Tài liệu thuế quốc tế

thue quoc te_tai lieu

Fiscal Harmonization in the Presence

goal that encounters major di¢ culties for its implementation Each countryfaces a particular trade-o¤ between …scal revenues generated by taxation andthe productive e¢ ciency loss induced by their respective tax code This paperprovides a quantitative measure of these trade-o¤s, for a number of taxes andfor the European Union member states, using a dynamic general equilibriummodel with public inputs Calibration of the model for the EU-15 memberstates gives us the following results: i) The maximum tax revenue level isnot far away from the current tax levels for most countries, ii) The cases ofSweden, Denmark and Finland are anomalous, as productive e¢ ciency can begained by lowering tax rates without a¤ecting …scal revenues, iii) In general,countries would obtain e¢ ciency gains without changing …scal revenues byreducing the capital tax and increasing the labor tax and iv) Capital taxharmonization to the average capital tax rate can be done with quite smallchanges in both …scal revenues and output for the majority of countries

Interna-tional Economics, Málaga, March 2007, X Jornadas de Economía Internacional, Madrid, June 2008, XXXII Simposio de Análisis Económico, Granada, December 2008 and Semi- nars at the European Central Bank for very useful comments and suggestions The authors acknowledge …nancial support from Instituto de Estudios Fiscales, SEJ-122 and Junta de Andalucía-Proyecto de Excelencia P07-SEJ-02479.

Universidad de Málaga ,Campus El Ejido s/n Spain Tel: 34-952-131247 Fax: 34-952-131299 e-mail: jtorres@uma.es

+00-1 Introduction

Fiscal harmonization for the European Union member states is a goal thatencounters major di¢ culties for its implementation Each country faces aparticular trade-o¤ between …scal revenues generated by taxation and theproductive e¢ ciency loss induced by the tax code Countries for which aparticular harmonized tax code requires more taxation will have to face anincreased productive e¢ ciency loss, whereas those required to decrease theirtaxes will have to face a loss in …scal revenues However, if we consider amenu of taxes, we can …nd some space for …scal harmonization by changingthe composition of the tax code By, say, increasing labor income tax in someproportion and reducing capital tax in some other proportion, we could keepconstant …scal revenues while increasing productive e¢ ciency This paperprovides a quantitative measure of these trade-o¤s for a number of taxes and

Fiscal harmonization is a very important question in the context of theEuropean Union, particularly with respect to capital income taxes for whichthere exist important di¤erences across EU countries Di¤erences in capitaltaxes will lead to competition to attract capital from abroad (the so-calledrace to the bottom), given the high capital mobility around the world This isparticularly important in the context of the European Union where there isfree capital mobility and it was the European Commission who stressed theneed to remove the corporate tax obstacles in order to promote the creation

of an integrated single market for doing business in Europe For instance,Tanzi and Bovenberg (1990) pointed out the need to harmonize capital taxeswithin the EU, given the existence of an uni…ed market with free capitalmovements However, it is not clear the way how harmonization should bedone First, the particular tax system implemented by each country re‡ectsdi¤erent objectives with di¤erent government expenditure patterns On theother hand, there are no clear reasons to think that a particular tax system ispreferable to another, and rises the question about the system around which

to harmonize the di¤erent tax systems

As pointed out by Tanzi and Bovenberg (1990), without harmonization

of capital income taxes, the allocation of capital across countries would beine¢ cient due to the fact that the capital returns would tend to be equal-

ized after and not before taxes as well as the existence of externalities onother countries Sørensen (2004) use a static general equilibrium model toanalyze corporate tax harmonization in the European Union, where harmo-nization is assumed to take place at the unweighted average corporate taxrate He obtain that the aggregate static e¢ ciency gain from corporate taxharmonization would be quite small.

In this paper we study the scope for …scal harmonization in the EU tries For it, we consider a highly aggregated dynamic general equilibriummodel similar to that of Conesa and Kehoe (2003) and Fernández de Cór-doba and Torregrosa (2005), to study the e¤ects of di¤erent tax codes for each

coun-of the countries in the EU-15 The main di¤erence between our model andthose of the literature is that we introduce in the production function a publicinput, where the stock of public capital is …nanced with …scal revenues Fol-lowing Feehan and Matsumoto (2002) we consider factor-augmenting publicinputs, that is, such inputs are considered as intermediate goods that a¤ectthe production function and give rise to increasing returns In the absence of

a public input in the production function, the tax code trivially associated

to full e¢ ciency is zero for all taxes Since we want to study the trade o¤between productive e¢ ciency and …scal revenues for a collection of countrieswith di¤erent public capital stocks, the introduction of a public input inducesthe need of some country-speci…c tax exaction in order to have production

In this line, the paper develops a DGE model calibrated to data from the EUeconomies to obtain e¤ective average tax rates, preference and technologyparameters to solve a set of question regarding the …scal policy in the EUcountries

To …nd the proportion in which each of the EU-15 countries should reduce

or increase taxes, is the quantitative question this paper aims to answer For

it, we have modelled the productive sector producing a single output out

of three productive factors, namely, private capital stock, labor, and thestock of a public input provided by the government This speci…cation of theaggregated production function allows us to model a public sector that oper-ates in two dimensions: redistributing income, and providing public capitalstocks, trough public investments, for the production process The aggre-gated production function will provide us with a measure of the e¢ ciencygains associated to di¤erent compositions for the income tax code

We compute the combinations of capital and labor tax rates (taking the

consumption tax rate as given) that maximize …scal revenues, i.e., we build

a bi-dimensional La¤er curve and compute its maximum in terms of thesetwo-dimensional …scal instruments to compare the current …scal revenue situ-

functions indicating the combination of capital and labor taxes that sponds to a certain level of aggregated output Assuming the same level of

corre-…scal revenues, we compute the combination of capital and labor taxes forwhich output is maximized In general, optimal taxation policies imply thereduction of the capital tax rate together with an increase in the labor taxrate

Four important facts arise from this comparison: First, the maximum

…scal revenue for each country is associated to relatively low values of thetax rates, and for most of the countries these values are very close to theobserved ones Second, the La¤er curve is very ‡at around the maximum.These two facts put together imply that the EU-15 countries studied here arenot very far from the maximal …scal revenue Third, the rate of substitutionbetween capital and labor taxes keeping …scal revenues constant is very large,i.e., a large decrease in capital tax can be compensated with a small increase

in the labor tax to keep a constant revenue This is a natural result due tothe relative participations in …scal revenues Since the rate of substitutionbetween capital and labor taxes that keeps production constant is in generallow, some space is open to modify the tax code so that revenues are keptconstant while increasing productive e¢ ciency Fourth, given the observedconsumption tax, the maximum productive e¢ ciency level is not far from azero income tax code level for most countries This implies that to maintainpublic capital stocks, …scal revenues obtained via the consumption tax areenough

These four features of the La¤er curve calculated for the EU-15 countries,suggest that a reduction in capital taxation may be the proper direction totake in an agreeable …scal harmonization We conduct a simulation exer-cise in which two possibilities are considered: i) following Sørensen (2004),harmonization is assumed to take place at the unweighted average capitaltax rate (0.26), and ii) harmonization is assumed to take place at the mini-mum capital tax rate, which corresponds to Ireland (0.14) When capital tax

also compute bi-dimmensional iso-revenue curves for the US and the EU-15.

harmonization is assumed to take place at the average rate, …scal revenuessu¤er only small changes in most of the countries However, output showssigni…cant changes When harmonization is assumed to take place at theIrish capital rate, …scal revenues are signi…cantly reduced for most countriesbut with large increases in output Alternatively, our approach of …nding theoptimal tax code for each country (pairs of capital and labor tax that keeprevenues at the observed level with increases in productive e¢ ciency) couldresult in a “convergence” of the tax codes If this is the case we would have

…nd the natural way to harmonize to some extent the European tax system.The measures we obtain from this simulated European tax system give us

an idea of the limits to …scal harmonization where gains are expected for allcountries

The paper is structured as follows In Section 2 we describe the model.Section 3 presents the data we use and the calibration procedure Section 4shows the …gures of the bi-dimensional La¤er curves Section 5 studies theoptimal tax code for each country The e¤ects of capital tax harmonizationare collected in Section 6 Finally, Section 7 presents some conclusions

2 The public inputs model

We consider a production function that relates output with three inputs: bor, private capital and public capital stocks Our choice of the productionfunction assumes that a positive level of public capital is necessary for pro-duction, which implies that for the output to be positive, there must be aminimum level of …scal revenues The government taxes private consumptiongoods, capital income and labor income to …nance an exogenous sequence of

Consider a model economy where the decisions made by consumers are resented by a stand-in consumer, who’s preferences are represented by thefollowing instantaneous utility function:

Private consumption is denoted by Ct:Leisure is NtH Lt;and is calculated

as the number of e¤ective hours in the week times the number of weeks in

The budget constraint faced by the stand-in consumer is:

t; l

t; k

that consumption and investment cannot exceed the sum of labor and capitalrental income net of taxes and lump sum transfers

The problem faced by the stand-in consumer is to maximize the value ofher lifetime utility given by:

(0; 1), is the consumer’s discount factor

The problem of the …rm is to …nd optimal values for the utilization of laborand capital given the presence of public inputs The stand-in …rm is repre-sented by a nested C.E.S with a standard Cobb-Douglas production function.The production of …nal output, Y , requires labor services, L, and two types

of capital: private capital, K, and public capital (public infrastructures), G.Goods and factors markets are assumed to be perfectly competitive The

and Klein (1996) use a isoelastic speci…cation of the tax schedule rather than a linear one

in order to capture the progressivity of income taxation.

…rm rents capital and hire labor in order to maximize period pro…ts, takingpublic inputs and factor prices as given The technology exhibits constantreturn to private factors and thus, the pro…ts are zero in equilibrium How-ever, the …rms earn an economic pro…t equal to the di¤erence between thevalue of output and the payments made to the private inputs We assumethat these pro…ts are distributed between the private inputs in an amount

by:

Finally, we consider the two-side role of the government: as a tax-levying and

as supplier of public inputs The government uses tax revenues to …nancespending in public investment (infrastructures) which rises total factor pro-ductivity and lump-sum transfers paid out to the consumers We assume thatthe government balances its budget period-by-period by returning revenues

The government obtains resources from the economy by taxing tion and income from labor and capital, whose e¤ective average taxes are

which do not contribute to either production or household utility, and thesetwo sources of expenditure plus the transfers to consumers, are the counter-

The government keeps a …scal balance in each period This assumption is

a single …rm and that all households receive equal ammounts of total pro…ts.

Span-done to highlight the distortionary e¤ects of taxes, mainly on capital

assume that the rate of depreciation of public stocks is identical to privatecapital, and therefore we write:

which is analogous to the private capital accumulation process Public vestments, such as railroads, airports, roads, law enforcements, etc., induce

in-a yein-arly ‡ow of nonproductive expenditures, in-and thin-at we will consider

Our model has three productive factors However, the third factor, the publiccapital, has no market price This implies that the rent generated by thepublic input must be assigned to the private factors

From the …rm pro…t maximization problem, the …rst order conditions are:

derivative of the pro…t function with respect to public capital we obtain that:

1= 1

t Gt 11: (7)Notice that equation (7) is not properly a condition of the model since there

is no agent to claim the income generated by the public input

From the above equations we can obtain the following relations that will

be useful for our calibration:

Cassou and Lansing (1998), among others They argue that this setup may represent

a closer approximation to actual constraints than one which allows the government to borrow or lend large amounts.

From private factor income ratios we obtain that RtKt=WtLt = =(1 ):The ratio of total private income to total public expenditures and privatefactors income to total public expenditures are:

The l.h.s ratio can be obtained from national accounts, whereas the r.h.s

is a transformation of the usual estimation of the output from an assumedaggregated Cobb-Douglas production function The …rm will produce ex-

amount is not inputted to the owner of the factor The government usuallydoes not charge a price that covers the full cost of the services provided withthe contribution of public inputs Therefore a rent is generated We assumethat this rent is dissipated and absorbed by the other factors as:

to the budget constraint at date’s t First order conditions for the consumer

Together with the …rst order conditions of the …rm (5) ; and (??), the

bud-get constraint of the government (4), and the feasibility constraint of the

economy, (9), characterize a competitive equilibrium for the economy

t=0,

i) The optimization problem of the consumer is satis…ed

ii) Given prices for capital and labor, and given a sequence for public

inputs, the …rst order conditions of the …rm, with respect to capital and

labor are satis…ed

iii) Given a sequence of taxes, and government investment, the sequence

of transfers and current spending are such that the government constraint is

satis…ed And …nally,

iv) The feasibility constraint of the economy is satis…ed

Notice that according to the de…nition of equilibrium for our model

econ-omy, the government enters completely parameterized, and …scal policy is

made consistent to the model and the data In other words, in our modelthe private sector reacts optimally to policy changes, and this policy changesare given exogenously.

3 Data and Calibration

Before simulating the model, values must be assigned to the parameters Theparameters of the model are:

In calibrating the model presented in the previous section we need three

parame-ters, ( ; ; ; ) and preference parameparame-ters, ( ; ; ) Following Kydland andPrescott (1982) we set in advance as many parameters as possible based upon

a priory information

Computational macroeconomic models of …scal policy depend crucially onrealistic measures of tax rates Agents decisions depend on marginal taxand therefore, e¤ective marginal taxes should be used in the calibration.However, the estimation of marginal tax rates is a di¢ cult task and, aspointed out by Mendoza, Razin and Tesar (1994), at an international level isoften an impractical task given limitations imposed by data availability anddi¢ culties in dealing with the complexity of tax systems Mendoza et al.(1994) proposed a method to estimate e¤ective average taxes and show thattheir estimated average tax rates are within the range of marginal tax ratesestimated in previous works and display very similar trends On the otherhand, Mendoza et al (1994) argue that their de…nition of e¤ective averagetax rates can be interpreted as an estimation of speci…c tax rates that arepresentative agent, in a general equilibrium context, takes into account.Sørensen (2004) also use empirical estimates of average e¤ective tax rates incalibrating a static GE model of …scal policy

In this paper we use e¤ective average tax rates, that we borrow fromBoscá et al (2005), who use the methodology proposed by Mendoza et al

(1994) to estimate e¤ective average tax rates7 Table 1 shows the estimatedaverage tax rates of Boscá et al (2005) for the year 2001 for the selectedcountries, including consumption tax rates, labor tax rates and capital taxrates However, the use of average e¤ective tax rates imply the use of con-servative values (smaller implied behavioral responses).

Strauss (1994) They obtain that the aggregate marginal tax rate is 1.8 times bigger than the aggregate average tax rate However, inspection of …gures from estimated average tax rates reveals this proportion to be very large.

lower than the standard deviation of both labor and capital taxes The labortax rate ranges from a minimum of 0.254 for UK to a maximum of 0.555 ofSweden Finally, capital tax rates ranges from the very low rate of Ireland(0.136) to the 0.388 of Denmark, with a variability similar to the one of thelabor tax.

Second, preference parameters are calibrated using data observations for theyears 2000-2001, taken from the OECD National Account Database From

function of data observations:

from 0.376 of Denmark and the Netherlands to 0.525 of Greece

Finally, we use data from national income and product account for the 14countries to calibrate technological parameters Data are taken from theNational Accounts OECD database First, in order to determine the value of

that is, non-sleeping hours of the working-age population, we assume thateach adult has a time endowment of 96 hours a week (H = 96) Populationaged from 15 to 64 years and average hours worked by year are obtained fromthe Corporate Data Environment OECD Database

Next, we compute the values for all the technological parameters in the

and Kehoe (2003) as unambiguous labor income divided the sum of biguous labor income and unambiguous capital income:

unam-1 = CE

where CE is the compensation of employees, GDP is the Gross DomesticProduct, N W I is non-wage income of the households, de…ned as the netoperating surplus plus the net mixed income of the household sector of theeconomy, and T S is taxes less subsidies The results obtained are consistentwith the ones reported in European Commission (1995) Aggregate capital

of the countries values are in the interval 0.30-0.34

The depreciation rate, , was chosen to match the depreciation-outputratio obtained in the data The capital stock was generated using a perpetualinventory method under the assumption of a geometric depreciation rate:

Capital series were generated for the period 1970-2001 The initial capitalstock was chosen iteratively to match the average capital-output ratio overthe period 1970-1979 In constructing the public capital stock we assumethat the depreciation rate is equal to the depreciation rate of the privatecapital stock Total public capital stock have been derived using series forgovernment consumption of …xed capital, given the computed depreciationrate Values for the depreciation rate go from 0.040 of Austria to 0.064 ofDenmark

The weight of public capital relative to private factors have been

calibrated to match the ratio of public capital to GDP Values range from

elasticity of substitution between public and private inputs is unity Note

economy Por instance, Cassou and Lansing (1998) introduce public capital stock using

a Cobb-Douglas production function In the calibration they consider a range of values for the public capital share of output between 0 and 0.12 Aschauer (1989) and Munnell (1990) estimate values of 0.39 and 0.34, respectively On the other hand, Aaron (1990) and Tatom (1991) estimate values that are not statistically di¤erent from zero Guo and Lansing (1997) obtain a value of 0.0525.

that this assumption implies that the production function given by (3) istransformed into a Cobb-Douglas:

4 The maximum of the La¤er curve

The model calibrated in the above section can be used to answer severalquestions about …scal policy in the EU countries The …rst natural question

in our context is related to the relationship between …scal policy in eachcountry and the La¤er curve How far are the current tax levels for eachcountry from the maximum tax revenue level? Is there any country to theright, the “dark side”, of the maximum of the La¤er curve? To answer thesequestions we …rst calibrate the model to identify the current situation foreach country This exercise will allow us to compute the maximum …scalrevenue level and the maximum productive e¢ ciency level, given the currenttax code Consumption tax rate are …xed and therefore, we focus on the role

of capital and labor taxes Thus, we build a bi-dimensional La¤er curve interms of labor and capital tax rates, as the locus of capital and labor taxrates that yield the same …scal revenues This bi-dimensional La¤er curvewill show the level of …scal revenues for each combination of capital and labortaxes From these calculations we can obtain a map of iso-revenue curves,indicating all the combinations of capital and labor tax rates which generates

a given …scal revenue

Figure 1(a-m) shows the iso-revenue curves for all countries In the ures, we plot the iso-revenue curve for the current (referred to 2001) level

…g-of …scal revenues for each country, indicating the current tax code in terms

of labor and capital income taxes and, the combinations of tax pairs thatproduce the same level of …scal revenues We also show the iso-revenuecurves corresponding to the 90%, 80% and 70% of the current …scal revenuesand the maximum level …scal revenues tax combination Several interestingresults emerge from these …gures First, the maximum …scal revenue levelcorresponds with relatively low tax rates values This means that, given thecurrent tax level, there is not so much space to increase capital and labor taxrates in the case of countries which want to increase …scal revenues Second,tax levels that maximize …scal revenues are fairly similar across countries in-dicating that the maximum of the La¤er curve is not very di¤erent from onecountry to another This is a consequence of the fact that preferences andtechnological parameters are quite similar across European countries Labortax rates at the maximum are very similar, around 49% for all the countries

A little more variability is found in the case of the capital tax rates, with aaverage value around 37%

Another important …nding is that for all the countries, the iso-revenuecurves takes the form of an ellipse but very vertical, representing capital tax

in the vertical axis and the labor tax in the horizontal axis This impliesthat …scal revenues are very sensitive to changes in the labor tax but not

to changes in the capital tax Several reasons can explain this result First,labor income is more important than capital income because it represents alarger share of national income Thus, …scal revenues are more sensitive tochanges in the labor tax than to changes in the capital tax Second, thisresult implies that distortionary e¤ects of capital taxes are larger than thecorresponding to labor taxes For instance, an increase in the capital taxrate provokes a very small change in …scal revenues, due to the fact that

such an increase a¤ects negatively, in an important proportion, the capitalaccumulation.

Results from this exercise are summarize in Table 3 Columns 2 and 3show tax rates that maximize …scal revenues and in bracket we show the dif-ference with respect to the current tax rates, while columns 4 and 5 computetax rates corresponding to the maximum productive e¢ ciency The last col-umn at the right shows the percentage deviations in terms of …scal revenues

of the current situation for each country with respect to the maximum …scalrevenues

Table 3: Maximal revenue versus e¢ cient tax codes

of the bi-dimensional La¤er curve Austria, Belgium, Denmark, Finland,France and Sweden are countries in which the current tax code is very closethe maximum tax revenue level On the opposite side, the countries that arefarther from the La¤er maximum are Portugal, UK, Ireland, and Spain

In particular, we observe three countries that are placed in the “dark side”

of the La¤er curve, where some taxes are above the maximum …scal revenuetax level; termed ”the prohibitive range” by La¤er (1981) These countries

that the capital tax is slightly above the maximum revenue capital tax Infact, Denmark is the country of the EU-15 with the larger capital tax rate.Simply, by reducing the capital tax rate, …scal revenues in Denmark wouldincrease The other two special cases are Finland and Sweden, where thelabor tax rate is above the maximum …scal revenues labor tax, particularly

in Sweden Therefore, in these two countries by reducing the labor tax ratewould obtain an increase in both, the …scal revenue and e¢ ciency

Finally, we also compute the maximum e¢ ciency tax code for each try, that is, the tax code corresponding to the maximum output level, giventhe consumption tax rate Without the existence of public capital in the pro-duction function, the maximum e¢ ciency tax code would be trivially zero,

coun-as it is obtained in Fernández-de-Córdoba and Torregrosa (2005) Not prisingly, the maximum productive e¢ ciency shows zero capital tax rates forall countries However, we …nd several examples with positive labor taxes,such as France, Ireland, Italy, Netherlands, Portugal, Spain and Sweden.This …nding shows that for these countries, …scal revenues obtained fromconsumption taxes are not enough to support the observed level of publicinput provision The largest values for the maximum e¢ ciency labor taxrates correspond to Ireland (8%) and Spain (6%), followed by France (5%)and Italy (4%) For Austria, Belgium, Denmark, Germany, Greece and UK,

sur-…scal revenues obtained from consumption tax are enough to support theobserved level of public input provision

This set of results show that the macroeconomic implications of the taxsystem in the European countries are very similar, both in terms of …scalrevenues and e¢ ciency We obtain that capital and labor tax rates cor-responding to the maximum of the La¤er bi-dimensional curve are similaracross countries Therefore, a natural way to achieve …scal harmonization inEurope would be the case if all the countries decide to move to the maximaltax revenues level If the objective of all countries were to maximize tax

Sweden is well to the right of the maximum of the La¤er curve for most of the tax instruments A similar result was found by Hansson (1984) using a static model.

revenues, then …scal harmonization would be almost perfect, with respect

to both labor and capital tax rates A similar result is obtained if all thecountries decide to implement a …scal policy with the objective to achievethe maximal e¢ ciency level In this case total harmonization of the capitaltax rates is obtained if all countries decide to use a maximal e¢ ciency taxcode

5 The optimal tax code

Next, we consider the optimal tax level for each country, …xing …scal revenues

at the current (referred to year 2001) observed level That is, given the

…scal revenues level and all the combinations of capital and labor taxes that

the maximum level of output The question we want to answer is, if it ispossible to increase productive e¢ ciency in the di¤erent European countries

by substituting one tax by the other without changing public revenues Tomaximize productive e¢ ciency given a level of …scal revenues implies to …nd

and labor tax that keeps production constant is equalized to the rate ofsubstitution that keeps …scal revenues constant For most countries thisimplies a substitution of capital by labor taxes, that is, government budgetbalance is maintained through adjustment in the tax rate on labor income.Table 4 shows the optimal tax code for each country together with thepercentage change in output, capital and labor that should be veri…ed inorder to attain the optimal tax schedule Additionally, Figure 2(a-m) com-bines the iso-revenue curves together with the iso-output curves, representingcombinations of capital and labor tax rates that produce the same level ofoutput We plot the iso-revenue curve corresponding to the current level of

…scal revenues together with the iso-output curves, normalized to 100 at thepoint of the current tax code For each level of …scal revenues, there existsonly one pair of tax rates that maximizes output, determined by the tangencypoint closest to the origin between the iso-revenue and the iso-output curves

As we can observe, the iso-output curves are concave and as we showed inthe previous section, the maximum e¢ ciency level corresponds to a non-zero

10

For most countries, optimal tax rates imply a reduction in capital taxrates and an increases in labor tax rates As a consequence, such a changewill increases capital stock and reducing labor This result is found in all thecountries except Finland and Sweden, that is, the two countries in which thelabor tax rate is above the maximum revenue labor tax.

Table 4: Optimal tax code