect tax system results in a more equitable after-taxdistribution than the Greek one. The same conclusion is reached by comparing distributionalcharacteristics of the 293 commodities and their respective tax rates under the Greek andthe UK system. The relevant correlation coefficients for various values of the inequalityaversion parameter are negative and statistically significant only under the UK tax simulationexperiment, suggesting that replacing the Greek with the UK tax system would representan improvement in the targeting of taxes in favour of the socially deserving.

The efficiency impact of the UK tax simulation can be judged by estimating marginalindirect tax rates as in Section 3. Results are presented for the ‘constant expenditure’scenario, but are qualitatively the same for the alternative scenario of ‘constant quantity’.Figure 5 shows the marginal indirect tax rates for the two representative household types of

0%

5%

10%

15%

20%

1stquintile(W-L)

2ndquintile(W-L)

3rdquintile(W-L)

4thquintile(W-L)

5thquintile(W-L)

LES

mar

gin

al t

ax r

ates

households type A households type B

Figure 5. Marginal indirect tax rates in UK tax simulation experiment—households type A and B.


Section 3 to be progressive across the income distribution, a result supported by econometricanalysis. Whether this pattern is optimal cannot be established given the current state oftheoretical knowledge and data.

4.3. Lessons from the UK Tax Simulation Experiment

The UK simulation experiment shows how one country’s rather successful policy experiencecan be useful for another country. The UK tax structure displays several features whichenhance its progressivity (though adding complexity), producing an unambiguously moreequitable after-tax distribution, as the relative gains of such a tax reform accrue in largerproportions to poorer households. Furthermore, efficiency arguments based on marginalindirect tax rate analysis no longer argue strongly for more simplification, as was the casewith the Greek indirect tax system.

The administrative (publicly carried) and compliance (privately carried) costs of taxationare an important dimension of the quality of a tax system (Slemrod, 1990). An indirect taxstructure which is complicated and in a state of flux, like the Greek one, is more costlyto administer, since information is processed slowly, audits are infrequent, resulting in laxtax enforcement and low compliance.8 Reducing the number of taxes and tax rates wouldreduce administrative complexity.

If total indirect tax revenue is to remain unchanged, the standard VAT rate would need tobe considerably raised. Higher VAT rates would create stronger incentives for tax evasionand would potentially call for higher spending on tax administration and enforcement mech-anisms. On the other hand, keeping the UK standard VAT rate (17.5%) would result in adecrease in total revenue collected by about 10–12%, depending on behavioural responses.Increased international tax competition of more mobile tax bases renders the shifting oftaxation towards certain other sources, e.g. capital, problematic. On the other hand, a de-crease in tax revenue would not be sustainable for Greece at least at present, when revenuecollection is crucial under increased pressures for restricting the public deficit, in line withthe Euroland Stability and Growth Pact. This does not diminish the value of the Britishtax experience for Greece. Rather it suggests that Greece could potentially benefit mostby replicating the much simpler and more transparent UK indirect tax structure, as wellas other desirable parts of the UK tax and expenditure system, notably an increased rolefor direct taxes and public expenditure. That would reduce the present reliance on indirecttaxation for redistribution and especially revenue collection.

5. Concluding Remarks

The indirect tax system in Greece has been the historical outcome of the efforts of gov-ernments to use it both as the main revenue-raising device and as a tool for redistribution.This has created a tax structure which is unnecessarily complicated and inefficient withoutachieving any noticeably beneficial redistribution. Marginal tax reform analysis revealedthat the efficiency element dominates the ranking of commodity groups unless government’saversion to inequality approaches Rawlsian values, although we also found that it was hard


to identify efficiency improving indirect tax reforms. That suggests that consumption pat-terns are not sufficiently differentiated to allow indirect taxes any redistributive leverage.Further analysis showed that marginal indirect tax rates are essentially constant across theincome distribution controlling for differences in demographic characteristics. That againsuggests that there is little distributional cost and possibly considerable administrative ben-efit in moving to a much simpler tax structure, while making taxes uniform reduces theiroverall disincentive effect.

This analysis of indirect taxation assumes that other tax instruments cannot be changed—an assumption that is most favourable for defending differential indirect tax rates. If thereform can be broadened to include other tax and transfer instruments, then the case fora simple and more uniform indirect tax structure is greatly strengthened. Various authors(Atkinson and Stiglitz, 1976; Deaton, 1981; Deaton and Stern, 1986) provide conditionsunder which uniform VAT on commodities is optimal. As Deaton (1987) points out, “a con-siderable part of the efficiency case for uniform indirect taxes is the lack of any empiricallyconvincing evidence that differential taxes could encourage more effort and less leisure”.Given the difficulty of rejecting this maintained hypothesis in distributionally relevant ways(Blundell and Walker, 1982; Browning and Meghir, 1991; Madden, 1995, 1997), it is clearlydifficult to defend non-uniformity on either efficiency or equity grounds. Judging from avail-able evidence for other European countries (Ebrahimi and Heady, 1988; Sandford, Godwinand Hardwick, 1989; Rapanos, 1997; Davies and Kay, 1985), the administrative gains of asimpler commodity tax structure would most likely exceed any efficiency losses. It thereforeseems that theory and the available evidence point towards a simplification of the presentindirect tax structure.9

The UK indirect tax structure in general conforms with the above principles and its simu-lation on Greek consumers has revealed a realistic alternative tax structure, which is simpler,more equitable and more efficient to implement and administer. Most recommendations forGreek tax reform depend on the proper functioning of other parts of the tax/transfer sys-tem. The same conclusion was reached in the UK tax simulation experiment. The role thatindirect taxation should play will depend on how effectively the other instruments are used.Simplifying the indirect tax structure would further benefit from rationalising and enforcingincome taxation, combined with a more active role for the transfer system. Comprehensivetax and transfer reform should have high priority for Greek policy makers.

Appendix A. Table A1

Appendix B. The Robustness of Marginal Tax Results

The robustness of the results can be tested by considering how much the ranking of com-modities varies with the choice of the demand system and the ranking by tax attractiveness(capturing equity). Table B1 gives the rankings for the demand system estimated by An-drikopoulos, Brox and Georgakopoulos (1992) (in the left panel), and by Alogoskoufis,Misoulis and Karavitis (1996) in the right panel. The rows are ranked by the efficiencydeterminant of taxation, θi/ fi , from equation (6), i.e. by 1 + �τkωkεki/ωi , in the columnheaded “Rank AIDS”.


Table A1. Calculation of marginal indirect tax rates—LES versus Working Leser models.

Linear expenditure system Working-Leser model

Dependent variable: EQTAX (a) (for LES) t-ratios in parenthesesDependent variable: TBURDEN (b) (for W-L) (significant coefficients in bold letters)Intercept 2,73 ( 0,00) 0.1056 (9.01)Expenditure level:

Non-durable expenditure per equiv. adult 0.1039 (23.13)ln(Non-dur expenditure per equiv. adult ) −0.0012 (1.20)

Age of head of household (ref. hous. (c): 51<age<65)Less than 25 years old 3605 (2.77) 0.0230 (6.07)Between 26 and 50 years old 803 (3.49) 0.0071 (4.81)Over 65 years old −933 (4.03) −0.0078 (4.76)

Household composition:Two adults in the household (ref. hous.: one adult) 1518 (3.14) 0.0167 (6.82)Three or more adults in the household 2080 (3.43) 0.0264 (8.69)One adult female in househ. (ref. hous.: no female) −829 (1.07) −0.0099 (2.86)Two adult females in the household −1421 (1.64) −0.0150 (3.63)Three or more adult females in the household −1958 (2.17) −0.0189 (3.82)Dummy for head female (ref. hous.: male head) 405 (0.84) 0.0009 (0.39)One child in the household (ref. hous.: no children) −658 (2.81) −0.0022 (4.76)Two children in the household −806 (3.14) −0.0043 (3.47)Three or more children present in the household −983 (2.41) −0.0059 (2.70)One retired person (ref. hous.: no retired people) −265 (1.09) −0.0044 (2.37)Two or more retired persons −367 (1.30) −0.0085 (3.36)

Degree of Urbanisation: (ref. hous.: urban area)Rural area −665 (4.01) −0.0092 (6.44)Semi-urban area −469 (2.37) −0.0039 (2.25)

Regional Location: (ref. hous.: E. Sterea and Islands)Peloponese and West Sterea −322 (1.63) −0.0057 (3.24)Macedonia 394 (1.69) 0.0057 (4.33)Crete −734 (3.78) −0.0100 (4.22)Thessalia 392 (1.06) 0.0053 (2.58)Islands of East Aegeon 611 (1.48) 0.0037 (1.40)Thraki 575 (1.61) 0.0052 (1.88)Ipiros 52 (0.18) 0.0029 (1.13)

Head’s occupation: (ref. hous.: empl. in public sector)Student −3675 (2.48) −0.0165 (3.01)Employee in the private sector 315 (1.08) 0.0040 (2.50)Employer (own business with employees) 756 (1.14) −0.0008 (0.31)Self-employed (without employees) −15 (0.06) −0.0021 (1.32)Unemployed 2211 (1.44) 0.0091 (1.42)Pensioner 387 (1.14) 0.0029 (1.23)

Level of education of head: (ref. hous.: no/basic ed.)High-school education (middle education) −142 (0.70) −0.0030 (2.35)Higher education −190 (0.52) −0.0079 (4.64)

Housing tenure: (ref. hous.: owner-occupier)Rent 624 (2.67) 0.0125 (9.46)

Residual sum of squares 2,57E + 11 9.84F-Statistic 123.26 24.45R2 (R2 adjusted) 0.3866 (0.3834) 0.1111 (0.1066)(a) EQTAX: paid taxes per equivalent adult;

excludes proportional car taxes.(b) TBURDEN: total indirect taxes/total expenditure;

indirect taxes exclude proportional car taxes.(c) Ref. hous.: reference household.

Source: 1987/8 HES data (NSSG, 1994), own calculations.


Tabl

eB

1.D

eter

min

ants

ofdi

rect

ions

ofta

xre

form

.

And

riko

poul

os,B

rox

and

Alo

gosk

oufis

,Mis

oulis

Geo

rgak

opou

los

(199

2)an

dK

arav

itis

(199

6)R

anki

ngof

Ran

king

ofef

ficie

ncy

com

pone

ntR

anki

ngby

effic

ienc

yco

mpo

nent

Ran

king

by

Goo

dA

IDS

LE

SIn

dire

ctTa

xE

quity

Tax-

Equ

ityG

ood

AID

SL

ES

Indi

rect

Tax

Equ

ityTa

x-E

quity

Rec

reat

ion/

11

35

9−4

entr

etai

nmen

t

Alc

ohol

26

63

7−4

Alc

ohol

12

33

7−4

Tra

nspo

rt3

32

211

−9

Tra

nspo

rt3

31

29

−7

Hou

sing

49

109

45

Hou

sing

48

88

44

Toba

cco

52

11

2−1

Toba

cco

21

21

2−1

Hea

ting/

light

ing

610

44

13

Hea

ting/

light

ing

54

44

13

Clo

thin

g7

49

610

−4O

ther

87

810

13−3

Oth

er8

77

78

−1Fo

od9

1312

113

8Fo

od6

99

93

6

Com

mun

icat

ion

108

77

52

Com

mun

icat

ion

76

65

50

Furn

iture

1111

58

62

Furn

iture

95

56

60

Hea

lth12

511

1212

0

Edu

catio

n13

1213

138

5

Not

e:L

ower

rank

ssu

gges

tlow

erin

gta

xra

tes,

high

erra

nks

rais

ing.


Table B2. Ranking of tax reform for inequality aversion of 1.

Andrikopoulos, Brox and Alogoskoufis, MisoulisGeorgakopoulos (1992) and Karavitis (1996)

Good AIDS LES Ind Good AIDS LES Ind

Recreation 1 2 6

Alcohol 2 9 9 Alcohol 1 3 4

Housing 3 4 5 Housing 4 7 7

Tobacco 4 1 1 Tobacco 2 1 2

Heating 5 3 2 Heating/lighting 3 2 3

Transport 6 7 3 Transport 7 5 1

Food 7 6 7 Food 5 6 8

Communication 8 5 4 Communication 6 4 5

Clothing 9 8 10

Other 10 13 12 Other 8 9 9

Furniture 11 10 8 Furniture 9 8 6

Education 12 11 11

Health 13 12 13

Note: Ranked by AIDS of Andrikopoulos, lower ranks should have taxes lowered

For each demand system, the ranking of the efficiency component is given, first for AIDSfrom (6), then for LES from (6a), and last for indirect additivity form from (6b).10 The nextcolumn gives the rank by taxation (lowest ranks have highest tax rates, which would bereduced in a move towards uniformity), and the next gives the rank by tax attractiveness(computed by ν = 1, though the ranking is relatively insensitive to ν, with the lowest rankleast attractive for taxation). The final column in each part of the table gives the tax rankless the tax attractiveness or equity rank, and indicates the extent to which these two criteriaare in conflict.

The table reveals the greatest conflict over transport and food according to bothAlogoskoufis and Andrikopoulos estimates. Housing is ranked considerably lower ac-cording to AIDS than according to the other two demand systems. Similarly, furnitureis ranked lower by (6b) than the other two, and health by the LES than the other two. WithAlogoskoufis’ estimates, the AIDS rankings of housing, furniture and communicationsdiffer from the other two (which are remarkably close to each other).

Finally, Table B2 shows the ranking of desirable directions of tax reform for a moder-ate degree of inequality aversion of unity, for the various demand specifications (rankedby the AIDS specification of (6)). This shows an alarming degree of disagreement formany goods, though reasonable agreement on the desirability of raising taxes on furni-ture and “other” (though that is a very wide category for Alogoskoufis’ estimates, andincludes many goods that are given a low rank by Andrikopoulos’ estimates). Food lookspossibly worthy of higher taxes, and tobacco and heating for lower taxes (though seethe qualifications in the text). The different estimates give ambiguous signals for the


remaining goods, suggesting the serious limitations of the tax reform approach in the presentcontext.

Acknowledgments

Financial assistance from the Commission of the European Communities under ContractERBFMBI-CT95-0291 of the Marie Curie Fellowship programme and from the BritishEconomic and Social Research Council under Contract R00429514173 of the PostgraduateResearch Studentships programme is gratefully acknowledged. We are indebted to Prof. V.Rapanos and the editor and referee for their useful comments. All views expressed are thesole responsibility of the authors.

Notes

1. For an extended analysis of the distributional effects of Greek indirect taxes, see Kaplanoglou (2000).2. The force of utilitarianism is that it is individualistic, that is, it respects individual wellbeing as measured

by the individual utility function, and it is consequentialist, in confining attention to outcomes, rather thanprocesses or rights (though these can be included by restrictions on either information or policy variables).When it comes to numerical quantification, individuals are weighted by the OECD equivalence scale. An agentis then defined as one equivalent adult, who is assumed to receive the equivalent share of total householdexpenditure.

3. This assumption can be defended theoretically for a competitive constant returns economy, or for an imperfectlycompetitive economy (Lockwood, 1988; Stern, 1987), and empirically in the case of Greece (Karageorgas,1973; Karageorgas and Pakos, 1988).

4. This approach is common in such studies: see for example the original article by Ahmad and Stern (1984),Madden (1989) and Decoster and Schokkaert (1989).

5. It should be noted that in all estimated equations yh corresponds as before to non-durable items. The parameterswere scaled up to reflect the share of non-durable expenditure in total expenditure.

6. All estimated regressions were tested for the presence of multicollinearity, heteroskedasticity and the existenceof outliers. Unless otherwise stated, the estimated equations passed the tests of multicollinearity and theexistence of outliers. Heteroskedasticity-consistent estimators were obtained using White’s method—seeMacKinnon and White (1985).

7. To simplify, we only consider a two-rate VAT, and assign goods in the lower (5%) rate to the zero rate band.8. For empirical evidence on the negative relationship between the number of VAT rates and VAT compliance in

OECD countries see Agha and Haughton (1996). For an excellent account of the relation between tax reformand tax administration in Greece, see Rapanos (1997).

9. Complete uniformity might be undesirable, given other distortions such as the CAP, while additional excisetaxes can be justified for commodities as explained in the paper.

10. The correct way to do this would have been to re-estimate the various specifications on the original data, butthat was not possible, and instead the own-price elasticities and the expenditure elasticities have been takenfrom the AIDS estimates.

References

Agha, A. and J. Haughton. (1996). “Designing VAT Systems: Some Efficiency Considerations,” The Review ofEconomics and Statistics 78, 303–308.


Ahmad, E. and N. Stern. (1984). “The Theory of Reform and Indian Indirect Taxes,” Journal of Public Economics25, 259–298.

Alogoskoufis, G., N. Misoulis and N. Karavitis. (1996). “Excise Taxes and their Effects,” Institute for EconomicPolicy Research, Athens (mimeo) (in Greek).

Andrikopoulos, A. A., J. A. Brox and T. A. Georgakopoulos. (1992). “A Short-Run Assessment of the Effectsof the VAT on Consumption Patterns: The Greek Experience,” Working Paper no. 92-03, Athens University ofEconomics and Business.

Atkinson, A. B. (1970). “On the Measurement of Inequality,” Journal of Economic Theory 2, 244–263.Atkinson, A. B. (1977). “Optimal Taxation and the Direct versus Indirect Tax Controversy,” Canadian Journal of

Economics 10, 590–606.Atkinson, A. B. and J. E. Stiglitz. (1976). “The Design of Tax Structure: Direct versus Indirect Taxation,” Journal

of Public Economics 6, 55–75.Blundell, R. W. and I. Walker. (1982). “Modelling the Joint Determination of Household Labour Supplies and

Commodity Demands,” The Economic Journal 92, 351–364.Browning, M. and C. Meghir. (1991). “The Effects of Male and Female Labor Supply of Commodity Demands,”

Econometrica 59, 925–951.Davies, E. H. and J. A. Kay. (1985). “Extending the VAT Base: Problems and Possibilities,” Fiscal Studies 6,

1–16.Deaton, A. (1979). “Optimally Uniform Commodity Taxes,” Economics Letters 2, 357–361.Deaton, A. (1981). “Optimal Taxes and the Structure of Preferences,” Econometrica 49, 1245–1260.Deaton, A. (1987). “Econometric Issues for Tax Design in Developing Countries.” In D. M. Newbery and N. Stern

(eds.), The Theory of Taxation for Developing Countries. New York: Oxford University Press.Deaton, A. and J. Muellbauer. (1980a). Economics and Consumer Behaviour. Cambridge: Cambridge University

Press.Deaton, A. and J. Muellbauer. (1980b). “An Almost Ideal Demand System,” American Economic Review 70,

312–326.Deaton, A. and N. Stern. (1986). “Optimally Uniform Commodity Taxes, Taste Differences and Lump-Sum

Grants,” Economics Letters 20, 263–266.Decoster, A. and F. Schokkaert. (1989). “Equity and Efficiency of a Reform of Belgian Indirect Taxes,” Recherches

Economiques de Louvain 55, 155–176.Ebrahimi, A. and C. Heady. (1988). “Tax Design and Household Composition,” The Economic Journal, Conference

Papers 98, 83–96.Feldstein, M. S. (1972). “Distributional Equity and the Optimal Structure of Public Prices,” American Economic

Review 62, 32–36.Gale, W. (1997). “What Can America Learn from the British Tax System?” Fiscal Studies 18, 341–369.Georgakopoulos, T. A. (1989). “The Indirect Tax System: Structure, Effects and Changes.” In G. Provopoulos

(ed.), Priorities of Fiscal Policy. Special Studies no. 18, Institute for Economic and Industrial Research, Athens(in Greek).

Grevenitis, P. and G. Sapounas. (1988). Taxation in Greece: Comparative Incidence on Socio-Economic Groups.Studies on the Agricultural Economy 32, Athens (in Greek).

Houthakker, H. S. (1960). “Additive Preferences,” Econometrica 28, 244–256.Kaplanoglou, G. (1999). “Distributional and Efficiency Aspects of the Greek Indirect Tax System: A Microsimu-

lation Analysis,” University of Cambridge, unpublished PhD thesis.Kaplanoglou, G. (2000). “A Microsimulation Analysis of the Distribution of the Indirect Tax Burden Among

Greek Households.” In L. Mitton, H. Sutherland and M. Weeks (eds.), Microsimulation Modelling for PolicyAnalysis: Challenges and Innovations. Cambridge: Cambridge University Press.

Karageorgas, D. (1973). “The Distribution of Tax Burden by Income Groups in Greece,” The Economic Journal83, 436–448.

Karageorgas, D. and T. B. Pakos. (1988). “The Distribution of the Tax Burden in Greece: 1982.” In Institute forRegional Development PASKE (eds.), In honour of Saki Karageorga. Athens.

Keen, M. J. (1997). “Peculiar Institutions: A British Perspective on Tax Policy in the United States,” Fiscal Studies18, 341–369.

Leser, C. E. V. (1963). “Forms of Engel Functions,” Econometrica 31, 694–703.


Lockwood, B. (1988). “Tax Incidence, Market Power and Bargaining Structure,” Birbeck College, London(mimeo).

MacKinnon, J. G. and H. White. (1985). “Some Heteroskedasticity-Consistent Matrix Estimators with ImprovedFinite Sample Properties,” Journal of Econometrics 29, 305–325.

Madden, D. (1989). “Indirect Tax Reform in Ireland,” Economic and Social Review 21, 27–47.Madden, D. (1995). “Labour Supply, Commodity Demand and Marginal Tax Reform,” Economic Journal 105,

485–497.Madden, D. (1997). “Conditional Demands and Marginal Tax Reform,” Oxford Bulletin of Economics and Statistics

59, 237–255.Ministry of Finance and Economy (2001). Main National Account Aggregates of the Greek Economy, 1960–1999

(ESA 95). Athens.National Statistical Service of Greece (1994). Data: Family Expenditure Survey 1987/88. Athens.Newbery, D. M. (1995a). Reforming Road Taxation. Group Public Policy, The Automobile Association. UK.Newbery, D. M. (1995b). “The Distributional Impact of Price Changes in Hungary and the UK,” The Economic

Journal 105, 847–863.Newbery, D. M. (1997). “Optimal Tax Rates and Tax Design During Systemic Reform,” Journal of Public

Economics 63, 177–206.Newbery, D. M. and T. Revesz. (2000). “The Evolution of the Tax Structure of a Reforming Transitional Economy:

Hungary 1988–98.” International Tax and Public Finance 7, 209–240.Newbery, D. M. and J. E. Stiglitz. (1981). The Theory of Commodity Price Stabilisation: A Study in the Economics

of Risk. Oxford: Oxford University Press.OECD. (2001). Economic Surveys 2000–2001: Greece, February 2001, Paris: Organisation for Economic Co-

operation and Development.Provopoulos, G. A. (1979). “The Distribution of Fiscal Burdens and Benefits by Income Groups in Greece,” Greek

Economic Review 1, 77–99.Rapanos, V. (1997). “Tax Reform and Tax Administration.” In G. Provopoulos and V. Rapanos (eds.), Functions

of the Public Sector: Problems and Prospects. Athens: Foundation for Economic and Industrial Research (inGreek).

Redmond, G. (1995). “POLIMOD: The Calculation of VAT and Excise Duties on Household Expenditure.”Research Note no. 11, The Microsimulation Unit, Department of Applied Economics, University of Cambridge.

Sandford, C., M. Godwin and P. Hardwick. (1989). Administrative and Compliance Costs of Taxation. Bath: FiscalPublications.

Slemrod, J. (1990). “Optimal Taxation and Optimal Tax Systems,” Journal of Economic Perspectives 4, 157–178.Stern, N. (1987). “The Theory of Optimal Commodity and Income Taxation: An Introduction.” In D. M. Newbery

and N. Stern (eds.), The Theory of Taxation for Developing Countries. New York: Oxford University Press.Stone, J. R. N. (1954). “Linear Expenditure Systems and Demand Analysis: An Application to the Pattern of

British Demand,” The Economic Journal 64, 511–527.Working, H. (1943). “Statistical Laws of Family Expenditure,” Journal of the American Statistical Association

38, 43–56.

indirect taxation in greece: evaluation and possible reform

Documents