contents articles · 2013-04-15 · of societal taste, requires a value judgement. yet it is not...

Contents3 Foreword

Mark Chandler

Articles

5 Income Taxation and Equity

Peter J. Lambert

21 Pollution Taxes in a Second-Best World

Gilbert E. Metcalf

35 Multi-dimensional Heterogeneity and the Design of Tax Policies

Helmuth Cremer

46 Tax Policy on the Outskirts of the European Union: Greenland

Carsten Vesterø Jensen & Søren Bo Nielsen

64 Pension Reforms and Taxation in Estonia

Ringa Raudla & Karsten Staehr

93 Automatic Fiscal Stabilizers in Estonia

Rasmus Kattai, Alvar Kangur, Tanel Liiv, & Martti Randveer

Roundtable: Tax Policy in the Baltic States on the Eve of Enlargement

116 Tax Policy in Latvia on the Eve of Enlargement

Daina RobeΩniece & Måris Jurußs

1

Content

121 Estonian Excise Policy on the Threshold of the European Union

Evelin Ahermaa

130 Approximation of Lithuanian Rates of Excise Duties on Tobacco Products

with the European Union

Book Review

133 Marketing Challenges in Transition Economies of Europe

(Editors: Gopalkrishnan R. Iyer and Lance A. Masters)

Evita Lune

2

Baltic Journal of Economics Autumn/Winter 2003

Foreword

This special issue of the Baltic Journal of Economics contains a selection of papers that werepresented during the conference on “Tax Policy in EU Candidate Countries: On the Eve ofAccession” held in Riga on 12-13 September, 2003. The conference was one of the results of aneffort by EuroFaculty to have an impact on the analysis of public finance issues in the Balticstates. This is a special priority due to the large problems with taxation and public administrationin accession countries, and particularly since they are such strategic inputs to further developmentof these economies.

The papers in this journal are a selection from the many fine papers that were presented at theRiga conference. Altogether, 32 papers were presented including country reviews for each of thethree Baltic states and invited papers by Helmuth Cremer, Gilbert Metcalf and Peter Lambert.There was also a session devoted to the special problems of excise taxation of cigarettes. Wecollect the papers here in the order in which they were presented.

The paper by Cremer reviews the recent literature on the optimal design of tax policies in anenvironment of multi-dimensional heterogeneity. Cremer demonstrates that this literaturechallenges the simple Atkinson and Stiglitz theorem prescription. Indeed the mix of variouspolicy instruments observed in real-world economies may well be necessary for optimal policydesign.

Metcalf presents a variety of policy relevant findings from the recent literature on environmentaltaxes and regulation. He addresses the double-dividend hypothesis and explains why this does notimply optimal income taxes are higher. Next he shows that knowledge of the direction of changesin optimal environmental taxation is not sufficient to know the effect on environmental quality.Reviewing his work on the Kyoto Protocol agreement, Metcalf shows that even a weak double-dividend may not occur when carbon taxes are used to replace non-energy related consumptiontaxes.

Lambert explores the relationship between value judgments and the design of tax systems.Turning to the proposal for an EU income tax he finds that an element of differentiation wouldenhance vertical equity without necessarily introducing horizontal inequity. Equity in this contextwould be served by proportional taxes differentiated in each country so that countries with highernational inequity have lower tax rates.

Kattai et. al. investigate the sensitivity of Estonian budget deficits to the economic cycle under thecurrent fiscal design. They find low sensitivity implying that even a substantial cyclical downturnwould not cause a breach of the Maastricht criteria as long as Estonia’s structural deficit waswithin 1% of gdp.

Hamalainen discusses the effect of tourism on optimal commodity taxation. If the welfare oftourists has zero weight, a simple uniform sales or value added tax would not be optimal. In fact,for goods that are consumed only by tourists it would be optimal to set the tax rate to maximisetax revenue, as long as there are no externalities. Where there are externalities, the Sandmo resultis modified due to the demand of visiting tourists, hence the marginal social damage of the

3

Foreword

commodity enters the optimal tax rule additively.

Jensen and Nielsen examine the tax system in Greenland and find its taxes are generally lowerthan the European average. They also provide several recommendations for reform. To improvethe profit tax they recommend eliminating over-generous depreciation allowances. They alsourge introduction of the full array of taxes observed in other European countries, including taxeson inheritance, gifts, property, general consumption, and energy. Lastly, they point to the heavyburden on public finances from Greenland’s subsidies of certain sectors.

Raudla and Staehr examine the reformed Estonian pension system, built on the three-pillar model.They argue that political compromises have resulted in an overly complex system. They suggestthat it may leave some groups vulnerable and represents a transfer from the public sector to theprivate sector. Since increased savings are not likely to affect long run growth, dynamic gains maycome mainly through greater labour supply.

The first country review paper gives a brief overview of the development of the Latvian taxationsystem. The next two papers concentrate the discussion on the development of cigarette taxationin Estonia and Lithuania. All three papers discuss the problems of making rapid changes not onlyin the rates but also the structure of the taxes on cigarettes. In order to smooth the shock to themarket, Estonia and Lithuania negotiated transition periods to meet EU requirements until 2009and 2010, respectively. One of the challenges is to increase tax rates while preventing the illegalmarket from booming, thus it is not deemed possible to reduce illegal trade in Estonia during thisprocess.

While these papers cover a variety of topics from both theoretical and applied perspectives, theyare all relevant to current policy making in countries joining and preparing to join the EU. HenceI commend them to the reader as a further contribution to this ongoing debate.

Mark Chandler

Stockholm School of Economics in Riga

Strelnieku iela 4a

Riga, LV-1010

Latvia

Email: [email protected]

4


INCOME TAXATION AND EQUITY

Peter J. Lambert*

Abstract This paper provides an exposition and explanation of the various ways in which value judgements can be

instilled into an income tax system, or, if inherent in a pre-existing one, can be drawn out and understood. A putative EU-

wide income tax, additional to the national income taxes of the Member States, is used as a vehicle for the analysis. When

the identification of equals is done using an appropriate ‘equivalent income function’, and the equal treatment command

modelled in terms of it, the resultant tax will in general be differentiated between countries. A supplementary command,

“equal progression among equals”, can be achieved if equals are defined as those at the same percentile point in the within-

country distributions, and if these distributions differ in logarithms only by location and scale. Differentiated proportional

taxes could even be equitable in this scenario, the flat rate being higher in less unequal countries. The value judgements

implicit in a given tax system can be exposed in terms of an equivalence scale which is in general “base dependent”.

1. IntroductionAs the unique tax instrument through which the government approaches its citizens directly, theincome tax is particularly subject to criteria of equity. Classical horizontal equity (HE) andvertical equity (VE), two basic commands of social justice, state respectively that equals shouldbe treated equally, and unequals appropriately unequally. HE can be seen as a minimal rule offairness, offering protection against arbitrary discrimination and reflecting the basic principle ofequal worth. VE requires differentiation among unequals, and its degree is a matter of societaltaste and political debate. See Musgrave (1990) and Steuerle (1999) for thoughtful discussion. Inthis paper, we outline the conceptual and measurement issues involved in characterizing thehorizontal and vertical stance of an income tax or tax and benefit system, and sketch someappropriate measurement procedures. We also discuss the issue of de novo tax design from anequity perspective, taking as a case in point a putative EU-wide income tax, formed as anadditional layer of tax rather than through harmonization of Member States’ existing direct taxsystems.

In thus exposing the equity (or, indeed, inequity) characteristics of an existing tax, alongside theprinciples which should govern the introduction of a new tax, the paper should provide a windowinto the methodologies economists have developed to analyze the value judgements governing thedirect taxation of a country’s citizens.

As simplistically described above, it might seem that HE is an absolute, and VE, being a matterof societal taste, requires a value judgement. Yet it is not this simple. Much recent analysis of thetax system’s horizontal stance is designed to expose horizontal inequity (HI), for example byquantifying the extent to which a personal income tax fails to be equitable as a family tax. Thisassumes that the tax designers were incapable of achieving the HE ideal. Household equivalencescales are typically invoked to identify the equals (and unequals), a value judgement whichbecomes centrally important for the analysis. (And not the only one; what ought “equal treatment”to mean?). Analysis of the HE stance of the tax system is, in almost all current literature,

5

Income Taxation and Equity

∗ Department of Economics, University of York, York YO10 5DD, UK

essentially normative. As for the vertical stance, distributional analysts tend to assume that thesocial decision-maker has selected, and enacted, a desired degree of progressivity, manifest in thetax system’s impact on inequality; if this belief is followed, the characterization of a tax system’sVE becomes an exercise in positive economics1.

The structure of the paper is as follows. In Section 2, we explain briefly the concepts underlyingvertical and horizontal equity. This involves discussion of progressivity and redistributive effect,the definition of equals and the meaning of equal tax treatment. We begin Section 3, onevaluating existing tax systems, by outlining the dominating strand of literature of the 1990s,according to which HE violations are captured by indices as loss of redistributive effect (verticalperformance) in a measurement system that attributes to the policymaker the same degree ofaversion to both horizontal and vertical inequality. We go on to describe a recent development inHI measurement which obviates this restriction.

In Section 4, we discuss some of the issues that would face the designer of a new income tax,taking as a “vehicle” for this analysis a putative EU-wide income tax, additional to the nationalincome taxes of the Member States, whose revenue would go directly to the centre. The questionof an EU-wide social welfare function (henceforth SWF) arises, in which a person’s domicile mayor may not be a relevant factor. By drawing on recent work in the regional context, we observethat if a common income tax were devised, applicable in all countries, then however equitable,the VE of the entire system (in fact, overall inequality and welfare) can be unambiguouslyimproved by allowing an element of differentiation in this tax, potentially admitting HI.

In Section 5, we explicitly allow for differences in the taxable capacities of persons or householdsin the Member States, and discuss the design issues this raises. A recently developed equitycommand, equal progression among equals, which could be said to mix vertical and horizontalaspects, may be achievable alongside HE and VE, and is so if the equals in different countries aredefined as those at the same given percentile point in their within-country distributions, and ifthese distributions differ in logarithms only by location and scale (as would be the case forlognormality). In such a scenario, we show, differentiated proportional taxes would be equitable,with the flat rate being higher in less unequal countries.

Finally, in Section 6, we discuss the detection of the value judgements that might underlie such afuture EU-wide tax, were it to have been imposed by the politicians without the advice of theeconomists. New work is explained which can draw out of the tax itself the implied valuejudgements of the decision makers about equals, through a base-dependent equivalence scale,turning horizontal tax analysis into a positive exercise, much as vertical analysis now is. Section7 concludes with an assessment of what has been achieved in the paper, namely, an expositionand explanation of the various ways in which value judgements can be instilled into an incometax system, or, if inherent in a pre-existing one, can be understood.

6


1 A third equity criterion, that of “no reranking” (NR), has coexisted with HE and VE in the measurement literature for almost 25 years,and is variously seen as an alternative to HE or a supplement to VE. According to this line of analysis, inequity is conceptualized as lackof perfect association between pre-tax and post-tax living standards and “addresses the fairness of a process of redistribution” (Plotnick,1981, p. 283). The classical view is that HE “enters as an end-state principle” (Musgrave, 1990, p. 120). We shall not dwell upon NR muchin this paper; see Lambert and Yitzhaki (1994) and Dardanoni and Lambert (2001) for further discussion. For discussion of equity issuessurrounding indirect taxes, which are not dealt with here, see for example Loomis and Revier (1988) and Decoster et al. (1997).

2. Vertical and horizontal equity

The simplest model of an income tax schedule is one in which the tax liability on an income of xis a pure function of that income; let us write it t(x).We might assume that t(x) is differentiable(almost everywhere), that 0 # t(x) # x and 0 # t’(x) # 1 for all x, and, for progression, that t(x)/xis increasing (i.e. t’(x) > t(x)/x). If net incomes n(x) = x - t(x) are plotted against gross incomesx, the relationship is typically upward sloping and concave, showing clearly that relative incomedifferentials get compressed in the transition from gross to net income (see Figure 1). As thisfigure also shows, negative income taxes, i.e. cash benefits, can be incorporated into the modelsimply by dropping the assumption that 0 # t(x). Then the graph cuts the 45Ε line at the break-even point between cash benefits and taxes. Letting g be the fraction of all income taken in tax,or ‘total tax ratio’, g = 3t(x)/3x, the distributive effect of the actual tax can be compared with thatof an equal-yield proportional tax (hence at rate g) on all income units: clearly the rich pay more,and the poor pay less, under the actual (progressive) schedule. It is as if, first, a flat tax at rate gwere imposed, with no exemption, and then rich-to-poor transfers undertaken. The redistributiveeffect (RE) of such an income tax schedule is measured by its inequality effect relative to that ofthe flat tax (which is, of course, neutral in inequality terms):

(1) RE = IN - IX[1-g] = IN - IX

where I is an index of relative inequality and the subscript indicates the distribution of incomeconcerned (N for net incomes n(x), X for gross incomes x, etc).

This model stood throughout the late 1970s and 1980s, following the seminal articles of Fellman(1976), Jakobsson (1976) and Kakwani (1977a,b)2. The model is good if the population inquestion is homogeneous in all attributes but income, but plainly inadequate if we wish to takeaccount of non-income attributes which differ between income units in the tax code. This is verycommon. For example, deductions and exemptions may be awarded for charitable giving, medicalexpenditures, mortgage interest (Johnson and Mayer, 1962), life insurance premia, workexpenses, childcare costs, etc. Differences in income tax treatment on non-income grounds may

7


2 But see Eichhorn et al. (1984) for a relaxation of some of the assumptions.

45o

(1-g)x

n(x)

Gross incomeGross income

Netincome

O

also arise through different sources and disposition of income (Gravelle, 1992), tax evasion andnon-compliance (Bishop et al., 1994), and the different treatment of urban and rural incomesespecially in developing countries. In such cases, we need to model the income tax code as amulti-attribute one.

The question arises, whether the tax deductions (etc) are equitable, and this is usually taken tomean horizontally equitable: do pre-tax equals get equal tax treatment? Towards the end of thepaper, we shall see how to impute a concept of equal treatment to the tax designer, which is suchthat an automatic response of “yes” can be given in answer to this question. More typically, wemight want to know if a given multi-attribute income tax can be judged equitable in its treatmentof families of different compositions and needs. The HE ideal, or aspiration, typically refers toindividuals. To extend it to families, equivalence scales are usually invoked.

The first step is to turn the business of identifying the equals - and the unequals, for that matter -into a unidimensional problem.3 We shall require income units’ pre-tax incomes, or livingstandards, call these x, to be measured on a scale which identifies the equals: equals will be thosehaving the same pre-tax income x. Manser (1979) discusses the modelling of household objectivesincluding different leisure times of their members, and Rosen (1978) demonstrates an empiricalprocedure which, given rich enough microdata, will “generate two vectors, one of family utilitiesbefore tax and one of family utilities after tax”, and he goes on to say that “the real problem inmeasuring horizontal equity is to summarize the differences between these vectors in ameaningful way” (p. 314). Gravelle (1992) quotes Steuerle (1983) as advocating an equivalizationprocedure to provide “a working definition of equity” across family sizes. As we shall see, ageneralized notion of equivalence scale is called for in order to explain some features of thetypical income tax system, and to articulate new equity criteria.

If the first problem is how to define the equals, the second, which follows hard upon it, is, whatdo we mean by “equal treatment”? A widely accepted equal treatment command is, equal post-tax living standards for pre-tax equals; other commands, equal average tax rates and equal taxesamong equals respectively, can be interpreted as variants on this (see Lambert, 2003).

The income unit may be the individual, the family, the household or that virtual person known asthe ‘equivalent adult’, advocated for welfare analysis when using equivalent income distributionsby Ebert (1997). 4

One equivalence scale which is popular with distributional analysts is the doubly-parametric scaleof Cutler and Katz (1992), in which household money income is deflated by a factor m = (NA +2NC)Ν, in which NA and NC are the numbers of adults and children in the householdrespectively, 2 0 [0,1] measures the relative importance of children and Ν 0 [0,1] representseconomies of scale. Figure 2 shows the relationship between gross equivalent income and netequivalent income for the UK tax and benefit system for 1993 when 2 = Ν = 1/2. As is apparent,there are plenty of instances of HI here: vertically aligned data points represent families with thesame pre-tax living standard but differing post-tax living standards.5

8


3 Many public economists would dispute the unidimensional form to which the HE problem is reduced by recourse to a living standardscriterion. Why, for example, should the tax treatment of a person whose income derives from his chosen holding of government bonds, withtax-privileged yield, be the same as that of another person whose living standard is derived from earnings or profits? This kind of objectionsuggests extending the dimensionality of the problem, possibly ad infinitum. If we would give full credence to questions of the form “Canwe ever find true equals?”, we would be led simply to inspect the tax code for society’s chosen form of tax treatment, and its inferreddefinition of equals; this would all but throw the baby away with the bath water, turning the HI question simply into one of tracking downassessment and payment errors.4 Suppose that z(n) is the equivalence scale value to be used to deflate the money income x of a family of size n. Ebert shows that if z(n)virtual persons are each allotted an equivalent income or living standard of x/z(n), then - and only then - will small money transfers fromhigher to lower living standard families necessarily be overall welfare-improving. Empirically, it is a matter of using the equivalence scaledeflators as sample weights.5 There are also plenty of instances of reranking (recall footnote 1).

A problem that has been perceived with equivalent income-based studies of HI is that theequivalence scale is typically selected by the analyst, from the outside as it were. Somecommentators have expressed the view that this amounts to “imposing HI from outside” if the taxis not, in fact, a family income tax designed to be coherent with an equivalence scale - or indeedif it is and the scale selected by the analyst is not the one being used by the policymaker. Wereturn to this point shortly.

3. Measuring VE and HIProcedures have been developed during the 1990s to measure the extent of HI in the personalincome tax by means of summary indices capturing HE violations as loss of vertical performance.Look at Figure 2. The vertical stance of the tax can be thought of loosely as its effect ‘on average’as between equals groups (i.e. between people with different living standards); according to theprogressive principle, this effect should be to reduce inequality. Within any given equals group,however, a tax system with HI in fact introduces new inequality - where there was none before(pre-tax). So there is a tension between the vertical and horizontal effects of a tax system with HI.This “between and within groups” insight invites the application of decomposable inequalityindices (more usually employed in a demographic context) to the problem; it is this which hasyielded up the HI measures of the 1990s. Papers by Aronson et al. (1994), Lambert and Ramos(1997) and Duclos and Lambert (2000) each provide a decomposition of redistributive effect intoorthogonal vertical and horizontal components:

(2) RE = V- H

9


Figure 2: UK Family Expenditure Survey 1993

.00

200.00

400.00

600.00

800.00

1000.00

1200.00

-500.00 .00 500.00 1000.00 1500.00 2000.00

Gross equivalent income

Ne

t E

qu

ival

ent

inco

me

Series1

in which V measures the inequality-reducing impact of the tax system on average, and H a lossdue to the ‘new’ inequality introduced by the presence of HI.6,7

All three of the studies just cited could be said to “impose HI from the outside”since they use theCutler and Katz (1992) equivalence scale for selected parameter values. A more defensibleprocedure would be to search for the HI-minimizing choice of equivalence scale parameters foreach regime, before undertaking any comparisons: this could both reveal the equivalence scalesmost nearly implicit in the tax systems under examination, and enable comparisons of residual HIgiven those scales. This kind of analysis has not been seen yet (but see Aronson et al. 1994, andvan de Ven et al. 2001 for steps in this direction).

Finally, we mention a very recent HI contribution, that of Auerbach and Hassett (2002), whichoffers new scope for understanding a tax system’s horizontal and vertical equity characteristics.The authors’ point of departure from the mainstream of the 1990s, as already described, is toinstitute a measurement system in which the SWF “need not evaluate ‘global’ (vertical equity)differences in after-tax income using the same weights as it applies to ‘local’ (horizontal equity)differences”. It is the very sameness of these weights, in the studies already cited, which leads tothe commensurate vertical and horizontal components V and -H of redistributive effect. Auerbachand Hassett allow that a social decision-maker could have a different aversion to inequalitybetween unequals from his aversion to inequality introduced by the tax system among equals. 8

With the measures these authors present, a search for the HI-minimizing values of the tworespective inequality aversion parameters could reveal the horizontal and vertical stance mostnearly implicit in the tax system.

4. A putative EU-wide income tax: VE and HESuppose, for the sake of argument, that a new layer of income tax were to be introduced in theEU, with the proceeds going directly to the center. What would be the issues facing the designerof such a tax? What would be the appropriate base for it? Should the tax be proportional, or oneembodying the progressive principle? One can imagine the new tax being levied at a flat rate onall disposable incomes, in order not to interfere with relative income differentials within countries.It is a small leap from there to suppose that a concessionary rate for the poorest countries mightbe instituted, and another small leap to a plethora of flat rates, negotiated country-by-country bythe politicians. (Note, though, that an EU-wide income tax has not been advised by economists,see e.g. Cnossen 2002, pp. 76-76 on this; however the spectre of such a development has beenmentioned from time to time in the popular press, in the UK at least).9 In this section of the paper,we make some pertinent observations, before proceeding, in the next section, to develop formallyan equity criterion taking into account, through an EU-wide SWF, the possibly differing taxable

10


6 In each paper, this loss is measured locally as inequality of post-tax income among pre-tax equals, and aggregated into a global indexusing a weighting scheme. In the first, which uses the Gini coefficient, the “- H” term is augmented by a measure of reranking. See alsoAronson and Lambert (1994) on this. Musgrave (1990) suggested the business of devising a local HI measure and then aggregating. The“identification problem” had beset many earlier studies of HI. If a sample contains few or no exact equals, how can we make it tell usanything about HI? One solution is to band people pre-tax, into ‘close equals groups’, and to measure inequality effects between and withinthese groups (see van der Ven et al., 2001 for more on this in respect of the Gini approach). The other solution is to use a kernel procedureto estimate non-parametrically and consistently the continuous population distributions of pre- and post-tax living standards from thesample joint distribution, and thence the indices V and -H (see Duclos and Lambert, 2000).7 In order to track the changing vertical and horizontal characteristics of a tax system through time, concomitant with an ever-changingincome distribution, point measures of vertical and horizontal performance at each percentile in the pre-tax distribution have beendeveloped, in Hayes et al (1994) and Duclos and Lambert (2000) respectively. Given suitable data over a run of years, it is possible toproduce 2-dimensional contour plots of the levels of progression and local horizontal inequity, at each (date,percentile) coordinate pair. 8 This construction clearly meets a view expressed by Steuerle (1999): “Many people strongly support horizontal equity even though theyreject the notion that government must adjust the position of any individual along any particular scale (for example, they may opposeredistribution). I do not think they are inconsistent in holding these two positions”. 9 To whit, consider the headline “”Now Britain faces single European tax system: France and Germany spearhead plan to control revenueand social security” in The Independent newspaper of 16th January 1997.

capacities of persons or households in the various Member States.

A theoretical construction of Cubel and Lambert (2002a,b) points the way. Even in the absenceof country-specific dimensions in the EU-wide SWF, it can be welfare improving, and inequalityreducing in the strongest sense (of Lorenz dominance), to impose country-specific income taxesrather than a common one across all countries. Specifically, these papers demonstrate that if two(or more) regions in a federation have very different levels of well-being then, whatever commonnew tax layer t(x) one might envisage, where x is a person or household’s disposable incomeregardless of domicile, be it proportional or progressive, VE can be enhanced by rewriting the taxcode to include an appropriate differentiation. Let n(x) = x - t(x) be a person or household’s netdisposable income after application of the EU-wide tax t(x). Let A and B be regions (groups ofcountries, for example) such that people in A are “generally poorer” than people in B (there is aprecise technical condition for this which can be found in either paper; it permits significantoverlap between the distributions). A welfare improvement and inequality reduction obtains if,instead of t(x) being imposed, a differentiated tax were instituted such that net incomes became(1+2)n(x) for a person with income x living in A, and (1-8)n(x) for a person living in B, where 2and 8 are small, and such that total revenue is maintained.10

But of course this recipe introduces HI: people with a given income (living standard) x who livein A will pay less tax than people with x who happen to live in B. It may seem strange that theintroduction of HI could improve matters, for the conventional wisdom says the opposite:whenever disparity is introduced, where there was equality before, welfare is reduced (Atkinson,1970). In the tax context, Kakwani and Lambert (1999, p. 28) put their finger(s) on it:“Discrimination can be interpreted as the loss of vertical equity attributable to group specificityof schedules in the tax code. If the code were to be replaced by the averaged schedule, there wouldbe a welfare increase”. There appears to be a conundrum here. Starting with a common schedulet(x), differentiation reduces inequality and raises welfare; but then averaging liabilities across thedifferentiated regions would also do that - and so on, re-differentiating, re-averaging.... welfareimprovements ad infinitum?

The conundrum is resolved by noticing that the informational requirements involved in this(continual) re-processing of tax liabilities place a limit on its feasibility. After averaging thedifferentiated tax, net income for somebody with living standard x would become (1+2)p(x)n(x)+ (1-8)(1-p(x))n(x), where p(x) is the proportion of income units having living standard x wholive in region A. This would bring demographics into the tax code (when p(x) ? 0 or 1),complicating the informational requirement and causing a loss of transparency, surely animportant consideration.

Differentiation of an income tax across countries can as well be rationalized if the EU-wide socialwelfare function contains country-specific dimensions, and in this case obviously need notinvolve HI. For example, the equal sacrifice model, but with a different utility-of-income functionin each country, could be invoked to rationalize different taxes.11 Atkinson and Bourguignon’s(1987) utilitarian SWF admits of dimensions other than income. For a population divided into“needs groups” i = 1,2,....n, a different utility-of-income function Ui(x) is attributed to each group,

11


10 In the case of proportional taxes, this says that rather than taxing all people in the EU at a common rate flat g, VE would beunambiguously enhanced by taxing people in the poorer group of countries at a lower rate gA, and in the richer group of countries at ahigher rate gB, where (1+2)(1-g) = 1-gA, and (1-8)(1-g) = 1-gB.11 According to equal (absolute) sacrifice theory, if U(x) is the utility-of-income function, the income tax t(x) should be designed to satisfyU(x) - U(x-t(x)) / uo ?x. Samuelson (1947) showed that if -xU”(x)/U’(x) > 1 ?x > 0 then such a t(x) is progressive. Ok (1995) demonstrateda reverse result, that if an income tax schedule t(x) satisfies t’(x) > 0 and t”(x) > 0 ?x, then there exists a U(x) with respect to which t(x) isequal sacrifice. See also Mitra and Ok (1996,1997) and D’Antoni (1999). Musgrave and Musgrave (1984) point to a perceived problemwith equal sacrifice taxes, that they provide no link between tax payments and the benefits received by income units from the proceeds:“This approach leaves the expenditure side of the public sector dangling” (ibid., p. 228).

with a hierarchy of needs specified by a systematic ordering of marginal utilities UiΝ(x), i =1,2,....n, at each fixed income level x. This model could perhaps be adapted to the equal sacrificeframework, but such an extension has not yet been made. In fact, it is not clear that such a modelwould be appropriate in the EU context. As between an income unit in Germany having $60,000p.a. and an income unit in Latvia also having $60,000 p.a., which is socially the more deservingof an additional dollar? Arguably the one in Germany suffers more relative deprivation than theone in Latvia, being further down its country-specific distribution of living standards (Runciman,1976); but would this merit a more lenient income tax in Germany? Does not one’s intuition gothe other way? There is an intricate issue here for income tax design, to which we now turn.

5. Designing a differentiated income tax: equity issuesFor simplicity at this point, let us confine attention to a population divided into two mutuallyexclusive and socially homogeneous subgroups, A and B say, and let us assume that groupmembership enters into the SWF as well as income level. Then we may adapt the equivalizingconcept, and posit an equivalent income function, which can be used to identify the equals acrossgroups in terms of their living standards. For group A, we could suppose that living standard isexpressed by money income, and then let S: + 6+ be the function which expresses the livingstandard of an income unit belonging to group B with income x. For n > 2 groups, n-1 suchfunctions would be needed. The equivalent income function has been proposed in this moregeneral setting by Donaldson and Pendakur (1999), and examined in detail in Ebert (2000). Themost obvious context for all of this is when A an B denote different household characteristics, e.g.A comprising singles and B couples. We shall use this example to interpret some of the resultswhich follow, but the setting is general enough for A and B to be two (groups of) countries. Thefunction S(x) need only be continuous and strictly increasing.

Thus a member of group A with income xA and a member of group B with income xB are equalsif and only if S(xB) = xA. Let the tax schedules for A and B be tA(x) and tB(x) respectively, andlet vA(x) = x – tA(x) and vB(x) = x - tB(x) be post-tax income functions. If by equal treatmentwe mean that those with the same pre-tax living standard should also have the same post-taxliving standard, this requires the following property: S(xB) = xA Ψ vA(xA) = S(vB(xB)); or,writing xB as x and substituting,

(3) S(vB(x)) = vA(S(x))

In words, the living standard after tax of a member of group B (e.g. a couple) having x before taxshould be the same as that of a member of A (single) having S(x) before tax.12

The equivalent income function for a constant relative equivalence scale is of the form S(x) = x/mwhere m is the deflator for the money incomes in B. This is the familiar scenario for equivalizinghousehold incomes. In a regional context, m could be a price deflator rendering region B moneyincomes into real terms as measured in A; then equals are those with the same real incomes (butsee on for other possible definitions of equals in this context). Substituting in (3), the horizontallyequitable tax for B, given a schedule tA(x) for A, must satisfy:

(4) vB(x) = m.vA(x/m)

In the context of families, this is precisely the quotient familial tax system, as used in France andLuxembourg and already anticipated by Vickrey (1947, pp. 295-6): “A more thoroughgoing and

12


12 If, on the other hand, ‘equal treatment’ would mean that pre-tax equals should experience equal average tax rates, the criterion would betA(S(x))/ S(x) = tB(x)/ x. If equal treatment is taken to mean equal tax payments, then tA(S(x)) = tB(x) is the criterion. For more on thesetwo, see Lambert (2003).

equitable procedure [than exemptions and credits] would be to set up some factor indicative of theneeds of the entire family, divide the total income by this factor, compute a per capita tax on this‘per capita income’, and multiply the tax so computed by the family size factor to obtain the totaltax for the family”. In the regional context, with m as the price deflator, it simply says that peoplein B should be taxed as they would be in A on the real value of their incomes.

The equivalent income function for a constant absolute equivalence scale is of the form S(x) = x-a, where a > 0 is a constant. Members of B (couples) need a fixed addition a to their income to bejudged equal to members of A (singles) at the same income level. Substituting in (3), thehorizontally equitable tax for B satisfies

(5) vB(x) = vA(x-a) + a

In the singles/couples context, the constant a is a married couple’s exemption or allowance:couples should receive the first a of their income tax-free and pay tax at the same rate as singleson the balance x-a (assuming x > a).

In the family context, if an income tax system is not of one of these two very straightforwardtypes, then it cannot be rationalized in terms of a constant relative or absolute equivalence scaleusing the equal treatment command in (3). Insofar as the British, Spanish and Canadian direct taxsystems are not of this type - and they certainly are not - the analyses of Aronson et al. (1994),Lambert and Ramos (1997) and Duclos and Lambert (2000) of HI in the UK, Spanish andCanadian direct tax systems, which are undertaken using constant relative equivalence scales, areindeed vulnerable to the accusation of “imposing HI from the outside”.

In order to design an equitable EU-wide additional layer of tax, an equivalent income function S:+ 6+ (or set of equivalent income functions Si: + 6+, 2 # i # n) would first be needed, torelate living standards in countries in group B at a given taxable income level with those in groupA (or to relate living standards in countries 2,3.... n with those in country 1, in the case of fullydifferentiated taxes). Then, setting tA(x) as the “reference” schedule which can embody anychosen degree of VE, tB(x) would have to be designed to satisfy (3) (or, ti(x) 2 # i # n would haveto be designed to satisfy (3) with respect to t1(x)) for full-blown equity. This kind of analysis hasnot been undertaken as yet, even in the family context,13 but a recent finding of Ebert and Lambert(2004) provides a potentially interesting way forward.

Suppose that the equivalent income function can be argued to take the isoelastic form, S(x) =(x/b)a where a > 0 and b > 0 are constants. Ebert and Lambert (2004) examine the consequencesfor the degree of progression faced by equals in this scenario. When the tax system tA(x), tB(x)is constructed to obey (3) and S(x) is isoelastic, members of group A with a given pre-tax livingstandard face the same degree of progression as members of group B with that same pre-tax livingstandard. Thus an extended concept of equity, equal progression among equals, is attainable inthis case (and in fact, in only the isoelastic case).14

Putting S(x) = (x/b)a into (3), and taking tA(x) as given, a formula for tB(x) results which is ingeneral complicated, but in the special case in which tA(x) is proportional, tA(x) = gAx say, we

13


13 In that context, if the Atkinson and Bourguignon (1987) SWF were adopted along with the equal sacrifice model, then the type-specific

utility functions UA(x) and UB(x) would define the equivalent income function S through the property UB(xB) = UA(S(xB)); that is, S =

UA-1 Β UB.

14 See op cit. for further details, and also Dardanoni and Lambert (2002). The progression measure is residual progression, defined for a

schedule t(x) as the elasticity of post-tax income v(x) = x - t(x) with respect to pretax income, i.e. as xvΝ(x)/v(x). In the family context, the

isoelastic form been recommended by Donaldson and Pendakur (1999) on positive grounds, as one which provides less restrictive

household demand functions than any constant equivalence scale m (viz. than S(x) = x/m for any m > 0), and can be uniquely estimated

from demand data.

find from (3) that for equity, tB(x) should also be proportional:

(6) tB(x) = gBx where gB = 1 - (1 - gA)1/a

In contradistinction to the result discussed in the previous section of the paper, we see from thisthat an EU-wide layer of differentiated proportional taxes could be supported as fully equitable ifa > 1. 15

An intriguing possibility arises if we suppose that the country-specific income distributions are alllognormal, or indeed, belong to any family of distributions which differ in logarithms only bylocation and scale.16 In precisely this case, an isoelastic function xB = S(xA) exists which matchesthe incomes in A and B position by position. If in this case we would assert that, in the differentEU countries, the equals are those at the same percentile point,17 then (3) can be used to specifyan EU-wide layer of additional income tax which both assures “equal treatment by percentile” inthe classical sense and also “equal progression by percentile”. Proportional EU taxes (which, aswe have said, have the advantage of not interfering with relative income differentials withincountries) would have to be differentiated to the extent that inequality differed between countries;the rate would be higher in less unequal countries, and lower in more unequal countries.18

6. Understanding a differentiated income tax It is more than likely that any eventual EU-wide layer of direct tax will be the outcome ofnegotiations by the politicians, each seeking to build in concessions for the country he or sherepresents. If such a tax package emerges, how may we infer its equity characteristic? In formalterms, the problem is this. Suppose that we the economists are presented, fait accompli, withdifferentiated schedules tA(x), tB(x) for countries classified into two groups A and B (or moregenerally with a bundle ti(x) : 1 # i # n of schedules, one for each country). Can we find anequivalent income function S(x) (or bundle of n-1 such functions Si(x) : 2 # i # n) such that (3)holds between the net income functions vA(x) and vB(x) (or the analogue of (3) between v1(x)and each vi(x), 2 # i # n)? If so, we can judge the package to be equitable, and explain to thepublic through an examination of S(x) the value judgement about equals across EU countries thatis implicit. If not, the EU-wide tax must be judged inequitable by the standard of our equaltreatment command (3).

In general, given tA(x), tB(x), no such function S(x) exists. A particular result of Lambert andYitzhaki (1997) deomonstrates this: in the family context, there exists no equivalent incomefunction S(x) compatible with the equal treatment criterion (3) for a tax system which corrects fordifferences in need by means of a tax credit.19

Of course, we can recognize the equity characteristic of a tax system tA(x), tB(x) satisfying (4),(5) or (6), since we drew these tax systems out of particular functional forms for S(x). We couldpush things a little further. If S(x) does exist for a tax system tA(x), tB(x), define m(x) and a(x)

14


15 It is clear from (6) that gB > gA if a < 1 and gB < gA if a > 1. The case a = 1 in the family context would be that of a constant relativeequivalence scale. In this case, if the singles are taxed proportionally, then for equity so should the couples be, and at the same rate. Thatis, a common proportional tax on the living standards of all households is equitable.16 Other examples include the Pareto and Singh-Maddala distributions.17 Clearly this would argue for the different tax treatment of an income unit in Germany having $60,000 p.a. and an income unit in Latviaalso having $60,000 p.a., a question we posed earlier. 18 If ln xA −N(:A, ΦA

2) and ln xB −N(:B, ΦB2) then S(x) = (x/b)a matches incomes position by position for a = ΦB/ΦA and b = exp (:A

- a.:B). If the Lorenz curves for A and B differ (so that ΦB ? ΦA, i.e. a ?1), and if tA(x) is proportional, then (as already noted) for “equityby percentile” tB(x) would be proportional too, but with a different rate. As observed in footnote 17, gB > gA if ΦB < ΦA and vice versa.19 The point is very simply shown. Let C be the married couple’s tax credit and substitute vB(x) = vA(x) + C into (3): S(vA(x) + C) =vA(S(x)). Supposing only that vA(x) # x ?x and that ?xo : vA(xo) = xo (i.e. that there are no single benefit recipients and there exists atleast one income level at which the tax liability for a single is zero), we find from (3) that S(xo + C) = vA(S(xo)), i.e. that vA(S(xo)) >S(xo), an immediate contradiction. Lambert and Yitzhaki (1997) reach the same conclusion in a more general model with n > 2 householdtypes.

by S(x) = x/m(x) = x - a(x). That is, regard the equivalizing procedure as generated by an income-related relative or absolute equivalence scale.20 Now apply (3):

(7) vB(x) = m(vB(x)).vA(x/m(x)) = vA(x-a(x)) + a(vB(x))

The interpretation of these forms is clearest in the family context. On the left, we have a quotientfamilial-type rule, in which the household is split into m(x) parts, each being taxed separately -but then the taxes are re-combined with a correction, for equity, to allow for the change in thescale value in the transition from pre-tax to post-tax income. On the right, we have an income-related married couple’s deduction: couples receive the first portion a(x) of their income x tax-free, and pay tax at the same rate as singles on the balance x-a(x), but again there is a correctionfor equity, an extra a(vB(x)) - a(x) of tax-free income being given to allow for the change in theabsolute equivalence scale value in the transition from pre-tax to post-tax income. If the putativeEU-wide tax system tA(x), tB(x)took either one of these two forms, it would be equitable forthe relevant income-related equivalence scale.

How should we analyze the VE and HI characteristics of an EU-wide income tax system forwhich we cannot recognize a function S(x) providing the definition of equals? We are left onlywith normative analysis. If the tax is in operation then, starting with sample data drawn from thejoint distribution of pre-tax and post-tax living standards across EU countries, we could impose aconstant relative scale (e.g. one rendering nominal incomes in B into real values in A’s terms),and assess the vertical and horizontal contributions to overall redistributive effect as in (2). Thiswould tell us to what extent the EU-wide tax departs from an idealized tax on citizens’ realincomes, and how unequally citizens at different real income values are dealt with on average.

In fact, a model of Ramos and Lambert (2003) may be appropriate here, which extends the HImeasurement framework of the 1990s by admitting some “deserving attributes” into the idealizedtax function, in addition to (real) income. The idea here would be to assess the performance of theEU-wide tax against that of an idealized tax schedule ϑ(e,d), where e is living standard and ddenotes domicile (in the EU context; but more generally d can be any attribute for which specialtax treatment has been sanctioned by the politicians or tax authorities). The measurement systempermits the inequity effect of sanctioned tax breaks to be distinguished from the inequity effect ofnon-sanctioned differences in tax treatment (if any), the former amounting to an authorized lossof VE, and the latter to a residual form of classical HI. Equation (2) becomes

(8) RE = [V - D] - J,

in this context, where, as in (2), V is the redistributive effect of the tax on average (in our case,on the real incomes of EU citizens), D denotes the loss of VE due to the country-specific factorsin the tax system, and J is the loss from non-domicile related differences in tax treatment apparentin the data (such as assessment and recording errors) at given real income levels.

Such a measurement exercise, intrinsically normative, can be said to “impose HI from theoutside” since it assumes a definition of equals known not to be implicit in the tax system, andmeasures departures of the tax from such an ideal. In a recent and very exciting development,Muellbauer and van der Ven (2002, 2003) have found a way forward for positive analysis, byvarying the definition of equals and the equal treatment command used in this paper to show that:“...tax and benefits systems are consistent with the equivalence scale methodology, even if theyare not designed in coherence with it”.

15


20 Income-related equivalence scales have been around in the equity context since at least Seneca and Taussig (1971), who remark that “themost interesting and important issues involving the application of equivalence scales to tax equity questions are intimately bound up withthe variation of equivalence scales with the level of income” (ibid, p. 255).

Thus far we have defined equals through the equivalent income function S(x), and equal taxtreatment by (3). Defining income-related equivalence scales m(x) and a(x) by S(x) = x/m(x) = x- a(x), we demonstrated how the equitable taxes (in (7)) would have to contain “equitycorrections” for the variations which take place in the scale value in the transition from pre-taxincome x to post-tax income vB(x). But if we think of m(x) and a(x) as constant for the couplewith gross income x – defined, for example, in terms of their ability or effort - that is, base-dependent rather than income-level-dependent21 - then a different equal treatment commandobtains. Specifically, the command becomes xA = xB/m(xB) Ψ vA(xA) = vB(xB)/m(xB) in therelative case and xA = xB - a(xB) Ψ vA(xA) = vB(xB) – a(xB) in the absolute case. These leadto a quotient-familial-type tax in the one case, and a tax with an income-related deduction in theother, but in each case no correction is required for equity:

(9) vB(x) = m(x).vA(x/m(x)) & vB(x) = vA(x-a(x)) + a(x)

(compare (7)).

The specifications in (9) cannot be represented by our (3). For members of group B, the samedeflator m* = m(xB) or subtraction a* = a(xB) is applied to create equals after tax as was usedto identify the equals before tax, which our (3) does not do. Muellbauer and van de Ven show thatthere exists a base-dependent equivalence scale which is implicit in (almost) any tax systemtA(x), tB(x) - the case of a family tax credit included - and they also provide an algorithm torecover that scale. 22

The Muellbauer and van der Ven construction thus rationalizes (almost) any tax system ashorizontally equitable, finding, by the use of positive analysis, the implied equivalence scale,which is base-dependent in general. According to this approach, HE is imposed from the outside,by assuming rationality on the part of the policymaker, and HI analysis all but ruled out.23

7. Concluding remarksIn this paper, we first explained the value judgements which underlie the concepts of VE and HE.We then described the measurement system that stood throughout the 1990s for capturing theextent of HI in a tax system, and showed that, in essence, invoking this methodology amounts to“imposing HI from the outside”. A common degree of aversion to both horizontal and verticalinequality is implicit in this methodology, and this is why the horizontal and vertical stances ofthe tax system are assessed commensurately. A more recent HI development, that of Auerbachand Hassett (2002), breaks the link between horizontal and vertical inequality aversion and bringspossibilities for deeper analysis.

We went on to discuss some of the issues that would face the designer of a new income tax, takingas a vehicle for this analysis a putative EU-wide income tax, additional to the national incometaxes of the Member States. By drawing on recent work in the regional context, we observed that,relative to a common tax on the (real disposable) incomes of all EU citizens, VE could beenhanced, without necessarily introducing HI, by admitting an element of differentiation in this

16


21 Muellbauer and van der Ven (2002) discuss this form of equivalizing in some detail in an optimal tax scenario, showing how it couldarise if the government’s objective is to relate needs to ability but cannot due to an unobservability constraint. 22 See Muellbauer and van der Ven (2003, pp. 86-90). As the authors remark, “..continuity, monotonicity and progressivity are sufficientfor the equivalence scale function [m(x) or a(x)] to be unique”.23 A residual term in Muellbauer and van der Ven’s econometric estimation procedure provides an upper bound for HI in their system. It isinteresting to note that the studies cited earlier, which “impose HI from the outside”, all assume VE, and detect it in terms of the stance ofthe tax “on average between unequals” (following Musgrave, 1990, in fact). In the NR strand of the literature, a suggestion of King (1983)is followed, such that the vertical stance of a tax system is given by breaking the disassociation between pre- and post-tax living standards,that is, by independently sorting pre- and post-tax living standards vectors and making a one-to-one mapping. In each case, analystseffectively “impose VE from the outside”. See Lambert (2001, chapter 10) for more on this

tax. By formally modelling the identification of equals in different countries through an‘equivalent income function’, we developed a criterion, in equation (3), for the equal treatmentcommand of HE, and discussed the design issues this raised. In particular, we showed that a newcommand, “equal progression among equals”, can be achieved if the equals in different countriesare defined as those at the same given percentile point in their within-country distributions, and ifthese distributions differ in logarithms only by location and scale. Differentiated proportionaltaxes would be equitable in this scenario, the flat rate being higher in less unequal countries.

Finally, we discussed in greater depth the detection of value judgements in an existing tax system,adducing recent work of Muellbauer and van der Ven (2002, 2003) which allows the analyst todraw out of the tax itself the implied value judgements of the decision makers, in terms of anequivalence scale which is in general “base dependent”. We observed that this development turnshorizontal tax analysis into a positive exercise, in essence “imposing HE from the outside”through the assumption of a rational tax designer.

ReferencesAronson, R., P. Johnson and P.J. Lambert (1994), “Redistributive effect and unequal income taxtreatment.” Economic Journal, vol. 104, pp. 262-270.

Aronson, J.R. and P.J. Lambert (1994), “Decomposing the Gini coefficient to reveal vertical,horizontal and reranking effects of income taxation.” National Tax Journal, vol. 47, pp. 273-294.

Atkinson, A.B. (1970), “On the measurement of inequality.” Journal of Economic Theory, vol. 2,pp. 244-263.

Atkinson, A.B. and F. Bourguignon (1987), “Income distribution and differences in needs.”Chapter 12 in G.R. Feiwel (ed.) Arrow and the Foundations of the Theory of Economic Policy.London: Macmillan.

Auerbach, A.J. and K.A. Hassett (2002), “A new measure of horizontal equity.” AmericanEconomic Review, vol. 92, pp. 1116-1125.

Bishop, J.A., K.V. Chow, J.P. Formby and C.-C. Ho (1994), “The redistributive effects of non-compliance and tax evasion in the US.” In J. Creedy (ed.) Taxation, Poverty and IncomeDistribution. Aldershot: Edward Elgar.

Cnossen, S. (2002), “Tax policy in the European Union: a review of issues and options.” WorkingPaper No. 758, CESifo, Munich.

Cubel, M. and P.J. Lambert (2002a), “Progression-neutral income tax reforms and horizontalinequity.” Journal of Economics, vol. 9 (supplement), pp. 1-8. (Special issue entitled Inequalities,Measurement and Applications, edited by P. Moyes, C. Seidl and A.F. Shorrocks).

Cubel, M. and P.J. Lambert (2002b), “A regional approach to income tax policy.” Public FinanceReview, vol. 30, pp. 124-143.

Cutler, D.M. and L. Katz (1992), “Rising inequality? Changes in the distribution of income andconsumption in the 1980’s.” American Economic Review, vol. 82, pp. 546-551.

D’Antoni, M. (1999), “Piecewise linear tax functions, progressivity and the principle of equalsacrifice.” Economics Letters, vol. 65, pp. 191-197.

Dardanoni, V. and P.J. Lambert (2001), “Horizontal inequity comparisons.” Social Choice andWelfare, vol. 18, pp. 799-816.

17


Dardanoni, V. and P.J. Lambert (2002), “Progressivity comparisons.” Journal of PublicEconomics, vol. 86, p. 99-122. Decoster, A., E. Schokkaert and G. van Camp (1997), “Horizontalneutrality and vertical redistribution with indirect taxes.” Research on Economic Inequality, vol.7, pp. 219-239.

Donaldson, D. and K. Pendakur (1999), “Equivalent-income functions and income-dependentequivalence scales.” Economics Discussion Paper No. 99-16, University of British Columbia,Canada.

Duclos, J.-Y. and P.J. Lambert (2000), “A normative and statistical approach to measuringclassical horizontal inequity.” Canadian Journal of Economics, vol. 33, pp. 87-113.

Ebert, U. (1997), “Social welfare when needs differ: an axiomatic approach.” Economica, vol. 64,pp.233-244.

Ebert, U. (2000), “Equivalizing incomes: a normative approach.” International Tax and PublicFinance, vol. 7, pp. 619-640.

Ebert, U. and P.J. Lambert (2004), “Horizontal equity and progression when equivalence scalesare not constant.” Public Finance Review, forthcoming.

Fellman, J. (1976), “The effect of transformations on Lorenz curves.” Econometrica, vol. 44, pp.823-824.

Gravelle, J.G. (1992), “Equity effects of the Tax Reform Act of 1986.” Journal of EconomicPerspectives, vol. 6, No. 1, pp. 27-44.

Hayes, K., P.J. Lambert and D.J. Slottje (1995), “Evaluating effective income tax progression.”Journal of Public Economics, vol. 56, pp. 461-474.

Jakobsson, U. (1976), “On the measurement of the degree of progression.” Journal of PublicEconomics, vol. 5, pp. 161-168.

Johnson, S. and T. Mayer (1962), “An extension of Sidgewick’s equity principle.” QuarterlyJournal of Economics, vol. 76, pp. 454-463.

Kakwani, N.C. (1977a), “Applications of Lorenz curves in economic analysis.” Econometrica,vol. 45, pp. 719-727.

Kakwani, N.C. (1977b), “Measurement of tax progressivity: an international comparison.”Economic Journal, vol. 87, pp. 71-80.

Kakwani, N.C. and P.J. Lambert (1995), “Vertical equity and horizontal inequity: a new approachto measurement” Working Paper No W95/2, Institute for Fiscal Studies, London.

King, M.A. (1983), “An index of inequality: with applications to horizontal equity and socialmobility.” Econometrica, vol. 51, pp. 99-115.

Lambert, P.J. (2001), The Distribution and Redistribution of Income (3rd edition). Manchester:University Press

Lambert, P.J. (2003), “Equivalence scales, horizontal equity and horizontal inequity.” Pages 75-84 in Dagum, C. and G. Ferrari (eds.) Household Behaviour, Equivalence Scales, Welfare andPoverty. Heidelberg: Physica-Verlag.

Lambert, P.J. and X. Ramos (1997), “Vertical redistribution and horizontal inequity.”International Tax and Public Finance, vol. 4, pp. 25-37.

18


Lambert, P.J. and S. Yitzhaki (1994), “Equity, equality and welfare.” European EconomicReview, vol. 39, pp. 674-682.

Lambert, P.J. and S. Yitzhaki (1997), “Income tax credits and income tax exemptions.” EuropeanJournal of Political Economy, vol. 13, pp. 343-351.

Loomis, J.B. and C.F. Revier (1988). “Measuring regressivity of excise taxes: a buyers index.”Public Finance Quarterly, vol. 16, pp. 301-314.

Manser, M.E. (1979), “Comparing households with different structures: the problem of equity.”American Economic Review, vol. 69, 222-226.

Mitra, T. and E.A. Ok (1996), “Personal income taxation and the principle of equal sacrificerevisited.” International Economic Review, vol. 37, pp. 925-948.

Mitra, T. and E.A. Ok (1997), “On the equitability of progressive taxation.” Journal of EconomicTheory, vol. 73, pp. 316-334.

Muellbauer, J. and J. van de Ven (2002), “Estimating equivalence scales for tax and benefitsystems.” Working Paper, Nuffield College, Oxford.

Muellbauer, J. and J. van de Ven (2003), “Equivalence scales and taxation: a simulationanalysis.” Pages 85-106 in Dagum, C. and G. Ferrari (eds.) Household Behaviour, EquivalenceScales, Welfare and Poverty. Heidelberg: Physica-Verlag

Musgrave, R.A. and P.B. Musgrave (1984), Public Finance in Theory and Practice (fourthedition). New York: McGraw Hill.

Musgrave, R. A. (1990), “Horizontal equity, once more.” National Tax Journal, vol. 43, No. 2(June), pp. 113-122.

Ok, E.A. (1995), “On the principle of equal sacrifice in income taxation” Journal of PublicEconomics, vol. 58, pp. 453-467.

Plotnick, R. (1981), “A measure of horizontal inequity.” Review of Economics and Statistics, vol.63, pp. 283-288.

Ramos, X. and P.J. Lambert (2003), “Horizontal equity and differences in income tax treatment:a reconciliation.” Research on Economic Inequality, vol. 10, pp. 45-63.

Rosen, H.S. (1978), “An approach to the study of income, utility and horizontal equity.”Quarterly Journal of Economics, vol. 92, pp. 307-322.

Runciman, W.G. (1966), Relative Deprivation and Social Justice. London: Routledge and KeganPaul/Penguin Books.

Samuelson, P.A. (1947), Foundations of Economic Analysis. Cambridge, MA: HarvardUniversity Press.

Seneca, J.J. and M.K. Taussig (1971), “Family equivalence scales and personal income taxexemptions for children.” Review of Economics and Statistics, vol. 53, pp. 253-262.

Steuerle, E. (1983), “The tax treatment of households of different size.” In R.G. Penner (ed.)Taxing The Family, Washington D.C.: American Enterprise Institute.

Steuerle, E. (1999). And Equal (Tax) Justice For All. Part Two: Horizontal and Vertical Equity.Washington D.C.: Urban Institute. URL http://www.urban.org A study in 8 parts, with IDs of1000150 to 1000143 respectively.

19


van de Ven, J., J. Creedy and P.J. Lambert (2001), “Close equals and calculation of the vertical,horizontal and reranking effects of taxation.” Oxford Bulletin of Economics and Statistics, vol. 63,pp. 381-394.

Vickrey, W.S. (1947). Agenda for Progressive Taxation. New York: Ronald Press.

20


Pollution Taxes in a Second-Best WorldGilbert E. Metcalf*

Abstract The purpose of this paper is to analyse recent issues involved in setting environmental taxes in a second-best

world. This is an area that has seen an explosion of research and new insights over the past decade and also an area with

which many EU countries (as well as candidate EU countries) have been grappling. The basic message of this paper is

that the policy prescriptions that most of us learned when studying environmental policy in isolation (that is, in partial

equilibrium) often must be significantly adapted once one moves to a general equilibrium framework with pre-existing

distortions. Put this way, there is nothing novel here; it is simply a restatement of the Theorem of the Second Best (Lipsey

and Lancaster (1956-1957)). This, however, risks trivializing the literature of the past decade. More important, the new

literature has clarified our understanding of environmental distortions and policy making in important ways and opened

up whole new areas of analysis.

1. Optimal Taxation of Environmental DamageLet me begin by noting a concept that gained considerable currency in the late 1980s and early1990s. It is the idea of the “Double-Dividend Hypothesis.” The “double-dividend hypothesis”suggests that increased taxes on polluting activities can provide two kinds of benefits. The firstdividend is an improvement in the environment, and the second dividend is an improvement ineconomic efficiency from the use of environmental tax revenues to reduce other distortionarytaxes.1 This is a relatively uncontroversial idea though it led some policy analysts to policyprescriptions that could not be supported by the theory. Some policy analysts argued that if theoptimal tax rate on environmental damages in a first-best world equals social marginal damages(the prescription due to Pigou (1932)), then it must be the case that the tax rate should be higherin a world with distortionary taxes if we could use that revenue to lower those taxes. After all, sogoes the argument, we’re getting an additional benefit from the tax. Therefore, we should relymore heavily on this instrument. This is incorrect for the simple reason that an environmental taxbrings with it its own distortions (aside from its environmental impact).

This can be illustrated with a simple diagram taken from Fullerton and Metcalf (1998). Themodel is a simple one. A single factor of production (labor) is used to produce a clean good (C)and a dirty good (D). The factor is paid its value of marginal product. Consumers can choose tosupply labor or consume leisure, using their labor income to purchase the two commodities.Consumers obtain lower utility as pollution rises. Finally government must raise a given amountof revenue for a fixed government use. Its tax instruments include a tax on wage income or a taxon the good producing pollution.

Figure 1 illustrates the equilibrium prior to the imposition of an environmental tax (here a tax onD). There is only a tax on wage income initially. In figure 1, labor is measured along thehorizontal axis. The real wage is measured along the vertical axis. Prior to any environmentalpolicy, the equilibrium occurs at L0 where Wn

0 = (1-tw)Wg. The triangle A measures thedeadweight loss arising from the tax on labor.

Now let’s consider the imposition of an environmental tax (here a tax on the dirty good). Twothings happen. First, the new environmental tax revenue allows for a reduction in the laborincome tax rate and a consequent increase in the real net wage. This revenue effect has been

21

Pollution Taxes in a Second-Best World

∗ Tufts University and NBER. Plenary Address for the Conference on Tax Policy in EU Candidate Countries On theEve of Enlargement. Riga, Latvia. September 12, 20031 Pearce (1991) appears to be the first person to use the “double-dividend” terminology in print. See Fullerton andMetcalf (1998) for a history of this concept dating back to Tullock (1967).

termed the revenue-recycling effect by Bovenberg and Goulder (2002), drawing on terminologyfrom Parry (1995). But in addition, the general price index (a function of the prices of the twocommodities) rises since the dirty good will now be more expensive, what Bovenberg andGoulder term the tax-interaction effect. Bovenberg and de Mooij (1994) show that this lattereffect more than offsets the revenue effect leading to a reduction in the real net wage from Wn

0

to Wn1 and a consequent fall in labor supply from L0 to L1. With that fall in labor supply comes

an increase in deadweight loss equal to the trapezoid B in the figure. The punch line is that whilethere are benefits from the environmental tax (reduced pollution) there are also costs (exacerbateddistortions in other markets). In rough terms, we’re adding DWL onto the wide end of a DWLtriangle.

In Metcalf (2003), I develop a simple analytic general equilibrium model to investigate theoptimal tax structure further. In particular, a representative agent obtains utility from two goods,one of which is associated with pollution (the dirty good). The agent also obtains utility fromleisure and disutility from pollution. Labor is the only factor of production and is used to producethe two private commodities and a fixed amount of a government good. Following much of theprevious research, I assume that utility over C and D is weakly separable from leisure andenvironmental quality and that this sub-utility function is homothetic. If the government raisesrevenues from commodity taxes, the optimal relationship between the optimal tax rates is

(1) t*D=t*C+(1-ετ*C)τ

where is the uncompensated labor supply elasticity and is the social marginal damages frompollution (measured in private income). First note that if there is no environmental problem, thenthe optimal commodity tax rates would be equal. This follows from our assumption of weakseparability and homotheticity in consumption. Therefore, we can view the difference betweentD and tC at the optimum as a Pigouvian tax increment. In other words, if the optimal tax rate onthe clean good is 20 percent and that on the dirty good 25 percent, then the Pigouvian taxincrement (required to achieve the optimal level of pollution) is five percent.

Equation (1) yields three useful bits of information in the presence of environmental damages.First, suppose that environmental tax revenues are sufficient to cover government expenseswithout a tax on the clean good (t*C = 0) . In this case, the tax on D (as well as the difference, tD-tC) exactly equals τ. This is the Pigouvian rule in a first-best situation. Second, even if a tax onC is required, the first best rule still holds so long as ε equals zero. Third, if neither of theseconditions hold, then the Pigouvian tax increment (t*D-t*C) falls short of τ so long as εtC ispositive. In other words, those who argued that the existence of a double-dividend means that thePigouvian tax should exceed social marginal damages were incorrect.

Bovenberg and de Mooij (1994) along with Parry (1995) were the first to point this out. AsBovenberg and de Mooij note

In this way, high costs of public funds crowd out not only ordinary publicconsumption but also the collective good of the environment. (p. 1088).

This statement is not precisely correct and I return to this point below. But the broader point thatBovenberg and de Mooij were making is very much correct: that the distortions associated withan environmental tax are of first-order importance in the face of other pre-existing distortions andcannot be ignored as they could if there were no other taxes in effect. Returning to the figure, the

22


DWL associated with the environmental tax is a trapezoid added on to the side of a DWL triangle.If the DWL triangle does not exist (no pre-existing distortion), there is no increment to the triangleto worry about. Note also the importance of the uncompensated labor supply elasticity (ε). AsBallard (1990) has noted, compensated elasticities are relevant for differential incidence analyseswhile uncompensated elasticities are relevant for balanced budget incidence analyses. The pointis that there are income effects with balanced budget policy changes. While we are in factundertaking a differential tax incidence analysis, the change in environmental quality has effectsanalogous to income effects making the uncompensated elasticity the relevant parameter ofinterest.

2. Prices versus Quantities: Some Comparative Statics2

The statement from Bovenberg and de Mooij that I quoted above has led to some confusionamong economists. Some have interpreted it to mean that in a second-best world, the optimalamount of environmental quality would fall. This, in turn, has led some analysts to worry thatpolicy makers might rely on this new second-best literature to weaken laws protecting theenvironment. One must draw a careful distinction between “price” questions (“Is the optimal taxon pollution higher or lower than social marginal damages?”) and “quantity” questions (“Is theoptimal level of environmental quality higher or lower in the presence of other distortions than itwould be in a world with no other distortions?”).3 The simple model that I sketched out abovecan be useful to understand that distinction. Imagine that government is setting tax ratesoptimally according to equation (1).4 And now imagine that the government needs someadditional revenue. How does the need for more revenue affect:

(1) the optimal Pigouvian tax increment t*D-t*C and

(2) the optimal level of environmental quality (E) which is a function of the aggregate supply ofthe dirty good?

Assuming that social marginal damages and the factor supply elasticity are (to a first orderapproximation) constant, we can differentiate equation (1) to get the answer to our first question:

(2) d(tD-tC)=-ετdtC

Under reasonable conditions and ruling out any Laffer tax effects, it is easy to show that sgn(dtD)= sgn(dtC) = sgn(dG) > 0. With dtC > 0, equation (2) indicates that the Pigouvian tax incrementfalls as G rises (so long as ε > 0). An increase in required distortionary tax revenues does notfavor increased taxation of the dirty good relative to the clean good.

The intuition underlying this result is quite simple. Sandmo (1975) showed that the optimal taxon a polluting good is a weighted average of a Ramsey efficiency component and marginalenvironmental damages (MED). As government revenue needs increase, the weight on theRamsey component rises and the weight on the environmental component falls. With separabilitybetween leisure and consumption goods, the optimal Ramsey components on the two goods areequal. Thus an increase in the Ramsey weight leads to a decrease in the difference between thetwo tax rates (i.e. the Pigouvian tax increment).

23


2 This section draws heavily on Metcalf (2003).3 The importance of this distinction has been pointed out in the public goods literature by Atkinson and Stern (1974).4 Equation (1) and the budget constraint are sufficient to pin down the actual tax rates.

Having answered the “price” question, I now turn to the “quantity” question. Note that thediversion of resources from the private to the public sector directly affects the environment to theextent that public services themselves may pollute more than the mix of private goods reduced.For example, if public services are entirely clean, the expansion of the government sector willlikely lead to a cleaner environment since the increased government output has no impact on theenvironment. To avoid this demand side effect, I assume that government spends its revenue onthe same mix of clean and dirty goods as does the private economy.5 In Metcalf (2003) I showthat environmental quality falls if

(3)

The parameter πC is the share of consumer expenditures on the clean good and σ is the elasticityof substitution between the clean and dirty good in consumption. Some simple algebra shows thatthe first expression in equation (3) is positive. Regarding the second term, labor supply will fallas the real wage falls so long as labor supply is not completely inelastic. In this model, the realwage will fall since the general price index rises (with the increase in the tax rate on both the cleanand the dirty good).

Now we can understand the forces that affect environmental quality. The first term in (3) is acommodity substitution effect. As the Pigouvian tax increment falls, consumers will substitutefrom C to D. The strength of this effect depends on the elasticity of substitution in consumption(σ). This substitution effect will work towards reducing environmental quality. The second termis a leisure substitution effect and reflects the fact that the increase in taxation will lead to asubstitution away from both produced goods towards leisure. Since leisure (in this model) is aclean commodity, this effect serves to improve environmental quality. Whether an increase ingovernment spending financed by increased taxes leads to a fall or rise in environmental qualitydepends on the relative size of the two substitution effects.

While the model here is quite simple, the basic point is more general. Decreases in the Pigouviantax increment as public revenue needs rise will affect environmental quality through thecommodity substitution channel. But additional channels also affect the supply of environmentalquality. In this model, the second channel is a leisure demand channel. A more realistic modelwould include other factor markets as well as additional commodity markets. Additional realismand complexity do not affect the central point that knowledge of the direction of changes inoptimal environmental tax rates due to changes in the economy is not sufficient for understandingthe impact on environmental quality.

Finally, lest I leave you with the impression that Bovenberg and his co-author did not understandthis distinction, I should point out that Bovenberg and van der Ploeg (1994) were careful tomeasure both price and quantity effects in an analysis of various environmental policies theystudy in a analytic general equilibrium setting.

24


5 This is the approach taken in Harberger (1962) to rule out demand side effects in his classic analysis of the incidenceof the corporate income tax.

σπ dLdt1+tC

CC

D

( (dt1+t L+ >0

3. Are Environmental Taxes Environmental Taxes?In the model above, there was a one-to-one correspondence between the amount of the dirty goodand pollution. In most cases, pollution is a by-product of production (or consumption) and thereare various ways to affect the ratio of pollution to output (or consumption). It is convenient forthe economist to model pollution as an input in production. If Z is pollution, and output (Q) isproduced with capital (K) and labor (L), then

(4) Q = f(K, L, Z)

Pollution abatement can be viewed as substitution out of Z and into K (and possibly L). This isa useful way to model pollution because it highlights the two channels by which a tax candiscourage pollution. If we levy a tax on Q, we will discourage the production of Q and thereforereduce the demand for all the inputs in production — including pollution. If we levy a tax onpollution directly, we will get this output effect as the price of Q inevitably rises in response tothe higher cost of production but we will get an additional substitution effect as firms substituteout of Z and into K and L. In other words, they have an incentive to engage in pollution abatementor avoidance strategies.

Unfortunately, most so-called environmental taxes are more often taxes on goods associated withpollution (taxes on Q) rather than taxes on pollution directly. Fullerton (1996) reviewsenvironmental taxes in the United States in the mid-90s and notes that not only are the U.S.environmental taxes not environmental taxes in the textbook sense, they are typically levied atvery low rates relative to the value of production and impose high administrative costs. Thisshould not make anyone feel sanguine about the importance of economic input into policymaking. A cursory review of environmental taxes in the EU suggests that the EU has not beenany more effective at targeting environmental taxes precisely than the United States. Todsen andSteurer (2002) note that transport and energy taxes comprise about 98 percent of EUenvironmental tax revenues of 228 billion euros in 1999. Specific pollution taxes account for lessthan 2 percent of EU environmental tax revenues and are only a significant tax instrument (interms of total collections) in the Netherlands, Denmark, and Belgium.

That this is so is not due to the failure to understand the issue. There are good reasons for whyactual taxes might miss the target. First, actual policy may not fully appreciate the importance ofhitting the target. Policymakers may have been concerned primarily with equity considerations,trying to ensure that polluting industries are made to pay for pollution — without realizing thatthe form of these taxes affect incentives to reduce pollution. Second, actual emissions may bedifficult or impossible to measure. In these cases, the best available tax may apply to ameasurable activity that is closely correlated with emissions. To reduce vehicle emissions, forexample, the gasoline tax may be the best available instrument. Third, the technology of emissionmeasurement is improving over time. Policymakers may be slow to adjust the tax base to reflectthe newly-reduced cost of measuring a particular pollutant.

If emissions cannot be monitored at reasonable cost, and policy is limited to a tax on the outputof the polluting industry, then how should that tax rate be set? One might think that theimperfection of this blunt instrument would reduce the optimal rate of tax.

In Fullerton, et al. (2001), we show that is not the case: the second-best output tax should be setto capture the exact same output effect that would have been captured by the emissions tax. If theunavailable emissions tax would have raised output price by 12 percent, for example, then theoutput tax should be set to 12 percent.

25


Finally, we calculated the incremental effects on welfare of slight increases in any pre-existingoutput tax or emissions tax, and we show the “welfare gap” of an imperfectly targetedenvironmental tax relative to a precisely targeted one. We found for plausible parameter valuesthat the welfare gain from an initial emissions tax was more than twice the gain from an initialoutput tax. This cost of missing the target did not depend on the size of the pre-existing outputtax, or on the size of the elasticity of substitution in utility, but it did depend on the elasticity ofsubstitution in production. A larger ability to substitute between emissions and other inputs inproduction substantially raises the importance of hitting the target.

Figure 2 illustrates this result. The horizontal axis indicates the level of an output tax on the dirtygood prior to a reform. (There is also a pre-existing labor tax.) The vertical axis then measuresthe welfare gain (or loss) from a small change in either the output tax (dashed line) or theemissions tax (solid line). There are three points to note. First, the emissions tax everywhereoutperforms the output tax. Not surprising since the emissions tax brings both the substitutionand output effects to bear on the pollution problem while the output tax only relies on the outputeffect. Second, the emissions tax is welfare enhancing across the entire range of pre-existingoutput tax rates. This is by no means a universal result but unless the tax system is seriously outof balance, a small environmental tax should be welfare enhancing. (More on this below.)Finally, there is a large range over which the output tax - levied to reduce pollution - is welfarereducing despite the environmental gains. In other words the increase in distortions in the taxsystem more than outweigh the environmental benefits.

4. Environmental Instruments More GenerallyAfter the initial spate of papers following Bovenberg and de Mooij, researchers began to directtheir interest more broadly from taxes to emissions quotas, tradable permits, and other regulatoryinstruments. Goulder, et al. (1997) studied revenue raising versus non-revenue raisinginstruments in the context of U.S. SO2 policy. Two results from that study are worth noting.First, they showed that raising revenue is not a sufficient condition for obtaining an efficiencygain over a non-revenue raising instrument.6 For example, an environmental tax is a revenue-raising instrument but if the revenues are given back lump-sum, it is conceptually identical in itseffects to a pollution quota scheme with permits which are given to polluting firms at no cost.(The conceptual equivalence requires a few qualifications: primarily the lump-sum distributionsmust be the same in both cases). Second, the efficiency advantage of taxes over grandfatheredpermits declines with pollution abatement. In the limit where there is 100 percent pollutionabatement, the two policies are identical. The reason is quite simple. With 100 percentabatement, the pollution tax raises no revenue!

Fullerton and Metcalf (2001) extended the Goulder, Parry, and Burtraw result to show that notonly was revenue-raising not sufficient to guarantee a welfare improvement, it was also notnecessary. In that paper we provided comparable analyses of various environmental policies andfound that the same welfare-raising effects of environmental protection could be achieved by atax that raises revenue, a CAC technology restriction that raises no revenue, and even a subsidythat costs revenue. Thus, raising revenue is not necessary for raising welfare. Instead, theexacerbation of the pre-existing tax distortion is associated with policies that generate privately-retained scarcity rents. Such policies include both the quantity-restricting CAC regulation and themarketable permit policy in which the permits are given to existing polluters. The problem withsuch policies is that the output price must rise by more than necessary to cover the cost ofabatement technologies; indeed we show that price must rise by an additional amount equal to

26


6 Strictly speaking this holds for new taxes only.

scarcity rents that arise as a result of the emissions restrictions. The higher output price reducesthe real net wage and exacerbates the labor tax distortion. That higher price is not such a problemif government captures the rents by using a pollution tax or by selling the permits, because thenthe labor tax can be reduced.

In the framework that Fullerton and I set up, the ability to identify and capture rents is key.Identifying who captures the rents in the absence of government capture is not alwaysstraightforward. Busse and Keohane (2003) have written a fascinating paper on the experiencefrom creating and giving away permits from the SO2 trading program arising from Title IV of the1990 Clean Air Act Amendments. The restrictions on sulfur dioxide emissions that result fromthis program create a barrier to entry that generates rents, rents that are capitalized in the value ofthe tradable SO2 permits. Busse and Keohane note that there are three candidates for the captorsof those rents: the electric utilities, the producers of low-sulfur coal, and the railroads thattransport low-sulfur coal. Phase I of the tradable permits program (1995-1999) coincided withthe opening up of a large reserve of low-sulfur coal in the western part of the country. That coalprovided a low cost alternative to burning high-sulfur coal that required the use of a large numberof SO2 permits. Busse and Keohane note that the western coal mines might have captured therents by raising the price of their coal. Alternatively, the railroad could capture the rents byraising rail rates. They argue that the railroads captured the rents through what was essentially atwo-part tariff consisting of an increase in fixed transport fees and reduced marginal transportcosts. While it may not be easy to identify who captures the rents, it is straightforward for thegovernment to capture the rents simply by auctioning the permits.

It should be noted that the choice with tradable permits is not limited to auctioning them or givingthem away. Bovenberg and Goulder (2001) carry out a computable general equilibrium analysisof a CO2 abatement policy carried out with tradable permits where industries are given enoughpermits to prevent a loss in the value of the firm (what the authors call “equity-value neutrality”)and sold the rest. The policy is designed to be equivalent to a $25 per ton carbon tax in the year2000. They find that only 15 percent of the permits need be grandfathered in the oil and gasindustry and only 4 percent in the coal industry. Equity-value neutrality can therefore be achievedat relatively low cost. Policy makers should be mindful of this when industry argues for completegrandfathering to preserve the equity value of the firm. Complete grandfathering is likely toovercompensate firms in a very big way.

5. Tax Distortions and Global Climate Policy7

Lastly I turn to the topic of policies to reduce global carbon emissions. At Kyoto, Annex B8

Parties committed to reducing, either individually or jointly, their total emissions of sixgreenhouse gases (GHGs) by at least 5 percent within the period 2008 to 2012, relative to thesegases’ 1990 levels.

The European Union (EU), as a signatory to the Kyoto Protocol, has accepted a quantitativeabsolute reduction of 8 percent from 1990 levels of its GHG emissions. Article 4 of the Protocolallows the EU to allocate its target among the Member States. A political agreement on thatredistribution was reached at the environmental Council meeting in June 1998, and is referred toas the “Burden Sharing” Agreement (BSA). Table 1 shows the BSA adopted at the environmentalCouncil meeting by Member States on June 1998. The sharing scheme specifies emissions targets

27


7 This section draws heavily on Babiker, et al. (2003)8 Annex B refers to the group of developed countries comprising of OECD (as defined in 1990), Russia and the EastEuropean Associates.

for each member country with the objective to reflect opportunities and constraints that vary fromone country to another, and to share “equitably” the economic burden of climate protection.

Working with colleagues at MIT, I investigated the degree to which implementation of Kyototargets through a carbon tax could lead to “weak” or “strong” double-dividends. Goulder (1995)distinguishes a “strong” and “weak” double dividend as follows. A strong double dividend occurswhen welfare is increased in response to an environmental tax regardless of the improvement inenvironmental quality. Given the great difficulties associated with quantifying the economicbenefits of an improved environment, a strong double dividend is appealing in that a case can bemade for an environmental tax without having to worry about the magnitude of the environmentalgains. It is possible for welfare (net of environmental improvements) to increase in response to agreen tax reform if the environmental tax revenues are used to lower a particularly egregiousdistorting tax. This simply points out the obvious fact that any tax reform to replace a highlydistorting tax with a less distorting tax is, in general, a good idea.

A “weak” double dividend occurs when the welfare improvement from a tax reform whereenvironmental taxes are used to lower distorting taxes is greater than the welfare improvementfrom a reform where the environmental taxes are returned in a lump sum fashion. A generalconsensus has emerged that the weak double dividend is an uncontroversial idea; it just says thatlowering a distorting tax is better than simply handing out the money. In my work with my MITcolleagues, I show, however, that in an economy with multiple distortions, a weak doubledividend need not occur. Moreover, we argue that climate policies under consideration inresponse to global warming will likely not provide a weak double dividend in a number ofEuropean countries.

The findings are obtained using MIT’s Emissions Prediction and Policy Analysis (EPPA) model,a recursive dynamic multi-regional general equilibrium model of the world economy that has beendeveloped for analysis of climate change policy. EPPA is built on a comprehensive energy-economy data set (GTAP4-E9) that accommodates a consistent representation of energy marketsin physical units as well as detailed accounts of regional production and bilateral trade flows. Thebase year for the model is 1995 and it is solved recursively at 5-year intervals.10 For our project,we used a new version of the model (EPPA-EU) including a breakdown for the European Unionso that we could model the BSA. A significant advantage of this approach compared withprevious work is that a common method and data set is applied to all countries and the cross-country results are thus comparable.

Table 2 provides our key results from that paper. The first scenario (labeled NRP) returns carbontax revenues to the representative agent in a lump-sum fashion. Not surprisingly, carbonreductions in this case reduce welfare relative to the reference scenario. The next two columnsprovide results for different tax reductions. In no case does welfare rise relative to the referencescenario. In other words, a strong double dividend is not possible in any of the EU countries orthe United States and Japan as a result of a carbon tax to achieve Kyoto. The use of carbon taxesto reduce labor taxes does give rise to a weak double dividend. Welfare losses under the LRPscenario are always lower than under the NRP scenario. This result is consistent with otherstudies that have found a weak double dividend when recycling carbon revenues to reduce labortaxes.

Interestingly, the weak double dividend does not hold in all cases when carbon tax revenues areused to lower non-energy consumption taxes (CRP). France, the Netherlands, Spain, and REU

28


9 For description of the Global Trade Analysis Project (GTAP) database see Hertel (1997).10 A full documentation of EPPA is provided in Babiker, et al. (2000).

are all better off with lump-sum recycling of the carbon tax revenues than if the alternative is toreduce non-energy consumption taxes. The failure of the weak double dividend to hold simplyreflects the existence of distorting energy consumption taxes that have not been reduced in thispolicy experiment. Intercommodity distortions are increased by a selective reduction inconsumption taxation; second best considerations mean that the weak double dividend is not auniversal phenomenon.11

6. ConclusionI have only touched on a number of interesting and important issues that have been investigatedover this past decade. In particular there are important distributional and political economyconsiderations that I have ignored. Let me end with a final point. My cursory review shouldsuggest that the complexity of instrument design in a second-best world means that economistswith computer models of the economies of countries and regions will play an important role inguiding policy makers to sensible environmental policy prescriptions. A further insight from mywork with the MIT modelers on global climate change has been that there is great variation acrosscountries in the efficiency impacts of different environmental policies. This suggests that oneshould be cautious about drawing conclusions from studies based on data from the United States,for example, on appropriate policies for Europe. Equal caution may be warranted about drawingconclusions from EU studies about appropriate policies for individual candidate EU countries.Economists studying the economies of the EU candidate countries should — as much as possible— carry out country specific analyses to guide them in devising country specificrecommendations. While tax harmonization in the EU can be a valuable goal, it is still importantto model and allow for the great variation within the EU, a variation that will only increase withthe expansion of the Union.

29


11 A carbon policy reduces the real net wage by raising the price of energy consumption goods. This reductionexacerbates pre-existing distortions in the labor market One key difference between the LRP and CRP policies is thatlabor tax recycling mitigates this reduction by explicitly reducing the tax burden on wages. The CRP policy mitigatesthe reduction by reducing other consumption taxes. This mitigation comes at the cost of higher intercommidity taxdistortions as the already large tax wedge between energy and non-energy consumption goods increases.

Figure 1. Welfare Losses From Distortionary Taxation

Figure 2. General Welfare Effects of Changes in Emissions or Output Taxes with Pre-ExistingLabour and Output Taxes

30


Wn1

Wn0

Wg

L' Lo

B A

D

S

Wage

L

-0.0004

-0.0002

0.0000

0.0002

0.0004

0.0006

0.0008

0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20 0.22 0.24 0.26 0.28 0.30

Pre-Existing Output Tax

Net

Effe

cts o

n W

elfar

e Change in Emissions Tax

Change in Output Tax

31


Table 1. Burden Sharing Agreement for 2010

Country Base 1990 = 100 Austria 87.0 Belgium 92.5 Germany 79.0 Denmark 79.0 Spain 115.0 Finland 100.0 France 100.0 United Kingdom 87.5 Greece 125.0 Ireland 113.0 Italy 93.5 Luxembourg 72.0 Netherlands 94.0 Portugal 127.0 Sweden 104.0 Total European Union 92.0 Source: Babiker, et al. (2003)

32


Table 2. Welfare Changes with Recycling

NRP LRP CRP USA -0.65% -0.49% -0.57% JPN -0.62% -0.56% -0.54% GBR -1.05% -0.97% -0.91% DEU -0.77% -0.69% -0.55% DNK -3.82% -3.54% -3.23% SWE -3.46% -3.27% -3.03% FIN -1.86% -1.67% -1.45% FRA -0.70% -0.64% -0.76% ITA -1.26% -1.08% -1.22% NLD -4.67% -4.45% -4.87% ESP -3.13% -3.01% -3.32% REU -1.27% -1.17% -1.44% OOE -1.96% -1.88% -1.84% Average -1.94% -1.80% -1.83% Welfare changes are relative to the reference scenario. Average is an unweighted average of the changes for the countries or country groups NRP - Lump-Sum Recycling LRP - Labor Tax Recycling CRP - Non-Energy Consumer Tax Recycling Source: Babiker, et al. (2003)

ReferencesAtkinson, Anthony, and Nicholas Stern (1974), “Pigou, Taxation, and Public Goods.” Review ofEconomic Studies 41: 119-28.

Babiker, Mustafa, Gilbert E. Metcalf, and John Reilly (2003), “Tax Distortions and GlobalClimate Policy.” Journal of Environmental Economics and Management 46: 269-87.

Babiker, Mustafa, John Reilly, Richard S. Eckaus, and Ian Sue Wing (2000), “The MITEmissions Prediction and Policy Analysis (EPPA) Model: Revisions, Sensitivities, andComparisons of Results.”. Cambridge, MA.: MIT Joint Program on the Science and Policy ofGlobal Change.

Ballard, Charles L. (1990), “Marginal Welfare Cost Calculations.” Journal of Public Economics41: 263-76.

Bovenberg, A. Lans, and Ruud de Mooij (1994), “Environmental Levies and DistortionaryTaxation.” American Economic Review 94: 1085-89.

Bovenberg, A. Lans, and Laurence Goulder (2001), “Neutralizing the Adverse Industry Impactsof Co2 Abatement Policies: What Does It Cost?” In Distributional and Behavioral Effects ofEnvironmental Policy, edited by Carlo Carraro and Gilbert E. Metcalf, 45-85. Chicago:University of Chicago Press.

Bovenberg, A. Lans, and Lawrence H. Goulder (2002), “Environmental Tax Policy.” InHandbook of Public Economics, edited by Alan J. Auerbach and Martin Feldstein, 1471-545.Amsterdam: Elsevier Press.

Bovenberg, A. Lans, and Fredrick van der Ploeg (1994), “Environmental Policy, Public Financeand the Labour Market in a Second-Best World.” Journal of Public Economics 55, no. 3: 349-90.

Busse, Meghan R., and Nathaniel O. Keohane (2003), “Pollution Control and InputMarkets:The Creation and Capture of Rents from Sulfur Dioxide Regulation.”. New Haven:Yale School of Management.

Deaton, Angus (1979), “The Distance Function in Consumer Behaviour with Applications toIndex Numbers and Optimal Taxation.” Review of Economic Studies 46, no. 3: 391-405.

Fullerton, Don (1996), “Why Have Separate Environmental Taxes?” In Tax Policy and theEconomy, edited by James M Poterba, 33-70. Volume 10. Cambridge: MIT Press for theNational Bureau of Economic Research.

Fullerton, Don, Inkee Hong, and Gilbert E. Metcalf (2001), “A Tax on Output of the PollutingIndustry Is Not a Tax on Pollution: The Importance of Hitting the Target.” In Behavioral andDistributional Effects of Environmental Policy, edited by Carlo Carraro and Gilbert E. Metcalf,13-38. Chicago: University of Chicago Press.

Fullerton, Don, and Gilbert E. Metcalf (1998), “Environmental Taxes and the Double DividendHypothesis: Did You Really Expect Something for Nothing?” Chicago-Kent Law Review 73,no. 1: 221-56.

——— (2001), “Environmental Controls, Scarcity Rents, and Pre-Existing Distortions.”Journal of Public Economics 80, no. 2: 249-67.

Goulder, Lawrence H (1995), “Environmental Taxation and the ‘Double Dividend’: A Reader’sGuide.” International Tax and Public Finance 2: 157- 83.

33


Goulder, Lawrence H. , Ian Parry, and Dallas Burtraw (1997), “Revenue-Raising Vs. OtherApproaches to Environmental Protection: The Critical Significance of Pre-Existing TaxDistortions.” RAND Journal of Economics 28: 708-31.

Harberger, Arnold C. (1962), “The Incidence of the Corporation Income Tax.” Journal ofPolitical Economy 70, no. 3: 215-40.

Hertel, Thomas W. (1997), Global Trade Analysis: Modeling and Applications. Cambridge,MA: Cambridge University Press.

Lipsey, R. G., and Kelvin Lancaster (1956-1957), “The General Theory of Second Best.”Review of Economic Studies 24, no. 1: 11-32.

Metcalf, Gilbert E. (2003), “Environmental Levies and Distortionary Taxation: Pigou, Taxation,and Pollution.” Journal of Public Economics 87: 313-22.

Parry, Ian W. H. (1995), “Pollution Taxes and Revenue Recycling.” Journal of EnvironmentalEconomics and Management 29: S64-S77.

Pearce, David (1991), “The Role of Carbon Taxes in Adjusting to Global Warming.” TheEconomic Journal 101: 938-48.

Pigou, Arthur C. (1932), The Economics of Welfare. Fourth ed. London: MacMillan and Co.,1932.

Sandmo, Agnar (1975), “Optimal Taxation in the Presence of Externalities.” Swedish Journal ofEconomics 77: 86-98.

Schob, Ronnie (1997), “Environmental Taxes and Pre-Existing Distortions: The NormalizationTrap.” International Tax and Public Finance 4, no. 2: 167-76.

Todsen, Steinar, and Anton Steurer (2002), “Environmental Taxes in the E.U., 1980-1999.”,, 1-7: Eurostat.

Tullock, Gordon (1967), “Excess Benefit.” Water Resources Research 3, no. 2: 643-44.

Viguier, Laurent, Mustafa Babiker, and John Reilly (2001), “Carbon Emissions and the KyotoCommitment in the European Union.”. Cambridge, MA.: MIT Joint Program on the Scienceand Policy of Global Change.

34


Multi-dimensional Heterogeneity and the Design of Tax PoliciesHelmuth Cremer*

Abstract This paper surveys some recent contributions which have revisited the design of optimal policy mix when

individuals differ in several characteristics. Because of the technical difficulties raised by multidimensional screening

models the specification of tractable taxation models has long been neglected. This is certainly a serious omission because

the underlying policy issues are of considerable importance for the policy design in today’s welfare states. The main lesson

that emerges is that the design of redistributive policy is not as “simple” as the Atkinson and Stiglitz theorem suggests.

Policy makers should not concentrate on a single instrument (labor income taxation). While the optimal design of the

income tax is certainly a significant ingredient of the overall policy mix, it is not the only relevant instrument. The optimal

policy mix may very well rely on a large number of instruments including capital income tax, social insurance and housing

policies.

1. IntroductionIn traditional optimal tax models, policy instruments are typically restricted in an ad hoc way.These restrictions may pertain to the instrument under investigation as in Sheshinski’s (1972)linear income tax problem. Alternatively, or in addition, they may be about the other availableinstruments. For instance, the Ramsey model (Diamond and Mirrlees, 1971) studies the design oflinear commodity taxes in a setting where other tax instruments (such as uniform lump sum ornonlinear income taxes) are not available. It is by now well known that these ad hoc restrictionscan lead to potentially misleading conclusions. For instance, the Ramsey tax rules have to beamended if income can also be taxed or if lump sum taxes are available. This issue will beextensively studied below. A second problem of the traditional approach is that the issue of theoptimal policy mix cannot be studied. This is, however, a fundamental problem for the design oftax and transfer policies in most countries. The welfare state which exists to a different degree indifferent countries is characterized by a large variety of public policies ranging from tax andtransfer policies to in-kind transfers, education policies, social insurance, etc. Quite obviously, allof these policies are interdependent in particular through their redistributive impact. Consequentlypolicy recommendation based on settings where only a single instrument is introduced have to beconsidered with great care.

Modern optimal tax theory has moved beyond the limitations of these traditional models. Itrecognizes that the central element in the theory of optimal taxation is information. Lack of publicinformation on personal characteristics prevents the government to levy optimum lump-sum taxesand forces it to impose taxes on income. More generally, instruments ought to be restricted solelyby informational considerations. This leads quite naturally to an integrated approach to thedetermination of the different instruments.

One of the first significant breakthroughs in that direction is due to Atkinson and Stiglitz’ (1976)seminal paper. These authors study the design of optimal commodity taxes in a setting where anoptimal general (possibly nonlinear) income tax is also available. They show that under certainconditions (which we shall review below) the availability of the income tax may make commodity

35

Multi-dimensional Heterogeneity and the Design of Tax Policies

∗ University of Toulouse. I would like to thank Philippe De Donder and Firouz Gahvari for their useful comments andsuggestions.

taxes a redundant instrument. Put differently, when the income tax is optimally designed,commodity taxes may not be needed in the first place. As a matter of fact the subsequent literaturehas shown that the implications of the Atkinson and Stiglitz theorem go well beyond the issue ofuniformity of commodity taxation. Under the conditions of this theorem most of the policyinstruments that characterize the welfare state effectively become redundant (as long as a welldesigned income tax is available).

Atkinson and Stiglitz’ analysis has had a deep impact on modern public economics. While theiroverall approach is often considered as providing the most suitable framework for an integrateddesign of tax and transfer policies, the specific conditions underlying their theorem have beensubject to a lot of controversies. Several lines of attack have been explored: non-separability, taxevasion, uncertainty, nonlinear technologies, etc. The most fundamental shortcoming of Atkinsonand Stiglitz’s analysis appears, however, to be of a completely different nature. They assume (likemost of the literature on nonlinear income taxation) that individuals differ in one singlecharacteristic, namely wage (earning ability). There are no taste differences nor any other sourcesof heterogeneity (like wealth differentials). This assumption appears to be motivated by technicalconsiderations rather than by economic or empirical arguments. There is certainly no reason tobelieve that individuals are alike in all respects but their earning ability; however, so far, thisassumption has been the price to pay for solving the optimal taxation problem. Multidimensionalheterogeneity (adverse selection) would have made the problem quite intractable.

This paper presents some of the recent contributions which have revisited the design of optimalpolicy mix when individuals differ in several characteristics. I shall illustrate the technical issuesthat are involved by presenting a brief overview of Cremer et al. (2001). This paper focuses onthe traditional direct versus indirect tax controversy. However, similar methodologies can be usedfor a variety of other instruments and I shall sketch the main applications. First, it is howevernecessary to have a closer look at the Atkinson and Stiglitz result and its implications.

2. The Atkinson-Stiglitz theorem and its implicationsThe Atkinson and Stiglitz theorem states that when preferences are weakly separable betweenlabor supply and produced goods, and when individuals differ solely in productivity nonlinearincome taxation is sufficient to implement any constrained Pareto efficient allocation.Consequently, commodity taxes whether linear or non linear are not needed.

To understand the intuition behind this result, one has to recall that the lack of public informationon personal characteristics prevents the government to levy optimum lump-sum taxes and forcesit to impose taxes on income. The optimal income tax problem amounts to determining aconstrained Pareto efficient allocation. In other words, one determines the best allocation that canbe achieved given the information that is available. Formally, the information structure enters theproblem through the incentive (or self selection) constraints. Roughly speaking these constraintsrequire that no individual would be better off by “mimicking” other individuals.

1

The policy mustbe such that each individual prefers his own consumption bundle to the other individuals’consumption bundles. When consumption levels (or transactions) of commodities are observable,these can also be taxed. However, this additional instrument is only useful if it improves thescreening. For this to be possible, the distortions associated with the new instrument must first ofall relax an otherwise binding incentive constraint. Second, the welfare gain associated with thisincentive effect must dominate the distortion’s direct (and negative) effect on welfare. Put

36


1 For instance, a high ability individual can always pretend to be of low ability by working less (and thus have a lowincome). Consequently, he cannot be taxed too “heavily”.

differently the distortion must hurt the mimicking individual more than the mimicked individual.Now, under the condition of the Atkinson-Stiglitz theorem, the mimicking and the mimickedindividual differ only in their labor supply. With separable (and identical) utilities this means thattheir preferences over goods are exactly the same (i.e., they have the same marginal rates ofsubstitution at any point). But then a distortionary tax on the goods hurts everyone in the sameway. It cannot be used as a basis for improved screening.

This theorem has far reaching implications for a number of other policy instruments. It impliesthat in-kind transfers which are effectively a special case of nonlinear commodity tax are alsoredundant; see Cremer and Gahvari (1997). Similarly, capital income should not be taxed; seeStiglitz (1985). This is because within a lifecycle model capital income taxation is equivalent toa tax on future consumption which in turn is a commodity tax. There is also no room for socialhousing or differential tax treatment of housing (Cremer and Gahvari, 1998) nor for redistributivesocial insurance (Cremer and Pestieau, 1996). The latter two instruments combine in-kindtransfers with subsidies (negative commodity taxes). Finally, the tax on polluting goods shouldfollow the Pigouvian rule (Cremer, Gahvari and Ladoux, 1998), while the Samuelson rule appliesfor public good provision even in second-best settings (Boadway and Keen, 1993).

The practical implications of Atkinson and Stiglitz’s result certainly hinge on the empiricalvalidity of their separability assumption, which has been questioned (see e.g., Browning andMeghir (1991)). However, the spirit of their result goes through even under non-separability aslong as labor supply has no “significant” effect on individuals’ marginal rates of substitution(between goods). A strong case for the various instruments must thus rest on a substantial impactof labor supply on the willingness to pay for the produced goods and to the best of our knowledge,such a case has not yet been made. Some alternative lines of attack have been explored like, forinstance, tax evasion in Boadway et al (1994)), and uncertainty in Cremer and Gahvari (1995)).More recently, Naito (1999) has shown that Atkinson and Stiglitz’s result also rests on thelinearity of the technologies.

Probably the most fundamental shortcoming of the Atkinson and Stiglitz theorem is, however, theassumption that individuals differ only in productivity. In reality, one can of course expect themto differ in other dimensions like taste, (inherited) wealth, health status, etc. When individualsdiffer in more than one dimension it is plain that one can expect the property to break down. Tosee this, recall the intuition provided above. Clearly, under multi-dimensional heterogeneity, allindividuals no longer have the same indifference curves over produced goods, even underseparability. But then it may be possible to design other policies which affect the mimicker morethan the mimicked individual. As a matter of fact, the difficulty is not so much to show thatAtkinson and Stiglitz’s result does not in general hold in such a setting. This can easily be shownwithout solving the overall problem and Mirrlees (1976) has already made this point (for the caseof taste differences). The challenging problem is to provide a characterization of optimal policies.This is the issue dealt with in the papers reviewed in the following sections.

3. Commodity taxesCremer, Pestieau and Rochet (2001) consider a setting where individuals differ not only inearning abilities but also in endowments. Intuitively, such a difference can be thought of asreflecting wealth inequality. In the general model, endowments are allowed to differ for all goods.However, the analogy with wealth is most compelling in the special case when endowments arepositive and different for only one good.2 Consequently, this case is given special emphasis here.

37


2 This is of course in addition to the individuals’ endowments in time.

3.1. General model and resultsThere are N types of individuals (i=1....N), in proportion πi who differ in labor productivities(wages) ni and initial endowments ωi in m consumption goods. Preferences are separable betweenproduced goods and labor; utility is given by u(Ci)-v(Li), where Ci is the consumption vector Li

and denotes labor supply. All goods are produced by a linear technology with constant marginalcosts (producer prices) normalized at one. Types, ni , ωi and labor supply are not observable.Before-tax (labor) income, Ii=niLi , is observable. Personal consumption levels, Ci, and personalnet transactions Z=Ci-ωi are not observable. The tax administration observes, however,anonymous transactions.

Within this fairly general framework the authors show that when the income tax is optimallydesigned commodity taxes are needed, except when µhi(ωh

k - ωik )=0 for all h,i=1...N and k=1...m

, where µhi is the Lagrange multiplier associated with the incentive constraint relating type h totype i. This condition is quite stringent. It holds in a trivial way, when all types have identicalendowment vectors. This does not come as a surprise for with identical endowments oneessentially returns to the original Atkinson and Stiglitz setting. However, the condition also holdsin a slightly more general setting: it is sufficient that all pairs of types which are linked by abinding incentive constraint have identical endowments. For instance, if all types except for, say,j have identical endowments the optimum implies zero commodity taxes if no incentive constraintinvolving type j (either from j or towards j) is binding. Though not impossible, such an outcomeappears to be more of a technical curiosity than an economically meaningful case.

This result restores a role for commodity taxes (except in very special cases). The intuition is quiteobvious. For a given vector of net demands, individuals do no longer have the same preferencesover goods so that the feature which drives the Atkinson and Stiglitz theorem is no longer present.To learn more about the structure of optimal commodity taxes, we now turn to a Cobb-Douglasillustration. It gives more clearcut analytical results and is also used to run simulations.

Cobb-Douglas illustrationThe illustration is based on the following additional assumptions. There are only two producedgoods (m=2) and initial endowments consist only of commodity 1 (the numeraire): ωi

2 =0. We canthen use the simplified notation: ωi =ω i

1 , p2=p where p denotes the consumer price, while is theper unit tax (consumer price minus producer price) on good 2. The subutility for the goods isgiven by u(C)=C1/4

1 C1/42 The authors show that the optimal tax rate on good 2 is then given by:

(1)

where λ is the Lagrange multiplier of the government’s budget constraint, while Ri denotes thedisposable income of individual i (after tax labor income plus market value of endowments), andV(p,R) is the indirect utility function associated with u(.) . Observe that the disposable income ofh who mimics individual i is given by

Rhi=Ri+p(ωh-ωi)

In words, the optimal tax rate (on good 2) has the same sign as a weighted sum of the differencebetween the mimickers’ endowment (“wealth”), ωh , and the mimicked’s endowment (“wealth”),ωi . Intuitively this makes a lot of sense. Individuals have no endowments in good 2, which has anincome elasticity of one. Consequently, wealthier individuals necessarily consume (and buy) a

38


larger quantity of that good. Taxing this good is thus desirable if the incentive constraints bindfrom high wealth to low wealth individuals. In that case, the commodity tax can be seen as anindirect way to tax the unobservable endowment. Unlike a direct taxation of the endowments, ithas of course the drawback that it creates distortions. Put differently, commodity taxation appearsas an imperfect (second-best) substitute for a (first-best) lump sum tax based on endowments.

The crucial question that remains open is to know whether incentive constraints effectively bindfrom high wealth to low wealth individuals. Roughly speaking this requires that the tax systemredistributes (or attempts to redistribute) from high wealth to low wealth individuals. Clearly, thiswill not necessarily be the case. One can expect such a pattern to occur when endowments andproductivities are positively correlated (the more productive tend to also have the higher wealth).The relative inequality of endowments as compared to productivity differentials may also beimportant.3 As well as on the correlation between endowments and productivity. In addition, theweights in the welfare function may be of crucial importance. These issues will be addressed inthe simulations to which we now turn.

SimulationsThey are based on the Cobb-Douglas illustration with the following specification. There are fourtypes (denoted by i=A,B,C,D) distributed on a rectangle (as depicted on Figure 1) with n1=5,n2=10 and ω1<ω2. Disutility of labor is quadratic

v(L)=L2

and the revenue requirement is . Furthermore we have a uniform distribution of types anduniform weights (utilitarian social welfare function).

Figure 1 : The distribution of types

39


3 When the correlation is negative but when wealth inequalities are much more significant than productivity (andincome) differentials we can expect a similar pattern to emerge.

A

B D

C

ω

ω2

ω1

n 1 n 2

n

Table 1 : Impact of wealth inequality

The results are shown in Table 1 which provides the optimal tax rate on good 2 as a function ofthe distribution of wealth (characterized by ω1 and ω2 , total wealth being the same in all cases.For the record, the table also provides the pattern of binding incentive constraints which is quiteintricate. It appears that for a given level of total endowments, the optimal commodity tax on good2 increases with wealth (endowment) inequality. This result is not surprising and it reinforces thepoint that the tax on good 2 is used as an indirect way to tax the unobservable endowment in good1. As expected, the optimal tax is zero when all types have identical endowments. When wealthdifferentials are small, the tax is positive, but “small”. Types who share the same productivitylevel are then “close” and the redistributive potential of the commodity tax is limited.

The authors present further numerical results. For instance they show that, the commodity taxtends to be higher the higher is the correlation between productivity and wealth. Recall that thecommodity tax is an indirect way to tax endowments. Now, taxing endowments (albeit indirectly)is more beneficial when a higher endowment tends to go hand in hand with a higher productivitylevel. Under positive correlation, a commodity tax is then a substitute not only for the taxation of(unobservable) endowments but also for the taxation of (unobservable) productivities.

4. Other policy instrumentsMulti-dimensional settings have also been used to restore a role for other policy instrumentswhose usefulness was undermined by the Atkinson and Stiglitz theorem. The formal models differquite significantly but the basic reason why the result breaks down is always the same. The majordifficulties are also similar to those encountered in the case of commodity taxes. While it is rathereasy to show that an instrument is not in general redundant in a multi-dimensional setting, it ismuch more difficult to study the optimal design of the various policies. I shall now sketch a fewrecent papers which have presented advances in that direction. They cover various instruments,ranging from capital income taxation to redistributive pricing in the public sector.

4.1. Capital income taxationCremer et al. (2003) builds on Cremer et al. (2001) and studies the implications of multi-

40


1 2 Binding IC constraints p

0 12 BA, CD, DA 1.14

2 10 CD, DA 1.07

3 9 AB, CD, DA 1.05

4 8 AB, BA, CD, DA, DB 1.02

5 7 AB, BA, CD, DA, DB 1.01

6 6 CA 1.00

ω ω

dimensional heterogeneity for the taxation of savings. The starting point is the observation that inreality, and for all sorts of reasons, not only ability but also part of inherited wealth may not beobservable. If this is the case, then an interest income tax might become desirable even withseparability. To be more precise, such a tax may then be an indirect way of screening for the partof inherited endowment that is not public knowledge. However, this essentially static argument isincomplete and leaves out some crucial elements when it comes to the taxation of savings. Tointroduce those, the paper departs from Cremer et al. (2001) in several respects. Mostsignificantly, it accounts for the fact that inherited wealth (unlike endowments in a static model)cannot be assumed to be exogenous and independent of the tax policy. The analysis shows thatthis is a very crucial feature; it does not in general tend to make the taxation of capital income aredundant instrument.

The paper considers a two-generations overlapping generations model in which individuals drawutility from present and future consumption, from leisure and from the prospect of leaving theirchildren a certain amount of wealth. Each individual is characterized by two parameters: hisproductivity and his initial endowment. It focuses on the steady-state solution. Capitalaccumulation is equal to saving and saving is motivated by two concerns: second periodconsumption and bequests. A particular bequest technology is adopted. There are two possiblelevels of bequest, low and high. The more the parent invests in this technology, the likelier hisheirs will inherit the high level. Bequests are motivated by a joy of giving argument also calledthe “warm glow effect” (Andreoni (1990)), as opposed to dynastic altruism or strategicmotivations.4

Within this setting, there are two reasons for departing from the zero capital income tax rule. Thefirst is the same as in Cremer et al. (2001): by taxing (or subsidizing) capital income, we indirectlyreach inherited wealth which by assumption escapes taxation. Not surprisingly, this reason isparticularly persuasive when there is some correlation between labor productivity and inheritedfinancial endowment. The second argument normally goes towards subsidizing capital incomebecause the effort toward bequeathing generates additional resources to the economy andhenceforth ought to be encouraged.

4.2. Social insuranceRochet (1991) and Cremer and Pestieau (1996) have studied the role of social insurance usedalong with optimal income taxation by a welfare maximizing government. They have considereda setting where individuals differ in two unobservable characteristics: ability and risk as measuredby the probability of incurring some loss. Because of this second source of heterogeneity there isa potential role for social insurance in this setting. The authors show that both tax progressivityand the suitable degree of social insurance coverage depend on the correlation between the twocharacteristics. In the empirically appealing case where high productivity individuals tend to be“good” risks, full social insurance is socially desirable. On the other hand, when high productivityagents are “bad” risks, the optimum can imply partial social insurance coverage and nodistortionary income taxation. In that case, social insurance alone is sufficient to achieve optimalredistribution.

41


4 This assumption is not uncommon in the literature on the dynamics of wealth distribution and on endogenous growth;see e.g. Glomm and Ravikumar (1992). Empirically, that type of bequests appears to be at least as realistic as the dynasticbequests; see Arrondel et al (1997).

4.3. Tax treatment of housingIn most countries government intervention in the housing market takes two forms; see Rosen(1985) for a survey. The first consists of certain provisions in the income tax codes which resultin preferential tax treatment of owner-occupied housing. The second concerns provision ofhousing to the “poor” at below market rents. Under the conditions of the Atkinson Stiglitztheorem, neither of these interventions would be necessary.

Cremer and Gahvari (1998) revisit the problem of the optimal tax treatment of housing in a settingof two-dimensional heterogeneity. They posit a two-group model in which preferences depend ontwo housing goods (high- and low-quality), n other consumption goods and labor supply.Consumers exhibit different tastes for the two types of housing goods and tastes for high-qualityhousing and wages (or more loosely, tastes for high-quality housing and incomes) go hand-in-hand. They prove that differential tax treatment of housing is a useful instrument for achieving a(constrained) Pareto-efficient allocation, and demonstrate that if, in comparison to the rich, thepoor have a higher marginal willingness to pay for low-quality housing, their consumption mustbe subsidized. On the other hand, their consumption should be taxed if their marginal willingnessto pay (for low-quality housing) is lower. The intuition is the same in both cases. When the taxsystem entails redistribution from high- to low-wage persons, the extent of redistribution islimited by the possibility of high- mimicking low-wage individuals. The suggested taxdifferentiation weakens this otherwise binding self-selection constraint and allows furtherredistribution.

4.4. Redistributive pricing and universal serviceAnother instrument which has been challenged by the Atkinson and Stiglitz result is the use ofnonlinear pricing by public (or regulated) utilities as a redistributive mechanism. This is acommonly observed practice which has received considerable attention in policy debates.Regulators often seem to be concerned with the idea of fairness towards “small” customers. Thefact that they label certain pricing schemes as “social tariffs” is indeed quite telling [seePhlips~(1983)]. The current debate on universal service in telecommunications and, in particular,its “support for low-income consumers” [see FCC (1996), section viii] provides anotherillustration. Programs like “lifeline” in the US, or the LUS (Light User Scheme) offered by theBritish Telecom are prominent examples of redistributive policies. They imply nonlinear pricingwith discounts linked to quantity and/or toll limits. [See FCC (1996) and OFTEL (1997) fordetails]. These examples raise the important question of what is the rational behind a policy ofdistorting public utility prices in order to help the poor. Why shouldn’t the government simply useincome tax instruments for this purpose? The Atkinson and Stiglitz’ theorem suggests that if thepoor’s electricity bill ought to be subsidized then this should be done in a lump-sum way(independently of their consumption level but conditional on their income) while leavingmarginal prices undistorted.

This role of nonlinear pricing as redistributive policy instrument is revisited by Cremer andGahvari (2002). They consider an economy with many types of persons who differ in twounobservable characteristics (earning abilities and tastes). They show that nonlinear pricing doeshave a redistributive role. Beyond that, they derive a number of interesting properties for theoptimal pricing schedules. The paper also shows that in this case, one can solve theimplementation problem in a very interesting fashion. The Pareto-efficient allocations areimplementable through two separate functions: a pricing function which depends only on thepublic sector output and a tax function which depends only on income. This implies consumersof the public sector output must be given the same tariff options regardless of their income levels.

42


Means testing is not necessary.

When tastes and earning abilities are positively correlated, the results imply that the consumptionlevel of the poor must be distorted downwards. This is what universal service policies typicallydo. It is done either (indirectly) through high marginal prices or (directly) through quantitylimitations. Lifeline and LUS impose quantity limitations. Similarly, consumers of electricity (inFrance) who do not pay their bills are not automatically disconnected. Instead, they receive alimited power supply (3 kw, which allows them to use a few light bulbs plus a single appliance atany one time). Social tariffs, on the other hand, impose high marginal price on the poor. The sameis true of one other alternative to disconnection offered in the electricity sector; a meter whichworks with coins or, in the future, with pre-paid cards.5

5. ConclusionThis survey has shown that the design of tax policies under multidimensional heterogeneity is achallenging problem. It raises interesting theoretical and methodological questions. Because ofthe technical difficulties raised by multidimensional screening models (see for instanceArmstrong and Rochet, 1999) the specification of tractable taxation models has long beenneglected. This is certainly a serious omission because the underlying policy issues are ofconsiderable importance for the policy design in today’s welfare states.

The main lesson that emerges from the contributions reviewed in this paper is that the design ofredistributive policy is not as “simple” as the Atkinson and Stiglitz theorem suggests. To be moreprecise, the policy makers should not concentrate on a single instrument. While the optimal designof the income tax is certainly a significant ingredient of the overall policy mix, it is not the onlyrelevant instrument. The optimal policy mix may very well rely on a large number of instruments.Interestingly this is what can be observed in most industrialized countries, but we do not know ofcourse a priory if all these instruments are used in the most effective way. Put differently, the useof a variety of instruments is necessary for the appropriate policy design but it is of course notsufficient.

What makes policy design even more complex is that different instruments are largelyinterdependent. And the overall problem of the determination of the policy mix is not an easy taskas the papers reviewed here have shown. While a few simple lessons have emerged, manyquestions remain unsolved and open for future research.

This certainly points to the need for simulations, possibly with “realistic” parameter values (seeCremer, Gahvari and Ladoux, 2003, for an example). Given the complexity of the problemnumerical calculations are often the only way to determine the pattern of binding incentiveconstraints. They also may provide an estimation of the extent of the welfare gain that a specificinstrument can provide—-an issue which conventional analytics cannot address. Last but notleast, simulations can lead to precise policy recommendations which go beyond the fairly generaland methodological questions addressed here.

43


5 This entails high marginal prices; otherwise other consumers may also want to use this option.

ReferencesAndreoni, J.,(1990), “Impure altruism and donations to public goods: A theory of warm-glowgiving”, Economic Journal, 38, 209-227.

Arrondel, L., A. Masson and P. Pestieau, (1997), “Bequests and inheritance : empirical issues andFrench-US comparison”, in G. Erreygers and T. Vandevelde (eds), Springer-Verlag Berlin, 89-125.

Atkinson, A.B. and A. Sandmo, (1980), “Welfare implications of the taxation of savings”,Economic Journal, 90, 529-549.

Atkinson, A.B. and J. Stiglitz, (1976), “The design of tax structure: direct and indirect taxation”,Journal of Public Economics, 6, 55-75.

Atkinson, A.B. and J. Stiglitz, (1980), Lectures on public economics, McGraw Hill, New York.

Boadway, R., M. Marchand and P. Pestieau, (1994), “Towards a theory of direct-indirect taxmix”, Journal of Public Economics, 55, 71-88.

Boadway, R., M. Marchand and P. Pestieau, (2000), “Redistribution with unobservable bequests:a case for taxing capital income”, Scandinavian Journal of Economics, 102, 253-267.

Boadway, R. and M. Keen, (1993), “Public goods, self-selection and optimal income taxation”,International Economic Review, 34, 463-478.

Cremer, H. and F. Gahvari (1998), “On optimal taxation of housing”, Journal of UrbanEconomics, 43, 315-335.

Cremer, H. and F. Gahvari (1997), “In-kind transfers, self-selection and optimal tax policy”,European Economic Review, 41, 97-114.

Cremer, H. and F. Gahvari, (1995), “Uncertainty, optimal taxation and the direct versus indirecttax controversy”, Economic Journal, 105, 1165-1179.

Cremer, H, F. Gahvari and N. Ladoux, (2003), “Environmental taxes with heterogeneousconsumers: an application to energy consumption in France”, Journal of Public Economics, 87,2791-2815.

Cremer, H, F. Gahvari and N. Ladoux, (1998), “Externalities and optimal taxation”, Journal ofPublic Economics, 70, 343-364.

Cremer, H., P. Pestieau and J.-Ch. Rochet, (2003), “Capital income taxation when inheritedwealth is not observable”, Journal of Public Economics, 87, 2475-2490.

Cremer, H., P. Pestieau and J.-Ch. Rochet, (2001), « Direct versus indirect taxation. The designof the tax structure revisited”, International Economic Review, 42, 781-799.

Diamond, P and J. Mirrlees, (1971), “Optimal taxation and public production” I—II, AmericanEconomic Review, 61, 8—27, 261—278.

FCC (Federal Communications Commission), (1996), “Recommended decision on universalservice”, November issue, CC Docket No. 96-45.

Gevers,L and P. Michel (1998), “Economic dynasties with intermissions”, Games and EconomicBehavior, 25, 251-271.

Glomm, G. and B. Ravikumar, (1992), “Public versus private investment in human capital:endogenous growth and income inequality”, Journal of Political Economy, 100, 813-834.

44


Hammond, P, (1988), “Altruism”, in J. Eatwell, M. Milgate and P. Newmanthe (eds) The NewPalgrave: A Dictionnary of Economics, Mac Millan Press, London.

Mirrlees, J., (1971), “An exploration in the theory of optimum income taxation”, Review ofEconomics Studies, 38, 175-208.

Mirrlees, J., (1976), “Optimal tax theory: a synthesis”, Journal of Public Economics, 6, 327-358.

Naito, H., (1999), “Re-examination of uniform commodity taxes under a non-linear income taxsystem and its implication for production efficiency”, Journal of Public Economics, 71, 165-188.

OFTEL (Office of Telecommunications), (1997), “Universal telecommunications services:proposed arrangements for universal service in the UK from 1997”, February issue, ConsultativeDocument.

Phlips, L. (1983) “The Economics of Price Discrimination”, Cambridge: Cambridge UniversityPress.

Rochet, J.C., (1991), “Incentives, redistribution and social insurance”, The Geneva Papers of Riskand Insurance, 16, 143-165.

Rosen, H. (1985), “Housing subsidies: effects on housing decisions, efficiency and equity”, in A.Auerbach and M. Feldstein (eds), Handbook of Public Economics, vol. 1, Elsevier, Amsterdam.

Sheshinsky, E, (1972), “The optimal linear income tax”, Review of Economic Studies, 39, 98-100.

Stiglitz, J.E., (1985), “Inequality and capital taxation”, IMSS Technical Report n° 457, StanfordUniversity.

Stiglitz, J.E., (1987), “Pareto-efficient and optimal taxation and the new welfare economics”, inA. Auerbach and M. Feldstein (eds), Handbook of Public Economics, vol. 2, Elsevier, Amsterdam

45


Tax policy on the outskirts of the EU: GreenlandCarsten Vesterø Jensen, Central Customs and Tax, DenmarkSøren Bo Nielsen, Copenhagen Business School, Denmark*

Abstract This paper provides an assessment of Greenland’s tax system and contemplates changes that may be undertaken

in the future to prepare for greater economic self-reliance and the country’s participation in the wider world economy. On

the outskirts of Europe, Greenland is an autonomous part of the Danish kingdom, though currently not a member of the

EU. However, its cooperation with European countries and its dependency on international trade renders it necessary for

the tax system in Greenland to be attuned to developments in the rest of the world. Drawing on a thorough international

benchmarking analysis of Greenland’s tax system, the paper’s special focus will be on the corporate tax system and its

interplay with personal taxation, as well as the system of import duties. In particular, we carry out computations of

effective marginal and average corporate tax rates, as well as average effective tax burdens on consumption, labour income

and capital income, and compare these to similar measures for EU countries. In addition, we outline how Greenland’s

economic policy in other areas interferes with tax policy. Especially fishery regulation, management of government-

owned companies, and housing policy have major implications for the tax system.

Key words: international benchmarking, effective tax rates, Greenland

JEL Classification: H20, H25

1. IntroductionThis paper gives an introduction to a number of tax issues in Greenland. It is based on a recentbenchmarking analysis initiated by the Greenlandic Home Rule government. The two authors ofthis paper were members of the benchmarking commission, and the results of the benchmarkinganalysis are presented in the report “Skatter og afgifter i Grønland, 2003”. The government inGreenland is currently considering possible changes in the tax system based on the report.

We begin with a short introduction to Greenland. Greenland is the world’s largest island with anarea of around 2.2 million km2, of which 410,000 km2 are not covered by ice. (The area ofDenmark is around 43,000 km2). The climate in Greenland is arctic, and the northernmostextremity of Greenland is situated less than 730 km from the North Pole. The distance from northto south is 2,670 km. Just over 56,000 people live in Greenland. The central parts of WestGreenland are the most densely populated. Most of the population - around 45,000 - live in thetowns, of which Nuuk (Godthåb) is the largest. 60 per cent of West Greenland’s population livesin the six largest towns; the rest in more than 120 settlements, trading posts and sheep stations.The official languages are Greenlandic and Danish.

Greenland is in a national union with Denmark together with the Faroe Islands. In 1979, however,Home Rule was introduced, after which the Danish parliament, Folketinget, transferred almost all

46


∗ We thank Kaare Hagemann, Jacob Janussen and Hans Christian Vinten for many stimulating discussions of

Greenland’s tax system. Further, we thank Michael Devereux, University of Warwick, and Diderik Lund, University of

Oslo, for discussion in connection with computations of effective tax rates. The address of the second author is:

Department of Economics, Copenhagen Business School, Solbjerg Plads 3, DK-2000 Frederiksberg, Denmark; tel.: (+45)

38 15 25 96, fax: (+45) 38 15 25 76, e-mail: [email protected].

legislation to the Greenlandic parliament, Landstinget. The Danish administration retains controlover some areas of government e.g. foreign policy, defence and administration of justice.

Greenland has two levels of public admininistration, Home Rule (central government) and 18municipalities. The municipalities collect income taxes to finance their activities. Most of them,however, are to a very high degree dependent on transfers from the Home Rule and from the (two)rich municipalities, namely the biggest local entity, the capital Nuuk with 14,000 inhabitants, andthe smallest entity, the navy base Ivittuut, with around 160 inhabitants. The other municipalitieshave between some 500 and 6,000 inhabitants. Another important feature is the big number ofcompanies owned, partly or entirely, by the Home Rule. Public companies are dealing withtraditional public services as well as activities which elsewhere would be undertaken by privatebusinesses. Public companies raise very limited revenue (dividends) for the Home Rule; and theyare first of all characterized by being heavily dependent on public subsidies.

Greenland is autonomous regarding tax policy and was so even before the introduction of theHome Rule in 1979. The Home Rule government as well as municipalities raise and collect taxes.

Greenland’s economy is dominated by shrimps and prawns, which account for about 2/3 ofGreenland’s exports, with halibut, crab and cod making up the rest of exports. Most everydaynecessities are imported, and international trade is dominated by the trade with Denmark.

The GDP of Greenland is estimated to be around 9.1 billion DKK, or 1.2 billion EUR (2001);payments and unilateral transfers from the Danish state make up around 3,5 billion DKK (almost0.5 billion EUR). Taking net factor flows abroad into account, disposable GNP amounts to 12.3billion DKK (1.65 billion EUR) in 2001, per capita not much lower than the level in Denmark.

The Home Rule government finances are peculiar in the sense that the block grant from Denmarkmakes up more than half of the total revenue, as shown below:

Source: Budget proposal 2003 of the Home Rule government

47

Tax policy on the outskirts of the EU: Greenland

Table 1

1.000 DKK Per cent of total revenue

2.793.300 51

115.766 2

321.398 6

776.259 14

671.622 12

212.619 4

220.833 4

242.552 5

87.877 2

5.442.226 100

Interest

Repayments on loans

Other income

Total

THE HOME RULE GOVERNMENT'S REVENUE, 2001

Block grants from Denmark

Other payments form Denmark

Payments from EU (fishing licenses)

Direct taxes

Indirect taxes

Dividends from public companies

As table 1 shows, no less than 53 per cent of total public sector revenue in Greenland consists ofpayments from Denmark. Payments from the EU for fishery licenses amount to another 6 per cent,while the two main tax categories, direct (income) and indirect taxes (import levies), contributewith only 14 and 12 per cent, respectively.

Section 2 provides a more detailed description of the tax system in Greenland. Section 3juxtaposes tax rates and burdens in Greenland to tax rates and burdens in other Nordic andEuropean countries. Then section 4 presents some considerations as to the design of the taxsystem and the relationship between tax policy and other economic policy in Greenland. Thebenchmarking committee put forth a number of recommendations as to changes in the tax systemin Greenland. Some of these are presented in section 5. Finally, section 6 offers concludingremarks.

2. The tax system in GreenlandThe overall structure of the tax system in Greenland is shown below:

Source: The Tax Directorate of Greenland

Personal income tax is the main revenue raiser in Greenland contributing around 75 per cent oftotal tax revenue. Local income tax makes up 75 per cent of the income tax revenue. Import dutiesaccount for nearly 20 per cent of total tax revenue. Other taxes and duties, including corporateincome tax, contribute with less than 10 per cent. Greenland has a traditional income tax systembased on the (global income) principle of taxing the total income on a yearly basis irrespective ofthe type or source of incomes.

Everybody living or staying in Greenland more than 6 months is liable to pay taxes in Greenland.Married couples are taxed jointly. Greenlandic income tax is a flat rate tax. The income tax ismade up of an 11 percent national tax, a 4 per cent joint municipal tax and a local municipal taxof between 25 and 29 per cent (the navy base municipality charging only 20 per cent, though).

48


Table 2

21

53

4

8

7

3

1

1

2

-

100Total tax revenue

TOTAL TAX REVENUE 2001. DISTRIBUTION BY TAXES

Import duties on tobaccos

Other import duties

Motor vehicle duties

Duty on off-shore prawn production

Provincial government income tax

Local income tax

Corporate income tax

Import duties on alcohol

Duty on freightage

Other duties

A personal allowance and a standard allowance ensure that the income tax has a progressiveeffect. The personal allowance is an annual 40,000 DKK for everyone. The standard allowance is8,000 DKK. Everybody can claim the standard allowance instead of deducting actual operatingcosts or the costs of working. Married couples get double allowances.

Greenland has a manual income assessment system and a withholding tax system for employees.The tax rates and allowances are shown below for the years 1993 and 2003. It is notable that verylittle has been changed over this 10-year period:

Source: The Tax Directorate of Greenland

Capital income and business income are taxed together with labour income, the only exceptionbeing that a capital income of 5,000 DKK annually is tax-free. This tax exemption covers everytype of business income, e.g. income from fishing and hunting, as well as interests and other typesof capital income.

Shareholder dividends are taxed at the same level as other income by the tax rate of themunicipality to which the company belongs. It is notable that the company is allowed to deductdividends when computing its corporate income tax liability. Capital gains are normally tax-freein Greenland. Capital gains are only taxed if they are earned via speculation or in trade.

There is no taxation of wealth or possession of capital, real property, inheritance or land use (allland is in the hands of the public sector, but no fee is charged for private use).

Companies pay a corporation tax of 35 per cent. For companies dealing with raw materialextraction, however, the tax rate is only 30 per cent. The tax base is the income of the companyless costs, including depreciation. The access to depreciation allowances on equipment andbuildings is very favourable compared to the rulings in European countries. First, there is a rathergeneral access to deduct investment costs in the year of purchase. Second, the tax-related rates ofdepreciation of equipment and buildings are relatively high, and third, if a company computes apositive taxable income, it is allowed to undertake further depreciation of assets corresponding toup to half of its calculated taxable income. In reality, the system approaches free depreciation,which was actually the rule in place until some tightening of corporate income taxation took placewith effect in 2001 and onwards.

49


1993 2003

Provincial income tax 11 per cent 11 per cent

Joint municipal tax 4 - 4 -

Municipal tax (average) 25,8 - 26,8 -

Corporate income tax 35 - 35 -

Corporate income tax for raw material extraction 35 - 30 -

Personal allowance 40.000 DKK 40.000 DKK

Standard allowance 6.000 DKK 8.000 DKK

Tax free business income 5.000 DKK 5.000 DKK

Table 3 TAX RATES AND ALOWANCES 1993 AND 2003

Greenland has no VAT, sales tax or other general tax on consumption.

Duties and excises are important revenue raisers for the Home Rule government contributing withnearly 50 per cent of the total tax revenue on provincial level. Import duties are the most importantindirect revenue raisers. First of all, duties on tobacco (cigarettes) and alcohol (beer) contributeconsiderably to revenue. Furthermore, there are import duties on vehicles, chocolate and on anumber of commodities of less fiscal importance. The import duties are like excises, but collectedat the import level only, rendering the importers liable to pay the duties.

Apart from import duties there are a number of other duties. The duty on freightage (primarily oncommodities coming from abroad) played a significant role in the past. However, the parliamenthas recently decided to abolish this duty. The duty on off-shore prawn production has raisedsignificant revenue in earlier years with high world market prices on prawns as well as high taxrates. The prawn duty is levied only on prawns produced on board fishing ships. The duty is aimedat giving incentives for local prawn production on shore.

Employers pay a labour market contribution making up 0.9 per cent of total wages. Finally, thereis a current tax on vehicles, as well as duties on lotteries, gambling and stamps.

When providing an overview of the tax system it is important to also include the so-called taxexpenditures which are the indirect subsidies caused by a favourable tax treatment of certainactivities or persons. There has not been made any calculation of the size of the tax expendituresin Greenland. However, some important examples could be mentioned: All business income lessoperation costs is taxable. However, deductions for depreciations are more favourable than theactual book depreciations. All types of income are in principle taxable, although a number of non-pecuniary incomes are not taxed e.g. free (time-limited) housing and free holiday-travel foremployed staff.

Finally, all capital income is in principle taxable. However, owners of private houses and users ofpublic land for private housing or production purposes are not taxed.

3. Tax burdens in an international contextThis section places the tax system in Greenland in an international context. Moreover, it computesvarious measures of tax burdens in Greenland and compares them to parallel measures forselected EU countries.

The tax system in Greenland and other countries

First of all, the Greenlandic tax system could be compared with the tax systems in neighbouringcountries: Denmark, Faroe islands, Norway, Sweden and Iceland. Such a comparison shows thatGreenland is peculiar in a number of ways: 1) Labour and capital income are taxed according toa common scheme. But in the other Nordic countries labour and capital income are taxedseparately and according to different schemes; 2) Income taxation is flat-rate. The other countrieshave progressive taxation schemes; 3) The tax rate on personal capital income is only slightlyhigher than the corporate tax rate in Sweden and the Faroe Islands. In Denmark, however, the toppersonal tax rate is around 30 percentage points higher than the corporate tax rate, and in Icelandand Norway the corporate tax rate is higher than the personal tax rate; 4) The tax rate on personalbusiness is only slightly higher than the tax on companies. However, in some of the othercountries, such as Denmark and Iceland, there is a huge difference in the level of tax. 5) There is,

50


as in Norway, no double taxation of dividends. But in the other countries, dividends are taxed ascompany income and subsequently as shareholder income. 6) Domestic and foreign shareholdersare taxed equally. But this is not the case in all neighbouring countries. 7) Greenland has no valueadded tax (VAT). However, all the other Nordic countries do have a VAT.

The most significant differences appear to be the Greenlandic flat-rate income tax system and theabsence of a general tax on consumption – VAT.

The total tax burden in Greenland amounts to around 35 per cent of GDP. That is low comparedto Denmark and Sweden with a tax burden amounting to more than 50 per cent of GDP. On theother hand, Greenland’s tax burden is close to the OECD average of 37.5 per cent and close to thelevel in neighbouring Iceland and Canada. The tax burden in Greenland is higher than in the USA,where it is 28.9 per cent of GDP.

Income tax amounts to nearly 25 per cent of GDP, which is very close to the level in Denmark,namely 26 per cent of GDP. The OECD average is only 10.2 per cent. The other OECD-countriesrely to a much larger degree than Greenland and Denmark on social security contributions.

It is significant that consumption taxation is very low in Greenland, corresponding to only 8 percent of GDP. In Denmark the figure is 16.9 per cent and the other Nordic countries are at almostthe same level. Only the USA has lower taxation of consumption, amounting to only 4.2 per centof GDP.

Effective corporate tax rates

The general corporate income tax rate in Greenland, 35 per cent, is a little bit higher than theaverage when compared with the tax rates of the EU-member countries. An exception inGreenland is the lower rate, 30 per cent, on companies dealing with raw materials extraction. Thenominal corporate income tax rates, however, do not give the full picture of the real tax burden.To that end it is much more fruitful to look at the effective marginal and average tax ratescalculated below.

Tax depreciation is a crucial element when assessing the tax base. When comparing depreciationrules in Greenland with the rules in the EU-countries, it is evident that the Greenlandic rules arefavourable. This is first of all due to very liberal rules allowing immediate write-off and access toreducing company surplus by means of extraordinary depreciation.

The so-called effective marginal and average corporate income tax rates measure the burden oftaxation on, respectively, marginal investments in existing firms, and wholesale investments innew firms. They are computed below for Greenland for the years 1999 and 2001, i.e. before andafter the reform of the country’s corporate income tax system (prior to 2001, there was acompletely liberal depreciation, but from 2001 maximum depreciation rates for different types ofcapital were introduced). The effective tax rates come in two versions: one without inclusion ofpersonal capital income tax rates, and one with. Further, the tax rates are calculated for differenttypes of assets (such as immaterial assets, buildings, machinery, financial assets and inventories)and for different modes of finance (retained earnings, new shares and debt). In principle thisyields fifteen different effective tax rates (and for marginal as well as average, and without/withpersonal tax rates). In order to provide a clearer picture of corporate tax burdens, these fifteenrates are aggregated into single measures below.

Once the effective tax rates have been computed for Greenland, it is possible to compare it withsimilar measures for EU countries on the basis of the computations made for the EU Commission

51


report on company taxation in the internal market (cfr. SEC 1681, October 2001). For the precisemethod for computing effective tax rates we will refer to the relevant Appendix of theCommission report (available from the authors upon request); we do wish to mention, though, thatcomputations are undertaken on the following assumptions as key parameters:

Shareholders’ real, before-tax return on alternative investments amounts to 5 percent, and theinflation rate is 2 percent. Economic depreciation rates are taken to be 15.35 pct. (immaterielassets), 3.1 pct. (buildings), 17.5 pct. (buildings), and 0 (financial assets, inventories). As acomparison, the maximum depreciation rates for these types of assets in Greenland are currently30 pct. (immaterial assets, equipment), 5 pct. (buildings), and 0 for the remainder (except for 10pct. for ships). Finally, equal weights are assumed in aggregating over types of assets, while thethree forms of finance (retained earnings, new issues of shares, and debt) are taken to account for55, 10, and 35 pct., respectively.

Table 4 below shows (1) formal corporate income tax rates; (2) weighted capital costs; and (3)weighted effective marginal corporate tax rates for all countries in 1999 and 2001 for all EUcountries plus Greenland. In addition, the underlying capital costs for individual asset types andfinancing modes are presented for 1999. The numbers in the table only reflect taxation atcompany level; personal taxes on corporate income are not included.

Greenland’s nominal corporate income tax rate lies in the middle of the pack in both years. Irelandis at the bottom with solely 10 per cent while Germany in 1999 had a total nominal corporate taxof 52.4 per cent. The German tax reform which was implemented in 2001 led to a markedreduction of the corporate tax rate, and in that year Belgium and Italy are at the top of the list withrates slightly above 40 pct.

The weighted capital costs for marginal investments in Greenland were 5.5 per cent in 1999 and6.3 per cent in 2001. The number in 1999 is the lowest on the list save for Italy, and the weightedeffective marginal tax rate in Greenland was only 9.2 per cent. in that year. The comparablenumbers for 2001 reflect the reform of depreciation rules in Greenland, since the weighted capitalcost rises, as does the associated weighted effective marginal tax rate – to 20.5 per cent. Nowmore countries have lower effective rates, but there are also many countries with higher effectiverates. For instance, Germany’s and France’s weighted effective marginal tax rates are 26 and 32per cent, respectively, in 2001.

52


Source: EU Commission (2001) and own computations

It is possible to demonstrate the tax burden for new firms by computing effective average tax ratesalong the lines of the EU Commission report (see also Devereux and Griffith, 2003). The intentionis to show the total tax burden in such firms, not just the tax burden on marginal investment asmeasured by the effective marginal tax rate. On top of the values of key parameters selected forthe effective tax rates it is now also necessary to specify the before-tax supranormal rate of returnfor investment in a new firm. A rate of return of 20 pct. is used in the calculations below.

53


Tab

le 4 C

OS

T O

F C

AP

ITA

L AN

D E

FF

EC

TIV

E M

AR

GIN

AL T

AX

RA

TE

S

2001 1999

Cost of capital 199

9 E

MT

R 1999

Country


Cost of capital

EMTR


Cost of capital

EMTR

Immaterial assets

Buildings

Equipment

Finansiel assets

Inventories

Retained earnings

New issues of shares

Debt

Retained earnings

New issues of shares

Debt

Austria

34,0 5,7

12,6 34,0

6,3 20,9

5,9 6,1

5,9 7,3

6,3 7,5

7,5 4,0

33,3 33,3

-25,0

Belg

ium

40,2 6,4

22,4 40,2

6,4 22,4

5,2 7,0

5,3 8,0

6,7 8,0

8,0 3,5

37,5 37,5

-42,9

Germ

any 39,4

6,8 26,1

52,4 7,3

31,0 5,4

7,2 5,8

10,0 7,9

9,7 7,6

3,2 48,4

35,5 -56,2

Denm

ark 30,0

6,4 22,4

32,0 6,4

21,6 4,2

8,0 5,4

7,1 7,1

7,4 7,4

4,4 32,7

32,7 -12,8

Spain

35,0 6,5

22,8 35,0

6,5 22,8

6,5 6,7

5,4 7,4

6,4 7,7

7,7 4,1

35,1 35,1

-21,9

Greece

37,5 6,0

16,9 40,0

6,1 18,2

6,8 5,1

6,1 5,1

7,4 7,6

7,6 3,4

34,2 34,2

-47,1

Fra

nce 36,4

7,3 31,8

40,0 7,5

33,2 5,2

8,5 8,4

8,0 7,4

9,0 9,0

4,6 44,4

44,4 -8,7

Finla

nd 29,0

6,4 21,3

28,0 6,2

19,9 6,1

6,1 5,6

6,8 6,8

7,2 7,2

4,5 30,5

30,5 -11,1

Italy

40,3 4,3

-15,9 41,3

4,8 -4,1

2,9 4,6

3,8 7,7

5,0 5,5

5,5 3,6

10,0 10,0

-38,9

Irela

nd 10,0

5,7 11,7

10,0 5,7

11,7 5,3

6,8 5,2

5,5 5,5

5,9 5,9

5,2 15,2

15,2 3,8

Luxemb

ourg

37,5 6,3

20,7 37,5

6,3 20,7

5,2 6,8

5,3 7,7

6,5 7,7

7,7 3,7

35,1 35,1

-35,1

Holla

nd 35,0

6,5 22,7

35,0 6,5

22,6 5,1

6,9 5,9

7,4 6,9

7,7 7,7

4,1 35,1

35,1 -21,9

Portuga

l 35,2

6,3 21,0

37,4 6,5

22,5 6,7

6,2 5,2

7,7 6,5

7,9 7,9

3,9 36,7

36,7 -28,2

Sw

eeden

28,0 5,8

14,3 28,0

5,8 14,3

5,0 6,0

5,0 6,6

6,6 6,7

6,7 4,3

25,4 25,4

-39,5

UK

30,0

6,7 24,8

30,0 6,6

24,7 5,5

8,2 5,6

6,9 6,9

7,7 7,7

4,8 35,1

35,1 -25,0

Gree

nland

35,0 6,3

20,5 35,0

5,5 9,2

4,2 4,2

4,2 7,5

7,5 6,5

6,9 3,5

23,1 28,1

-41,3

The result of the computations of effective average tax rates for Greenland as well as for the EUcountries are shown in Table 5 below. It lists (1) the formal corporate tax rate; (2) weightedeffective average tax rate for 1999 and 2001. In addition, effective average tax rates for singleasset types and for modes of finance are shown in the table.

That the formal corporate tax rate in Greenland is not very different from the rates in EU countrieshas already been noted. This is in fact the reason why the weighted effective average tax rates inGreenland are at the level of those in EU. With a high pre-tax rate of return (as the 20 pct.), theeffective average tax rate approaches the nominal corporate tax rate. Implicit subsidies inadvantageous depreciation rules play a lesser role for average than marginal tax rates. Still, thefree depreciation in Greenland’s corporate tax system in 1999 does drive effective average taxrates for immaterial assets, buildings and machinery somewhat below the nominal rate.

Source: EU Commission (2001) and own computations

54


Table 5 EFFECTIVE AVERAGE TAX RATES

1999 1999

Country

Corp

ora

te inco

me

tax

2

001

EA

TR

2

001

Corp

ora

te inco

me

tax

199

9

EA

TR

1

999

Imm

ate

rial

assets

Build

ings

Equip

ment

Fin

ansie

l assets

Invento

ries

Reta

ined

earn

ings

New

issues o

f share

s

Debt

Austria 34,00 27,9 34,00 29,8 28,6 29,2 28,4 33,2 29,9 33,9 33,9 22,3

Belgium 40,17 34,5 40,17 34,5 30,7 36,1 31,0 39,2 35,5 39,1 39,1 25,8

Germany 39,35 34,9 52,35 39,1 33,9 39,0 34,9 46,8 40,8 46,1 40,1 27,7

Denmark 30,00 29,0 32,00 28,8 21,3 34,3 25,3 31,4 31,4 32,3 32,3 22,3

Spain 35,00 31,0 35,00 31,0 31,1 31,8 27,4 34,2 30,7 35,2 35,2 23,3

Greece 37,50 28,0 40,00 29,6 35,5 30,4 33,4 11,6 37,1 34,4 34,4 20,8

France 36,43 34,7 40,00 37,5 30,6 40,6 40,1 39,0 37,1 42,1 42,1 28,8

Finland 29,00 26,6 28,00 25,5 24,8 24,8 23,1 27,3 27,3 28,8 28,8 19,3

Italy 40,25 27,6 41,25 29,8 24,9 29,8 27,4 36,1 31,1 31,8 31,8 26,1

Ireland 10,00 10,5 10,00 10,5 8,9 15,8 8,2 9,8 9,8 11,7 11,7 8,2

Luxembourg 37,45 32,2 37,45 32,2 28,6 33,7 29,2 36,6 32,9 36,6 36,6 24,0

Holland 35,00 31,0 35,00 31,0 26,7 32,4 29,2 34,2 32,5 35,1 35,1 23,3

Portugal 35,20 37,0 37,40 32,6 33,2 31,8 28,6 36,5 32,8 37,0 37,0 24,5

Sweeden 28,00 22,9 28,00 22,9 19,6 23,4 19,7 25,7 25,7 26,0 26,0 17,1

UK 30,00 28,3 30,00 28,2 24,2 33,7 24,7 29,3 29,3 31,8 31,8 21,6

Greenland 35,00 35,0 35,00 32,7 29,0 29,0 29,0 38,2 38,2 32,5 36,5 27,1

Average effective tax rates on consumption, labour and capital income

A relatively simple and rough way to compare the tax systems in different countries is to calculatethe size of the tax burdens on the three primary objects of taxation: consumption, labour income,and capital income. Below we compute in table 6 average effective tax burdens on the threefactors and compare them to similar measures for the EU countries.

Our point of departure is the report by Martinez-Mongay (2000), which introduces the averageeffective tax burdens and provides computations for EU countries. It builds on the methodsuggested by Mendoza et al. (1994) (for an early application to EU countries see Lassen andNielsen (1996)).

The average effective tax on consumption in Greenland can be gauged to 11 per cent in 2001. Thatcan be compared to tax burdens on consumption of 30.4 per cent in Denmark in 1998 (and anestimated burden of 29.6 pct. for 2001). For EU-15 the numbers are, respectively, 20.5 per centand 20.9 per cent It emerges clearly that the tax burden on consumption in Greenland is very lowin an international context. All EU countries have average effective tax rates on consumptionmarkedly above Greenland’s.

For labour income it is possible to derive an average effective tax rate in Greenland of some 28per cent in 2001. As can be seen in the table, this number is also somewhat lower than similarstatistics for Denmark and the EU-15. For Denmark, Martinez-Mongay has estimated 42.7 percent in 1998 and, for 2001, 43.0 per cent ,while the averages for EU-15 are 37.3 per cent in 1998and an estimated 36.2 per cent in 2001. In the table, only Ireland, Greece and UK have loweraverage effective tax burdens on labour income than Greenland.

The average effective tax rate on capital income can be calculated at 22 per cent for Greenland in2001. The table again contains the similar statistics for Denmark and the rest of the EU countriesin 1998 and 2001 (estimated). EU-15 has rates of 22.9 and 23 per cent in the two years, while theDanish tax burdens on capital are somewhat higher at 28.4 pct. and 29 per cent. The upshot ofthese figures is that Greenland does not deviate from the EU countries when it comes to theaverage effective tax burden on capital, although its rate is significantly lower than that ofDenmark.

In conclusion, Greenland’s tax system entails a much more lenient taxation of consumption thanis the case in EU and also a relatively low tax burden on labour income. For capital income,though, the average effective tax rate is at the level of the EU countries.

55


Source: Martinez-Mongay (2000) and own computations. Figures for 2001 are estimates.

4. Tax policy and other economic policyIn the design of tax policy and in the evaluation of its impact it is important to realize that taxpolicy does not stand alone, but instead interferes with other lines of economic policy. In thissection we first list some general aims in tax policy, which, to a large extent, also constituteguidelines for the design of other economic policies.

Then we look closer at some of the specific policy areas with which taxes and duties interfere. Webegin with some considerations as to the possible contributions of natural resources to the revenueside of the public budget. Resources are here both renewable and exhaustible resources. Next, weconsider industrial, labour market, social and housing policies. Importantly, the more resourcerents in fishery that are captured by the public sector; the more stringent the scrutiny of fundsspent on industrial development; the less housing is subsidized, and the more revenue can beharvested by publicly owned companies, the smaller is the burden placed on the tax system.

The most important criteria for a well-functioning tax system are well recognized and concern (i)efficiency; (ii) distribution; (iii) horizontal equity and fairness; (iv) administration; (v)compliance; and (vi) incentives for tax avoidance and evasion. The best taxes are those that

56


Table 6

1998 2001* 1998 2001* 1998 2001*

Belgium 20,4 20,8 45,1 43,6 23,8 23,5

Germany 17,0 18,4 43,7 42,0 15,4 14,8

Spain 16,8 18,3 29,6 30,0 18,5 18,5

France 24,8 23,9 41,3 41,8 21,6 21,8

Ireland 24,0 24,7 23,2 22,6 20,5 19,5

Italy 23,2 22,9 35,3 35,1 25,5 25,4

Luxembourg 25,3 26,7 31,7 30,4 34,3 32,2

Holland 18,6 20,2 35,9 32,9 24,3 23,3

Austria 23,2 24,1 41,0 39,3 19,2 17,3

Portugal 22,0 23,8 24,8 28,7 22,1 26,0

Finland 24,3 24,9 43,8 42,8 24,3 24,3

Greece 19,3 20,8 28,7 29,3 20,3 17,8

Sweeden 25,8 24,2 51,9 50,6 28,5 27,3

UK 18,1 18,4 25,4 24,7 34,1 34,0

Denmark 30,4 29,6 42,7 43,0 28,4 29,0

EU-15 20,5 20,9 37,3 36,2 22,9 23,0

Greenland 11,0 28,0 22,0

AVERAGE EFFECTIVE TAX BURDENS

CountryConsumption Labour incomencome Capital income

interfere as little as possible with private sector activity; the distributional consequences of whichare acceptable to the public at large; that imply that people with comparable incomes paycomparable taxes; that are easy to administer and to comply with; and that provide few incentivesfor evasion.

It is essential to realize that all the criteria in some form can be transferred to other economicpolicies. As an example, social transfers should not interfere too much with behaviour on the partof recipients by, e.g., keeping them away from the labour market. They should be paid to theespecially needy, and the way they are paid out should counter cheating. As another example, ifthe tax system has gone out of its way to secure a reasonable distribution of consumptionpossibilities in society, a housing policy with no transparency and perhaps perverse distributionaleffects can throw everything overboard. And if the taxation of interest is kept low to induce peopleto save for old age, the effect will diminish, if people with low savings are awarded relatively highpensions when old.

Finally, if the opportunities for capturing resource rents in the exploitation of natural resourcesare not used, the government has to turn to less efficient revenue sources, whence taxes inflicthigher than necessary deadweight losses on the economy.

Natural resource (fishery) policy

Natural resources can be renewable or exhaustible. They enter the economy’s production togetherwith physical capital and labour (human capital). Some societies have only few natural resources,and their living standards hence rest on the use of physical and human capital. Most countries,though, utilize natural resources in agriculture, fishery, oil production, mining, etc.

The management of natural resources is the responsibility of society at large. At the outset, theseresources are common, and their exploitation would benefit the entire population. But the actualdecisions made by politicians in practice determine the character of resource use and thus who inthe end stands to gain from the resources.

One possible policy is ‘free for all’, the entirely unregulated use of a resource. This normally leadsto the ‘tragedy of the commons’. Each individual user of the resource has no incentive to take intoaccount that his/her own exploitation of the resource diminishes the opportunities of others toutilize the resource. As a consequence, the exhaustible resource will be exploited too rapidly, orthe renewable resource is harvested so intensively that it prevents future resource use - or in theworst cases, the resource becomes extinct. And unregulated exploitation of the resource leads tolarge losses for society, especially if those using the resource have invested in too much capacity.

There are many examples of minor and major tragedies in resource use, not least in fishery. Thetechnical development in fishery equipment has proceeded so far that it has enabled theeradication of many species and such intensive harvesting that it takes many years to bring thestock back to a level that renders utilization of the resource possible again.

Another possible policy is to hand over the rights to exploit a given quota of the resource to agroup in society for free. Initially, the resource in question represents value to society at large, butthis is broken by limiting the use of the resource to only a few members of society. Only they havethe opportunity to secure additional income from the resource. What could have been an incomesource for the entire country ends up as an income source for only a small group.

There are likewise many examples of this kind of policy. The aim typically has been to preventtragedies and excess investments in capacity, and to facilitate increasing returns to scale. This

57


management of the resource is likely to be superior to the free-for-all situation, but the possibilityof the entire society gaining from the resource use is lost.

A third possibility is to hand over the rights to utilize the resource to private agents at a cost. Theprice of access to the resource should ideally reflect the supranormal income which the user canderive from exploiting the resource, once inputs of capital and labour are accounted for. This pricecan, in principle, be secured by auctioning off rights to utilize the resource, if enough independentinterested parties bid for the rights. The revenue from such auctioning off of quotas would accrueto the public fiscs and finance the provision of public services and transfers. Only for higher levelsof public expenditures does it become necessary to resort to taxation.

In the case of Greenland the rights to exploit prawns and fish stocks have been handed out forfree. In fact, there are even many examples of fishery being subsidized rather than taxed by thegovernment in Greenland. Systems of regulating the use of natural resources vary immenselyacross countries, and ideal regimes will be hard to find. Fishery regulation has been discussedextensively in many countries. For interesting accounts of regulation in Iceland and the FaroeIslands we refer to Grótinum (2003) and Gylfason and Weitzman (2003). The former points outthat it will be possible to earn a ‘double dividend’ in the Faroe Islands, if regulation shifts fromthe present fishing-day system to a system, where the exploitation of the resource is subject to atax (replicating an auctioning system). On the one hand, management of fish stocks is improvedupon; on the other, revenue from quotas or taxes lowers the need for tax revenue, reducingdistortions in the tax system. Hence, the wording of double dividend. Gylfason and Weitsmanargue that resource depletion charges (RDCs), while under ideal conditions being equivalent toauctioned-off quotas, may in practice have a number of advantages.

Interestingly, as was mentioned in the Introduction, Greenland obtains a payment of more than300 mill. DKK primarily for fishery licences from EU. What the government needs to do isessentially derive similar license fees from internal agents in fishery. There are many (free)licenses in Greenland’s fishery today; they are with or without quotas, and limited or not limitedin time. But it should be legally feasible to take back these licenses to Greenland’s governmentafter a period of 5-7 years or so, provided the current users are given sufficient notice. Thereafter,they can be auctioned off, or resource depletion charges can be introduced, so that the revenuescan be collected by government.

Industrial policy

As for industrial policy, a sound general objective is to leave private and public firms tothemselves. One caveat concerns those firms that are natural monopolies. Here, there is anargument for exploiting increasing returns to scale, but the extent to which profits should bedriven below zero and the firms in question receive subsidies must be determined by the scarcityand dearth of public funds.

In many countries there is a pronounced tradition for public subsidies for a plethora of purposes.There are two important consequences of this. First, firms have an incentive to ‘make themselvesready for a subsidy’, i.e. conduct excessive investments, and incur excessive costs etc., so that theneed for a subsidy becomes more pressing. Second, it is necessary to levy additional taxes in orderto finance the subsidies. These taxes generate distortions, and the cost of these distortions mustbe taken into account in any decision to concede subsidies. Higher labour income taxes, forinstance, may lead to lower labour supply and to higher wages, and in some cases those firms thatbenefit from the subsidies at the same time suffer from having to pay higher wages to their

58


workers.

Criteria for offering subsidies must be stringent. Subsidies should only be given for activities thatare socially desirable and privately unprofitable. If an activity is not profitable to undertake by thefirm in question, there is a pronounced risk that it also is not desirable for society as a whole. Andif the activity is privately profitable, there is no reason for the subsidy in the first place.

When would a project be privately unprofitable, yet welfare-enhancing? One possibility is that theproject generates positive spillovers; other firms learn from production activities in the firmwithout having to pay for the productivity increase they subsequently are able generate. Anotherpossibility is that the firm faces prices, which do not reflect values to society. An example is theprice of labour. While the firm will have to pay the going wage, the social opportunity cost ofusing workers in the firm may be lower than the wage, if the workers alternatively would havebeen unemployed. A second example concerns market prices, which do not reflect society’s costof producing them. If, for instance, the supplier of an input is a monopoly, its price will in generaldiverge from the marginal cost of producing the input.

A third possible argument for subsidizing might be the desirable distributional consequences ofsome activity. Here, though, one must always contemplate whether there are better-targeted waysof bringing about an improvement in distribution. Public transfers and the tax system arepresumably superior in this respect.

What this discussion boils down to is that industrial policy in Greenland should be subjected tostringent tests before decisions as to handing out subsidies are made. For the moment, a host ofpublicly owned companies generate no revenue to the government, but instead require continuedinjection of subsidies. Only a minority of public companies is able to pay dividends to the publicfiscs. If this situation can be reversed, it would mean a significant reduction in the need to levydistortionary taxes to finance public expenditures, including subsidies to firms.

Housing policy

An area which in all countries is tantamount to huge public subsidies, is the housing area.Greenland is no exception here. In fact, housing subsidies, direct plus indirect, are probably biggerthan in any other country in Europe. The reasons are many. First, all land is public, and despitethe fact that inhabitable land is scarce, users of land pay no fee for their use to the government.Second, many individuals receive personal housing subsidies. Third, owner-occupied housing issubsidized via the fact, that there is no taxation of imputed rent, yet interest on mortgages is tax-deductible. Fourth, the return to capital employed in other parts of the housing market likewise isnot taxed. Fifth, part of housing loans receives public subsidies.

An example may clarify the mechanisms and point to the possible magnitude of housingsubsidies, i.e. the gap between the private and social cost of housing. For this purpose we take theso-called 10-40-50 arrangement related to owner-occupied housing in Greenland. A house costing1.4 mill. DKK to construct is financed by the 10-40-50 arrangement. This implies that theprospective owner presents a down-payment of 10 percent; 50 percent of the cost is financed onthe market; and 40 percent is financed via public loans for which there are no interest paymentsnor any payback of the loan during the first 15 years. In the calculation it is further assumed thatthe value of the land occupied by the house is 10 percent of the cost of the house; that the marketinterest rate is 5 percent, and that inflation is 2 percent. Finally, economic depreciation of thehouse is assumed to be 4 percent per year. Annual private user cost in the house can now beapproximated by Q[(1-a)[i(1-t)-p]+d], where Q is the cost of the house; a the loan subsidy share;

59


i the nominal interest rate; t the capital income tax rate; p the inflation rate; and d the rate ofdepreciation. The social user cost of housing amounts to Q[r+d]+grQ, where r is the real interestrate and g the value of the land relative to the cost of the house. Inserting the relevant numbersone can derive that the effective housing subsidy in the 10-40-50 arrangement amounts to around40 percent of the social cost!

While owner-occupied housing is heavily subsidized in Greenland, rental housing is also(indirectly) subsidized by rent regulation and by personal housing subsidies. The situation is,however, clearly unsustainable; as in other countries with rent regulation and extensive subsidies,it is exceedingly difficult to find an apartment or a house almost anywhere in Greenland. Adetermined movement towards a more freely functioning market seems inevitable. This will entaileasing rent regulation and reducing all housing subsidies, probably except for personal housingsubsidies. Such a process towards a more flexible housing market will sharply reduce the need tolevy taxes to finance housing subsidies and provides our last example of how economic policiesin other areas are intimately related to tax policy in Greenland.

5. Recommendations for reformThe benchmarking commission referred to in the Introduction recommends a number of(alternative) changes in the Greenlandic tax system.

First of all the commission recommends maintaining the present flat rate income tax system whichis very simple and transparent. The commission does not at this stage support the idea ofintroducing a progressive tax system with a graduated tax rate scheme. That would createadministrative problems and make the tax system less transparent. According to the commission,however, there is a need to strengthen the redistribution effects of the tax system as well as of theeconomic policy in general. The commission, therefore, recommends introduction of a currenttaxation on real property, a tax on inheritance and gifts, and a tax on the use of land. Thecommission also recommends a higher personal allowance, financed e.g. by introducing VAT oranother form of general tax on consumption and by broadening the income tax base including allincome from labour and capital in cash or kind.

The commission recommends keeping the present business and company taxation systemincluding the present taxation of shareholders’ income. There is no need to lower the corporateincome tax rate, which is very close to the rate of individual income tax.

There is also no need to change the existing system of taxation of shareholders’ income, accordingto which shareholder income is taxed at the same level as other capital and labour income. Themost pressing issue regarding business income is to amend and strengthen the depreciation rulesin order to ensure a higher degree of accordance between tax depreciations and book depreciationswith respect to equipments and buildings. In particular, the liberal rules allowing immediate write-off and allowing extraordinary depreciation with positive taxable corporate income should beremoved.

The commission has gone through a number of existing ideas and suggestions in Greenland thataim at using tax regulations to promote certain business activities such as conversion of businessinto a company; generational change; employee share options schemes; and raw materialextraction. The commission has rejected all these ideas from the viewpoint that equal taxtreatment of all forms of businesses, individuals and economic transactions is economically themost healthy and efficient way forward for Greenland. Special regulations for selected types ofbusiness or transactions will as indirect subsidies merely lead to distortions and inefficient use of

60


resources.

The commission has gone through all the existing duty regulations and puts forth a number ofamendments and proposals. The most significant ones are mentioned here.

Greenland should attempt to get rid off all tax regulations discouraging free competition andinstead focus on raising revenue. It is the impression that Greenland has inefficient competitionin goods markets (not least retail business). This is partly due to regulations (e.g. licensingsystems) originated in the earlier colonial system and now administered by the Home Rulegovernment. It is also due to the introduction of a number of trade protection measures. Especiallyregarding cigarettes, beers, soft drinks and lamb meat it is evident that government tax regulationsare causing higher consumer prices (and profits) rather than tax revenue.

The existing duties could become more efficient revenue raisers. This is first of all obvious as faras concerns the duties on vehicles, lotteries, cigarettes, alcohol and soft drinks.

The commission also recommends the introduction of more general taxes on consumption.Greenland is one of the few countries in the world that does not tax energy. The commissionrecommends introducing a tax on energy. In the first instance, a tax on fuel used for vehicles andboats, and as the next step a tax on fuel for heating as soon as the existing subsidies for heatingpurposes are done away with.

Greenland is also one of very few countries with no value added tax system or general sales taxes.The commission recommends introducing VAT as an efficient revenue raiser with a big potentialfor government to achieve a means to finance lowering or abolishing other existing taxes. AGreenlandic VAT could be adapted to the fact that Greenland has a very narrow local productionbase and that most commodities are imported from Denmark by ship.

6. Concluding remarksThis paper has reviewed the tax system in Greenland on the basis of a recent internationalbenchmarking analysis undertaken as part of Greenland’s structural action plan and as preparationfor greater self-reliance of the country in the future. We have argued that because of a big blockgrant from Denmark that accrues directly to the public sector in Greenland, the role of the taxsystem in securing financing of public expenditures is reduced relative to that of other Europeancountries.

Indeed, we also found that the tax burden on the prime factors of taxation, consumption, labourincome and capital income is smaller than the European average. More specifically, the taxburden on consumption is much smaller, and the labour income tax burden somewhat smaller thanthis average. A main reason for this is that there is no general consumption tax in Greenland (noVAT), and personal income is taxed according to a flat-rate income tax with a modest rate.

As to capital income taxation, the corporate income tax in Greenland has a fairly high nominalrate. However, because of very generous depreciation rules, there is less ‘bite’ in corporateincome tax.

On the basis of the account of the existing tax system and some considerations as to the majorobjectives of the tax system we went on to list a series of recommendations for altering the systemin the future. We did stress, first of all, that the existing income tax system appears quite sensiblefor a country like Greenland with modest revenue needs. The motto is then ’keep it simple’.Retain the existing rates of corporate and personal income tax, but strengthen the corporate tax by

61


eliminating overly favourable depreciation allowances. Likewise there is no reason to change theexisting system of dividend taxation, where a rebate is given for already paid corporate tax.

To complement the flat-rate personal income tax and to improve the distributional profile of thetax system we recommended inheritance and gifts taxes, as well as taxes on property and land use.

As for indirect taxes we recommended that Greenland considers adopting some form of generalconsumption tax. Either a standard VAT or a similar tax, which exploits the fact that Greenland’sdomestic production base is very narrow, and most of imports flow from Denmark. We alsoargued that Greenland should introduce general energy taxation as found in all other Europeancountries.

A major point in our analysis was that tax policy and other lines of economic policy in Greenlandare interrelated. In particular, the burden on the tax system depends heavily on policy in thenatural resource, industrial, housing, social and labour market areas.

At the moment the management of Greenland’s resources of fish and prawns entails a netexpenditure for the public fiscs. This does not have to be the case. A system of auctioning offlicenses or resource depletion charges would give the public sector an opportunity for securingfunds to replace distortionary taxes. Further, the public sector ownership of a series of companiesimplies a sizeable net expense. Instead, insisting on the invested capital yielding at least somereturn to the public sector also would lessen the burden on the tax system. Finally, it will be hardto find another country in which housing is more subsidized, directly and indirectly, than inGreenland. Again, moving towards a free housing market and less subsidies, for instance viataxation of property, imputed rent and land use, will yield an opportunity to cut other taxes oraccommodate additional public service provision.

ReferencesCommission of the European Union (2001), Company taxation in the internal market, SEC(1681) final, October

Michael Devereux and Rachel Griffith (2003), ‘The taxation of discrete investment choice’s,Working Paper No. W/98/16, Revision 2, Institute for Fiscal Studies, London; forthcoming,International Tax and Public Finance, 2003

Johnny i Grótinum (2003), ‘Ressourceøkonomiske problemstillinger på Færøerne’ (Ressourceeconomic issues in the Faroe Islands, in Danish), Master’s thesis, University of Copenhagen.

Grønland 2001-2002 (Greenland Statistical Yearbook 2001-2002, in Danish), GreenlandStatistical Office, 2002

Thorvaldur Gylfason and Martin Weitzman (2003), ‘Icelandic fisheries management: Fees versusquotas’, CEPR Discussion Paper No. 3849, March

David Dreyer Lassen and Søren Bo Nielsen (1996), ’Er skattebyrden i Danmark højere end iandre europæiske lande?’ (‘Is the tax burden in Denmark higher than in other Europeancountries?’ in Danish), Nationaløkonomisk Tidsskrift 134, 209-222

Carlos Martinez-Mongay (2000), ‘ECFIN’s effective tax rates. Properties and comparisons withother tax indicator’s. Economic Paper No. 146, ECFIN, October

Enrique Mendoza, Assaf Razin and Linda Tesar (1994), ‘Effective tax rates in macroeconomics’.Journal of Monetary Economics 34, 297-323

62


Hans-Werner Sinn (1991), ‘Taxation and the cost of capital: The “old” view, the “new” view, andanother view’, in Tax Policy and the Economy, National Bureau of Economic Research,Cambridge, Massachusetts

Tax Directorate of Greenland (2003), ‘Skatter og afgifter i Grønland’ (‘Taxes and duties inGreenland’, in Danish), report from the benchmarking committee, March 2003

63


Pension Reforms and Taxation in EstoniaRinga Raudla & Karsten Staehr*

Abstract: Estonia completed its pension system reforms in 2002. The new 3-pillar system features a first pillar of

universal state pension, a second pillar of funded supplementary pension and a third pillar of independent pension savings.

The paper reviews the new pension system and the early experience. It also brings up a number of unsettled issues, in part

stemming from the interplay between the pension and taxation systems. Albeit the new pension scheme has been

successfully implemented, it is excessively complex and non-transparent. Certain groups could experience inadequate

pension coverage. The public finances are in the short term adversely affected by lower payroll and income tax revenue.

Any longer-term effects on economic performance are very uncertain and likely to stem primarily from changes in the

labour supply.

Keywords: Pension reform, payroll taxes, savings, incentive effects.

JEL classification: G23, H55, O16.

1. IntroductionEstonia regained its independence from the Soviet Union in August 1991. This small Baltic nationof only 1.4 million inhabitants embarked immediately on economic reforms with the explicit goalof creating an open and market-based economy (Laar 2002). Prices, industry and trade wereliberalised at an early stage. Estonia restored its currency, stabilised inflation, privatised state-owned firms and embarked on market-supporting reforms, e.g. restructuring the financial sectorand enacting property and bankruptcy laws (Berengaut 1998). Throughout the reform periodEstonia has been among the most pro-market transition countries, witnessed by a virtual absenceof trade restrictions, free capital movements, encouragement of foreign direct investment, a lowflat-rate income tax and no corporate tax on reinvested corporate profits.1 Growth has been onaverage 5-6% per year since the mid-1990s, while inequality is higher than in most WestEuropean countries (OECD 2000:152-155).

After having established a market-based economy, Estonia has recently embarked on “deeper”institutional reforms of e.g. education, health care, taxation and pensions. The pension system wascompletely overhauled by the introduction of a 3-pillar system, implemented in stages from 1998and completed by mid-2002. The first pillar is a “state pension” financed from currentcontributions, i.e. the pillar relives on Pay-As-You-Go (PAYG) financing. The payouts from thefirst pillar are partly based on the pensioner’s work and earnings history. The second pillar is afully funded defined-contribution pension. Contributions based on the individual’s earnings are

64


* Ringa Raudla, Department of Public Administration, University of Tartu, Estonia, e-mail: [email protected]. Karsten Staehr,EuroFaculty, University of Tartu, e-mail: [email protected]. We would like to thank conference participants in Rigaand Quebec City, in particular the discussants Helge Pedersen and Niek Schoeman, for valuable comments to previousversions of the paper. This research was supported by a grant from EuroFaculty.

1 The “Index of Economic Freedom” published by the Heritage Foundation and the Wall Street Journal ranked theEstonian economy as among the world’s 10 freest in both 2002 and 2003 (Heritage Foundation 2002, 2003).

assigned to his or her pension account, administered by private funds. Participation is compulsoryfor persons born in 1983 or later, but voluntary for older persons. The third pillar comprisesadditional voluntary retirement savings encouraged by preferential tax treatment.

The Estonian pension reforms are widely considered a success, mostly because participation inthe second pillar has been much larger than anticipated (see e.g. IMF 2003). Still, participationrates are at best a rather narrow success criterion and provide little information about the societalwelfare effects.

Pension systems are a challenging object of analysis. The systems are often very complex and amultitude of issues within social policy and economics are spanned.2 The Estonian pensionreforms affect society in numerous ways; inter alia, the well-being of the retirees, the inter- andintra-generational distribution, the sustainability of the fiscal position and the long-term growthprospects. A pivotal point is the interplay between the pension and tax systems. In Estonia, e.g.,the new funded pensions are partly financed via redirection of payroll taxes into individualpension accounts administered by private funds. Clearly the effects of the pension reform can onlybe analysed in conjunction with the induced or accompanying changes in the tax system.

This paper has three purposes. First, it gives an overview of Estonia’s pension reforms puttogether from legal acts, regulations and other source materials, some of which are only availablein the Estonian language. Second, it discusses the early experiences of the reforms, in particularthe uptake and the risk profile of the preferred private pension funds. Third, the paper raises anumber of issues stemming from specific design elements, warranting further exploration. Publicpolicy and public administration issues like consistency, simplicity and transparency of the entirepension reforms package are addressed. Other issues include the overall effectiveness of thereform in ensuring adequate pension coverage for future pensioners, the possible effects ongovernment finances, as well as derived effects on growth, savings etc.

This study is mainly of a descriptive and exploratory character. This is partly dictated by the veryshort time span since the inception of reforms and consequently a lack of data. Still, we believethat our detailed presentation and examination of the reforms adds to the current debate onpension reforms – in Estonia as well as in other countries. The pension system’s basic functioning,sustainability and distributional impact is determined by a myriad of detailed provisions and theinteraction between these and the tax system.

The issues are of interest beyond the continuing policy debate within Estonia. Estonia is amongthe first countries in Eastern and Central Europe to introduce a 3-pillar pension system and theexperiences of Estonia might be of interest in these countries. We also believe that the issues areof importance for a wider set of countries. Estonia’s reforms are characterised by partiallyvoluntary participation, diversion of tax revenues from the public pension system to private fundand relatively “high-powered” incentives. Lessons from these particular design choices are alsoapplicable for high-income economies facing crucial policy choices when reforming their pensionsystems.

A number of Latin American countries have more than a decade of experiences with fundedpension systems, but the background in these countries is somewhat different from that intransition economies. Transition countries started out with a universal and relatively generousPAYG pension, while this is usually not the case in Latin American countries. Also, thepopulation in most Latin American countries is younger than in transition countries. The reformsin Latin America started out at an earlier stage of the demographic transition towards a “greying”

65

Pension Reforms and Taxation in Estonia

2 Muller (1993:3) states: “[R]etirement pension schemes are among the most complex and multidimensional arrangementsin modern societies”.

society. In these respects the transition economies, including Estonia, might be in a morechallenging situation than the Latin American countries when reforming their pension systems,but at the same time in a position that more closely resembles the situation in the high-incomecountries.

The 1990s saw renewed interest in pension and pension reforms in academic as well as policycircles. The World Bank strongly advocated pension reforms in which traditional PAYG schemeswere to be supplemented by funded systems (World Bank 1994). The main focus was initially ondeveloping countries, but turned towards transition economies in the second half of the 1990s, cf.numerous contributions from the World Bank (Lindeman et al. 2000, World Bank 2000) and theInternational Monetary Fund (Branco 1998, Cangiano et al. 1998).

The need for reform and the consequences of the World Bank recommendations are, however,debatable. Orszag & Stiglitz (2001) argue that many of the touted benefits from privately runfunded pension schemes do not stand up to closer scrutiny.3 A number of case studies have soughtto analyse the pros and cons of various pension schemes (see e.g. contributions in Murphy &Welch 1998 and Feldstein & Siebert 2002).

Only little has been published on the Estonian pension reforms in the English language. TheWorld Bank has made public the institution’s view on the policy choices of Estonian pensionreforms (World Bank 1999). Toots (2000) describes changes in the pension provision during the1990s, the reforms up to 2000 and the plans for a second pillar. Reiljan & Kulu (2003) discuss thebackground for the choice of pension reform strategy in Estonia, but consider only briefly theimpact of the reforms. Likewise, Tavits (2003) discusses factors influencing the decision inEstonia to draw up and implement the pension reform independently instead of adopting a foreignblueprint.4

This paper is organised as follows. Section 2 discusses briefly the rationale for public involvementin pension provision and the arguments for switching to a funded scheme. Section 3 describes thedesign of the pension reforms in Estonia. Section 4 considers the early outcomes, i.e. until the endof 2003. Section 5 raises some unsettled issues regarding the design choices of the pensionreforms. These contentious issues include the simplicity and transparency of the reforms, theprospect for adequate pension provision in the future, the effect on public finances and possibleeffects on growth etc. Section 6 offers some conclusions.

2. Pension reformsGovernments in most countries are in different ways involved in the provision of retirementpensions. This involvement can be corroborated using efficiency, distributional and/orpaternalistic arguments (Stiglitz 2000:358-263). Information asymmetries can lead to adverseselection resulting in missing markets. Moral hazard incentives (e.g. for excessively earlyretirement) might also lead to certain groups being excluded from private retirement insurance.Furthermore, private pension contracts might not adequately insure against social risks like waror inflation. High transaction costs can also render private pension provision inefficient. Ifpension saving is left to the market, income inequalities from peoples’ working lives will becarried into their old age, i.e. to a stage in life when people are often vulnerable and with few

66


3 See also the comments in Holzmann & Stiglitz (2001), Barr (2000), Murphy & Welch (1998) and the survey of thecritique of the “new pension orthodoxy” in Muller (1999:24-33).

4 The webpage of the Pension Centre (2003) provides a useful overview (in the English language) of the Estonian pensionsystem after the reform.

options for taking care of themselves. The result could be social deprivation and poverty forgroups of elderly people. Finally, a paternalistic government might want to protect people againsttheir own myopic behaviour and resulting inadequate pension savings.

Although there are compelling arguments for government involvement in pension provision, thegovernment still faces the question about the specific form of involvement. The “traditionalchoice” has been a PAYG system, often with all or a substantial part of the pension payouts beingin the form of defined benefits. As a rule, both high-income and transition economies haveextended this system to almost the entire population, while other countries have only extended thebenefits to specific groups, e.g. former government-employees. The defined benefits PAYGsystem can provide adequate pensions in a predictable and equitable way, thus addressing manyof the arguments for government intervention. However, a number of drawbacks of traditionalPAYG systems have recently been emphasised in policy-oriented research (World Bank 1994,Lindeman et al. 2000):

A major problem is related to the long-term fiscal sustainability of PAYG schemes in light ofchanging demographics. Longer expected lifespan and reduced fertility will increase the pensioncosts as a percent of GDP in the future. Table 1 presents data for present and projected pensionexpenditures in transition economies. The second row shows the pension expenditures in 1997-98, the third row shows the expenditures in 2050 assuming that current fertility and mortality ratesremain constant, while the fourth row shows the expenditures in 2050 assuming that the fertilityand mortality rates gradually converge towards West European levels. Among the samplecountries, the projected pension expenditures – assuming an unreformed PAYG system –apparently explode in Poland, but the projected increase is also striking in the case of Estonia.

Table 1. Pension expenditures as % of GDP in a sample of transition economies

Source: Lindeman (2000:9).

Although modest compared with projections for the future, current pension expenditures arealready a substantial fiscal burden in many transition economies. This is especially the case asthese countries experienced substantial drops in measured output in the early and mid-1990s,corresponding employment and labour supply retrenchments and less effective tax collection(Reiljan & Kulu 2003).

The long-term fiscal sustainability problem is aggravated by the fact that PAYG systems arefinanced from current tax receipts. As the pension burden increases, taxes will have to beincreased, essentially increasing the wedge between the pre-tax wages and attained consumptionpossibilities. The result of these incentive effects might be reduced labour supply (if thesubstitution effect dominates) and, thus, further difficulties generating sufficient tax revenue tofinance the PAYG pension.

67


Estonia Lithuania Poland Russia Ukraine In 1997-98 9.7 6.1 14.2 6.9 9.3 In 2050 with unchanged fertility and mortality 17.6 10.6 27.1 13.6 14.5 In 2050 with higher fertility and lower mortality 19.4 11.1 34.5 15.5 17.6

The incentive problems of unfunded PAYG systems are even more pronounced for personsreaching the legal pension age (which in some countries is relatively low, especially for women).The implicit tax on working could be close to 100% if the PAYG pension is reduced when thepensioner has labour income besides his or her pension. Higher explicit taxes would likely leadto an even further reduction of the pensioners’ labour supply.

The fiscal sustainability problem can to some extent be addressed by incremental changes of thepension schemes, e.g. by increasing the legal pension age, by gradually lowering the PAYGpension payouts or by making the payouts depend more closely on previous labour marketparticipation. While these steps can reduce the sustainability problem, they can only be applied toa certain degree without undermining the basic premise of the PAYG pension, namely to provideadequate pension coverage for elderly people.

Based on the experiences of Chile and other Latin American countries, the World Bank suggestedin World Bank (1994) a multi-pillar pension scheme. The first pillar provides a minimum pensionto all pensioners and can be PAYG financed. The second pillar is a mandatory funded schemewith individualised savings accounts. A third pillar with a voluntary private pension can be added.The underlying idea is to balance the concerns about fiscal sustainability and labour marketincentives with the desire to provide adequate pension coverage for most people. The multi-pillarapproach ensures a minimum provision for all pensioners, while at the same time isolating thepension provision – at least partially – from the effects of demographic changes.

The increased funding of pensions (second and third pillar) has been envisaged as generating anumber of other potential benefits, including:5

• Higher pension payouts if the interest on the funded pension savings is higher than theimplicit interest in a PAYG system.

• Development of domestic financial markets, and the creation of domestic financialinstitutions, especially if pension funds are privately run. Pension funds with a longinvestment horizon might contribute to greater financial stability.

• Potential for higher domestic savings if household savings increases and the governmentbudget balance does not deteriorate. Higher domestic savings could lead to a lower interestrate and higher growth if capital mobility is incomplete.

• Increased labour supply if the partial funding of pensions restrains tax pressures.

The list of the derived effects is incomplete and some of the effects are mutually inconsistent.While the list of potential benefits from switching from a pure defined benefit PAYG system to a3-pillar scheme appears impressive, the benefits are by no means certain. Recent debates havequestioned the theoretical and empirical foundation for the multi-pillar approach and have raisednumerous problems related to funded pension schemes (Barr 2000, Orszag & Stiglitz 2001).Interestingly, while more and more countries introduce 3-pillar systems, this debate intensifies.

Some areas of consensus have, however, emerged. First, pension systems have importantimplications not only for social conditions, but also for economic development and fiscalsustainability. Second, although private entities can offer pension savings schemes, governmentintervention is required to ensure the adequacy of pensions and regulate and supervise the privatepension funds. Third, the social and economic impacts of a country’s pension scheme depend in

68


5 Some authors – confusingly – label these additional benefits as “externalities”.

large part on the detailed rules and regulations governing the scheme. To rephrase the point;within the same framework, e.g. a multi-pillar scheme, different specific implementation-schemescan bring about rather different outcomes. Fourth, although somewhat overlooked in thediscussion it is evident that the interplay between the pension and the tax systems are of crucialimportance.

Having decided on a multi-pillar pension system, the policy makers still face a large number ofdecisions concerning each of the pillars and the overall functioning of the system (Fox & Palmer2001, Muller 1999:9-13): Who should be eligible for coverage of each of the pillars? Shouldparticipation be mandatory or voluntary? How should each of the pillars be financed? To whichextent should the first pillar be pre-funded? How should taxes be used to influence participationand financing? How should benefits be determined (age-criteria, defined benefit vs. definedcontribution, etc.)? Should management and ownership of pension pillars be public, private or amix of both?

This list is not exhaustive and some combinations are mutually exclusive. Still, within a 3-pillarscheme a very large number of combinations are possible, each combination denoting specificpolicy choices. The decision of reform elements may depend on the existing pension system,country characteristics, the scope of the perceived demographic shock, etc. The choice might alsobe influenced by the policymaking process and more generally by the political economyconsiderations (Muller 1999, James & Brooks 2001). Importantly, however, it will also reflectpolitical preferences with respect to distribution and societal values (Muller 1999, chp. 2). TheEstonian reforms are no exception to this picture; economics, demographics, value judgementsand politics shaped the Estonian pension reforms described in the next section.

3. The design of the Estonian pension reforms 3.1 The pension system inherited from Soviet times and the early reformsFrom Soviet times Estonia inherited a PAYG system with benefits linked to former wage and withhigh replacement rates reaching 80%. The statutory retirement age was 55 for women and 60 formen. Furthermore, there were several special provisions for persons with disabilities and selectedoccupations, which reduced the average effective retirement age to about 57 for men and 53 forwomen (Fox 1994). It has been argued that the main economic effect of this relatively generoussystem was to inflate the dependency burden on the working population (Fox 1998:372). Thenumber of pensioners in Estonia was growing throughout the Soviet period as a result of theretirement age remaining the same over 40 years while the average survival age increased (Leppik1998).

As in other Eastern European countries, at the beginning of the transition Estonia had the percapita incomes and tax-collecting capabilities of middle-income countries, but the old agedemographics and spending on pensions (as a proportion of GDP) of high-income countries(World Bank 1994:42).

In January 1991 a payroll tax (“social tax”) was introduced to collect resources for health services,old-age benefits and child benefits, and most social benefits were tied to the minimum wage(Toots 2000). In April 1993 the Act on State Living Allowances was implemented, revising theeligibility conditions and benefits of social security pensions. The Act became the basis forregulating old age pensions during the whole transition period of the 1990s. The retirement agewas scheduled to increase gradually until 2003. According to the Act of 1993 public pensionsconsisted of two parts: a flat-rate “national pension” connected to the minimum wage and a

69


supplementary part, which took into account the length of employment. Neither salaries of theSoviet period nor the transition period were taken into account in calculating the pension (Toots2000). In 1994, an amendment to the Act established that the amount for a basic pension wouldbe set annually by the Parliament.

In the early phase of the transition, changing the pension system was mainly seen as a social issueisolated from economic and demographic developments (Leppik 1998). At the beginning of the1990s, the main issue that the reform was supposed to address was the necessity to increasepensions. Only in the second part of the decade the problem of the sustainability of the pensionsystem came to the fore.6

3.2 Introducing the 3-pillar systemDesigning the 3-pillar system commenced in 1997 when the Cabinet decided to form a jointworking group from the members of the parliamentary Social Committee and the experts from theMinistry of Social Affairs to prepare a comprehensive overhaul of the pension system (Tavits2003:646).

According to the “Conceptual Framework of the Pension Reform” (Government of Estonia 1997)approved by the Estonian Government in June 1997, the main goals of the pension reform were:to guarantee the political stability and legal certainty of the Estonian pension system and convincethe public that the new system will continue without major changes. The system was to betransparent and as simple as possible, and the different pillars of the pension system and otherforms of social insurance were envisioned to form a coherent and consistent wholecomplementing each other. It was also emphasised that the new pension system has to guaranteea minimum level of pension provision for all Estonians, and ensure the desired degree ofredistribution in a transparent way. Other goals set out in the paper were to ensure the long-termsolvency of the pension system, to encourage economic growth, to decrease the proportion of theunderground economy, to avoid fiscal imbalances and to enhance the functioning of the financialmarkets.

The Conceptual Framework envisaged a 3-pillar pension system, and its underlying philosophywas very much influenced by the suggestions put forth by the World Bank in e.g. World Bank(1994) (Bank of Estonia 2003). The reform was subsequently elaborated over the next 4 years byadopting the acts changing the first pillar and introducing the second and third pillar. Interestingly,the overall reform of the pension system actually started with the introduction of the third pillar,followed by changes to the first pillar and then the introduction of the second pillar. So, forexample in 1998 the first and the third pillar of the 3-pillar system pension system were alreadyfunctioning, while the debates on the second pillar continued until 2001.

70


6 The choice not to pursue further reform at that stage was partly based on the realisation that no pension insurance systemcould have worked under the extreme inflation experienced in the early 1990s (Tavits 2003:655).

Figure 1. The Estonian old-age pension system, 2003

* indicates that the amount is adjusted annually by the average of the annual increase in theconsumer price index and the annual increase in total first pillar contributions.

71


= x

= x

1st pillar 2nd pillar 3rd pillar

Tax10 %

13 % social tax to health fund

2 %

4 %

Base amount*

Service year related pension

# years of pensionable service

Value of one service year*

Contribution related pension

# implicit service years from 1999

Value of one service year*

20 % social tax

16%

Max. 15 %

Tax 26 %, 3 x usual tax-free deductible

The first pillar of the new scheme is PAYG financed, but the pension paid out is in fact calculatedfrom three separate parts, viz. a base amount, working time-based component and a contribution-based component. The second pillar provides funded coverage based on contributions to privatepension funds, and additional voluntary retirement savings. The second pillar is compulsory forpersons born in 1983 or later, but voluntary for older persons (up to the upper age-limit). Whenindividuals have joined the second pillar, a fraction of their wage earnings will be directed intotheir individual accounts. The third pillar refers to voluntary pension savings encouraged by alenient tax treatment. We will now consider the three pillars in more details, cf. also Figure 1.

3.3 Reform of the first pillar: goals and design The main law regulating the first pillar is the State Pension Insurance Act (adopted in 1998),which pursues several policy objectives. First, it is aimed at decreasing the labour marketdistortions by strengthening the link between contributions and benefits. Second, it strives toinhibit the increase of the system dependency ratio by modifying the eligibility criteria ofpensions. Third, it aims to guarantee the compliance with the European Union requirements onequal treatment of men and women (EUS 2000:28).

The state pension system offers universal coverage for all employees and self-employed personsand is administered state by the Ministry of Social Affairs and the National Social InsuranceBoard. Also, the State Pension Insurance Register was established as a new unit of the NationalSocial Insurance Board. The register keeps individual level data about the amounts of social taxpaid on their behalf and keeps track of the individual contributions needed to calculate thecontributions-related part of the old-age pension.

The first pillar operates according to PAYG principle: current pensions are financed from theearmarked social tax paid by employers and the self-employed. The social tax contributionamounts to 33% of income, of which 13%-points is health insurance contribution, and theremaining 20%-points is earmarked for pensions, cf. also Figure 1.7 However, when a person hasjoined the second pillar, 16%-points instead of all 20%-points goes to the first pillar, while theremaining 4%-points is directed to the second pillar individual accounts.

An important change of the first pillar is the introduction of the notional defined contributionprinciple – in addition to the defined benefit principle. The old-age pension paid out of the firstpillar consists of three components: a base amount, a working-time part based on number ofaccumulated years of pensionable service and a contribution-related part.

The base amount is paid to all retirees who are eligible for state pension. The base was determinedin the State Pension Insurance Act in 2002 and is indexed annually, cf. below.8 The baseamounted to EUR 37 in 2003.

The working-time component is calculated as the number of accumulated years of pensionableservice attained before 1999 multiplied by the value of one service year. The number of years ofpensionable service comprises the time during which the person had a job for which social taxwas paid, was a member of an artistic association or trade association, was in military service, wasenrolled in an education institution, etc. The value of one service year was originally also

72


7 The pension contribution part of the social tax is used not only for old-age pensions, but also for disability pensions andsurvivor’s pension.

8 It should be noted, however, that the indexing formula has not been strictly adhered to. For example, in July 2003 thebase amount was increased by EUR 6 in addition to the increase that took place as a result of indexing in April 2003.

determined in the State Pension Insurance Act and is indexed, cf. below. The value of one serviceyear was EUR 2 in 2003.

The contribution-related component depends on the contributions paid into the first pillar onbehalf of the employee, or by themselves in case of self-employed persons, after 1 January 1999and is found by multiplying the sum of a person’s annual factors (or “implicit years”) by the valueof one service year. One annual factor or “implicit year” is a person’s pension-earmarked socialtax as a fraction of the average pension-earmarked social tax paid by all contributors in the givencalendar year.

The base amount and the value of one service year is found according to an indexation formula,cf. the amendments to the State Pension Insurance Act enacted in January 2002.9 The index iswritten up by one half of the increase of the consumer price index and one half of the increase ofthe pension parts of the social tax contributions.

In order to qualify for old-age pension an individual must have completed at least a total of 15qualifying years, which comprise years of pensionable service before 1999 and years in which theperson has earned at least 12 times the monthly minimum wage since 2000. Persons reaching theage of 63, but who do not qualify for any other type of pension, and who have lived in Estonia atleast 5 years before claiming the pension can claim the national pension. The national pensionrate was fixed by the State Pension Insurance Act and is indexed in the same way as the old-agepensions. In 2003 the national pension amounted to EUR 60.

Other changes stipulated by the State Pension Insurance Act of 1998 include the equalisation ofthe pension age at 63 for both men and women by the year 2016, the opportunity of earlyretirement with actuarially reduced pensions and the opportunity for delayed retirement resultingin higher pensions.10

3.4 The second pillar: goals and design The working group with a purpose of designing the second pillar started their work in 1999. Itdrew on the experiences of Chile, Poland, Hungary and Latvia (Erlemann & Oone 2002). TheFunded Pensions Act, which is the main act regulating the second pillar, came into force 1October 2001 but since the registration of the pension management companies and the pensionfund took time, it was only possible to start joining the second pillar from 4 May 2002.

As brought out in the explanatory note to the Funded Pensions Act and in the Progress Reportsubmitted to the European Commission (EUS 2000) the aims of the second pillar are: to preventa drop in the standard of living after retirement, to avoid the reduction in the average replacementrate due to unfavourable demographic developments, to make the pension system less susceptibleto political pressures, thus to diversify the risks endangering the pension system. Also, it aims to

73


9 Before the introduction of the indexation formula, the determination of the level of pension benefits was morediscretionary. The State Pension Insurance Act passed in 1998 stipulated that the current pension payouts are re-calculatedeach year, depending on the expected social tax revenue and following the rule that 35% of the pension expenditure shouldbe used for base amounts and 65% for the other components. The base amount, however, was to be determined by thelegislature at the passage of the annual state budget whereas the value of one service year was to be set by the Cabinet.

10A person can retire with an early-retirement pension up to three years before the legally stipulated retirement age, butin such case the amount of pension is reduced by 0.4% for each month falling short of the legally stipulated retirementage. As to the postponed retirement pension, the pension is increased by 0.9% for each month by which a person postponesretirement.

increase the responsibility of individuals in the pension-system and facilitate redistribution alongthe life-cycle.

The participants in the second pillar have to pay 2% of the gross salary to the pension fund andin addition to that the state adds 4%-points from the pension-earmarked part of the social tax.11

Thus, in total 6% of the person’s income is transferred to the pension account of the person, whilethe person has paid directly only 2%. In return for their contributions, individuals acquire units ofpension funds. Thus, the second pillar is fully funded, providing defined contribution typepensions.

In designing the second pillar throughout 1999-2000 it was debated whether the second pillarshould be compulsory or not. Initially designed as compulsory for all under 50 years old, then forunder 35 years old, it then became a target of political battles where part of the governing coalitionwanted to make it completely voluntary since they wanted to fulfil their electoral promise to theelectorate of not raising taxes.12 As a compromise, it was finally decided that participation in thesecond pillar would be compulsory for those born after 1982, while it would be voluntary forothers. Thus, the second pillar will become compulsory for everybody in about 40 years (Reinap2000). However, once an individual decides to join the scheme and acquire units of a mandatorypension fund, and submits a subscription application, the contract becomes a binding obligationthat cannot be given up. The person will have to contribute to the second pillar throughout his orher lifetime.

The person who has subscribed to the second pillar can freely choose between the differentpension funds run by private pension management companies, and change the funds bytransferring the assets from one fund to another. In order to keep down the administration andtransaction costs, the Central Register of Securities administers the payouts and databasescentrally. The Register keeps track of the applications, the funds chosen, the contributions paid,the pension fund units acquired, and benefits paid.

The unit-holder is entitled to payouts from the funded pension, when the he or she has reached thestipulated retirement age, is being paid a state pension, and at least five years have passed sincethe person started paying contributions. The payouts depend on the size of the accumulatedcontributions, and on the rate of return of the pension fund. The unit-holder can enter into aninsurance contract, paying the insurer, as a single premium, the redemption price of all theredeemed units of pension funds, and the payouts will be made periodically, on the basis of thecontract. Payments are made in the form of annuities, i.e. periodically payable amounts on thebasis of the insurance contract (annuities can be equal amounts or increasing amounts). However,if the monthly annuity exceeds three times of the amount of the national pension a person canchoose periodic payments from the pension fund without entering into an insurance contract tothe extent of the amount that exceeds three times of the national pension amount. Also, if the totalamount contributed to mandatory pension funds is less than twice the amount of the nationalpension, a person will be entitled to withdraw the whole amount in lump sum.

As for the taxation of the pension payouts, the total payouts out of the first and second pillars are

74


11 The logistics of payments look like the following: The employer of a person who has subscribed to the second pillarhas to withhold 2% of the person’s salary and transfer it to the Tax Board. The Tax Board has to verify that contributionsare made correctly and transfers the contributions together with the 4%-points from the social tax contributions to the bankaccount of the registrar of the Estonian Central Register of Securities in the Bank of Estonia.

12 It is questionable whether making people pay into individualised pension accounts should be called raising taxes, sincethe contributions remain a property of the individual (Potisepp 2000).

calculated, and the amount of triple the income-tax free minimum is not taxed. (See also Figure1.) The payouts above that amount are taxable as personal income, currently at 26%.

The investment activities of the pension funds are regulated by the Funded Pensions Act and thesupervisory authority over the pension management companies is exercised by the FinancialSurveillance Authority. It is not permitted to invest the assets of the pension funds in shares to anextent that is greater than 50% of the market value of the assets of the fund. The Act mandatesrisk-spreading by stipulating that the value of securities issued by legal persons belonging to thesame group cannot total more than 5% of the market value of the assets of a mandatory pensionfund. Similarly, the assets of a pension fund deposited in a single credit institution or in creditinstitutions belonging to the same group cannot total more than 5% of the market value of theassets of the pension fund.

If a pension management company violates the requirements, it does not render such transactionsvoid, but the company has to compensate the unit-holders of the pension fund for any damageresulting from the violation. The size of the loss will be determined on the basis of all theproprietary damage caused, including any loss of profit compared to the situation where suchviolation would not have occurred and where the assets of the pension fund associated with theviolation would have been invested similarly to the other assets of the pension fund. The FinancialSurveillance Authority will require a pension management company to acquire units in suchquantity that the total value of the units held by the company would be at least equal to the sizeof the loss.

According to the Guarantee Fund Act passed 20 February 2002, the Pension Protection SectoralFund is established out of the contributions of pension management companies. If there is a loss,and the pension management company has not compensated unit-holders before the deadline setby the Financial Surveillance Authority or if the pension management company does not havesufficient resources, the loss will be defrayed by the Pension Protection Sectoral Fund. A unit-holder would be compensated in full for losses up to EUR 10,000 per specific loss event of theunit-holder. Any loss exceeding EUR 10,000 will be compensated for to the extent of 90%.However, the state does not give any guarantees regarding the value of the units. The only lossesthat are compensated for are those arising from violating the stipulated investment requirementsand other rules set out in the acts and regulations.

3.5 The third pillarThe third pillar of the pension system – the privately managed voluntary funded pillar or thesupplementary funded pillar – is regulated by the Pension Funds Act (entered into force 1 August1998). Individuals can participate in the third pillar in two different ways (EUS 2000): they caneither opt for pension insurance policies offered by licensed private insurance companies or forunits of pension funds managed by private asset managers. Thus, the third pillar participants havea choice between the defined-benefit and the defined-contribution type schemes.

The participation in the voluntary private pension schemes has been encouraged by tax incentivesstipulated in the Pension Funds Act. The contributions can be deducted from the taxable income,with a total of up to 15% of the income. As for the taxation of payouts, if the pension savings istaken out as a lump sum after the person reaches retirement age, a rate of 10% applies. The regularpayouts or annuities, based on the insurance contract, however, are not taxable. The pension ageis decided in the contract between the individual and the insurance company. However, theminimum age for which the tax exemptions apply is stipulated by the Act, and is set at 55 years.

75


4. The early outcome of the pension reformJoining the second pillar became possible from 4 May 2002. The first phase of joining lasted until1 June 2002 at which time 44,600 people had joined the second pillar. Those who had joined bythat date could start making contributions from July 2002. The second phase lasted until 31October 2002 with contributions to the second pillar starting from January 2003. By the end ofthe second phase 207,200 people had joined the second pillar, which amounted to 36% of thetaxpayers and was a significantly larger number than expected. The third phase of joining lasteduntil 31 October 2003 with contributions starting from January 2004. Figure 2 depicts the numberof people having joined the second pillar. The different phases are clearly distinguishable.

Different age-groups have joined the second pillar in different strengths. Table 2 shows theproportion of persons in various age-groups having joined the second pillar by 31 October 2002and 31 October 2003, respectively. The participation rate of the oldest group remains broadlyconstant at approximately 25%. This reflects the fact that for those born 1942-56, the secondphase, ending 31 Oct. 2002, was the last chance to join the second pillar. For those born 1957-61,the third phase ending 31 Oct. 2003 was the last chance to participate. This group saw asignificant jump in participation during the third phase. Other age-groups have also seensubstantial increases in the proportion of persons joining, especially the group of those born after1982 where participation is compulsory for persons having their own income.

Table 2. Proportion of persons in different age-groups having joined the second pillar, %

Source: Own calculations based on Erlemann & Oone (2002), Oone (2003b) and StatisticsEstonia.

76


There are six pension fund management companies running a total of 15 different funded pensionfunds. There are essentially three types of funds that the participants can choose from: aggressivefunds investing up to 50% of assets in shares, balanced funds investing at least 75% of the assetsin debt instruments and up to 25% in shares, and conservative funds investing 100% of the assetsin debt instruments.

The risk-preferences of different age-groups are depicted in Table 3. For all age-groups takentogether, approximately 2/3 have opted for aggressive funds. The preference for aggressive fundsclearly decreases with age. According to Oone (2003a), the preferred risk-profiles wereinfluenced by the performance of the different types of funds.13 In spring 2003 the performanceindex for conservative funds was increasing, while the index for aggressive funds was decreasing.In autumn 2003, however, the index for aggressive funds was increasing again, leading significantnumbers of people to opt for it.

Table 3. The risk preference of different age-groups, % of all funds, Oct. 2003

Source: HEX Tallinn (2003) and Oone (2003b).

As for the gender profile of those joining the aggressive fund, in the second phase 69% of menand 60% of women opted for the aggressive funds (Erlemann & Oone 2002). By the end of thethird phase 72% of men and 64% of women men had chosen the higher-risk fund (Oone 2003a).

The concentration of participation can also be brought out as one of the trends: 86% of theparticipants have chosen funds managed by the two biggest banks in Estonia. 52% have joinedthe funds managed by Hansapank and 34% the funds managed by Ühispank (HEX Tallinn 2003).The total value of assets in the second pillar funds amounted to EUR 63.4 million as of 31December 2003 (Pension Centre 2003).

Table 4. The number of people joining the third pillar, end of period

Source: Sormunen (2003b).

The number of people who have joined the third pillar has been gradually increasing, cf. Table 4.By October 2003 the total value of assets was EUR 5.2 million in third pillar pension funds andEUR 34 million in third pillar pension insurance funds (Pension Centre 2003).

77


13 The performance of the different funds is tracked by the Estonian Central Register of Securities and the informationpublished on the webpage of the Estonian Pension Centre: www.pensionikeskus.ee/?id=631.

5. Points of argument Ideally the choice of pension system should be derived from the maximisation of society’s socialwelfare function (Stiglitz 2000:93-116). In practice, however, pension systems are very complexand affect society in a host of ways, which cannot be adequately incorporated in a formalmaximisation procedure. The policy-makers’ choice of various design elements must be guidedby specific (partial) objectives.

The Estonian pension reforms were aimed at addressing a large number of objectives, cf. section4. The objectives were in some cases closely connected, but also exhibit potential conflicts. In thissection we bring up a number of areas where it is arguable to which extent the goals can be met.The discussion is mostly indicative and focuses chiefly on pointing out areas where more analysisis needed. Subsection 5.1 focuses on the public policy and also administrative objective ofcreating a simple and transparent pension system. Subsection 5.2 considers the basic objective ofadequate pension provision, which naturally entails a discussion of distributional aspects.Subsection 5.3 considers the derived effects on the public finances. Finally, subsection 5.4discusses the broader effects on the overall performance of the Estonian economy.

5.1 Transparency and administrative simplicityThe “Conceptual Framework of the Pension Reform” (Government of Estonia 1997) states thatthe new Estonian pension system should be as simple and transparent as possible. The simplicityand transparency of the system would imply that the pension system can be easily understood byindividual citizens and in cases where they have to make decisions (e.g. whether to join the secondpillar or not), they can make the decision without having to go through an excessively complexreasoning process. It is, however, arguable whether these goals have been achieved; cf. also thedescription of the new pension system in section 3. The complexity stems from several sources,viz. the individual pillars, the interplay between the three pillars and finally the interplay betweenthe pension system and the tax system.

The first pillar of the pension system is complicated as the calculation of the working-time andcontribution related parts are relatively involved. The working-time component will eventuallydisappear, but it will take many decades. The indexing scheme for the payouts is also relativelyFurthermore, having joined the second pillar affects the calculation of the contribution related partof the first pillar pensions since the annual factors are based on the ratio of the individualcontribution and average contribution for each year. The main effect here is that those who havejoined the second pillar accumulate smaller annual factors and receive lower pension payout fromthe first pillar when they retire compared to the scenario if they had not joined. Thus, the implicitinterest rate of the first pillar contribution is endogenous to the system but cannot be influencedby the individual.

The second pillar in itself is perhaps easier to understand, but the investment risks are not easilyassessed. The interaction between the first pillar and the second pillar is also complex and theimplied value of the inherent taxation postponement and tax reduction is difficult to estimate indeciding whether to join the second pillar. In fact, the interaction between the two pillarsregarding the influence of joining the second pillar on the calculation of the contribution-relatedpart of the first pillar pension was not clearly brought out by the politicians, the pension funds orthe media during the introductory phase of the second pillar. While this could be understandablefrom the viewpoint of trying to attract as many people as possible to join the second pillar, it couldbe argued that there was no sufficient or easily available information regarding the interaction

78


between the pillars necessary for making an informed choice. Also, putting together the payoutsfrom both the first and the second pillar in order to find whether they exceed the triple of incometax free minimum could be seen as adding complexity to the system.

The third pillar is less complex. Still, the interaction between the first and second pillars and thethird pillar is not entirely clear. For example: when would it be beneficial to the individual torefrain from joining the second pillar and instead invest the saved amount in the third pillar?Obviously the future taxation of the payouts from the different pillars is one of the importantfactors in this context.

It might also be argued that the second pillar and the contribution-based part of the first pillar toa large extent duplicate each other, adding complexity with little apparent gain. Acounterargument is that the risk profiles of the two pillars are different; while the first pillarcarries a large amount of political risk, the second pillar carries mainly economic risks. Thisargument is only partly valid. All future pension payments are subject to substantially politicalrisk and policymakers can change e.g. the accumulation rules and the taxation of the two pillarsin the future.

The pension reforms have led to a system, which does not provide a clear and transparent basisfor making the decision of whether or not to join the second pillar. Arguably all pension systemsare complex and their functioning to some extent opaque. While this is unfortunate in all systems,it is especially unfortunate in the 3-pillar system implemented in Estonia where most individualsmust make decisions concerning whether to join the second pillar or not and whether to contributeto the third pillar.

5.2 Adequate future pension provision? The fundamental goal of all pension system is to ensure adequate pension coverage for retirees asalso emphasised in the Estonian government’s “Conceptual Framework of the Pension Reform”(Government of Estonia 2003).

Having different pillars may help to provide greater security by diversifying the pension payouts,since the first and the second pillars are subject to risks that are not perfectly correlated. The firstpillar is exposed to risks like an ageing population and increasing unemployment, whereas thesecond and third pillars are vulnerable to financial market downturns and long-lasting inflationshocks. However, the emphasis on contribution-related benefits in all three pillars may result ininsufficient pensions for some parts of the population.

Table 5 presents the Estonian government’s forecasts of the payouts from the first and secondpillars, cf. the “Preaccession Economic Programme of 2003” (Government of Estonia 2003).Although such forecasts are very uncertain and sensitive to changes in the assumptions, it isnoticeable that the average replacement rate is expected to be falling markedly between 2005 and2030. Only at the very end of the prognosis period is the replacement rate higher than in 2005.

79


Table 5. Prognosis for pension payouts from first and second pillars, %

Source: Government of Estonia 2003, own calculations.

From approximately 2030 the expected payouts from the second pillar are significantly higherthan the payouts from the first pillar. It follows that the pension benefits of those who have notjoined the second pillar may turn out to be rather low. Also noticeable is the substantial differencebetween the expected average replacement rates (including those who have not joined the secondpillar) and the replacement rates for second pillar participants in 2020 and 2030. As was broughtout in section 4, almost half of those born 1957-61 (and who cannot join anymore after 31 October2003) did not join the second pillar. For this group the average replacement rates may becomparatively low.14

Although the expected average replacement rates for second pillar participants generally show anincreasing trend, merely looking at the averages may not adequately describe the replacementrates for people with non-traditional working-paths (e.g. housewives, long-term unemployed etc)or low-income earners, who have contributed relatively little into the first and possibly the secondpillar. Since a significant part of the first pillar depends on previous earnings and the payouts fromthe second and third pillar are also earnings-related, a significant number of people might notmanage to accumulate enough earnings-related contributions during the working-time.

The level of pensions might also end up being determined by international requirements thatEstonia has chosen to follow. For example, Article 12 of the European Social Charter (ratified byEstonia in 2000) requires that a minimum level of social insurance have to be guaranteed, cf. ILOConvention no 102. The ILO Convention sets out that the pension for a person who has workedfor 30 years has to be at least 40 % of the average wage earned by a male blue-collar worker.

It could be argued that those joining the second pillar are taking on a significant risk regardingtheir pension provision, especially if they opt for the more aggressive funds that invest in shares.The payouts or annuities depend on the redemption price of all the redeemed units of pensionfunds, which reflects the situation in the stock markets when the person reaches the retirementage. Since there is no government guarantee against the investment risks, and the Guarantee Fundcompensates only for losses arising from the violation of the prescribed investment rules andlimitations, the annuities received on the basis of the contract may prove to be inadequate. Indeed,even though the regulations concerning the investment activities of the pension funds are ratherstringent and aimed at spreading the risks as much as possible, they do not protect the unit-holders

80


14 On the other hand, for those who did not join the second pillar, the contributions-related part of the first pillar pensionswould be somewhat larger because of the 20%-points of the social tax contribution (instead of 16%-points as it is in caseof the second pillar participants) being used for finding the annual factors. The extent to which that would “compensate”for not having joined the second pillar is difficult to ascertain.

from downturns on the financial markets. Also, there could be significant variation between thepayouts received by different individuals depending on how well the stock market is doing at themoment when the insurance contract determining the size of annuities is signed.

High administrative costs of running the private pension funds would, ceteris paribus, lead tolower returns (Orszag & Stiglitz 2001). In the case of Estonia, the administrative costs arecertainly decreased owing to the centralised system where the Tax Board verifies that thecontributions are made correctly, and transfers funds received upon the payout of contributionsand the necessary data into the bank account of the registrar of the Estonian Central Register ofSecurities where the number of pension fund units acquired for the contribution are calculated andtransferred to the chosen pension fund. Thus, the data-processing costs should be quite low for thepension management companies compared to a more decentralized system. In addition, theFunded Pensions Act sets out ceilings for service fees that can be charged from unit-holders.15

During 2002-03 competition between the pension management companies has hold theadministrative fees below the stipulated maximum levels. However, it is not certain that thecompanies continue charging such low fees in the future. Also, the competition between thepension funds in attracting customers either to join the second pillar or change fund once theyhave joined has given rise to extensive advertising campaigns, which can impact on the feescharged from the participants.

In addition to the adequacy of pensions in absolute terms, the question of relative adequacy andthe issue of redistribution brought about by the pension system should be addressed. The“Conceptual Framework of the Pension Reform” (Government of Estonia 1997) also brought outthat the reform should ensure the desired degree of redistribution in a transparent way. Thedistributional impact of the reforms is difficult to estimate at this stage, in part because of theabsence of background data on the persons contributing to the second and third pillar but somepreliminary insights can be brought out. The first pillar payouts are partly linked to thepensioner’s previous connection to the labour market and, hence, the payouts from all three pillarsare highly correlated with the pensioner’s earnings history. In case of the first pillar pensions, theinequality in terms of the pension benefits is gradually increasing because of the growingimportance of the contributions-related part in relation to the working-time component.

When looking at the differences in the first pillar contributions-related part, then the differencesin the accrued “implicit years” is already evident. Those earning the minimum wage haveaccumulated the sum of annual factors amounting to 1.35 in 1999-2003 (around 0.34 per year).At the same time, the largest annual factor for 2002 was 59.7 (Taliga 2003). Table 6 shows thenumber of persons having achieved certain ranges of annual factors during the period 1999-2002.A factor below 1 means that the annual pension contribution was less than the averagecontribution in that year. Approximately 70% of all contributors pay less than the averagecontribution. The accumulated effect on the pension payouts could thus be rather significant.

81


15 For example, according to the Funded Pensions Act, the rate of the redemption fee for a unit of a pension fund cannotexceed 1% of the net asset value of the unit.

Table 6. “Implicit years” (yearly coefficients) accrued, 1999-2002, thousands

Source: Taliga (2003).

The taxation of the pension payouts is lenient and, anyway, not very progressive. In addition, it islikely (although not verified by statistical data ultimo 2003) that primarily high-income earnerswill take advantage of the third pillar and, hence, receive the rather substantial tax rebates inherentin this pillar. This selection bias would further tilt the system toward regressiveness. Takentogether, it is clear that the reformed pension system brings about only limited interpersonalredistribution in the long term.

The new pension system also impacts on the redistribution between current pensioners and currentcontributors. The introduced indexing formula means that the pension payouts increase by onehalf of the inflation and one half of the increase in the pendion part of the social tax contributions.This implies that pension payouts to current pensioners are likely to increase at a lower pace thanwould have been the case if the indexing formula were written up using e.g. wage growth or socialtax intake. Because of the large uptake of the second pillar participation in 2003, the effect of thelevel of pensions predicted for 2004 is already significant and amounts to receiving a pensionwhich is 10% less than they would have received without the introduction of the second pillar(Sormunen 2003c). Thus, the current pensioners are in some sense financing the transition coststhrough the new indexing formula, which make the pension increases fall behind the wagegrowth.

5.3 Public financesPublic pension systems are important for public finances in the short (and medium) term as wellas in the long term. The implementation of funded pension is partly meant to address potentiallong-term solvency problems, cf. section 2. However, the shift from a full PAYG system to apartially funded system entails transition costs impacting on the public finances in the short andmedium term.

Pension reforms impact on the public finances via both revenues and expenditures. In the case ofEstonia, the short-term net effect on the public finances is mainly the result of a negative effectfrom lower contributions and a positive effect from the new indexing scheme. An assessment ofthe budget effects from the comprehensive pension reforms is however complicated by the factthat it is difficult to specify a realistic counterfactual.

The second pillar is operated so that 4%-points of the payroll tax (the social tax) is redirected toindividual accounts. Furthermore, the 2% of the wage-bill paid by second pillar participants is tax-exempted. Contributions to the third pillar are also tax-exempted, with a total of up to 15% of theincome of the taxable period. The foregone revenue stemming from the third pillar was EUR 3.8million in 2002 (Leppik 2003) amounting to somewhat less than 0.1% of GDP. Naturally, this

82


amount would also increase if third pillar savings outpaces GDP growth in the future.

Thus, by design the reforms have costs in the form of lower revenues to the state pension fund aswell as lower general tax revenues. This is the transition cost from a PAYG to a partially fundedsystem showing up in the form of a deteriorating budget balance when measured – asconventionally – on a yearly basis (Mackenzie et al. 2001).

However, the reforms also introduced an indexing scheme giving equal weight to both increasesin the pension part of the social tax contribution and the consumer price index. This implies inperiods with high real wage growth (and more persons joining the second pillar), that the pensionpayouts are outpaced by the wages.

The overall impact of the pension reforms on the budget was initially expected to be positive(Government of Estonia 2002). This assumed however only 80,000 would have joined the secondpillar in 2003 whereas the number of participants actually exceeded 350,000. The PreaccessionEconomic Programme of 2003 (Government of Estonia 2003) brings out more realisticpredictions, reported in Table 7, assuming that around 300,000 people join the second pillar.

Table 7. Expected net impact of the pension reforms on the government budget, millionsEUR

Net impact = impact on revenues – impact on expenditures.

Source: Ministry of Finance prediction in August 2003, cited in Government of Estonia 2003:49.

The budget for the year 2003, adopted in December 2002, incorporated expected costs forfinancing the second pillar equal to EUR 38 million (1.5% of the budget, and 0.6% of GDP).16

Until the summer 2002 it had been stipulated that the central government budget be balanced.However, in anticipation of the financing requirements of the pension reforms, amendments weremade to the Law on State Budget allowing the state budget to exhibit a deficit.

Since the take-up of the second pillar was significantly larger than expected, reaching over350,000 by the end of 2003, the impact on the budget of 2004 is even more substantial. It isexpected that the amount of social tax contributions diverted to the second pillar will be aroundEUR 57 million (Sormunen 2003c). In the explanatory note of the budget for 2004 (Ministry ofFinance 2003:420) it is brought out that the predicted pension expenditure will exceed the revenuefrom the social tax contributions earmarked for pensions by EUR 19 million.

Although the contributions to the second pillar are smaller than in several Central Europeantransition countries, the short-term budget costs are still noticeable (see also Tali 2002). In thisrespect Estonia resembles these Central European countries, which are experiencing a substantialdeterioration of their budgets, partly as the result of pension reforms (Gimbel & Eddy 2003).These public finance problems make it more difficult for the countries to fulfil the criteria for

83


16 According to the estimates of the Bank of Estonia (2003), during the next ten years the estimated annual cost of thesecond pillar may amount to 0.5% of the GDP. In light of the larger than expected uptake, this estimate could be ratherlow.

joining the EMU.

The short-term impact on the budget balance gives only a partial picture of the public financeimplications of pension reforms as the reforms also impact on future revenue and expenditureflows. A full picture of the fiscal stance would need to consider a suitably discounted sum offuture net public obligations, both explicit and implicit (Mackenzie et al. 2001). Several factorsinfluence the government’s long-term expenditure and revenue stream:

First, the second and third pillar pension payouts taking place in the future will be subject totaxation. However, as previously argued the tax treatment of the pension savings is lenient in mostcases. If these rules are not changed in the future, the tax intake from the private pension payoutswill be relatively small.

Second, the costs of first pillar pension payouts might change. The background for the reform wasa future increase of the ratio of pensioners to persons active in the labour market, cf. section 3.The rationale for the pension reform was to avoid spiralling PAYG payouts and, hence, highertaxes. Still, this requires that the implied downward adjustment of pension payouts from the firstpillar will be retained. There are reasons to believe that this could be politically difficult.Experience from high-income countries shows that it is difficult to cut public pension payouts(e.g. Bonoli & Palier 2000). This might also be the situation in the case of the transitioneconomies in which there is a long tradition for relatively generous and equalitarian pensions. Alarge number of Estonians may end up with little or no funded pension coverage, as they do notcontribute to the second or third pillars. In this situation it might be politically unacceptable toretain the stipulated first pillar payouts to an absolute minimum. The result could be relativegenerous payouts via the first pillar combined with foregone revenues from the second and thirdpillar.

Third, derived effects on the economy in the form of larger production would lead to higher taxintake and, hence, an improved fiscal stance in the long term. This channel therefore hinges onthe size of the labour supply response and possibly also on effects stemming from higher domesticsavings. The dynamics of this revenue source is difficult to estimate.

5.4 Economic performanceThe Estonian pension reforms had wider goals than “just” providing adequate pension coveragewithout risking short-term or long-term solvency problems in the pension system. This subsectiondiscusses some of these broader objectives of the reforms and their underlying rationale.

Household savings

A pivotal issue is how funded pension schemes impact on private household savings. Thequestion is of importance for at least three reasons. First, one of the goals of the pension reformis to entice the households to take greater responsibility for future labour income cessation andhigher private savings would indicate that this policy objective has been achieved. Second, theEstonian pension reforms entail that private pension savings is “sweetened” by transfer of taxrevenue (second pillar) and substantial tax concessions (third pillar) presumably giving greaterincentive for private savings. Third, household savings constitute in many countries an importantpart of overall savings.

84


Mackenzie et al. (1997) provide a survey of the theoretical and empirical literature. Evidently itis crucial how the pension system is financed and how the private sector reacts to the pensionscheme. Empirical work for the USA and some other countries shows that the introduction ofprivate pension schemes increases savings, i.e. the increased pension savings is not fully offset bya reduction of other forms of private savings. Lavi & Spivak (1999) show that the result holds forIsrael using macro data as well as household data.

Estonians contributing to the second pillar pay only approximately 1/4 of the savings out of theirown pocket, while the redirection of part of the payroll tax and the saved income tax make up thebalance. If the second pillar contribution does not offset other forms of household savings, thenthe households’ net savings would – ceteris paribus – increase by the entire contribution.However, the second pillar contribution might offset other forms of household savings, and theoverall pension system reforms might also influence the households’ savings decisions. Still, thenet effect on household savings is likely to be positive. The very short period since the inceptionof the Estonian reforms and the absence of micro data implies that no empirical testing of the issuecan be undertaken.

National saving

The effect on overall savings of the new 3-pillar pension scheme is of interest, not least from amacroeconomic viewpoint. Overall national savings stems from the government, the householdsand the corporate sector (including the financial sector). Numerous case studies (Mackenzie et al.1997) and cross-countries econometric analyses (Samwick 2000) generally conclude thatswitching to a funded pension system has indiscernible effects on overall savings.

The Estonian pension reforms impact the overall government balance via several channels, cf.subsection 5.3. The direct effect of the diverted payroll tax is to worsen the government balance,but this is moderated by the new indexing scheme for pension payouts. The net effect has beenclearly negative. Households are likely to increase their savings, cf. above. The effect on thecorporate sector depends mostly on derived effects, e.g. changes in production, and this effect isuncertain. While government savings fall, household savings increase while corporate savings isuncertain. It appears likely that the switch to a funded system is indeterminate and anyway ratherinsignificant. This is in accordance with international experience.

Labour market participation

Another stated goal of the pension reforms is to increase labour market participation. Theofficially measured participation can be influenced by substitution between the official and theshadow economy as well as by changes in the total labour supply.17 The introduction of thefunded pillars improves the incentives to participate in the official labour market as theconventionally measured tax pressure is reduced.

Vork (2002) shows using household data that the labour supply in Estonia is very sensitive tochanges in the tax rate, but does not determine the relative importance of changes in the overallsupply and substitution from the informal to the formal sector. Thus, the stronger incentives forformal work introduced by the reforms of the pension and tax system are likely to lead to largerlabour market participation. The improved incentives are not restricted to those joining the secondpillar as the first pillar payouts to a large extent depend on the lifetime contributions.

85


17 A reduction in the underground economy was also considered a goal in itself.

Functioning of financial markets

Deeper and more efficient markets are often touted as a goal of introducing funded pension runby private financial entities, and Estonia is not an exception, cf. subsection 3.2. As described theuptake of the second pillar has been strong and the private funds appear well functioning. Oneshould notice, however, that the pension funds are almost entirely investing its assets abroad. Thisinvestment strategy surely reflects a prudent diversification of risks. Still, it also implies that theEstonian financial markets have been little influenced by the pension reforms and any positiveeffects are likely to be indirect, e.g. by an upgrading of fund management know-how.

Economic growth

A stated objective of the pension reforms was to increase economic growth, an area of greatimportance for Estonia with its per capita production substantially below neighbouring countrieslike Finland and Sweden. A number of possible sources generating higher income have beenconsidered in this subsection. In a closed economy, a higher national savings rate will lead to alower domestic interest rate, increased capital accumulation and higher growth (at least for aperiod of time). However, Estonian is a small and extremely open economy and the interest ratelevel is essentially determined from abroad. Thus, changes in the savings rate influence thecurrent account balance, but should have no impact on growth.

Increased labour market participation would lead to higher recorded production. The welfareeffect however might not fully reflect the increased recorded production. To the extent that theincreased labour supply is caused by substitution from the shadow economy to the officialeconomy, the increased production largely reflects a statistical reclassification. To the extent thatoverall labour supply increases, the higher production goes together with longer working hours.

Based on recent empirical literature, a broad-based view is that deeper and more efficientfinancial markets have a positive impact on economic growth (Ross 1997). Whether there will beany discernable effects in Estonia might depend on the actual developments in the financialsector, specifically whether the financial intermediation within Estonia improves.

In sum, it appears very uncertain whether the pension reforms have any discernable effects oneconomic growth in the medium term. If indeed there will be a positive effect, it is most likely tocome through increased labour market participation.

6. Final commentsThis paper has considered the recent reforms of Estonia’s pension system. The reforms meant anoverall tightening of the eligibility requirements. As discussed in e.g. section 3, the design of thereforms has often represented political compromises leaving the new system with a number ofcharacteristic “birthmarks”:

• A high degree of voluntary participation, especially at the early stages of reform.

• High-powered incentives: Payouts from the first two pillars are closely correlated withpast earnings.

• Substantial incentives for participation, especially redirection of payroll taxes for thesecond pillar and tax exemption for the third pillar.

86


• Lower payroll and income tax revenue, but revenue loss “counterbalanced” by pensionpayouts only partly indexed to wage increases.

• Individual freedom to choose between the different private pension funds and their riskprofile.

These design choices exhibit a certain internal consistency. For example, the voluntarycontributions essentially necessitated generous incentives to encourage participation and quicklyreach a “critical mass”. The design choices also show that the pension reform continues themarket-based and individualistic approach characterising Estonia’s economic policies since thecountry regained its independence in 1991.

Our discussion focused on a number of contentious points, where it is possible to question theexpediency of the reforms. The political process and many interests to be embraced have led to avery complex pension system. Informed decisions with respect to the participation in the secondand third pillars require a thorough understanding of the system and meaningful estimates offuture economic developments and policies. For most Estonians the choices will be verychallenging and it could be feared that aggressive advertising, past fund performance and marketrumours have undue impact on peoples’ pension decisions.

We also argued that the “high-powered” incentives could lead to inadequate pension coverage forpersons with non-traditional careers, e.g. housewives not working, away from home and otherpersons with very low lifetime income. One might conjecture that the rates and indexationschemes will be changed in the future as this problem becomes apparent.

The redirected payroll tax and deferred income taxation, ceteris paribus, lead to a substantialgovernment revenue loss. Rapid wage growth in combination with the changed indexationscheme has prevented serious public finance problems during the first couple of years. Still, theincentive schemes of the second and third pillars – in the form of payroll tax redirection andincome tax concessions – are notable. A substantial part of the new private pension savingsessentially reflects a transfer of resources from the public sector to parts of the private sector. Theappropriateness of this choice hinges in large part on the private and the public sector’s reactionto this transfer.

Derived effects in the form of improved economic performance are difficult to access. Any effectsfrom savings to economic growth are unlikely, but the pension reforms might lead to larger laboursupply and hence a period of higher growth.

In sum, the implementation of a 3-pillar pension system in Estonia has proceeded smoothly andthe uptake of the funded second pillar has been much greater than anticipated. Still, many issuesremain unsettled. The interplay between the different pension pillars and the tax system iscomplicated and appears in some cases unfounded. The high-powered incentives might haveinappropriate distributional consequences and could lead to poverty problems. The publicfinances have weakened in the short-term.

Estonia has undergone more than a decade of economic reforms creating a “normal” market-basedeconomy closely integrated with especially the European economies. The pension reformsdiscussed in this paper are another step in the direction of structural change and market-conformity. Still, our paper suggests that there is room – and reasons – for future changes to theEstonian pension and taxation systems. Estonia has also in this respect come to resemble otherEuropean countries.

87


ReferencesAndrews, Emily & Mansoora Rashid (1996), “The Financing of Pension Systems in Central andEastern Europe: An Overview of Major Trends and Their Determinants, 1990-1993”, World BankTechnical Paper Social Challenges of Transition Series, no. 339.

Bank of Estonia (2003), “Eesti pensionireform ja selle mõju eelarve tasakaalule” [The Impact ofthe Pension Reform on the Budget Balance], Rahapoliitiline ülevaade, March, p. 18-20.

www.eestipank.info/pub/et/dokumendid/publikatsioonid/seeriad/ylevaade/_2003_01/_2mar_taust.pdf

Barr, Nicholas (2000), “Reforming Pensions – Myths, Truths, and Policy Choices”, IMF WorkingPaper, no. WP/00/139.

Berengaut, Julian et al. (1998), “The Baltic Countries. From Economic Stabilization to EUAccession”, IMF Occasional Paper, no. 173.

Bonoli, Giuliano & Bruno Palier (2000), “Pension Reforms and the Expansion of PrivatePensions in Western Europe”, in Heyen, Erk (ed.): Pension Systems for Public Servants inWestern Europe (19th/20th c.), Nomos Verlagsgesellsschaft, p. 153-174.

Branco, Marta (1998), “Pension Reform in the Baltics, Russia, and other Countries of the FormerSoviet Union (BRO)”, IMF Working Paper, no. WP/98/11.

Cangiano, Marco, Carlo Cottarelli & Luis Cubeddu (1998), “Pension Developments and Reformsin Transition Economies”, IMF Working Paper, no. WP/98/151

Erlemann, Jaanus, Kaidi Oone (2002), “Kogumispensioni reformi vahekokkuvõte ja liitujatestatistika” [The Interim Summary and Statistics of the Funded Pensions Reform].

www.pensionikeskus.ee/et/kodanikule/uudised/index.php?news_content_id=2598&news_fid=5024

EUS (2000), “Estonia’s Progress Report for the Commission Review”, European UnionSecretariat, Estonia, June 2000.

www.eib.ee/files/report.pdf

Feldstein, Martin & Horst Siebert (2002), Social Security Pension Reform in Europe, NationalBureau of Economic Research and Chicago University Press.

Fox, Louise (1994), “Old-Age Security in Transitional Economies”, World Bank Working Paper,no. 1257.

Fox, Louise (1998), “Pension Reform in the Post-Communist Transition Economies”, in Nelson,Joan, Charles Tilly, & Lee Walker (eds.): Transforming Post-Communist Political Economies,Task Force on Economies in Transition, National Research Council.

Fox, Louise & Edward Palmer (2002), “New Approaches to Multipillar Pension Systems: Whatin the World is Going On?”, in Holzmann & Stiglitz (2002).

Gimbel, Florian & Kester Eddy (2003), “Hungarians Look to Pension Pay-off”, Financial Times,13 October, FT Fund Management Supplement, p. 8.

Government of Estonia (1997), “Pensionireformi kontseptsiooni põhipostulaadid” [ConceptualFramework of the Pension Reform].

88


www.ensib.ee/toetused/pensionireform.html

Government of Estonia (2002), “Preaccession Economic Programme of 2002”.

www.tan.ee/tan/et/doc/Documents/1031667618.24/document_info

Government of Estonia (2003), “Preaccession Economic Programme of 2003”.

http://keskus.parnumaa.ee/files/85.pdf

Heritage Foundation (2002), “The 2002 Index of Economic Freedom”.

www.heritage.org/research/features/index/2002/

Heritage Foundation (2003), “The 2003 Index of Economic Freedom”.

www.heritage.org/research/features/index/2003/

HEX Tallinn (2003), “Kogumispensioni statistika” [Statistics on the Funded Pensions].

http://files.hex.ee/pensionikeskus/dokumendid/Kogumispension_05112003.pdf

Holzmann, Robert & Joseph Stiglitz (eds., 2001), New Ideas About Old Age Security, The WorldBank.

IMF (2003), “IMF Concludes 2003 Article IV Consultation with the Republic of Estonia”,International Monetary Fund, Public Information Notice, no. 03/127.

www.imf.org/external/np/sec/pn/2003/pn03127.htm

James, Estelle & Sarah Brooks (2001), “The Political Economy of Structural Pension Reform”,in Holzmann & Stiglitz (2001).

Laar, Mart (2002), Little Country that Could, CRCE New Series, no. 19

Lavi Yaacov and Avia Spivak (1999), “The Impact of Pension Schemes on Savings in Israel”,Applied Economics, vol. 31, pp. 761-774.

Leppik, Hele (2003), “III sammas tõi tagasi ligi 60 miljonit krooni” [The Third Pillar BroughtBack 60 Million Kroons], Eesti Päevaleht, 8 May.


Leppik, Lauri (1998), Eesti pensionireform: kolm sammast [Estonian Pension Reform: The ThreePillars], Jaan Tõnissoni Instituut.

Levine, Ross (1997), “Financial Development and Economic Growth: Views and Agenda”,Journal of Economic Literature, vol. 35, no. 2, p. 688-726.

Lindeman, David, Michal Rutkowski & Oleksiy Sluchynskyy (2000), The Evolution of PensionSystems in Eastern Europe and Central Asia: Opportunities, Constraints, Dilemmas andEmerging Practices, The World Bank.

Mackenzie, George, Philip Gerson, Alfredo Cuevas & Peter Heller (2001), “Pension Regimes andSaving”, Occasional Paper, no. 153, IMF.

Mackenzie, George, Philip Gerson, Alfredo Cuevas & Peter Heller (2001), “Pension Reform andthe Fiscal Policy Stance”, IMF Working Paper, no. 01/214.

Ministry of Finance (2003), “2004 aasta eelarve eelnõu seletuskiri” [Explanatory Note to the

89


Budget of 2004].

www.fin.ee/index.php?id=7963

Muller, Katharina (1999), The Political Economy of Pension Reform in Central-Eastern Europe,Edward Elgar.

Murphy, Kevin & Finis Welch (1998), “Perspectives on the Social Security Crisis and ProposedSolutions”, American Economic Review, Papers and Proceedings, vol. 88, no. 2.

Neudorf, R. (2001), “Inimesed kardavad riskeerida pensionifondi investeerides” [People AreAfraid to Risk by Investing in Pension Funds], Eesti Päevaleht, 29 August.

http://leht.epl.ee/artikkel.php?id=116340

OECD (2000), Economic Surveys 1999-2000. The Baltic States. A Regional EconomicAssessment, OECD.

Oone, Kaidi (2003a), “Tõsised kampaaniad on lõppenud” [Serious Campaigns have Finished].Äripäev, 6 November.

www.pensionikeskus.ee/?id=2178

Oone, Kaidi (2003b), “Kogumispension arvudes” [Funded Pension in Numbers].

http://files.hex.ee/pensionikeskus/dokumendid/031024_rahakompass_IIsammas_EVK.ppt

Orszag, Peter & Joseph Stiglitz (2001), “Rethinking Pension Reforms: Ten Myths about SocialSecurity Systems”, in Holzmann & Stiglitz (2001).

Pension Centre (2003), Pension Centre Web page.

www.pensionikeskus.ee

Postimees (2003), “Tänavu liitub fondidega kuni 100,000 inimest” [This Year 100,000 People areExpected to Join the Funds], 5 May.

www.postimees.ee/index.html?op=lugu&id=99429&number=797&rubriik=3

Potisepp, Priit (2000), “Pensionireform – kas oleme ikka õigel teel” [Pension Reform – Are Weon the Right Track?], Äripäev 15 Sept.

www.aripaev.ee/1763/arv_kolumn_176302.html

Puss, Tiia, Mare Viies & Reet Maldre (2000), “Pensionikindlustuse reformi majanduslikud jasotsiaalsed aspektid Eestis ning selle mõju rahvastiku arengule ja perepoliitikale” [The Social andEconomic Aspects of the Pension Reform in Estonia and Its Impact on the Development of thePopulation and Family Policy].

www.riik.ee/rahvastik/tpyss.htm

Reiljan, Janno & Liina Kulu (2003), “Analysis of the Incentives for the Implementation of theWorld Bank’s Multi-Pillar Approach: Pension Reform in Estonia”, mimeo, University of Tartu.

Reinap, Aivar (2000), “Riik ei sunni tulevikule mõtlema” [The State Does not Force People toThink of the Future], Postimees 27 Oct.

www.postimees.ee/index.html?number=840&op=lugu&id=2250

Reinap, Aivar (2003), “Pensionitõus sõltub paari tuhande meestöölise keskmisest palgast” [The

90


Pension Increase Depends on the Average Wage of Some Thousand Male Workers], Postimees 6November


Rumessen, Inge (2002), “Pensionifondid võivad tulevikus suuremaid riske võtta” [The PensionFunds Can Take Larger Risks in the Future], Äripäev 13 Dec.


Samwick, Andrew (2000), “Is Pension Reform Conducive to Higher Saving?”, Review ofEconomics and Statistics, vol. 82, no. 2.

Sormunen, Raivo (2003a), “Kolmanda samba atraktiivsus väheneb” [The Attractiveness of theThird Pillar is Decreasing], Äripäev 25 April.


Sormunen, Raivo (2003b), “Kolmas sammas liigub ühtlaste tuuridega edasi” [The Third Pillar isMoving on with a Constant Speed], Äripäev 25 July

http://www.pensionikeskus.ee/et/kodanikule/uudised/index.php?news_content_id=2796&news_fid=0

Sormunen, Raivo (2003c), “II sammas närib esimest sammast väiksemaks” [The Second Pillar isMaking the First Pillar Smaller]. Äripäev 28 November.


Stiglitz, Joseph (2000), Economics of the Public Sector, W.W. Norton & Co, 3rd ed.

Tali, Veiko (2002), “Riigieelarve defitsiit ja teised võimalikud riskid” [The Deficit of the Budgetand Other Possible Risks].


Taliga, Harry (2003), “Eesti pensionisüsteemi probleemid töötajate vaatevinklist”, presentation atthe seminar “Pension Strategy and Participation of the Elderly in the Labour Market”, organisedby the European Commission and the Estonian Ministry of Social Affairs, 12 November.

http://files.hex.ee/pensionikeskus/dokumendid/Seminar-Taliga[1].ppt

Tavits, Margit (2003), “Policy Learning and Uncertainty: The Case of Pension Reform in Estoniaand Latvia”, Policy Studies Journal, vol. 31, no. 4, p. 643-660.

Toots, Anu (2000), “Effect of Estonian Social Security Reform on Old-Age Poverty:achievements and problems”, mimeo, Tallinn Pedagogical University.

www.nispa.sk/news/toots.rtf

Vork, Andres (2002), “The Effect of Social Benefits on Labour Supply in Estonia. AnEconometric Analysis”, mimeo, University of Tartu.

World Bank (1994), Averting the Old Age Crisis. Policies to Protect the Old and PromoteGrowth, Oxford University Press.

World Bank (1999), “Pension Reform Policy Workshop. Aide Memoire”, published after World

91


Bank mission to Estonia.

www.worldbank.org/wbi/pensions/courses/march2000/ proceedings/pdfpaper/week1/estonia.pdf

World Bank (2000), Balancing Protection and Opportunity: A Strategy for Social Protection inthe Transition Economies, The World Bank.

[All Internet links valid 6 January 2004.]

Statutes

Funded Pensions Act, State Gazette RT I 2001, 79, 480; amended RT I 2002, 44, 284 and RT I2002, 23, 131.

Guarantee Fund Act, State Gazette RT I 2002, 23, 131.

Pension Funds Act, State Gazette RT I 1998, 61, 979, amended RT I 2001, 48, 268.

State Pension Insurance Act, State Gazette RT I 1998, 64/65, 1009, consolidated text RT I 2001,9, 42, and amended Act RT I 2001, 100, 648.

State Budget, State Gazette RTI, 07.01.2003, 3, 18.

Law on State Budget State Gazette RT I 1999, 55, 584.

European Social Charter

http://conventions.coe.int/Treaty/EN/Treaties/Html/035.htm

ILO Convention No 102 Concerning Minimum Standards of Social Security.

http://ilolex.ilo.ch:1567/english/convdisp1.htm

92


Automatic Fiscal Stabilisers in Estonia Rasmus Kattai, Alvar Kangur, Tanel Liiv, Martti Randveer *

Abstract The paper discusses the functioning of automatic fiscal stabilisers in Estonia. The aim of the research is to

evaluate government budget sensitivity to economic fluctuations and thereby assess the importance of automatic fiscal

stabilisers in Estonia. Specifically we are interested in whether the functioning of automatic fiscal stabilisers might under

certain circumstances create difficulties for the fulfilment of the Maastricht deficit criterion according to which the public

deficit is not allowed to exceed 3% of GDP.

Keywords: Automatic fiscal stabilisers, structural budget balance, economic cycle, fiscal policy.JEL classification: E32, H2, H5, H6, H87

1. IntroductionThe purpose of this paper is to analyse automatic fiscal stabilisers (AFS) in Estonia, i.e.fluctuations in budget revenue and expenditure affected by change in economic activity that,through influencing domestic demand, smooth business cycles. The paper focuses only on generalgovernment budget balance reactions — the size of the automatic fiscal stabilisers, which isdefined as a difference between actual budget balance and the cyclically adjusted balance(structural balance). AFS stabilising capacity is not observed in this particular research.

Fiscal policies and the analysis of fiscal stabilisers has been added to the agenda mainly becauseEstonia is applying to become a member of the European Economic and Monetary Union (EMU)necessitating the evaluation of the governments ability to maintain the balance of budget withinthe boundaries set by the Maastricht Treaty, i.e. less than 3% of GDP. As AFS makes thebudgetary position more volatile, it is necessary to observe how big these effects are. The role ofAFS is primarily seen as the course for the future – fiscal stabilisation should be based on AFSfunctioning mainly in EMU. Discretionary fiscal policy should be the exception rather than theusual means. In an analysis of discretionary fiscal policy the deficiencies of these methods are asfollows:

- Temporary external inertia1. According to Buti and Sapir (1998) the time that it takes toacknowledge changes in the economic circumstances and to make the necessary decisionsis from 6 to 18 months.

- Loss of output. Fiscal factors usually reduce economic output2. At the same time

93

Automatic Fiscal Stabilisers in Estonia

∗ Bank of Estonia: [email protected], [email protected], [email protected], [email protected]. Theviews expressed are those of the authors and do not necessarily represent the official views of the Bank of Estonia.

1 Under temporary external inertia we mean the time that it takes for fiscal factors to influence domestic demand. Usuallythis is shorter in cases of cutting income tax, profit transfers and liquidity constraint for the management of domesticeconomies (Hemming et al., 2002).

2 There have been a number of empirical studies in this field. Focusing on works published lately (improved methodology,better statistic base) for example Bassanini, Scarpetta and Hemmings’ (2001) argue that direct tax increases on the revenueside and increased income transfers adversely affect the growth of GDP.

democratic election principles allow the easier management of budget deficits than thebalance or surplus of the budget. During an economic downturn, most governments tendto support domestic demand through a general increase in the current consumption, whichin turn, through the expansion of dependent groups, makes consumption cut backs in thegrowth phase of the economy almost impossible3.

- Instability. Government’s discretionary actions make central bank decision-making morecomplicated (Taylor, 2000). This is due to the fact that discretionary actions are ofteninconsistent, and each new coalition seems to have a different view of the implementationof given steps.

These findings explain the research results in the European Commission’s 2000 report. Accordingto this, discretionary steps worked in a pro-cyclic manner for EU states from 1970-1990,increasing the fluctuations of the GDP gap. The European Commission has found that ifdiscretionary steps can be justified at all, then it is only in the context of a very strong asymmetriceconomic downturn affected by a decrease in demand or in the instance of economic overheating.The situation is different if the economic downturn is caused by permanent shifts in the supplyside, which may reduce potential GDP. In such cases automatic stabilisers can be destabilising(Meyermans, 2002). The problem here is that it is impossible, in the short term, to distinguishbetween negative demand side impulses and supply shock.

Because of this, the Maastricht Treaty set a 3% limit on public sector budget deficits and a 60%limit on debts in relation to GDP. In 1997 in Amsterdam, this vision was even more clearly statedwithin the Stability and Growth Pact (SGP). According to this agreement, the member statesshould keep their budgets stable or in surplus in the long term and let automatic fiscal stabilisersoperate freely during the entire business cycle. Also, fines4 are imposed for those who exceed thebudget deficit limits.5 Empirical research has shown that the 3% deficit leaves more than enoughroom for automatic fiscal stabilisers to work without the necessity to use discretionary measureswhen the structural budget is balanced.

To determine the size of automatic fiscal stabilisers and the structural balance, we have used thewidely accepted two-step method. The first stage of this method is to measure GDP gap. Thesecond is to identify the cyclical sensitivity of all budget components (on revenue and expenditureside) — their sensitivity to the GDP gap. Finally the budget’s cyclical component and structuralbalance are calculated based on sensitivity estimates and the GDP gap.

The only drawback of the two-step method is possible over-estimation of the sensitivity, mainlybecause it does not take into account the mutual impact of the fiscal position and domesticdemand. Also, some steps on the expenditure side (for example changes in the health insurance

94


3 For example, Fölster and Henrekson (2000) found in their research that although previously missing or weakconnections had been found between budget sizes and the growth of GDP, the research done was not reliable, especiallyfrom the technical point of view. The more these studies have solved econometric problems the more clearly do thenegative connections between the size of the government sector and the growth of GDP appear. According to the OECD(2000) a raise in taxes of 1% in the long run reduces the productivity to the employed citizen by 0.6 to 0.7%.

4 The size of the fine consists of the fixed portion, which is 0.2% of annual GDP and the changing portion, depending onby how much the boundaries are exceeded. The maximum fine is 0.5% of GDP.

5 Exemption from paying the fine is given when GDP decreases by more than 2% (a situation which during the last 40years has only happened in Finland (EMU member state) and the United Kingdom and Sweden (non-members)).Depending on the provisions, exemption from paying the fine can be considered in cases when the GDP falls between 0.75to 2%.

system) can be pro-cyclic and shifts in economic structure may affect the estimation results. Analternative would be to use the SVAR (Structural Vector Auto-Regression) method to calculatestructural budget balance. SVAR analysis also provides an estimation of the influence ofdiscretionary political steps on the budget, and the corresponding component will be separatedfrom the revenue/expenditure time series (see for example Höppner, 2002). Many factors renderthis method unreliable in our context. One of these is the shortness of the data series but also themodest size of the Estonian economy and Estonia’s position in the transition phase all limit theuse of SVAR. The main deficiency of the SVAR method is that it is useless in cases of structuralchanges in the economy, which are present when dealing with a transition economy. Also, whenassumptions change, results will also change remarkably.

The structure of this paper is as follows. The second section contains a brief overview of the roleof automatic fiscal stabilisers in EU member states in order to give a comparison with Estoniandata. According to the two-step method, estimation of the potential GDP and GDP gap isdescribed in section three. In the fourth section the budgetary cycle sensitivity of the Estoniangovernment sector is calculated, followed, in the fifth section, by a presentation of the Estoniangovernment sector structural budget estimates and the size of the automatic fiscal stabilisers.

2. The role of automatic fiscal stabilisers in the EUThe European Commission uses structural budgets as inputs for budgetary analysis. According tothe SGP these are one of the main indicators according to which the fiscal policies of memberstates are planned. It must be mentioned that structural budget estimates are not totally adequate— since even if the separation of the budget’s cyclical components was not dependent upon thecalculation method used, the business cycle expressed by the GDP gap is not directly measurable.Different international institutions use different methods to define the GDP gap and this is whythe results differ quite significantly. But structural budget estimates are still used.

According to different sources, the average cyclical sensitivity of budgets among EU memberstates is approximately 0.5, implying that when the GDP gap grows by 1 percentage point, thereis a shift in the budget balance of 0.5% of GDP. Generally, in southern EU member states thesensitivity is smaller. This group includes Greece, Spain, France, Portugal, Italy, Austria butLuxembourg and Ireland as well. The indicator is highest in Denmark, Sweden, Great Britain, theNetherlands and Finland (see Figure 1).

Figure 1. Budgetary sensitivities of EU countries (Bouthevillain et al., 2001; EuropeanEconomy, 2002; Valtioneuvoston Kanslia, 2000).

95


0.2

0.4

0.6

0.8

1.0

1.2

BEL GER GRE SPA FRA IRL ITA LUX NED AUT POR FIN EUR12

DEN SWE GBR EU15

EC OECD IMF ECB

It is noteworthy that budget revenues are more sensitive to business cycles. The influence onexpenditure is relatively small. This means, that the fluctuations of the business cycle influencetaxes more than government transfers and consumption (see European Central Bank (ECB) datain Table 1).

Table 1. The cyclical sensitivity of budget revenues and expenditures in EU member states

(% of GDP when the GDP gap grows by 1 percentage point)

Source: Bouthevillain et al., 2001 (ECB).

The size of the cyclical components depends on budget cyclical sensitivity and the present stateof the economy. The results differ depending on the research method. Thus the results should beinterpreted as an interval of where the actual figure potentially stands. In states that have largerautomatic stabilisers, the effect of the budget cycle can be up to 3% of GDP. The EMU averageis about 0.5% of GDP (see also Appendix 1).

One of the reasons why the cyclical components vary so much is the difference in the abovementioned sensitivity assessments. The other reason is that international organisations usedifferent methods to assess the GDP gap. The OECD, IMF and European Commission (EC) usethe production function6 while the European Central Bank uses the Hodrick-Prescott’s filterdefined GDP trend7. The results of these two methods are out of step with each other.

The AFS functioning makes revenues and expenditures more volatile and thereby it is moredifficult to keep the budget balanced. Nevertheless, several studies have shown, that theprescribed 3% limit allows sufficient room for automatic stabilisers to work without risking thestated boundaries, assuming that the structural budget is balanced or adequately overbalanced(Artis et al., 2000; Barrell et al., 2001; Dalsgaard et al., 1999; Dury et al., 2000). Artis and Buti(2000) indicate that Belgium, Denmark, Spain, Ireland, Luxembourg, the Netherlands, Portugaland the United Kingdom should keep their structural deficit between 0% and 1% of GDP.Germany, Greece, France, Italy and Austria could have a deficit that is even larger than 1%.Budgets in Finland and Sweden are relatively cycle sensitive, and so in these countries thestructural budget should be in surplus. Similarly, Barrel, Hurst and Pina (2002), claim that onlyAustria should keep its structural budget in balance in order to prevent automatic stabilisersincreasing the deficit in excess of 3%. Most EMU states may experience a deficit as large as1 %of GDP.

96


6 The European Commission has used the production function to measure potential GDP since 2001. Previously they alsoused the Hodrick-Prescott filter.

7 Bouthevillain et al., 2001 have written about using the HP filter for the measuring of potential GDP.

BEL GER GRE SPA FRA IRL GBR ITA Total 0.56 0.45 0.38 0.40 0.53 0.42 0.65 0.48 Revenue 0.49 0.40 0.38 0.35 0.48 0.33 0.43 0.47 Expenditure -0.07 -0.05 0.00 -0.05 -0.05 -0.09 -0.22 -0.01

LUX NED AUS POR FIN DEN SWE EU15 Total 0.33 0.69 0.47 0.50 0.55 0.67 0.75 0.53 Revenue 0.30 0.45 0.50 0.42 0.48 0.56 0.61 0.44 Expenditure -0.03 -0.24 0.03 -0.08 -0.07 -0.11 -0.14 -0.09

3. Estonia’s potential GDP and GDP gap.The first step in finding a structural balance is to estimate the GDP gap, which is needed for bothestimating budget sensitivity and extracting budget’s cyclical component. As mentioned in thefirst section, in the context of EMU states the correct GDP gap is most important determinant ofthe size of the cyclical component of the budget, and that is why its measurement in the currentcontext is critical. It should also be mentioned that it is very difficult to measure the GDP gap inEstonia because available GDP time series cover only a little more than one business cycle.

Determining the GDP gap is based on an estimate of the potential GDP. There are two mainoptions to estimate the potential output — using the Hodrick-Prescott filter or the productionfunction, both of which are used in this paper.

3.1. Measuring the GDP gap using the Hodrick-Prescott filterThe main argument for using the Hodrick-Prescott filter is that with this method there is no needto assess the production inputs. Because we lack highly reliable data on capital stock, a suitableapproximation must be found for the factors used in the production function which, as a result,reduces the reliability of the estimate of the potential GDP. Following these considerations, theHodrick-Prescott filter is used for estimating potential GDP. The positive characteristics of thefilter are its simplicity and transparency. According to Hodrick and Prescott (1997) a given time-series for real GDP (Y r

t ) is a sum of long run (Y *t ) and cyclical (Y c

t ) components8:

(1) Y rt =Y *

t +Y ct t=1...T

and the smoothness of the long run growth path of real GDP (potential GDP) (Y *t ) is the sum of

the squares of its second difference:

(2)

The estimate of the potential GDP and thereby also GDP gap is determined by the value of thefilter’s smoothing parameter (λ), the selection of which is subjective. The greater the value of thesmoothing parameter the bigger is the extracted cyclical component (GDP gap) — as λ reachesinfinity (λ⇒∞), long term component approaches a linear trend. Estimated trend follows thedynamics of time series when λ is zero.

Hodrick and Prescott suggest that the value of the λ should be 1600 when using quarterly data.But it might be better to take the value of the λ smaller in Estonian case, because there have beenmany structural shifts in Estonian economy and thereby the structural changes are taken intoaccount while generating the trend. This is why two different values of the λ have been used inthis paper: 400 and 1600. There is no approved way to check the correctness of the results; theonly possibility would be to see how, while increasing and reducing the value of the λ, theestimate of the GDP gap changes.

Changing the smoothing parameter from 400 to 1600 does not influence the estimates of the GDPgap substantially; the difference between the GDP gap, depending on the parameter used, hasbeen varied between 0.003 and 0.3 percentage point in 1996 to 2001. The adequacy of theparameters (the extent to which the estimate of the gap is accurate) can be determined when we

97


8 See appendix 1 for the list of acronyms.

have the gap values calculated also on the basis of production functions.

3.2. Measuring the GDP gap using production functions.Smoothing time series is a simple and often used method in estimating the potential output level.Nevertheless, from the point of view of macro-economic analysis it is too mechanical – it doesnot take into account the structural peculiarities of the economy and the limits imposed byproduction and other endogenous factors. Keeping the above-mentioned in mind, from the pointof view of economic theory the most favourable way of estimating the non-inflationary level ofoutput is through an analysis of the production function. What follows in this chapter drawsdirectly on the methodology of OECD (see for example Giorno et al., 1995).

The method used here is based on estimation of simple two-factor Cobb-Douglas productionfunction for business sector using the average functional distribution of factor incomes. Theobtained error term is smoothed and, as a result, provides an estimate of trend total factorproductivity. The potential output of the private sector is then calibrated, using the samefunctional structure of the production function together with obtained estimate of total factorproductivity, actual capital stock and previously calibrated estimate of full employment. The latterconsists of an employment adjustment according to the gap between actual unemployment and theestimated NAWRU (Non-Accelerating Wage Rate of Unemployment). This is the main difficultywith production function analysis, because estimating the level of full employment is extremelydifficult in every state, while in the case of Estonia the problems are intensified due to structuralshifts and shortness of the time series.

In detail the method mentioned above follows next steps. The estimated business sectorproduction function takes the following form:

(3)

where Yp,r

t is the business sector output in constant prices, Lp

t is the actual employment level of thebusiness sector, Kt is actual capital stock of the business sector, Gt is total factor productivity,and α is the labour share (the functional distribution of factor incomes). The natural logarithms ofrespective variables are given using small letters.

(3’)

Given the value of α, which according to the period’s average labour share is 0.6 – the gt seriesis calculated and smoothed using the Hodrick-Prescott filter. The obtained time series has beenused as an estimate of total factor productivity (gt

*). In estimating the production function it isextremely important to constructing the time series of capital stock that has been done using theperpetual inventory method. The estimate of the end-point of the capital stock draws oncomparisons of international research findings in developed and developing countries. Byaccumulating data on fixed capital formation from the Summers-Heston database, it is possible toshow that the capital-output ratio in developed countries is more than twice the respective ratio indeveloping countries (Mankiw, 1995). Taking into account that the capital-output ratio in

98


industrial countries can be a maximum of over 3, in Estonia’s case the indicator can be amaximum of 1.5, which has been used in this article.9 The capital stock of the business sector isthen constructed in retrospect according to the above-mentioned perpetual inventory method.Leaving aside the problems originating from the identification of constant amortisation rate, netinvestments derived from the national accounts have been used in capital accumulation.

The next step is to find the estimate for the business sector of full or potential employment (Ltp*

).The following formula is used for this purpose:

(4)

where LFSt* is the smoothed workforce (defined as the product of working-age population and

the trend of the participation rate), NAWRUt stands for the non-accelerating wage rate ofunemployment, and Lt

g

is employment in the public sector.

The estimate for NAWRUt is derived from an equation used by the OECD that assumes thechange in wage inflation to be proportional to the gap between actual unemployment andNAWRUt :

(5) a > 0

where ∆ is the first difference operator, Wtb

and Ut are wage and unemployment levelsrespectively. Assuming that NAWRUt is constant between two discretionary consecutive timeperiods, the estimate of a is given by:

(6)

that allows to express NAWRU with the following formula:

(7)

The obtained time series has been smoothed and is then used in the calculations10 of the potentialemployment of business sector. Inserting the estimates of log full employment (lt

p*

) and totalfactor productivity (gt

*

) to the initial production function and assuming that the capital stock is at

99


9 During the analysis other K/Y ratios were also experimented with, which, however, didn’t gave so reliable results.

10 As Giorno et al., (1995) mentions, the obtained short-term estimate for NAWRU follows the actual unemploymentdynamics and can differ from the long-term NAWRU that would have been calculated based on constant unemploymentrate (see for example Kearney et al., 2002).

its potential level, we can find the potential output of the business sector (ltp*

):

(8)

The potential output of the economy can be found by adding the government sector’s output tothe business sector’s potential output11. In the figure below two GDP gaps, estimated with theproduction function approach, are compared to those obtained with HP smoothing. The firstestimate treats government and the agricultural sector as exogenous (or operating at their potentiallevels, see GDP gap (PF 1) on Figure 2). The second estimate only takes the government sectoras exogenous (see GDP gap (PF 2) on Figure 2). The two GDP gaps estimated using theproduction function method are relatively similar, their dynamics and magnitudes of changes arelogical when confronted with actual data.

Figure 2. Estimates of GDP gap based on HP filter and production function.

The variation in the GDP gap estimates is noticeable. In comparing the gaps obtained with the HPfilter with those based on production function approach we can see that the estimates differ on anannual basis of about 1 percentage point - this can be considered a reasonably small variation12,caused by differences in the method. Since we lack criteria for favouring one estimate aboveanother, all derived GDP gaps are used in calculations of structural balance. This allows us todetermine the interval for the actual structural budget balance.

4. The cyclical sensitivity of the budget componentsAbout 90% of the Estonian government’s gross revenue comes from taxes and thus the remaining10% is non-tax revenue. The government sector budget revenue (Rt) comes from personalincome tax (Tt

P), corporate income tax (TtC), social tax (Tt

S), excises (TtE), value added tax

(TtV) and non-tax revenue (NTRt):

100


11 In other words, the presumption was that the government operates at its potential level; in some studies the sameassumption also goes for the agricultural sector.

12 As mentioned in section two, estimates of the cyclical components of budget in EMU member states differ dependingon the method used by over 2% of GDP.

(9) Rt=TtP+Tt

C+TtV+Tt

S+TtE+NTRt

Public sector expenditures (Et) are as follows: purchased goods and services (CtG) government

sector transfers to households (TRtG), capital expenditures (It

G) and interest payments (itG):

(10) Et=CtG+TRt

G+ItG+it

G

Budget balance (Bt) equals revenues minus expenditures of the same period:

(11) Bt=Rt-Et

Similarly to GDP time series also government sector revenue and expenditure consists of longterm and cyclical components. The actual budget balance is the sum of the long run balance, i.e.structural balance (Bt

S) and cyclical changes in it (BtC). Structural and cyclical components can

be separated the same way in revenue ( RtSand Rt

C accordingly) and expenditure items (EtS and

EtC ):

(12) Bt=BtS+Bt

C=( RtS-Et

S ) +(RtC-Et

C)

Structural balance is represented as follows (Hagemann, 1999):

(13)

where Btc,j is the cyclical part of the budget’s j-s component, dependent on macro indicator gap

(vtc,j) and the sensitivity of the budget’s j-s component to the influencing macro indicator (εBj,vj):

(14)

The following graph is to illustrate how business cycle influences government budget revenue andexpenditure and thereby overall budgetary balance. The expression Bt

j /Y*t p

Yt denotes the share

of budget’s j-s component (in nominal terms) in nominal GDP’s long run growth trend, which isdefined as a potential (real) GDP (Y*

t ) multiplied by GDP deflator ( pYt ). The term Y*

t pYt is used

instead of nominal GDP because the latter includes cyclical changes and thereby is notappropriate when assessing budget’s cyclical component. A-j shows the share of budget’s j-scomponent in nominal GDP’s long run growth path when output gap equals zero (Yr/Y* =1)

101


Figure 3. The sensitivity of budget items’ share in nominal GDP long run growth path to theoutput gap.

The curve of budget revenue (Rt) is upward sloping showing that growth in GDP gap causes anincrease in revenue GDP ratio. This is mainly because of a change in the effective tax rate. Theslope angle of the curve Rt indicates the aggregate sensitivity of revenue components. Theaggregate sensitivity equals a/b and satisfies the condition:

(15)

where εRj, vj is the sensitivity of revenues’ j-s component to the GDP gap. In case of expenditures(Et), the situation is vice versa. For expenditure components condition (16) should hold, meaningthat the share of expenditures in nominal GDP’s long run growth path falls when GDP gapincreases.

(16)

where εRj, vj denotes the sensitivity of expenditures’ j-s item to the GDP gap. The effect comesfrom cyclical changes in unemployment, which increases in recession phase and decreases whenthere is a boom in economy. Government expenditure on unemployment compensation grows anddiminishes respectively, in counter-cyclical manner.

It can be shown that the structural budget is balanced if the equity (17) is valid (C on Figure 3),otherwise not:

(17)

Specifically we are interested in the spread between Rt and Et when — budget’s cyclicalcomponent, depending on slope angles of the curves. This is why the equations of the curves mustbe defined. Equation (18) describes budget’s j-s component’s curve in general form, i.e. both forexpenditure and revenue items. Again, εRj, vj should satisfy conditions (15) and (16).

(18)

102


Multiplying equation (18) by Y*t p

Yt , we get:

(19)

As real GDP multiplied by GDP deflator equals nominal GDP (Y*t p

Yt=Yt), we can write:

(20)

The first part in equation (20) shows the long run growth path of budget’s j-s item, the second onecyclical component, affected by the gap between nominal GDP and its long run growth path.

4.1. Sensitivity of budget revenuesTaxes can be divided into two groups according to their natural macroeconomic base: the first andlargest group has a nominal base (VAT, income tax, etc.), while the second category includestaxes with a real base (excises) – these are taxes that are imposed on a physical quantity of goods.Taking into account the unity of estimates and leaving aside possible price effects, this paperfollows the standard approach in using a nominal base on all taxes.13

4.1.1. Personal income taxThe revenue from personal income tax rose from 4.5 billion kroons in 1996 to 7 billion in 2001.During this period, personal income tax made up about 23% of the government sector’s taxationrevenue. Incoming tax payments to the government sector can be shown as follows:

(21) TtP=trP(Wt

B-TFtP)Lt-ϕ

where Lt stands for employment, WtB is the average gross wage during the period, TFt

P is thetax-free income, trP is the tax rate and ϕ is other permissible deductions.

In order to find credible sensitivity estimates, the true reaction to changes in economic activity,we have to clear the time-series data of the influence of changing taxation policies, i.e. theinfluence of discretionary fiscal policy should be eliminated. This means that we have to find anestimate of what the tax inflow would have been like if the changes had not been put into practice(the structural element would have grown along its potential long term trajectory). Taking intoaccount the Estonian government sector’s gross revenue, the most influential discretionary policyhas been the raising of tax-free income for personal income tax.14

Raising the tax-free minimum and the subsequent change to received income tax can berepresented as follows (presuming that the number of employed and the tax rate are constant and

103


13 See for example Bouthevillain, 2001.

14 From 1996 till 1999 the tax-free income was 500, in 2000 it was 800 and since 2001 it has been 1000 kroons a month.

the division of wealth in the given period does not change):

(22) ∆TtP=-Lt∗∆TFt

P*trPt

The influence of raising tax-free income in equation (22) was limited by fixing it to the year 2000.The resulting figure for the period in question equals 800 kroons. So if the taxation policy hadbeen as it was in 2000, then from 1996 till 1999 inflow would have been 12% lower (15% at thebeginning of the period, 8% at the end). In 2001, on average, the tax inflow would have been 5%higher.

In Estonia proportional personal income tax has been implemented. In most cases the lack ofprogressive taxation means that this tax does not work as an effective automatic fiscal stabiliser.But here we have to take into account the influence of personal tax-free income on the applicablesize of the revenue.

In determining the sensitivity of adjusted personal income tax we used regression analysis.Ordinary least squares method is employed to estimate A

-P

and in equation (23):

(23)

We haveused the Hodrick-Prescott filter based potential GDP time series, with the smoothing parameter400 in the estimation process. There was no noticeable difference when comparing these resultswith the ones got with the parameter 1600. This is due to the functional form, clearly seen inequation (18), where the long run growth of nominal GDP depends on potential GDP and therebythe estimate of sensitivity is not affected by the method used for potential GDP calculation (seeAppendix 3). The sensitivity of personal income tax is 0.078, implying that an increase in GDPgap by one percentage point causes a raise in personal income tax revenues by 0.078 per cent ofGDP.

4.1.2. Social taxThe largest tax revenues for the government come from social tax. On average this amounts toabout 34-35% of all income. Income from social taxes grew from 7 billion to 12 billion kroonsbetween 1996 and 2001. Social taxes have remained stable. Changes to social policies have beenrelatively small and need not be separated from the data. Equation (24) is used to estimate thecyclical sensitivity:

(24)

It appears that the sensitivity of social tax is 0.116 – when GDP gap increases by one percentagepoint, social tax revenue raise by 0.116 per cent of GDP.

104


4.1.3. ExcisesIncome from excises grew from 1.7 billion in 1996 to 3.5 billion kroons in 2001. Excises accountfor about 10% of government sector tax income. The regulations concerning excises havechanged a great deal during this period. Describing the changes is problematic, mainly because ofthe number of changes and because they usually have a seasonal structure. In further analysis wehave to assume that changing these taxes has had a little or no affect on their inflow to the generalgovernment budget. Again the equation based on formula (20) is applied:

(25)

The cyclical sensitivity is 0.042 in case of excises.

4.1.4. Value-added taxWhile in 1996, the inflow to the government budget from VAT was 5 billion kroons, in 2001 ithad risen to 8.5 billion kroons, making up 26% of budget income from taxes. During this periodthere were changes made to the list of goods that attract VAT, there were also changes to tax rates,but the influence on incoming taxes was relatively small and there is no need to eliminate thefactors resulting from discretionary policies to correct the sensitivity estimation (see Equation 26).

(26)

The value of has been 0.096 during 1996-2001.

4.1.5. Corporate income taxBudgetary revenue from corporate income tax has to be divided into two periods. The linebetween them is January 1, 2000. Starting from that day no tax was required to be paid onaccumulated profits, only dividends were a subject of taxation. Since then, budget income fromcorporate income tax has decreased to only 2% of government sector tax revenues. In 1998 taxrevenues coming from corporate income tax were at their highest, about 2 billion kroons. By theyear 2001, this income had decreased to 0.7 billion kroons. We applied equation (27):

(27)

but no cyclical pattern in corporate income tax revenues could be detected. The conclusion hereis that the cyclical sensitivity of corporate income tax is zero.

105


4.1.6. Non-tax revenuesThe cyclical sensitivity of non-tax revenue is estimated the same way as it was done on theoccasion of taxes (see Equation 28).

(28)

But again, as already seen in case of corporate income tax, there was no evidence on clear cyclicalimpact on non-tax revenue.

4.2. Government sector expenditureHalf of all government sector expenditures are on goods and services. Out of this 40 per cent arewages and 60 per cent on other goods and services. Government expenditures on goods andservices are not bounded to economic growth directly and this is why they do not changeautomatically as a reaction to a shift in economic activity but the change is caused by politicaldecisions, i.e. discretionary fiscal policy.

Transfers and subsidies make up about 40 percent of total expenditures, out of which transfers tohouseholds constitute a little more than 70 per cent. About 70% of transfers are pensions, 13% arechild support and 7% are illness compensations. Similarly, as these expenditure items were notbounded to economic growth during 1996-2001, they were not dependent on business cycleeither. The only components of expenditure, which react to business cycles, are unemploymentcompensation and living allowances, but added together these only make up about 5% of alltransfers. Even if unemployment compensation was highly dependent on the GDP gap, it’s verysmall share in transfers indicates that the impact on total expenditures must be even less thanmodest. This is the reason why it can be concluded that cyclical fluctuations should only influencehousehold transfers marginally. Since in the case of government expenditure and transfers it canbe assumed that sensitivity equals zero; this has not been measured separately.15

5. Calculating the structural balance and the size of automatic fiscalstabilisersThe basis for calculating the structural balance was given previously by equations (13) and (14).Since the only macro indicator in use is the GDP gap, the cyclical components of the revenueitems (taxes and non-tax revenue) can be expressed as follows:

(29)

where Rtc,j is the cyclical component of j-s revenue item, Rt

c is it’s actual value in period t, isrevenue’s j-s item’s sensitivity to the GDP gap. Cyclical component of all revenues isas follows:

106


15 According to European Commission opinion the government sector expenditure in Germany, Greece, Spain, Italy andin Austria is non-elastic towards cyclic fluctuations too (European Economy, 2000).

(30)

Because, as previously shown, expenditures do not depend on GDP fluctuations, the cyclicalsensitivity of budget forms on the basis of cyclical sensitivity of revenues:

(31) BtC=Rt

C+EtC=Rt

C EtC=0; t=1...T

The sum of revenues sensitivity ( ΣjεRj, vj ) is about 0.35, showing that there is a cyclical change

in budget balance amounting to 0.35 per cent of GDP when GDP gap increases by one percentagepoint. Using this indicator in regard to the business cycle, we are at the same level as EMUcountries with a less sensitive fiscal position (the EMU average is 0.5). It appears that thedifference between the structural and actual budget balance hasn’t exceeded 1.35% of GDP onany sample year. Even in 1999, when according to PF 1 GDP gap reached -3.9%, the cyclicalcomponent of the budget was found to be only –1.31% of GDP (see Table 2).

It can be said that because of the low budget sensitivity in relation to GDP the threat of exceedingthe SGP 3% deficit limit is low, this is true even if the structural deficit reaches 1% of GDP (thenthe GDP gap should exceed 5% and the criteria would be in danger).

Table 2 The cyclical components of general government budget (% of GDP)

The variation of the cyclical component, depending on the research method, can be regarded asrelatively small. The difference between minimum and maximum values is, on average, 0.35 %of GDP. This is also why the estimates of the structural budget balance are rather similar. Thedifference between the actual and structural budget balance is formed according to whether theGDP gap is positive or negative and whether discretionary steps have caused a structural budgetdeficit or surplus. In a growth phase the functioning of automatic fiscal stabilisers improve theactual fiscal position when compared to the structural balance. When the structural budget is indeficit, then the current positive GDP gap reduces the actual budget deficit (see Figure 4, 1998).In case of structural budget surplus the actual budget surplus turns out even greater (1997).Conversely, in an economic downturn the actual budget position worsens when compared to thestructural position (1996, 1999 and 2000).

107


Figure 4. GDP gap (%), the actual and structural balance in case of different smoothingparameters and production functions (% of GDP).

From Figure 4 it can be concluded that the operation of the automatic stabilisers has madeEstonia’s government sector budgetary position more volatile (resulting directly from the natureof the stabilisers). This has been intensified by counter-cyclical steps of discretionary fiscalpolicy. The standard deviation of the actual budget balance in 1996-2001 was 2.4% of GDP, thestandard deviation of the structural balance was at the same time 1.7% of GDP. In 1997, when theactual surplus was 2.7% of GDP, the structural surplus, depending on the method used, wasbetween 1.8 and 2.1% of GDP. In 1999, when there was a remarkable deficit and governmentsector expenditure exceeded revenue by 4.6% of GDP, the structural deficit between 3.3 and 3.6%of GDP was somewhat smaller.

Following the dynamics of the structural budget, we can clearly identify the tendencies in fiscalpolicy, i.e. whether the discretionary fiscal policy supports or hinders total demand. We cananalyse fiscal policy tendencies by comparing current policy decisions with previous periods. Forthis reason the annual difference of structural budget balance is used for identifying the effects ofdiscretionary fiscal policy. These differences in structural budget for different estimates of GDPgap are shown in Figure 5.

108


Figure 5: Differences in the structural budget balance compared to previous years (% ofGDP) and the growth rate of actual GDP.

As we can see, in 1998 and 1999, the annual difference in structural balance was negative,inferring that, in comparison to previous years, the government had directed more resources to theeconomy and carried out expansive fiscal policies. Conversely, in 1997, 2000 and 2001 the fiscalpolicy was restrictive in nature. Comparing these results with the dynamics of GDP gaps we canconclude that during the sample period Estonian fiscal policy has been mainly counter-cyclicaland directed at smoothing cyclical fluctuations in an economy.

6. ConclusionIn this paper we confirm the assumption that the Estonian tax framework and budget expendituredetermine the modest size of Estonia’s automatic fiscal stabilisers. Based on data from 1996-2001, and using the two-step method, the cyclical sensitivity of the budget was found to be only0.35 (in EU countries the average is between 0.3 and 1.1). This means that when the GDPdiverges from its potential by 1%, then the budget balance will accordingly change up by 0.35%of GDP. Because this method does not take into account reactions between the budget and theeconomy, the actual cyclical sensitivity may be even smaller than the figure calculated. Thismeans that, for example, in 1999 the economic downturn (negative gap 3.9%) caused a budgetdeficit of only 1.3% of GDP.

The positive aspect of such a low cyclical sensitivity is the limited danger of exceeding the SGPbudget deficit limit of 3% of GDP. Calculations show that even in an extreme case, when the GDPgap reaches 5%, the Estonian government sector structural budget deficit could be up to 1% ofGDP without placing the given criteria at risk.

109


-4

-2

0

2

4

6

8

10

12

1997 1998 1999 2000 2001

structural balance (=400) structural balance (=1600)structural balance (PF1) structural balance (PF 2)real GDP growth

ReferencesArtis, M. J. and M. Buti (2001), “Close to Balance or In Surplus – A Policy Marker’s Guide tothe Implementation of the Stability and Growth Pact”, Journal of Common Market Studies, 38(4).

Barrel, R. and K. Dury (2001), “Will the SGP Ever Be Breached?”, The Stability and Growth Pact– The Architecture Of Fiscal Policy in EMU, Palgrave.

Barrel, R., I. Hurst, and A. Pina (2002), Fiscal Targets, Automatic Stabilisers and Their Effectson Output, National Institute of Economic and Social Research, London.

Bassanini, A., S. Scarpetta and P. Hemmings (2001), “Economic Growth: The Role of Policiesand Institutions. Panel Data Evidence from OECD Countries”, OECD, Economic DepartmentWorking Papers No. 283.

Bouthevillain, C., Cour-Thimann, P., van der Dool, G., de Cos, P.H., Langenus, G., Mohr, M.,Momigliano, S., Tujula, M. (2001), “Cyclically Adjusted Budget Balances: an AlternativeApproach.” European Central Bank, Working Paper no. 77.

Buti, M. and A. Sapir (1998), “Economic Policy in EMU”, A Study by the European CommissionServices, Clarendon Press, Oxford.

Dalsgaard, T., de Serres, A. (1999), “Estimating Prudent Budgetary Margins for 11 EU Countries:a Simulated SVAR Model Approach.” OECD Economics Department Working Papers, No. 216.

Dury, K., Pina, A. (2000) “European Fiscal Policy after EMU: Simulating the Operation of theStability Pact.” EUI Working Papers, ECO.

European Economy. Public Finances in EMU – 2000. Directorate-General for Economic andFinancial Affairs, No. 3/2000.

European Economy. Public Finances in EMU - 2001. Directorate-General for Economic andFinancial Affairs, No. 3/2001.

European Economy. Public Finances in EMU – 2002. European Commission, Directorate-General for Economic and Financial Affairs, No. 3/2002.

Fölster, S. and M. Henrekson (2000), “Growth Effects of Government Expenditure and Taxationof Rich Countries”, Stockholm School of Economics, Working Paper No. 391.

Giorno, C., Richardson, P., Roseveare, D, van den Noord, P. (1995), “Estimating PotentialOutput, Output Gaps and Structural Budget Balances.” OECD, Economics Department Workingpapers, No. 152, Paris.

Hagemann, R. (1999), “The Structural Budget Balance. The IMF Methodology.” IMF WorkingPaper.

Hemming, R., M. Kell and S. Mahfouz (2002), “The Effectiveness of Fiscal Policy in StimulatingEconomic Activity – A review of the Literature”, IMF Working Paper 02/208.

Hodrick, R. J., Prescott, E. C. (1997) “Postwar U.S. Business Cycles: An EmpiricalInvestigation.” Journal of Money, Credit and Banking, Vol. 29, No. 1, pp. 1-16.

Höppner, F. (2002), Fiscal Policy and Automatic Stabilisers: A SVAR Perspective. Institute forInternational Economics, University of Bonn.

Kearney, I., McCoy, D., Duffy, D., McMahon, M. and Smyth, D. (2002) Assessing the Stance ofIrish Fiscal Policy, ESRI.

110


Mankiw, N.G. (1995), “The Growth of Nations”, Brookings Papers of Economic Activity.

Meyermans, E. (2002), Automatic Fiscal Stabilisers in the Euro Area: Simulations with the NIMEmodel. Fourth Workshop on Public Finance, Banca d’Italia, Perugia, 21-23 March 2002.

Taylor, J. (2000) “Reassessing Discretionary Fiscal Policy”, Journal of Economic Perspectives,14(3).

Valtioneuvoston Kanslia (2000), “Economic Policy in the Framework of a Single MonetaryPolicy.” Prime Minister’s Office Publications Series.

111


AppendicesAppendix 1. Cyclically adjusted budget balance in EU countries.

(comparison of cyclical components calculated based on the methods of ECB, EC, OECD andIMF)

112


Appendix 2. List of acronyms

113


114


Appendix 3. Results of sensitivity estimation

115


Tax Policy in Latvia on the Eve of EnlargementDaina RobeΩniece and Måris Jurußs *

The Latvian tax system is very young – only in its twelfth year of implementation – however,Latvia has succeeded in creating a viable and effective tax system in such a comparatively shortperiod of time, in which the main fiscal aims (ensuring state budget revenue), are balanced withthe use of taxes as an instrument of economic policy.

Latvia has succeeded in doing so because, during the reorganization of the tax system in the mid-90’s, new tax laws were developed according to the experience of developed democratic countriesand taking into consideration advice from experts representing a number of internationalorganisations. From this point of view it could be argued that Latvia, creating its tax system fromthe start, has had certain advantages over the “old” tax systems. After all, Latvia had theopportunity to become acquainted with best practice and avoid the mistakes that other countrieshad made earlier. This, however, does not mean that there have been no mistakes in creating theLatvian tax system. The rest of this paper will analyse the different types of taxes introduced inLatvia, and the impact of the impending EU accession.

Taxes on Income

There are two taxes on income in Latvia – personal income tax and the enterprise income tax,which form a united system. This means that:

• in general, any person deriving income in Latvia is either taxable with personal incometax or enterprise income tax;

• the same income is never subjected to both the income taxes at the same time.

Latvian income tax laws have been formed according to the principle that:

• the range of taxable persons and the tax base have to be wide;• the rate has to be comparatively low.

The Latvian income tax system is comparatively neutral in terms of resources reallocation, as thelaws create no special taxation advantages to one particular economic sector or another.

The fact that quite a large number of reliefs has been granted under the enterprise income tax,furthermore, that the efficiency of the reliefs in terms of attracting investment or solving socialproblems is low, could be mentioned as a negative tendency o some extent. Therefore, in order toencourage entrepreneurship, it has been envisaged to gradually decrease the enterprise income taxrate to 15 per cent by 2004, and at the same time cancel separate former reliefs.

116


* Daina RobeΩniece, Director of Tax Policy Department and Måris Jurußs, Deputy of Director of Tax Policy Department,Ministry of Finance, Latvia.

The personal income tax rate of 25 per cent is also uniform to all payers and all types of income.It facilitates the administration of this tax. At the same time the problems characteristic toprogressive income tax rate are absent – taxing income at a higher rate as a result of inflation-initiated increase in income, tax planning for applying a lower rate to income etc.

Latvian income tax laws have been designed to eliminate traditional tax avoidance methods –there are provisions in the law regarding incomplete capitalisation, transfer pricing, governancein respect of transactions with low-tax and zero-tax states and jurisdictions. However, taking intoconsideration Latvia’s progress towards accession to the European Union, these provisions mustbe in compliance with the single market.

In terms of personal income tax, it is necessary to address within the budget capacity the issue ofincreasing the exempt minimum, as the current level is the lowest not just among the Balticcountries but also among all EU candidate countries. Therefore its functioning in accordance withthe envisaged objectives is not ensured.

The generally accepted principles of both the European Union and OECD countries in terms ofdrafting laws have been taken into consideration during the writing of Latvian income taxes laws.The requirements of European Union directives are gradually incorporated into them in the sphereof direct taxation – both in respect of taxation of dividends and the application of income taxeslaws during enterprise reorganization.

In order to settle taxation issues at an international level, to avoid double taxation and preventfiscal evasion, intense work has been carried out in concluding tax conventions. Latviacommenced work on concluding double taxation conventions together with both the other Balticcountries in 1992, and until the present moment negotiations with 47 countries have taken placefor concluding tax conventions. At the present moment conventions with 44 countries have beeninitialed, 30 conventions have been signed and 26 conventions applied.

Real Estate Tax

The real estate tax reform that commenced in 1997 is still ongoing. The purpose of this reform isto ensure the simple and fair taxation of real estate, which means introducing a uniform tax rateof 1 percent, introducing a uniform tax base – the cadastral value of real estate – and covering awide scope of payers of real estate tax - all persons that own real estate in Latvia. Tax reliefsprovisioned for in the law are minimal and can only be granted by the municipalities, the budgetsof which receive the tax income (the tax is payable into the budgets of municipalities at the fullamount).

In order to implement the reform in accordance with the above-mentioned principles, a real estateassessment authority has been created in Latvia. The analysis and recording of real estatetransactions are made on a regular basis, and the data obtained is used in valuing real estate.

The administration of the real estate tax has been fully transferred to municipalities since the2000. Taking into account the fact that for purposes of calculating the real estate tax the cadastralvalue of real estate is determined by the State Land Service, the exchange of information betweenthe State Land Service and municipalities is very important. The exchange of information isexpanding and has also been considerably improved.

The postponement of implementation of several stages of the reform is, however, a weakness of

117

Tax Policy in Latvia on the Eve of Enlargement

the reform. The postponement has been necessary, taking into account the planned reduction ofmunicipality budget revenue, due to the reduction of the tax rate, and exemption of certaincategories of real estate from the tax. This means that the tax on buildings and constructions islevied on the book or inventory value, and not on the cadastral value of the buildings andconstruction. A 1.5% tax rate is still applicable, and the tax is not levied on private houses andreal estate belonging to institutions financed from the state budget.

The legal aspects of the formation of dwelling rights have also not been completely settled in law,as well as some matters connected with the maintenance of blocks of flats. The mentioneddrawbacks could create problems in the calculation and collection of the tax in 2004, when flatswill also be subjected to real estate tax.

Value added tax

Value added tax and excise taxes in Latvia are based on EU requirements. Work on harmonizationof these taxes is ongoing, and it is planned to bring the Latvian law into conformity with EUrequirements in the remaining time period remaining before the EU accession.

In compliance with 6th Directive (77/388/EEC) Article 12 3(a) a Member State may apply one ortwo reduced rates for goods and services specified in Annex H.

For the moment, the Latvian Law on Value Added Tax provides two rates of value added tax: thegeneral rate of 18 per cent and the reduced value added tax rate of 9 per cent for certaintransactions. A VAT rate of 0 per cent is applied to export and transit deals.

A reduced VAT rate of 9 per cent was introduced as of 1st January 2003 in order to harmonizethe Latvian value added tax system with the 6th Directive (77/388/EEC) wherewith certain goodsand services which are not compliant with the above directive are excluded from exemptions. AVAT rate of 9 per cent is applied to the following supplies of goods and services:

• supplies of veterinary medicine;

• special food for infants;

• supplies of mass media;

• accommodation provided by hotels and similar establishments;

• supplies of water;

• sewerage services;

• collection, transportation and disposal services for household waste.

As of 1st January 2004, a reduced VAT rate of 9 per cent will apply to the supply of books as wellas certain approved medicaments. It should be noted that by providing single VAT procedures onbooks and medicaments produced in Latvia and those imported the distortion on competition willbe eradicated.

The increase in the cost of the above-mentioned goods and services will not have seriousconsequences on the purchasing power of the population. The VAT taxpayer has the right todeduct input VAT from the tax amounts indicated in tax invoices received from other taxablepersons regarding goods and services for ensuring their own taxable transactions.

118


It should also be mentioned that the VAT refund system for all non-resident legal persons,according to the requirements of the 8th and 13th VAT Directives, was introduced on 1st January2003. By introducing this system Latvia has successfully fulfilled commitments undertakenwithin the pre-accession process to the EU.

Problems connected with the administration of VAT are a continuing weakness. Cases of VATfraud are mainly related to the refund of VAT overpayments. In light of this, the Finance Ministryand the State Revenue Service prepared amendments to the Law On Value Added Tax, which areenvisaged as measures against VAT fraud. Amendments were effective from July 16, 2003.

The so-called carrousel schemes, well-established in EU Member States, can also be seen inLatvia. The schemes involve not just individuals residing in Latvia, but also those residing inneighbouring countries. Thus the quick and effective exchange of information is very importantin counteracting the use of such schemes. Work on improving information systems and thecreation of the necessary databases to EU requirements has begun at the State Revenue Service,in order to be ready to become a part of a united system of exchange of information within theEU.

Continuing work in the interpretation and implementation of the European VAT acquis, Latviahas also prepared a draft for amendments to the Law On Value Added Tax providing for a VATapplication in European Single Market (intra-Community trade) that would be in line with theprovisions of the 6th Directive (77/388/EEC).

To meet the needs of the single market, the objectives of the new VAT system are as follows:

- to be simple and modern;

- to ensure equal treatment for all transactions and uniform application of provisions of 6th

Directive;

- to guarantee effective taxation and controls to maintain the level of VAT revenue;

- to move to destination-based taxation (i.e. charging the tax in the country of destination);

- to make it possible to abolish controls at tax frontiers.

The European Single Market changed the way in which VAT is accounted for on goods movingbetween Member States of the EU. Thanks to the existing EU system, the free movement of goodswithin the Community could no longer be treated as import or export. VAT on import and exportwill be abolished and replaced by the new concept of VAT on intra-Community acquisitions ofgoods and on intra-Community supplies of goods. Because of that, before Latvia’s accession tothe EU the most important priority for Latvia is the introduction of the VIES system. This systemenables the authorities in each EU Member State to ensure that intra-Community transactions areproperly recorded and accounted for.

Excise Taxes

One of the taxes that has undergone the biggest changes in Latvia is the excise tax, which washarmonized in respect of oil products, alcoholic beverages and tobacco articles and according toEU requirements by introducing a system of excise goods warehouses, the suspended payment oftax, implementing tax guarantees and other requirements set by the directives.

Excise tax reform began in 1997 when work on drafting the new excise tax law began. Althoughall the necessary legislative documents had entered into force by 2000, the excise tax reform

119

Tax Policy in Latvia on the Eve of Enlargement

cannot yet be considered complete, as the harmonization of tax rates is ongoing, and will becompleted only in 2010 (in respect of cigarettes).

The entire excise tax system is being improved. It is becoming more consistent with the principlesaccepted in the EU, and it will allow Latvia to integrate in the single market and administer thetax according to the same principles present in other EU member countries, at the same time asensuring definite tax income to the budget. The main EU requirements already introduced inLatvia are a system of excise goods warehouses, excise tax suspension arrangements, excise taxguarantee system. and other requirements.

However, at the same time some problems in the harmonization of excise taxation have to benoted. First, excise tax rates, as in the majority of other candidate countries, are lower than therates set in the EU. This means that Latvia has to significantly increase the excise tax burden inthe near future. As a result, the prices of these goods will increase, expenses will increase forenterprises as will the desire to avoid tax. An increase in illegally traded goods is anticipated.

It should also be mentioned that the situation concerning the introduction of EU rates of taxationis different in each group of excisable goods. In general the rates have already been introduced onalcoholic drinks, although problems are expected in this sphere. As concerns excise tax on oilproducts, the rates will substantially increase on diesel fuel. This is largely down to the growth invalue of the euro over recent years.

The biggest problems connected with excise taxes for Latvia, as in other candidate countries,are those regarding the implementation of EU requirements on excise tax on cigarettes, sincethe requirements apply not only to the rate of tax, but also to the structure of the tax, procedureof price determination, and even trade with cigarettes. Latvia is the last of the three Baltic Statesto have introduced the combined rate of excise tax on cigarettes. The combined rate came intoforce on July 1, 2003. A lot of effort was taken in explaining and clarifying the new system tothe tax administration and local businesses, in order to ensure that the tax is properly calculated,the maximum retail price is correctly determined and reflected, and the regulations on tradewith cigarettes are observed by both the tax administration and businesses.

120


Estonian Excise Policy on the Threshold of the European Union (based on cigarette Policy)

Evelin Ahermaa*

Abstract The purpose of this paper is to evaluate Estonian excise policy concerning cigarettes in the following way: to

compare the price level and tax rates between Estonia and other countries (the EU, the EU candidate states, EFTA) and

give an overview of excise policy in terms of the European Union (EU). The influence of Estonian excise policy on the

price level, consumption, legal / illegal market and the state budget are presented. Estonia as an applicant country of the

EU has to harmonize its tax policy (including excise duties). The harmonization process of excise duties will be finished

upon accession to the EU (01.05.2004), except with tobacco products. The latter group has a transition period until 2010;

this is due to the wide presence of illegal goods in Estonia. Therefore, it is important to evaluate changes in excise policy

concerning the requirements of the EU. The following paper analyses different possibilities in reaching these requirements

and their influence on the Estonian economy. Changes in tobacco tax rates mostly influence the cigarette market, which

are the most consumed tobacco product in Estonia, therefore the paper is especially concentrated on cigarettes.

Keywords: excise duty, illegal market, taxation

JEL Classification: E62, K42

1. IntroductionExcise duty is the indirect tax on specific goods such as tobacco products. These taxes are payablefor the consumption or use of final products, whether they are manufactured within a country orimported. Excise duties are tightly connected with the state budget, price policy, environmentalaspects, health care and other fields. Different tax rates of excise duty reflect a dilemma ofgovernments: are excises just a source of revenue or do they play a role in influencing consumers’behaviour? This question has driven many discussions. As Chaloupka and Warner (1999) show,excise duties are a good source of revenue and therefore popular; this is because of the relativelyinelastic demand of excise goods (such as tobacco products). But excise duties should also, tosome extent, compensate the negative effects of consuming such goods.

Excise policy can be closely connected with tax fraud. Due to the big difference between the salesand cost price of excise goods, illegal trade can spread widely. Excise duty increases the price,but does not change the quality of the goods. Unstable excise policy and high excise rates fosterillegal trade.

Smuggling is widespread in Estonia and causes loss of revenue in the state budget. Due to the factthat there is an active illegal trade in Estonia, the loss in tax revenue amounts to hundreds ofmillions of Estonian kroons (EEK). This is a very important fact, because excise duties are treated

121

Estonian Excise Policy on the Threshold of the European Union

∗ Estonian Institute of Economic Research, 6 Rävala Street, 19080 Tallinn, Estonia. e-mail: [email protected]

as an essential source of revenue in Estonia.

The following paper concentrates on price and tax comparisons between Estonia and otherEuropean countries, and excise tax harmonization in Estonia up to 2010. The hypothesis is thatEstonia should reach a higher level of the overall excise duty (set as a % of the retail selling price)by 2010 than is actually required by the EU’s minimum rate.

2. Comparison of Excise Policy between Estonia and European CountriesEstonia has been an EU candidate country since 1998. Estonian tax policy has already beenlargely harmonized, including excise policy. The tax rates of tobacco products have beenincreased and restructured. For example, the excise duty of cigarettes consists of two parts:specific and proportional (ad valorem) rates (see table 1). Estonia also uses a special fiscalwarehouse system. The fiscal warehouse is a place where goods subject to excise duty areproduced, processed, held, received or dispatched. It guarantees supervision of the manufacturingof tobacco products and excise payments.

Table 1. Dynamics of The Excise Duty on Cigarettes in Estonia 2001-2004

The taxation requirements of the EU as concern cigarettes are the following (see CouncilDirective 92/79/EEC):

the excise duty consists of the specific duty per unit of the product and proportional (advalorem) duty calculated on the basis of the maximum retail selling price;

the overall minimum excise duty (specific duty plus ad valorem duty excluding VAT)shall be set at 57% of the retail selling price (inclusive of all taxes) and shall not be lessthan 60 EUR per 1000 cigarettes for cigarettes of the price category most in demand (64EUR from 1 July 2006);

if the overall minimum excise duty is levied at a minimum 95 EUR per 1000 cigarettes forcigarettes of the price category most in demand, the previously mentioned 57% minimumrequirement may be not applied;

the specific rate shall be between 5-55% of total sum of the overall excise duty and VAT;

all cigarettes have the maximum retail selling price, and the excise calculation based on it(if the retail trader sells cigarettes at the lower price, there are no reduction of the excisepayment).

EU Member States can establish their own excise policy according to the rules mentioned above

122


Indicator Unit 01.07.2001 01.07.2002 01.07.2003 01.07.2004

Specific excise per 1000 cigarettes EEK 150 175 210 240

Proportional excise on the basis of the maximum retail selling price % 21 23 24 25

1 EUR= 15,64664 EEK

(see fig. 1). Some countries have chosen a larger specific excise (for example, Ireland, Denmark,Portugal), while others have chosen a proportional one (for example, Greece, Italy, Spain,France). The Estonian choice has been a more balanced system, i.e. almost equal specific andproportional rates.

Figure 1. Taxation of Cigarettes in Estonia and the EU in 2002

Source: Excise Duty Tables (2002)

Excise revenue is the highest in the United Kingdom (UK), followed by Ireland and Nordiccountries. This is due to differences in the price levels of cigarettes (see fig. 2). There is atendency within the EU to tax cigarettes higher in the northern countries and vice versa in thesouthern countries. In comparison with EU Member States Estonia is at a very low level, andcigarettes are twice as cheap as in the cheapest EU country.

123


0 20 40 60 80 100

Estonia

Luxembourg

Sw eden

Spain

Germany

Greece

Netherlands

Belgium

Austria

Italy

France

Finland

Portugal

Ireland

UK

Denmark

% of the retail selling price

Specific excise duty Ad valorem excise duty VAT

Figure 2. The Price Category of Cigarettes Most in Demand in Estonia and the EU in 2002

Source: Excise Duty Tables (2002)

It is also important to use price level indices in cross-country comparisons. This index iscalculated as the ratio between Purchasing Power Parities (PPP) and exchange rates for eachcountry, in relation to the average of other countries (to the EU average in this paper). In figure3, if the price level index is higher than 100, the country concerned is relatively expensivecompared to the EU average and vice versa.

The following comparison includes EU Member States, the European Free Trade Associationcountries (EFTA, excl. Liechtenstein) and EU candidate countries. The Estonian price level has amedium place among EU applicant countries, but is higher than the two other Baltic States ofLatvia and Lithuania. Estonia can be divided into one group with Slovak and Hungarian pricelevels for tobacco products. At the same time, in Estonia those products are available at about60% below the average EU price level.

124


0,95

1,85

1,95

2,07

2,08

2,35

2,69

2,82

2,83

3,16

3,60

3,85

4,00

4,03

5,00

7,01

0,00 1,00 2,00 3,00 4,00 5,00 6,00 7,00 8,00

Estonia

Spain

Portugal

Italy

Luxembourg

Greece

Austria

Belgium

Netherlands

Germany

France

Sw eden

Finland

Denmark

Ireland

UK

EUR per pack

Figure 3. Cross-Country Comparison of GDP and the Price Level for Tobacco Products in2001

Source: Purchasing Power Parities … (2002); Eating, Drinking, Smoking …(2002)

As figure 3 shows, northern countries still have higher price levels for tobacco products thanothers. This is mostly due to efforts to prevent smoking through high excise duties. Unfortunately,higher excise duties and, as a result, higher price levels have not proved themselves. Consumptionof tobacco products is a little bit lower than in other countries but northern countries also sufferfrom a high level of illegal trade. The latter have been the negative side effect of such excisepolicy. For example, it has been estimated that every fifth cigarette in the UK is illegal, while thecustoms of Sweden, Denmark and Finland each confiscated ca 30-50 millions cigarettes per yearin 2001 (Nordisk …, 2002; The Black …, 2000). It can be concluded that a very strong excisepolicy has not gained the stated goals. A strong excise policy alone cannot be effective; it needsan improved system of controlling and punishing, which in turn requires increased resources andinvestment.

Although Estonian excise rates and price levels for cigarettes are not comparable with the above-mentioned countries, the illegal market has spread. With the purpose of increasing income intothe state budget, Estonian excise policy at the end of the nineties was focused on increasing thetax rates. Unfortunately, an absence of effective control tools and a relatively low standard ofliving, as well as widespread illegal trade has caused a continued loss of tax revenue (see table 2).

125


0

50

100

150

200

250

Luxe

mbo

urg

Irel

and

Den

mar

kN

ethe

rland

sA

ustr

iaB

elgi

umFi

nlan

dG

erm

any

Italy

Fran

ce UK

Swed

enSp

ain

Port

ugal

Gre

ece

Nor

way

Switz

erla

ndIc

elan

dC

ypru

sSl

oven

iaC

zech

Rep

.H

unga

rySl

ovak

iaEs

toni

aPo

land

Lith

uani

aLa

tvia

Bul

garia

Rom

ania

Tur

key

Mal

ta

GDP volume index price level index

0

50

100

150

200

250

GDP volume index per capita in PPS (EU-15=100)

Index of the price level for tobacco products in PPP (EU-15=100)

EU-15 EFTA EU candidate countries

Table 2. Indicators of Estonian Illegal Cigarette Market during 1999-2002

Source: Ahermaa (2003)

Table 2 indicates a decreasing share of illegal trade and market. This positive tendency can beexplained by improved control activities and an improved economic situation. After all, the lossin tax revenue amounts to hundreds of millions EEK and every fourth or fifth cigarette smoked inEstonia is illegal. Therefore the impact of EU accession on Estonian excise policy and thecigarette market needs to be examined.

3. The Future Tax Perspectives of Estonian Cigarette MarketAs autlined above, Estonian excise policy on tobacco products has already been largelyharmonized. Final harmonization calls for continuously increasing tax rates.

Estonia has a transition period until 2010 for harmonization of excise duty rates for cigarettes. Thetransition period was applied for because of the specific situation in Estonian tobaccoconsumption and the market. The transition period allows excise duty to be increased step by step.Otherwise, the sharp increase in the tax rate could not be effectively administrated. This is due tothe relatively low standard of living (Estonian smokers are very price sensitive) and the largeillegal cigarette market in Estonia. The latter is generated by price differences between Estoniaand its neighbouring countries (especially Russia). The sharp price increase (in comparison withthe surrounding economic space) will not be administrated, because of the tax revenue loss and itwill not correlate with growth in the standard of living. The transition period offers more stability.Still, the question of which is the most effective way of dividing the tax increase over thetransition period will arise.

Due to the fact that the Estonian income level is not comparable with any EU Member State, theirpractise concerning excise policy cannot be entirely applied. Moreover, several Europeancountries have trouble with the illegal tobacco trade.

The following analysis will take into account the position of the Ministry of Finance:

- reaching the overall excise duty (specific plus proportional duty excluding VAT) 64 EURper 1000 cigarettes by 2010;

- increasing the tax rates once every year (the growth of the excise duty should be

126


approximately 1-2 EEK);

- maintaining the balance of the specific and ad valorem excise duties (equal shares).

As previously described there is one more EU requirement – the overall minimum exciseduty has to be set at 57% of the retail selling price for cigarettes of the price category mostin demand. Therefore, there are several possible scenarios for Estonia. Table 3 representstwo of them, where the forecast of the price categories bases on the following exponentfunctions (p marks the retail selling price and t marks time):

- in the case of the 57% and 64 EUR requirements p = 6,4036 e 0,8512 t ;

- in the case of the 60% and 64 EUR requirements p = 6,7407 e 0,7999 t .

Table 3. The Forecast for the Price Category of Cigarettes during 2003-2009

Source: Tubakaaktsiisi …(2002)

If the aim is to reach 57% and 64 EUR at the same time, it will put pressure on the increase of theretail selling price. Past experience indicates that it will be more likely that the lower price levelwill be attained by 2010 but that there will be a higher overall excise duty share from the retailselling price. The scenario of 60% and 64 EUR can be achieved and therefore choosing thecorresponding excise duties from table 3 is advisable. A higher share of overall excise duty(>60%) does not reduce the price level because it would not provide a sufficient share tomanufacturers, importers and traders.

According to the development of the market situation and the price level during the transitionperiod, final corrections of the tax rates should be made in 2009. The forecast can be affected byseveral factors e.g. joining the EU, euro-zone and the single market. Accelerated economicconvergence is likely to take place as well as an increase in the income level after EU accession

127


(see Tubakaaktsiisi …, 2003). Nevertheless, the planned excise policy will put relatively strongpressure on the illegal cigarette market, which will be partly reduced by the expected growth ofthe income level. It will be important to guarantee administration of the growth of the tax rate, i.e.the illegal market will exist but not increase. According to the above-mentioned data, there willnot be any economic assumptions for reduction of illegal trade; the share of the illegal market canbe foreseen at approximately the same level as in 2002-2003 if border control issues are secured(Estonia as the EU Member State).

In conclusion, the transition period will allow the acceptance of moderate tax increases, allowtheir administration and also their acceptance by the consumers.

4. Concluding RemarksExcise duties are one priority of the tax harmonization process with the EU requirements inEstonia. The excise duty levied on cigarettes is one of the most problematic issues because thereis a big difference in the tax rate between Estonia and the EU minimum rate. In comparison withEuropean countries it can be said that the Estonian price level for cigarettes is on the medium levelamong the EU applicant countries but it is ca 60% below the average EU price level.

With the purpose of enlarging the state budget, Estonian excise policy at the end of the ninetieswas to increase tax rates. Supervision institutions were not able to administrate sharp increases ofthe tax rates and the illegal trade of cigarettes spread. It is problematic to increase the tax rate tothe EU level in a situation where there is an active illegal trade, high price sensitiveness ofsmokers and the surrounding economic situation (lower price levels in the neighbouring non-EUcountries). Therefore, Estonia has a transition period until 2010 for applying the requiredminimum excise rate on cigarettes.

The aims for the transition period are to increase tax rates once every year, maintain almost equalspecific and proportional excise duties and reach the overall excise duty 64 EUR per 1000cigarettes by 2010. Overall excise duty should be set at 60% of the retail selling price forcigarettes of the price category most in demand. This will decrease pressure to increase the retailselling price.

References

Ahermaa, E. (2003), Illegaalse tubaka ja alkoholi turg. – Konjunktuur, 2 (145), 87-93.

Council Directive 92/79/EEC of 19 October 1992 on the Approximation of Taxes on Cigarettes.– Official Journal L 316, 31.10.1992, p. 8.

Chaloupka, F.J. and K.E. Warner (1999), “The Economics of Smoking.” NBER Working PaperSeries. Cambridge: National Bureau of Economic Research.

Eating, Drinking, Smoking – Comparative Price Levels in EU, EFTA and Candidate Countriesfor 2001. – Statistics in Focus: Economy and Finance, 2002, no. 42.

Excise Duty Tables (2002), Brussels: EC.

Nordisk tobaksstatistik 1970-2001 (2002), Vällingby: Statistical Bureau VECA.

Purchasing Power Parities and Related Economic Indicators for EU, Acceding and Candidate

128


Countries and EFTA (2002) – Statistics in Focus: Economy and Finance, no. 56.

The Black Market in Tobacco Products. (2000). Pieda Consulting.

Tubakaaktsiisi harmoneerimisega kaasnevad muutused Eesti majanduses (2002). Tallinn: EestiKonjunktuuriinstituut.

129


Approximation of the Rates of Lithuanian Excise Duties on TobaccoProducts with the European Union

Council Directive 92/79/EEC provides for a minimum excise duty on cigarettes and that theincidence of which shall be set at 57 % of the average retail selling price for the most popularcigarettes. In addition, the directive provides that starting from July 1, 2002 the general exciseduty rate should be set at not less than EUR 60 per 1000 cigarettes and should be increased toEUR 64 per 1000 cigarettes by July 1, 2006.

With a view to protecting consumers from a steep increase in the excise duty rate along with anincrease in the price of cigarettes, which would determine a great difference between thepurchasing power and an increase in the price of tobacco products as well as from loss of revenueand the possible spread of smuggling, Lithuania has negotiated a transition period until December31, 2009 in order to implement all the requirements for the excise duty levied on cigarettes laiddown in Council Directive 92/79/EEC.

At the moment a preliminary timetable for an increase in the rates of excise duties has been drawnup in co-ordination with the European Commission. According to this timetable the beginning ofthe period will see an increase in excise duties at a smaller rate whereas at the end of the saidperiod excise duties will grow at a faster rate. On average, the rate of excise duty will grow by15% each year. This timetable is adjusted annually taking into account the cigarettes of the pricecategory most in demand at that time and the macroeconomic indicators (See Figure 1).

The rate of excise duty on cigarettes comprised of the specific component of LTL 42.6 per 1000cigarettes and the ad valorem component of 10% (in percentage from the maximum retail sellingprice) was established on March 1, 2003. On January 1, 2004 the specific component rate per1000 cigarettes was increased from LTL 42.6 to LTL 47.5 whereas the ad valorem component ratewas increased from 10% to 15 %.

According to calculations made by the Ministry of Finance, following an increase in the specificcomponent rate from LTL 42.6 to LTL 47.5 per 1000 cigarettes and an increase in the ad valoremcomponent rate from 10% to 15 % the retail selling price of cigarettes will increase from LTL0.24 to LTL 0.35 per 20 units depending on the cigarette brand. The average retail selling priceof lower class filtered cigarettes manufactured in Lithuania will increase by 11%, the averageretail selling price of the first class filtered cigarettes imported to and manufactured in Lithuaniawill increase by 9% and the average retail selling price of non-filtered cigarettes manufactured inLithuania will increase by 12% (See Figure 2).

As of May 1, 2004 taking into account Council Directive 2002/10/EC, Lithuania will begin toapply the following minimum rates of excise duty laid down in the said Directive: LTL 38 perkilogram levied on cigars and cigarillos and LTL 111 per kilogram levied on smoking tobacco.

130


Figure 1

Lithuanian excise tax harmonization with European Union (MPC-Red&White)

Figure 2

Lithuanian excise tax harmonization with European Union (MPC-Red&White)

131

Approximation of the Rates of Lithuanian Excise Duties on Tobacco Products with the European Union

42.647.5 57.4

75.092.4

117.4

147.6

10% 18% 18% 20% 20,5% 22,5%

0.0

50.0

100.0

150.0

200.0

250.0

2003 2004 2005 2006 2007 2008 2009

LT

L

Ad Valorem (% of RSP) Specific (LTL/000)

55,969,9

88,2110,6

138,6

173,0

221,0

15%

12,3 13,816,6

21,726,8

34,0

42,8

22,5%20,5%20%18%18%15%0,0

10,0

20,0

30,0

40,0

50,0

60,0

70,0

2003 2004 2005 2006 2007 2008 2009

EU

R

Ad Valorem (% of RSP) Specific (EUR/000)

16,220,2

25,632,0

40,1

50,1

64,0

10%

Red&White retail price increase and excise incidence

Red&White retail price increase and excise incidence

132


2,65 2,99 3,423,96

4,625,42

6,52

42,2%46,8%

51,5%55,9%

60,1%63,8% 67,8%

0,001,00

2,003,004,005,00

6,007,00

2003 2004 2005 2006 2007 2008 2009

Lt

40,0%

50,0%

60,0%

70,0%

80,0%

90,0%

100,0%

RSP/pack, LTL Excise Incidence (% of RSP)

0,77 0,870,99

1,151,34

1,57

1,89

42,2%46,8%

51,5%55,9%

60,1%63,8%

67,8%

0,00

0,50

1,00

1,50

2,00

2003 2004 2005 2006 2007 2008 2009

EU

R

40,0%

50,0%

60,0%

70,0%

80,0%

90,0%

100,0%

RSP/pack, EUR Excise Incidence (% of RSP)

Book ReviewMarketing Challenges in Transition Economies of Europe (1999), (Editors: Gopalkrishnan R.Iyer and Lance A. Masters). Haworth Press Inc, 221pp. $42.95 ISBN: 0-7890-0979-X

This book consists of eight separate articles by international teams of researchers from the USA,Estonia, Germany, The Netherlands, Sweden, Finland and United Kingdom and elsewhere. Thepapers compiled in the book are rather diverse both in terms of content, geographical coverage aswell as the methodology applied. There are three studies on the Russian market, two on Romania,one on Estonia, one comparative study on four CEE countries (Bulgaria, Hungary, Poland,Slovenia) and another that deals with general and theoretical issues related to the development ofmarkets in the region. This is discussed by the editors of the book Gopalkishnan R. Iyer and LanceA. Masters.

The book gives an interesting overview of the diversity of processes happening in the region,demonstrates the challenges and unpredictability of developing markets and provides an in-depthinsight in the situation of selected markets in the time period of major reforms taking place in theregion. This can be viewed as useful reading material and a practical guide for readers outside theregion, as well as potential investors who have not been exposed to the problems in Central andEastern Europe before and after the fall of the centrally planned economies. The best example isthe Chapter written by Amjad Hadjikhani and Martin Johanson (“Expectation as the DrivingForce for Entry and Exit”) which deals with the strategic challenges of a Swedish company thathas entered the turbulent but huge Russian market, but failed to achieve the expected results.

The chapter by Olga Kuznetsova and Andrei Kuznetsov “Why Is the Progress of MarketisationSlow in Russia?” discusses the evolution of the Russian market in the late eighties and ninetiesand the behavior of entrepreneurs and enterprises in these new circumstances. “StrategicOrientation of Russian Managers: Is the “New” Russian Firm Market Driven?” written by BrendaRichey, Peggy Golden, Jarmo Nieminen and Denise Johnson complements the discussion onrecent developments in the Russian market by providing the results of their survey on the strategicorientation of Russian managers attending Marketing Executive Development Programs at the St.Petersburg University of Economics and Finance.

The pace of market development over the recent decade has been so high that by 2004 some ofthe data used for analysis in the book is dated. This refers not just to the statistical informationpresented in a number of the chapters, but also to the qualitative information that describesprocesses taking place in the market. The editors of the book have foreseen this possibility andhave focused their article on theoretical aspects; however this cannot be said for the authors of theother chapters. In order to succeed in marketing, the company (potential investor) should use“future” rather than historical data and therefore look for more recent sources of information.

Urmas Varblane’s and Terri L. Ziacik’s chapter “The Impact of Foreign Direct Investment on theExport Activities of Estonian Firms” is undoubtedly very relevant, especially for small economieslike the Baltic states. However, this article does not seem directly related to marketing and thusnot obviously associated with the efforts of the other authors to tackle the challenges of companiesentering CEE markets. Moreover, at this time it would be interesting to discuss the impact of EUenlargement on investment and foreign trade in Estonia and the other accession countries.

The only cross-country study presented in the book was authored by Marin Alexandrov Marinovand Svetla Trifonova Marinova and is titled “Foreign Direct Investment Motives and MarketingStrategies in Central and Eastern Europe”. The chapter is based on a survey of foreign companies,

133

Book Review

host governments and host companies in Bulgaria, Hungary, Poland and Slovenia.

The chapters on “Small Business Retailing in Privatizing Economies: The Influence of Managers’Individualistic and Collectivist Values” and “How the Fall of the Iron Curtain Has AffectedCosnumers’ Perceptions of Urban and Rural Quality of Life in Romania” provide the reader withan insight into selected areas of Romanian socio-economic life. Both of the studies aremethodologically justified and highlight the peculiarities and challenges of entering the market inEstonia.

Should the authors of the articles try to generalize their findings to the whole CEE region, anumber of issues could be given more attention. In particular, the uniqueness of the differentgroups of countries should be addressed. For example, Estonia, Latvia and Lithuania not only hadto reform their economies after the fall of the centrally planned system, but to set up newindependent states. The countries of the former Yugoslavia had, and still have, to deal with theconsequences of serious political conflicts. The Czech and Slovak Republic’s have becomeseparate countries. Marketing is a discipline that requires a rather sophisticated approach to itssubjects. Companies entering the region have to be aware not only of the historical roots of thecurrent situation, but to identify numerous peculiarities related to standards of living, lifestyles,ethnic origin and the perceptions in the life of inhabitants in each particular country and itsregions. The editors’ concluding remarks should have suggested the potential application of thefindings presented in the book.

Overall, the book is very pleasant and easy to read, contains a good mix of theoretical,methodological and empirical material, demonstrates the knowledge and experience of theauthors and the editors and contributes both to the development of research and entrepreneurshipin Central and Eastern Europe.

Evita Lune

Research Associate

Stockholm School of Economics in Riga

E-mail: [email protected]

134
