Taxing the Working Poor

7/27/2019 Taxing the Working Poor

1/165


2/165

Taxing the Working Poor


3/165


4/165

Taxing the WorkingPoorThe Political Origins and Economic

Consequences of Taxing Low Wages

Achim Kemmerling

Postdoctoral Fellow, Jacobs University, Bremen, Germany

Edward ElgarCheltenham, UK Northampton, MA, USA


5/165

Achim Kemmerling 2009

All rights reserved. No part of this publication may be reproduced, stored ina retrieval system or transmitted in any form or by any means, electronic,mechanical or photocopying, recording, or otherwise without the priorpermission of the publisher.

Published byEdward Elgar Publishing Limited

The Lypiatts15 Lansdown RoadCheltenhamGlos GL50 2JAUK

Edward Elgar Publishing, Inc.William Pratt House9 Dewey CourtNorthamptonMassachusetts 01060USA

A catalogue record for this bookis available from the British Library

Library of Congress Control Number: 2009922747

ISBN 978 1 84720 778 4

Printed and bound in Great Britain by MPG Book Group, UK


6/165

v

Contents

List of Figures viList of Tables viiList of Abbreviations viiiList of Country Abbreviations ixPreface xi

1 Introduction 12 A comparative wel are state analysis o tax mixes 113 The economics o taxing labour 254 Political economy applied to tax mixes 595 Empirical evaluation 846 Conclusion: employment and redistribution are not

incompatible 120

Bibliography 126Index 145


7/165

vi

Figures

1.1 Tax progressivity and service employment 51.2 Tax burden, progressivity and representativity 62.1 Temporal evolution o the tax mix 164.1 Employment protection and the tax mix 765.1 Real wages, union members and strikes in the nineteenth

century 1025.2 German and British tax mix compared 1046.1 Marginal efective tax rates o three countries 123


8/165

vii

Tables

2.1. Country comparisons o tax-to-GDP and efective tax rates 182.2 Indicators o tax structure 213.1 Time-series results or three countries 463.2 Cross-section results or six periods 483.3 Pooled results or employment and unemployment 513.4 Pooled results or sectoral data 523.5 A synopsis o theoretic results 544.1 Correlation o diferent measures or the voter space 674.2 Mean comparisons o taxes and political institutions 715.1 Cross-section results 885.2 Pooled estimates 905.3 Simultaneous equations 955.4 Comparison o tax schedules 105


9/165


10/165


11/165


12/165


13/165


14/165

1

1. IntroductionFreedom rom taxation bred laziness and lack o ingenuity, declared the Comtede Maurepas, 18th century president o the French Navy Board in a letter to thegovernor o Canada. (Weaver, 1914: 746)

But not only would the right to assistance breed laziness among the workingclasses, so too would it extinguish in the upper classes the sweetest and most

ruit ul virtue, charity, a good which, trans erred into State taxes. Louis-RenVillerm, French eighteenth chronicler o the British poor laws (Smith, 2000:101011)

Do excessive wel are states cause crises in labour markets, or do crises inlabour markets cause growing wel are states? This hen-and-egg questionhaunts both academic and political debates alike. In academia apologistso the ree-market would con ront those who uphold the insurance andequity character o the wel are state. Ideologically extreme politicians see

in the wel are state either a social hammock which lures workers intoidleness or a bulwark against capitalist attacks o workers real wages. Inthe turmoil o these battles it is sometimes orgotten that both the wel arestate and the labour market depend on each other. In this book I want to

ollow this idea in a specifc domain o the wel are state: labour taxation.In particular, I will argue that the real question in contemporary wel arestates is not whether, but how wel are is fnanced. How does the structureo taxation a ect (low-wage) workers, and why does politics in di erentcountries lead to di erent tax structures? Both questions depend on each

other and are, in act, no recent phenomena.

HOW TO TAX LABOUR AN OLD QUESTION?

Di erences in the unding o wel are states have been debated ever since thevery beginning o wel are statism. Bismarcks decision to organize Germansocial insurance as a contribution-based scheme was observed early on byLasalle and others as a major attempt to produce social security withoutsocialism (Tennstedt and Winter, 1993). Ironically, the contribution-basedrevenues o Bismarckian wel are states were not the intellectual o spring o Bismarck himsel . On the contrary, Bismarck avoured a tax-fnanced social


15/165


16/165

Introduction 3

no-politics idea argue that most o the empirically discernible variation inthe tax structure is due to in ation, growth or other socioeconomic un-damentals. Hence, diverging national trajectories are driven by automaticfscal responses (Volkerinket al ., 2002). Historical institutionalists enhancethis argument. They claim that the important tax policy decisions weremade some 100 or more years ago. From then on, the tax system evolvedas a consequence o institutional path-dependencies and created politicaldynamics o its own (Alber, 1982; Steinmo, 1993; Pierson, 1996).

There are scholars who disagree. Early on Titmus (1974) claimed thattaxation is a undamental part o the wel are regime which in turn is o political origin. Le t versus right positioning still has explanatory powerin peoples attitudes on public spending versus taxation (Kitschelt, 1994;Rehm, 2005). Governments ollow this logic, once we control or changesin the economic and strategic environment (Beramendi and Rueda, 2007;Kato, 2003). I there are deviations rom a partisan pre erence, theseare due to the role o veto points such as Supreme Courts or Federalism(Gangho , 2006; Hettich and Winer, 1999). Hence, all in all the le t versusright is still an important description o the underlying di erences in tax-policy pre erences o governments (Wagschal, 2003). So, who o the twosides is right?

THE ARGUMENT

There is one crucial di erence between nineteenth-century Germany andcontemporary OECD countries. Back then costs could not be completelyrolled-over to labour, since wages were close to what workers needed interms o basic nutrition and accommodation. Hence it may well be that theparliament discussed two di erent orms o taxation on capital rather thanon labour. Note that Bismarck himsel was not worried about workers,

he was worried about the industry. Today, however, many countriesprovide substantive levels o public social security and have, above all,comparatively high real wages or most workers. This implies some level o decommodifcation,3 that is wages are considerably higher than the level o subsistence. The economic implication o this is that incidence o major tax

orms such as income, payroll and indirect taxation is largely on labour (seeChapter 3). For this reason I ocus exclusively on taxing labour, or it is,perhaps unintentionally, the modern wel are state that is a key reason whymost o these taxes today all on labour. With the long-term shi t towardstaxing labour came complaints that labour taxation is excessive and causesdisequilibria in the labour market.

Against this historical background, the no-politics idea is clearly not


17/165

4 Taxing the working poor

wrong, but somewhat misleading. Path dependencies are at work andhave led to a growing wel are state. But the underlying logic di ers. Thecon ict o interests in the tax structure is nowadays one within labour,less between labour and capital. This does not mean that tax competitionon mobile capital income, globalization or the con ict between the veryrich and the poor has become meaningless. But it is not the only con ict,maybe not even the decisive one, i you ocus on tax structure as opposedto tax levels. Rather, there is a con ict between low-skilled and medium-skilled workers on employment and income generation. To illuminatethis potential con ict, I will not analyse the employment e ects o generalredistribution, that is I will not engage with the old question o an optimallevel o the public sector. Rather I am interested in the implications o di -

erent mixes o taxation and its consequences or the tax structure in termso progressivity, holding the level o taxation constant. Why do countrieschoose di erent orms o labour taxation, and why do they change theirstrategies across time?

The politics-still-matters idea is neither alse, but has usually played theheterogeneity o workers down ( or a similar point see Rueda, 2005, 2006).Interests o low-wage workers and medium-wage workers do not neces-sarily coincide. Both segments will at times be in con ict with each otheron the optimal design o tax structure, and in particular on progressivity.

The reason is that tax progressivity has only straight orward redistributivee ects in a per ect market economy. Only in this case higher progressivitytranslates into higher redistribution between rich and poor voters, andonly here the pre erences or progressivity ollow a clear le t-versus-rightpattern. In labour markets with some degree o regulation, progressivityalso translates into redistribution o employment and income probabili-ties. Lack o progressivity will strongly harm low-skilled workers, whereasmedium-skilled workers are le t in an ambivalent situation. I there is com-petition between medium- and low-skilled workers, the ormer will have an

ambiguous stance on progressivity.For such a line o thought, I must presume that progressivity and taxstructure will have some impact on employment, albeit this impact is weakand di ers rom country to country. Figure 1.1 shows evidence or this.Let us assume or the time being that the ratio o income taxes to payrolland indirect taxes is a good measure or the progressivity o a tax system.Later on I will deal with this operationalization in more detail (Chapter2), but or now we will simply assume it. Let us urther assume, as is re-quently done ( or example Scharp , 2000; Kemmerling, 2003), that thelow-wage sector is concentrated in categories 6 and 9 o the InternationalStandard Industrial Classifcation system (ISIC) which contains servicesector workers. Then Figure 1.1 tells us that there is indeed a negative


18/165


19/165


on the le t shows the relationship between the overall tax burden measuredas total taxation as a proportion o GDP and my index o tax progressiv-ity.4 Again, the justifcation or such an operationalization will come at alater stage, here it is enough to see that the relationship between the two isnot clear-cut. We see that some traditionally conservative Anglo-Americanwel are states have a high proportion o income taxes, and similar thingsapply to traditionally social-democratic wel are states. This is not new,and scholars have ocused on the di erent characteristics o income taxesbetween these countries (Steinmo, 1993). I do not dispute this, but I point

at the act that political clusters are not visible in such a graph. Otherwiseconservative countries should be in the lower western corner o the graph,whereas social-democratic countries in the upper eastern corner.

The right panel o Figure 1.2 shows the relationship between tax pro-gression, as measured be ore, and an indicator o the surplus coverage o trade unions. For this purpose, I compiled data on bargaining coverageand union density. The indicator merely subtracts the latter rom the

ormer. The idea is that a trade union is less representative, i (a) uniondensity is low, and (b) the outcome o their negotiation a ects many non-members, hence bargaining coverage is high. Accepting this defnition orthe moment, we can see a negative relationship. France has the ewestrepresentative unions and a tax system in which non-progressive tax orms

AUL

AUT

BEL

CAN

DEN

FIN

FRA

GER

GRC

ICE

IRL

ITA

JPN

LUX

NLD

NOR

NZL

PRTSPA

SWESWI

TURUKD

USA

CZEHUN

KOR

MEXPOL

SLK

0

0 . 5

1

1 . 5

2

I n d e x o

f p r o g r e s s

i v i t y

10 20 30 40 50Total tax ratio

AUL

AUT

BEL

CAN

DEN

FIN

FRA

GERITA

JPN

NLD

NOR

NZL

PRTSPA

SWESWI

UKD

USA

0

0 . 5

1

1 . 5

2

I n d e x o

f p r o g r e s s

i v i t y

0 20 40 60 80Surplus coverage

Figure 1.2 Tax burden, progressivity and representativity


20/165


21/165


22/165


23/165


NOTES

1. For reedom is nothing but death! I you dont want to reach into your pocket and thetreasury, you will not achieve anything. To impose the whole burden on industry, I dontknow whether it could stand that. It would hardly work or all industries. For some itmay work: or those in which the wage o workers is just a small component o the entirecosts o production. [. . . The problem however] exists or those, in which wages amountto 80 or 90 per cent o all costs, and whether those could survive, I dont know. (owntranslation)

2. Though clearly not equivalent I will use both terms low skilled and low wageinterchangeably.

3. Esping-Andersen (1990: 37) defnes decommodifcation as a readiness to enable indi-viduals and amilies to uphold a socially acceptable standard o living independent o market participation.

4. I use 30-year averages or all country observations. Hence the fgure shows long-term

correlations between the three variables.


24/165

11

2. A comparative wel are state analysiso tax mixes

to tax about 1300taxen to assess, put a tax on; borrowed rom Old Frenchtaxer, learned borrowing rom Medieval Latintaxare, rom Latin, and bor-rowed directly into English rom Latintaxare evaluate, estimate, assess, handle,probably a requent orm o tangere to touch. (Barnhart, 1988: 1118)

(be-)steuern steuern mit dem Steuer lenken, [. . .]stiuren lenken, leiten,(unter)sttzen, hel en, beschenken, ausstatten, eine Abgabe entrichten, [. . .]stiurenin eine Richtung bringen, lenken, [. . .]stiuren/stierensteuern, richten.(P ei eret al ., 1993: 1359)1

The ultimate question o this book is simply why national governmentsdi er in taxing (low-wage) labour. For an answer we frst need to knowwhat choices governments have when they decide how to tax labour.Second, we need to know what shapes political actors pre erences or di -

erent orms o taxation. In a political economy tradition, the latter willdepend on the basic unctions o taxation in a society, and the di erencesbetween di erent tax orms in per orming these unctions. This calls orconceptual work, and or taxonomy in particular. A look into the etymo-logical history o the word taxation may hence be doubly illuminative.

Comparing the English and German etymology o the word tax, taxa-tion, as the two quotes rom above show, has two di erent roots. In oneperspective, taxes are used to levy resources in order to pursue other poli-cies. Hence, taxes are merely the means or achieving di erent goals, and

should be considered as political and economic budget constraints. Thisis the traditional notion o tax in English. Such a notion o tax policy isvisible among both economists and political scientists who have collapsedthe analysis o expenditure and fnancing o wel are states into one modeo explanation ( or example Esping-Andersen, 1990; Meltzer and Richard,1991).

The second perspective is prevalent in the origins o the German conceptor taxation Steuer. In this conception, taxes have a much more direct

impact on pre erable outcomes. Taxes are used to induce certain typeso behaviour (say smoking less) or outcomes (lower inequality). Titmus(1974) was one o the frst comparative researchers to acknowledge theinterrelationship between taxation and social policy. I will start this


25/165


26/165


27/165


28/165

A comparative wel are state analysis o tax mixes 15

across countries (OECD, 1999b; Sachverstndigenrat zur Begutachtungder gesamtwirtscha tlichen Entwicklung, 2003). A whole cottage industryhas evolved to construct better empirical indicators than primitive tax-to-GDP ratios. The frst landmark contribution in this direction was o Mendoza et al . (1994) who proposed a simple, easy-to-calculate measureo e ective tax rates on actor incomes and consumption. This measureshowed reasonable consistency with micro-based estimates (Mendozaetal ., 1994: 316). Yet, Carey and Rabesona (2002) and Volkerinket al . (2002)argue that Mendozas methodology underestimates the e ects o exemp-tions on some orms o sensitive income. Both contributions propose anew measure o their own correcting or this and other problems in theMendoza methodology.7

All the macro-aggregate measures di er somewhat (Haanet al ., 2003),but Carey and Rabesona fnd signifcant deviations only in a ew cases.8 Cusack and Beramendi (2006) have tested several data sets in multivariateregressions and have also ound ew di erences. Haanet al . (2003), fnally,show high correlations among all macro measures o tax rates, whereas thecorrelation between the OECD micro indicator and the macro indicatorsis only reliable across countries but not along time series (Heady, 2004:284).

For the purposes o this book, I will stick to aggregate data on tax-

to-GDP ratios and Carey and Rabesonas e ective tax rates (Carey andRabesona, 2002).9 The term taxation ollows the general OECD defni-tion o taxes used in the Revenue Statistics and includes social securitycontributions (OECD 1998: 30). Income taxes include all taxes with theOECD label 1000, that is taxes on income, profts, and capital gains.Payroll taxes combine two categories: 2000 consists o compulsorysocial security contributions and 3000 consists o any additional taxeson payroll and the work orce. Finally, indirect taxes correspond to thecategory 5000 o the OECD, that is taxes on goods and services like VAT,

excises, and sales taxes. The three tax variables are a percentage o GDP(t1 gdp, t23 gdpand t5 gdp).The e ective rates are based on those o Carey and Rabesona. The

authors only report taxes on capital income, labour income and consump-tion, whereas I am interested in splitting the tax on labour income into apayroll and an income tax component and merging the latter with taxes oncapital income. I do not dispose o all the necessary in ormation to makethese trans ormations. So my indicators or an e ective income tax ratet1ef and an e ective payroll tax ratet23ef are only roughly correct givenCarey and Rabensonas methodology, but I expect the resulting deviationsto be marginal. The e ective consumption tax ratet5ef is theirs.10For mostOECD countries, e ective tax rates start in 1975 and end in 2000, whereas


29/165


tax-to-GDP ratios are available or longer periods. With both tax-to-GDPratios and Carey and Rabensonas measures o e ective tax rates at hand,we are now able to compare the size o tax revenues and the respective ratesacross countries and time.

Temporal and Cross-Sectional Variation

Tax-to-GDP ratios have risen considerably in the last 40 years. Today, theaverage total tax ratio or OECD countries is around 40 per cent o GDP.Given that the three orms o taxation amount to 75 per cent o publicrevenue in most countries they orm the backbone o wel are state fnance.In contrast to total labour taxation, the tax mix has changed. In particular,

payroll and indirect taxes have outpaced income taxation in most coun-tries. Figure 2.1 shows these trends or the mean o all OECD countries.Payroll taxes, including social security contributions, nearly doubledbetween 1965 and 2002. Income taxation has remained airly stable sincethe 1970s (le t panel). In comparison, the e ective income tax rate reachedits high point in the 1980s (right panel o Figure 2.1) and declined therea -ter. This shows the impact o a series o tax re orms in most o the OECDcountries beginning with the US in 1986 (Blundell and MaCurdy, 1999).From the 1980s to the 1990s indirect tax rates have increased by more than

4

6

8

1 0

1 2

1 4

1960 1970 1980 1990 2000 2010year

Groupmean t1gdpGroupmean t23gdpGroupmean t5gdpP

1 4

1 5

1 6

1 7

1 8

1 9

1960 1970 1980 1990 2000 2010year

Groupmean t1effGroupmean t23effGroupmean t5eff

Figure 2.1 Temporal evolution o the tax mix


30/165


31/165


32/165


33/165


very similar to an income tax (Levy, 2000: 315).13 The base o incometax base is larger than that o the payroll tax, since the income taxationincludes not only wages but also other sources o income. It also includesmore people, since sel -employed and non-employed people have to payincome taxes, whereas payroll taxation is usually restricted to dependentemployees. In an ideal-type world, the tax base o general consumptiontaxes should be even broader, as they a ect all people. In the real worldthere are exceptions to this rule due to economic openness and exemptionsin indirect taxation. On basis o a head count the situation is much clearer:general consumption taxes indeed a ect all people, whereas income taxesare restricted to those with income, and payroll taxes, even more narrowly,to those with wage income.

Are these assumptions empirically correct? Estimations o the tax baseon a head-count basis are very sensitive due to huge administrative di er-ences (OECD, 2007b). One can assume without loss o generality thatgeneral consumption taxes have a base o 100 per cent. For social securitysystems it is very di cult to get comparable in ormation, since the numbero contributors di ers between insurance schemes. It is sa e to assume,however, that the number o those covered by public social insurance inany OECD country is smaller than the number o those in the labour orce( or example Montanari, 2001). For personal income tax there is some

direct in ormation available.14

The OECD (2007b: 124) collects data onthe number o registered tax payers as percentage o total labour orce.Table 2.2 shows these data or OECD countries. We see that countriesdi er markedly in the relative extension o their income tax base, ranging

rom a Korea with merely 10 per cent o the labour orce paying incometaxes to 250 per cent in New Zealand. At the bottom o the table there arecorrelations or the three tax orms with the measure o the tax base. Notsurprisingly, the tax base is higher in those countries in which income taxa-tion in general is higher. There is hence some evidence that countries with a

relatively dominant income taxation have more inclusive tax bases.Progressivity is the second major characteristic that sets income, payrolland indirect taxes apart. Progressivity is politically important or equityreasons, since it shi ts the tax burden towards richer people. It may alsobe important or e ciency reasons, since it a ects the decisions to take upwork, especially or the low-wage sector. In general, progression can beachieved by two means: indirectly with exemptions such as basic allow-ances or deductions or those with less income, or directly with an increas-ing marginal tax rate. Both orms play a role or labour income, but it ispredominantly indirect progression which a ects the labour orce partici-pation in the low-wage sector, whereas marginal rates are more important

or higher wages and or the number o working hours (see below).


34/165


Table 2.2 Indicators o tax structure

Country Tax base Progressivity Insurancecomponent

AUL 167.1 16.0 94.3AUT 135.9 1.5 99.9BEL 130.4 2 3.9 95.6CAN 135.4 1.5 95.0CZE 48.6 6.5 DEN 158.6 20.0 93.4FIN 184.2 9.2 99.9FRA 124.3 2 2.0 99.9

GER 69.8 6.2 99.5GRC 223.3 19.7 98.2HUN 107.3 9.0 ICE 143.8 4.8 IRL 104.6 19.9 94.4ISR ITA 6.4 99.9JPN 69.5 11.8 96.6KOR 9.4 11.9 LUX 44.3 17.3 98.7

MEX 19.5 18.3 NLD 84.9 3.6 94.5NOR 152.5 11.9 96.7NZL 242.9 18.0 93.1POL 163.4 0.0 PRT 121.8 9.7 98.7SLK 18.9 0.0 SPA 186.1 1.0 99.9SWE 157.7 11.8 95.9SWI 10.3 94.5

TUR 12.8 3.5 UKD 96.9 2 8.9 94.9USA 149.7 9.3 95.1Average 116.6 17.9 96.8Correlationst1e 0.54*** 0.21* 2 0.45***t23e 2 0.12 2 0.46** 0.70***t5e 0.12 2 0.03 0.1

Note: Tax base is the ratio o registered taxpayers to the labour orce (OECD, 2007b),Progressivity is defned asMTR167% 2 MTR67% (c . text and OECD, 2004b). Insurancecomponent is 100 minus Disney (2004)s calculus o the pension tax.


35/165


Typically, income taxation is the only progressive tax orm. For low tomedium income brackets payroll taxes are proportional (Wagsta et al .,1999). People whose earnings are above a certain threshold usually donot pay contributions so that, by and large, payroll taxes lie somewherebetween regressive to proportional. Again there are exceptions to this rule.Some countries have no ceiling on payroll taxes so that the ormal statuso these taxes is much more akin to income taxation (Goerke, 2002: 238).Moreover, some countries such as Austria and Switzerland have progres-sive elements in their payroll taxes (Messere, 1993). Thus the degree o progressivity or each tax orms shows some variation across countries(Wagsta et al ., 1999). Finally, the progressivity o (general) consumptiontaxes is similar to payroll taxes. I all individuals consume their li e-timeearnings until their deaths, consumption taxes should be proportional toincome (Homburg, 2003: 157). Many wel are states exempt certain ormso consumption such as housing rents or ood rom consumption taxes.The consequences o these exemptions are not clear. For instance, it isvery much debated whether a VAT is slightly progressive, proportionalor regressive.15 By and large, it is hence justifed to assume that generalconsumption taxes are proportional.16

There are several ways to measure the progressivity o the tax system.One way is to use the a orementioned OECD data on the taxation o high-

and low-wage earners (Heady, 2004). In Table 2.2 I present such data orthe year 200203. I have calculated the di erences in per cent o the taxburden or those earning 167 per cent and 67 per cent o the average pro-duction wage. Using this indicator one can see that there is some positivecorrelation with the size o the income tax ratio, strongly negative withpayroll taxation, and no relationship with indirect taxes. Taken together,these fndings imply that payroll taxation does make it di cult to have atax system that is progressive, whereas a high proportion o income taxesis a necessary, but not always su cient condition o having a progressive

tax mix. This is also intuitively plausible, since loopholes and exemptionsin the income tax code beneft the rich rather than the poor.Finally, the insurance component o taxes is the most intricate charac-

teristic o the tax mix. According to the OECD (1998: 30) taxes are com-pulsory, unrequited payments to general government. Hence, the defningattribute o taxes is their loose connection to public services. This principleis, o course, violated in the case o social security contributions, and mostclearly in systems o defned benefts where contributions ensue a legallybinding contingent entitlement. In this sense social security contributionsare more akin to orced savings (OECD, 1995). It is or this reason that Ihave included this characteristic in the discussion o the tax mix. Becausesocial security contributions are earmarked or special purposes, such as


36/165


insurance against the risk o becoming unemployed, sick or old, it dependson the degree o equivalence between payments and benefts whether theyare taxes or insurance premiums ( or example Schmidet al ., 1987: 93). I the equivalence is close to unity, contributions are reliably perceived asinsurance and should be less distortive (OECD, 1995: 10).

It comes as no surprise that countries that have a Beveridge wel aresystem have ar higher tax components in their tax systems than wel arestates with a Bismarckian legacy. According to Disney (2004), tax compo-nents are up to 10 per cent or the UK or New Zealand, whereas they areclose to zero or Germany, Austria and Italy (Disney, 2004: 293). I showthese data or the latest available year (1995) in Table 2.2. Although thedata are scarce one fnds a strong relationship between the mix and theinsurance component. Countries with high income taxes have a high taxcomponent in their social security systems. Countries with high payrolltaxes also have high insurance components. Again there are exceptionsto the rules as the a orementioned French social security contributions(CSG) show, since they are earmarked or social trans ers. In addition,most wel are states have used various orms o discount actors demo-graphic, business-cycle adjustments or plainly redistributive ones whichlead to a varying degree o equivalence between contributions and benefts.All things considered, however, a wel are state that is based on ideal-type

social security contributions should clearly have the highest degree o equivalence between taxes and social trans ers.Let us summarize the major points o this chapter. First, we have seen

that tax and social policies are closely knit together. I we ocus on taxesonly, we might orget unctional equivalents. For this reason I will returnto these equivalents at a later stage. Second, we have seen that in the taxmix there is a remarkable shi t away rom income taxation. We have alsoseen that some countries such as Sweden or the US resisted this trend.Third, we have seen that the tax mix reveals underlying structural charac-

teristics. Countries in which income taxation dominates have on average more inclusive and progressive tax systems as well as lower insurancecomponents. With this in ormation in mind, we can now turn to the e ectsthese taxes have on the labour market.

NOTES

1. Translation: be-steuern to steer to direct with a steer, stiuren to direct, lead, support,help, grant, endow, to pay a duty, stiuren to orce into a direction, direct. The ety-mological relationship between both German roots contribution and steering wheel remains unresolved so ar (Seebold, 2002: 882).

2. Issues o democratic mobilization and taxation go back to the roots o early democracies


37/165


where taxation meant representation. Historically, it was the growing tax burdenthat made people demand specifc prerogatives such as parliamentary control o budgetdecisions. In the contemporary literature on comparative political economy, this rela-tionship has been turned upside down. The tax system and level is a consequence o the degree o representation o di erent electoral segments. I , say, turnout is higheramong poor voters, one would expect higher levels o taxation (Franzese, 2002; Roemer,2001). The idea that representation causes taxation is pervasive and has in uenced otherareas such as the theory o democratization (Meltzer and Richard, 1991; Boix, 2003).While this notion is certainly not wrong it neglects the potential or reverse causality.

3. The Dutch fnance ministry, or instance, claws back $6295 rom a total o $24 717 o gross unemployment benefts. Net benefts are only slightly higher than or an equivalentrecipient in Austria (Adema, 2001: 16).

4. An empirical proo would go beyond the scope o this book. For a crude test I haveregressed Ademas data on hidden versus universalistic tax-based social policies on ameasure o median voter pre erence or spending versus taxation (?). Controlling or thetax burden, I fnd the ollowing results or 14 countries: The more a state taxes benefts,and the less it uses implicit social policies, the more popular is spending versus taxation(results available on request).

5. I one accounts or di erences in taxation o benefts, and the size o tax cuts and similarinstruments in the public budget, the rate o the social expenditure o the USA rises

rom 15.8 (gross) to 23.4 (net) per cent o GDP whereas it drops or Denmark rom 35.9(gross) to 27.5 (net) per cent.

6. Homburg (2003: 90) gives an example. The frst tax rate o the 2004 German incometax schedule applied only to a short income range between 7000 and 13 000 euro perannum. This allows politicians to cut the frst rate ostentatiously without losing toomuch public revenue.

7. The Eurostat (2004), produces another e ective tax measure, but its temporal and cross-country coverage is limited.

8. Serious deviations were ound or Canada on consumption taxes, the Czech Republicon labour taxation, Netherlands on capital taxation, Portugal on labour taxation, andSwitzerland also on labour taxation (Carey and Rabesona, 2002: 143).

9. I thank the authors or sharing their data with me.10. See Carey and Rabesona (2002: 133) or a detailed description o their method. I split

labour income in the two sources OECD 1000 and 20001 3000, and adjusted therespective tax bases.

11. The results are available on request.12. There are many di erent approaches to the concepts o path dependency and hysteresis

in social sciences ( or example Pierson, 2004), but ew o these ever talk about howto operationalize these phenomena on grounds o a quantitative, political economyapproach. Though hysteresis and nonstationarity are not the same as path dependence

they are clearly related concepts.13. Some authors claim that the introduction o the CSG may even be interpreted as thebeginning o path-switching rom a Bismarckian to a Beveridge system o fnancingsocial protection (Kato, 2003: 105).

14. Among other things, countries di er to the extent with which tax fling is obligatory.15. For the German debate see Bedauet al . (1998) who argues that the German VAT is by

and large proportional.16. A major problem with this materialistic perspective is, o course, that economic models

do not ascribe utility to the act o saving, but characterize it as de erred consumption. I saving plays a role in the ormation o status or social security, normative implicationso indirect taxation may well be more in line with the typical gut reactions o manypeople: it avours saving relative to consumption and arguably benefts richer peoplemore than the poor.


38/165

25

3. The economics o taxing labourThe aim o this chapter is to investigate the role o the tax mix in the deter-mination o employment and unemployment. It may come as a surprisethat the answer is not immediately in the a rmative, but depends on thespecifc conditions under which a given labour market works. This chapterwill there ore provide a survey o the theoretical and empirical economicsliterature on the link between taxation and (un-)employment.1 It will applythe key insights in this literature to the question o the tax mix and will givesome empirical evidence on an aggregate basis. The chapter will not onlyprovide in ormation about normative consequences what is an optimaltax mix? but it will also give an account o the economic tradeo s andrestrictions policy-makers ace when choosing tax rates.

The frst section describes the choice menu o di erent theoreti-cal approaches to unemployment, as selecting a particular approachalready entails consequences or the link between taxation and the labour

market. The second section deals with the overall problem o high totaltaxation and unemployment. As will become clear, the theories lead tocontradictory answers once you allow or imper ectly unctioning labourmarkets. The third section extends the survey to issues o the tax struc-ture: progressivity, tax base, and the insurance component. The ourthsection takes a closer look at empirical studies which have been per-

ormed on the impact o labour taxation. It will become clear that one o the crucial empirical issues is that the quantitative response o unemploy-ment towards changes in tax policy is di erent in each country. Some

simple regressions illustrate these problems, and show where to look ordi erences in the impact o the tax mix on labour markets. These fndingssubstantiate the claim that some tax orms matter more than others, andthat this insight especially holds true or sectors with lower productivity.The fnal section summarizes the chapters major theoretical, empiricaland normative fndings.

INTRODUCTION: MODELS OF UNEMPLOYMENT

In the ollowing I could never pretend to provide an adequate summary o all acets o economic approaches to the phenomenon o unemployment.


39/165


Readers with a good understanding o labour economics are well advisedto skip this section. However, a quick sketch o these approaches is inevi-table since the state-o -the-art on the issue o labour taxation is dividedalong many bi urcations o di erent basic models or unemployment.The choice o the model, as is to be shown, is decisive or evaluating theemployment e ects o di erent tax orms. But be ore I deal with these di -

erent approaches a disclaimer seems appropriate.

General Disclaimer

A disclaimer has to be made or what ollows. Most contemporaryapproaches to either labour supply or unemployment, and their relation-ship to taxation, consist o partial micro- oundations or the aggregatesupply side o an economy. They admit that there is a certain level o cycli-cal or demand-side unemployment, but assume that the bulk o the taxe ect lies in the labour market itsel . The ip-side o this implicit assump-tion is, o course, that these orms o unemployment can be easily di er-entiated rom the orm o stable structural unemployment that is due tosupply-side actors. Such a distinction invites critique rom many di erentdirections (Bluestone, 2001; Franz, 1993). In act, most macroeconomistswould stress the role o both aggregate supply and demand shocks or the

evolution o unemployment rates across time (Layardet al ., 1991).Moreover, most o the approaches that will be reviewed have also beencriticized or being partial partial analyses (Atkinson, 1993: 23), in thesense that they not only ocus on the labour market, neglecting generalequilibrium e ects, but also on work instead o wel are. Some labourmarket experts (and politicians) argue that work itsel has an intrinsicvalue (ibid.). Correspondingly, I will ocus on partial equilibrium models,but give some hints when results may be changing because o aggregate

eedback e ects. I will only analyse two out o the three major problems

identifed by the OECD regarding the relationship between taxes and thelabour market (OECD, 1997b): taxes as an unemployment trap and as aproblem o increasing labour costs. Where adequate, however, I will alsobrie y deal with the third, that is taxes as a poverty trap or the low wagesector. This is o importance or cases when employment and wel are con-sequences o changes in tax policies diverge.

Having said this, there is substantial reason to believe that taxation hasthe potential to a ect (un-)employment through the aggregate supply side.As is to be shown taxes are among the strong suspects or cross-countryvariation and longer time periods. The a orementioned caveats obviouslylimit the ability to make analytical and normative generalizations. The

ocus on static versus dynamic issues, employment instead o wel are


40/165

The economics o taxing labour 27

and supply versus aggregate demand issues seems justifed, I am moreconcerned with the reactions o politicians towards problems, and theseare arguably stronger in the short run and in cases where policies directlya ect employment.

A Quick Glance at Diferent Models

To understand how taxation in uences (un-)employment, a quick primerin di erent economic explanations o labour market per ormance is neces-sary. It is necessary because channels, and ultimately, predictions di erdramatically between these approaches.

In the competitive model (CM) the labour market should behave as anyother per ect market, that is it should clear. Firms are proft-maximizingentities, treat labour the same as as any other actor and take wages as given.They will hire workers as long as the marginal productivity o workersequals the (marginal) wage. Wages themselves are set competitively asa result o labour demand and supply. Workers balance their utility o working, that is receiving wages or consumption, and o not working, thatis leisure and out-o -work income such as social trans ers. Taxes createa di erence between nominal wages and real purchasing power. The so-called tax wedge can, but need not have an impact on the input o labour.

It is an empirical question depending on the responsiveness o supply anddemand towards taxation. The usual hunch o the CM is that the primee ect o taxation lies in its potential to reduce employment. I unemploy-ment occurs at all it is o rictional nature and arise as a consequence o transaction costs during the job search or as a consequence o structuralbarriers in the market causing mismatch between supply and demand.

Starting with the latter Mismatch Theory (MT) models a relationshipbetween unemployment and a set o variables accounting or structuralimbalances across economic sectors or social groups.Prima acie, these

approaches merely provide a descriptive image o the situation, unless theyare enriched by an explicit sociological or economic theory o structuralbarriers or instance in the so-called theories o labour market segmen-tation. Alternatively, mismatch approaches recur to theories o the frm,o migration or o e ciency wages or wage bargaining (Layardet al .,1991: 302). Mismatch theories provide an understanding o the problemo heterogeneity among the work orce. Though mismatch is a sectoralphenomenon o unemployment, it necessarily pushes up the overall levelo unemployment (Layardet al ., 1991: 47).

Search Theories (ST) explain turnover and unemployment spells byocusing on the process o taking-up (in ow) and leaving jobs (out ow).

The in ow rate depends on a set o actors exogenous to the theory,


41/165


42/165


unemployment workers would not have to ear the negative consequenceso being caught. Unemployment is a disciplinary device, as it makes shirk-ing more costly (Blanchard and Fischer, 1989: 460). In EWT taxationreduces the net wage o workers and there ore decreases workers e ort.In the short run the EWT is ambiguous about the normative implicationso taxation, since employment and wel are e ects could work in oppositedirections. Yet, in the long run, assuming zero frm profts, tax and wagechanges work in the same direction and lead to higher unemployment(Goerke, 2002: 59).

E ciency wage models assume that frms set wages and employmentlevels. However, in many countries unions may exert an in uence on wagesby bargaining with employers (associations). Wage-Bargaining Theories(WBT) investigate the possibilities or unions rather than frms to bid upwages. As later chapters show WBT is most closely related to politicaleconomy accounts (Calm ors and Dri ll, 1988; Iversen, 1999) so that I willdedicate some more time to their description. Approaches within the set o WBT di er in two crucial aspects: frst, in the utility unction o the unions;second in the structure o the wage bargaining (Oswald, 1982).

As or the ormer, older approaches mainly discuss whether unions ollowa process o maximizing the utility o the pivotal mean or median member.Newer approaches endogenize union membership, since wage bargaining

also has an impact on the incentive or individuals to join or not to join.In essence, the union has to overcome a problem o collective action, sincewage increases o ten cover all workers whether they are members or not.These models include certain benefts such as a reputation e ect, which isonly available to union members. In its most general orm the objective

unction o a utilitarian union may be denoted as (Goerke, 2002: 11)

Utility 5 n(v(w) 1 re) 1 r (l 2 n)v(R) (3.1)

This is a weighted utility o those who are employed (n) the frst termo the RHS and those that are unemployed (l 2 n 5 u). An employedperson derives utilityv rom (net) wages and the reputation e ect o beinga member re. The unemployed benefts rom the reservation wageR,usually some kind o a social beneft. I r , the degree o utilitarianess o the unions objective, equals zero, then the union only represents employedpeople the classic example o pure insider-ship.

Given the utility unction o unions and the proft maximization o employers, the result o wage bargaining can be modelled by the Nashproduct between trade unions and employers associations (see Equation3.2). The Nash-product is a weighted distribution o gains rom bargainingbetween the two sides. Since both sides have a reservation strategy, which is


43/165


not to bargain, both sides must derive some net gains. Firms want to makea proft P 0, whereas unions maximize the di erence between the wagewand the reservation wageR. The distribution, however, might be in uencedby someb, a measure o bargaining power between employers and unions.I b is equal to one, we have a monopoly union dictating the bargaining, atthe other extreme (b 5 0) and the zero-proft condition (P 5 0) we obtainthe result o the CM with zero rent or the union.

NP 5 [(12 r )n(w 2 R)]b(P )12 b (3.2)

Thus, the specifc orm o this product depends on the kind o bargainingbetween employers and unions. In the case where a monopoly union acesmany frms, the union can set both wages and employment levels. In theso-called Right-to-Management models (RTM) the union sets the wageunilaterally (as in the monopoly union case) or bargains over it, but themanagement decides on employment. In this case it is usually assumed thatthere are many unions and frms, and that taxes and trans ers are given.There are other bargaining models (Oswald, 1993), but these are enough toshow that the objectives o unions and the structure o bargaining matter

or taxation. I , or instance, unions care about all employed people, thenthey are likely to include the general e ects o taxation in their own cal-

culations. Otherwise, the interests o insiders are more important and theissue depends on how taxation a ects members relative to non-members.This e ect interacts with the bargaining structure. In RTM the uniononly bargains on wages and the tax e ect will be larger than in modelso a monopoly union which also determines employment. I taxationalso a ects union membership the tax e ect could almost vanish, but it isdi cult to yield general predictions (see below).

As mentioned in the disclaimer, all these models share a number o prob-lematic assumptions about the role o aggregate demand, frms as price-

takers or the importance o wel are state institutions (Layardet al ., 1991:Chapter 7). On empirical grounds it is obvious that none o the di erentapproaches emerges as the best explanation or unemployment (Layardetal ., 1991; Bean, 1994). Yet unemployment shows some persistence, thoughnot necessarily per ect hysteresis, and is there ore, by and large, in linewith a broad notion o insider-outsider models (Heijdra and Ploeg, 2002:205). The act that persistence in some countries is much smaller than inothers adds plausibility to the claim that the institutional structures o labour markets have a deep impact. Given this ambiguity it is probablybest to elaborate the tax e ect on employment with a comparison o di -

erent models predominantly EWT and WBT. This is what the next twosections intend to do.


44/165


45/165


46/165


to go up; second, do higher wages lead to higher labour costs, which frmsare unable to roll-over?

As or the frst issue taxes and wage increases simple versions o ST,EWT and WBT models provide a similar rationale. I both wages and thereservation income are equally taxed, changes in the tax parameters willnot change unemployment rates (Pissarides, 1998). To take one exampleLayard et al . (1991: 108) show the irrelevance o taxation or the case o union bargaining with unemployment benefts as the only available reser-vation wage. In their model, benefts are a fxed proportion o net-o -taxwages, that is with tax parameters changing wages and reservation wageschange equallyR 5 (1 2 t )B . In this case, the Nash product (see Equation3.2) becomes

Wr 5 [wi (1 2 t ) 2 B (1 2 t ) ]bS bi P ei (3.3)

Optimizing this Nickell and Layard (1999) yield the ollowing result:

wi (1 2 t )

wi (1 2 t ) 2 B (1 2 t )5

bh sp 1 (1 2 sp)sp

(3.4)

Here, S i represents the ratio o labour to capital input in production,whereas s

P is the proportion o profts in value added.h is the wage

elasticity o demand or labour, that is how labour demand responds tomarginal changes in wages.sP and h act as weights or the distributiono the Nash product in addition to the bargaining powerb. Since on theLHS o Equation 3.4 (12 t ) may be cancelled out, optimization leads toa frst-order condition independent o t and equilibrium unemployment isuna ected.

This is a very stylized version o a reservation wage, since it abstractsrom in ormal or capital income, and in our context most crucially, rom

other orms o wel are state trans ers. For the low-wage sector, the res-ervation wage might be much more akin to a basic means tested socialassistance. For people with higher skills trans ers or housing, childrenor education are requently excluded rom taxation. In both cases there isno clear relationship betweenT and B , that is there is redistribution. Letsassume thatR depends o two orms o trans ers, one o which is untaxed(B 0):

R 5 B 0 1 (1 t ) B 1 (3.5)

Under such circumstances the terms including taxation o Equation 3.4do not cancel out, and taxes raise unemployment, since they change therelative proportion between wages and reservation wage (see Appendix). As


47/165


trivial as this example may seem, it is not innocuous, since it implies that anysocial security system with a certain degree o redistribution induces a ormo incentive problem through high levels o taxation. Stiglitz (1999) reachesa similar conclusion using an e ciency wage explanation o unemployment.In general, Pissarides (1998) shows that such a result holds or a variety o di erent models, including search, bargaining and e ciency wage models.

O course this model is still crude. In the case o a utilitarian union, orexample, Oswald (1982: 566) shows that the unambiguous e ect o taxeson unemployment only holds or a su ciently high elasticity o labourdemand, or no other unearned income and or workers with relative riskaversion. At least or low-wage workers, there is good reason to believethat these assumptions hold and that taxation has an impact on unemploy-ment or non-competitive labour markets with a system o redistributivesocial trans ers.

To know whether this e ect is not only a necessary but also su cientexplanation o higher unemployment, one has to know whether frmsthemselves are acing higher labour costs because o increasing taxation.This question has created a cottage industry or empirical analyses onlabour demand and the incidence o taxation (OECD, 1990; Hamermesh,1993). The basic insight o this research tradition is that no matter whichmarket side is o cially taxed, the burden itsel alls on the market side with

ewer exit options, or in technical terms: a lower elasticity (Homburg,2003: 155). As with the question o labour supply, the question o the inci-dence o taxation is hence an empirical one. Hamermesh (1993), havingreviewed numerous studies on labour demand, concludes that demandelasticities are typically positive,4 whereas elasticities o labour supplyare much smaller (see previous section). Correspondingly, in the long runmost o the burden alls on workers, either through lower nominal wages

or given prices, or through lower real wages as a consequence o higherprices.5 This shows that the ocus on these three tax orms (see Chapter 2)

is justifed. Yet, shi ting never works per ectly, not even in the long run.Both Hamermesh (1993) and the OECD (1990) conclude in their surveysthat some 60 to 70 per cent o taxes are typically shi ted towards workers,though in some studies shi ting may indeed be higher.6

Finally, there are some actors that make shi ting more complicated.First o all, in a EWT model the shi ting o taxation does not work per-

ectly (OECD, 1990: 154), since taxes reduce e ort and frms may even beinduced via higher taxation to pay higher wages. Shi ting might notwork i there are other labour market institutions, such as a minimum wageor a basic income scheme. Finally, shi ting the tax burden might be entirelydi erent or di erent segments o the work orce. Workers with highly spe-cialized human capital may respond totally elastically, and there ore the


48/165


49/165


50/165


51/165


Even i general consumption taxes have the largest tax base, their e ectsare hard to predict in models with unemployment (Goerke, 2002: 202),since they not only depend on the question whether the reservation wage istaxed, but also on whether frms actually manage to shi t taxes on to prices.I no shi t occurs, a VAT is equivalent to payroll taxes in its employmente ects, i ull shi ting occurs it is akin to (proportional) income taxes. Witha certain degree o shi ting the question o whether VAT is harm ul is anempirical one. For the case o EWT Goerke (2002: 259) states that a switch

rom payroll to general indirect taxes is only benefcial or employment i benefts are not indexed to in ation. For the case o an open economy, ithas been argued that tax competition is smaller or indirect taxes than orincome or payroll taxes, i consumption is taxed where it occurs (Homburg,2003). This is usually the most orce ul argument or shi ting the basistowards indirect taxation (Kato, 2003). I , however, consumption itsel ismobile or elastic, or instance i it disappears into the shadow economy,the case or indirect taxes as a response to international competition losessome o its lustre.

One might also propose a fnal re orm scenario: rom payroll to incometaxation. In act, Atkinson (1995) shows or the case o a segmented labourmarket and a ST ramework that such a switch is benefcial due to its taxbase e ects. Goerke (2002) reaches a similar conclusion via another expla-

nation. Since the reservation wage is either not taxed or is only taxed as aras employees social security contributions and income taxes are concerned,a shi t rom employers to employees taxation decreases unemployment.In this sense Goerke rejects the old Invariance o Incidence Propositionaccording to which it does not matter whether ormally workers or employ-ers have to pay the taxes.

All things considered, the case or indirect taxation as the best taxorm or employment is strong on the basis o e ciency considerations,

in particular in an open economy, but is less so in a non-competitive

labour market. The possibility o exempting sensitive sectors such as thelow-wage sector may prompt calls or a tax mix based on income taxeswith high basic allowances. It remains an empirical question whether thise ect will dominate the general tax base e ect. I indirect taxes are indeedemployment-e cient, we are once again in the traditional e ciency versusredistribution tradeo .

Progressivity and Redistribution

How does progressivity a ect an imper ect labour market? Van der Ploeg,or instance, argues that In a second-best world the distortions arisingrom non-competitive labour markets are partially o set by the distortions


52/165


53/165


market exemplifed by white collar workers and those or a non-com-petitive market such as the one or organized blue collar workers. Thisis also consistent with the fndings o Disneys (2000) survey. He arguesthat labour supply, and unemployment, depends on marginal rates in

our instances: or low income workers i progression interacts with socialbenefts; or high income earners i marginal rates reduce e ort and invest-ment in human capital; or older workers i it increases the probability o them retiring early; and or younger workers i it makes their entry into thelabour market more di cult.

All things considered, the balance o evidence seems to be tilted towardsadvocates o tax progression. O course, liberal economists will alwaysargue that it is better to remove the sources o the distortion rather thantrying to minimize its costs.11 Moreover, the theoretical debate could surelyspin the wheel o policy advice, once worker heterogeneity and long rungeneral equilibrium e ects have been accounted or. Nevertheless, thefnding that progression and employment are reconcilable seems to bolsterthe case o scholars who argue the case or di erent institutional systemsachieving similar employment outcomes. In this respect it is important tonote that progression does not only imply redistribution, but also insur-ance (Sinn, 1995).

Insurance and Contingency

Most orms o taxation entail explicitly or implicitly an insurancecomponent. This is most obvious in the case o payroll taxes defned astax2000 in the OECD revenue statistics including social security contri-butions (see Chapter 2). This contradicts the narrow defnition o the termtaxation. The tendency to use the border between taxes and earmarkedcontributions has led to an overly negative perception o payroll taxes(Disney, 2004). But even Disneys conception o the insurance component

o taxation is only part o the picture, since any kind o tax-based redis-tribution be it in the orm o progression or in the orm o at beneftsand proportional taxes leads to insurance or people whose income isuncertain. Hence the debate on tax-based insurance has two dimensions.First, a narrow one ocusing on earmarked taxes; second, a broader oneon the wel are implications o taxation which holds or all major orms o taxation, as long as there is an element o redistribution attached to it.

Starting with the frst, Disney (2004) shows that the negative impact o payroll taxes decreases considerably i payroll taxes are divided into aninsurance and a tax component, that is those contributions or which thereis no adequate contingent claim on social trans ers or in-kind benefts. Hefnds that more women seek employment when the insurance component


54/165


55/165


fnd ways o maximizing their payo s rom insurance schemes by eithercontributing less, or receiving benefts worth more than their actuarially

air value. The massive trend o early retirement might be read as sucha problem o en orcement (Ebbinghaus, 2002). The crucial problem o apublic insurance system is there ore that it could undermine itsel and withit its employment- riendliness.

Moreover, the insurance issue is truly at the heart o much con usionabout wel are state e ects and, more specifcally, the debate about labourmarket institutions. The question o whether it is rent or insurance seekingthat explains the origins and the maintenance o these institutions, is hardto answer without any prior knowledge (Agell, 2002). One needs to imposemore restrictive assumptions about the pre erences and the behaviour o people to derive normative conclusions about the e ects o institutions.Given these caveats it is di cult to directly compare the insurance e ects o income, payroll and indirect taxes.Prima acie, payroll taxes are attractive

or enhancing employment i they guarantee social benefts. Income taxesalso provide some insurance because o their progressiveness. Insurancee ects should be smaller or indirect taxes only, but all these claims arecontingent on how they interact with the system o social expenditures andthe willingness to invest in human capital.

I have completed the observations on the role o the tax mix in labour

markets and ound instances in which any o the three tax orms may emergeas a winner. We have seen, however, that tax progressivity is not incompat-ible with low-wage employment. To the contrary, direct progressivity alle-viates the tax burden on low-wage earners whereas indirect progressivityexempts income below a certain threshold and thereby shows that it is notalways a larger tax base which is employment- riendly. Insurance, fnally,legitimizes the use o social security contributions, but, again, this is o lessinterest or low-wage earners who need insurance via redistribution. Takentogether, it implies that o all three major orms o taxation, income taxes

are the least detrimental or (low-wage) employment. This is an importanthypothesis I will investigate empirically in the ollowing.

EMPIRICS

Cross-country studies on the impact o taxation on employment have amajor beneft: they account or e ects not visible on the micro level. Themajor downside o aggregate analysis is that there are huge methodologicalproblems in such comparisons. This section frst addresses these particu-lar problems in the hope o improving our understanding o why there isso much diversity in the empirical fndings. Against this methodological


56/165


background, I test the impact o the tax mix on both employment andunemployment. Obviously, I do not want to reinvent the wheel, since somany studies have already been done. But three issues mentioned abovehave received inadequate attention in most o these studies: di erences inthe impact o income, payroll and indirect taxation; the conditionality o fndings on country-specifc e ects; and their particular relevance or low-wage sectors.

Methodological Problems in Empirical Studies

Some recent empirical studies based on international comparisons and theuse o aggregate data reveal a puzzling diversity. Some authors (Daveriand Tabellini, 2000; Volkerinket al ., 2002) fnd that i the tax mix matters,payroll taxes are most problematic, while consumption taxes are gener-ally considered to be less harm ul, and income taxes range somewherein between (OECD, 1995: 69). For others (Nickell, 1997; Disney, 2000;Hutton and Ruocco, 1999) the mix either does not matter or matters di -

erently. In act, there is a remarkable range o di erent fndings that createtoo much ambivalence or use ul policy recommendations. The questionthat arises is why this is the case. Daveri (2001) mentions three problemsprevalent in all empirical studies omitted variable bias, aggregation,

endogeneity to which I will add a ourth: poolability. I will brie y discusseach o them.To begin with omitted variable bias, complex issues such as the con-

nection between taxation and unemployment always create problems o model specifcation. I important variables are le t out, taxes may load highsimply because o their correlation with other variables (Disney, 2000). Yet,truly exogenous variables and correctly specifed models would need moretheoretical input. For example, one may include indicators or individu-als pre erences or public spending. Some scholars have tried to correct

or omitted variables and to account or endogeneity by the use o suchinstruments ( or example Kato, 2003). It is likely that these variables areendogenous to taxation. Such approaches neither do justice to the problemo complementarities between certain types o public institutions, nor dothey give a thorough account o the intricate relationships between taxesand indicators o spending. Other instruments may simply be inadequate.Empirical instruments such as government partisanship are well known inthe comparative politics literature, but are not very good proxies or, say,government expenditure (Kittel and Obinger, 2003). Given these problemsit is probably best not to exaggerate the ear o omissions.

The second problem o aggregation, is perhaps the most mundane, butcertainly not the most trivial. Empirical problems o aggregation loom


57/165


58/165


channels run in both directions and taxation and unemployment eedeach other. Summerset al . (1992), or example, argue that the distortionalimpact o taxation on labour markets is lower in coordinated economies.I this premise is correct, politicians in these countries can raise taxes to ahigher level than in other economies, since the optimal tax rate in terms o employment consequences is higher. There is, according to Daveri (2001),no easy way out o this problem, but there are some possibilities one o which is to endogenize politics in theory (see Chapter 4).

Simple Estimates o Tax Mixes and Employment

Given the previously discussed set o methodological problems, I haveto be very cautious when conducting statistical in erences or aggregatedata on taxation and employment. The method used here is there oreakin to Kittel and Obingers (2003) step-by-step procedure when theyanalysed the link between economic integration and social expenditure. Ibegin this subsection with an analysis o the time-series and cross-sectionproperties o the tax-employment relationship. In the ollowing subsec-tion I estimate pooled data. For all estimations I use three di erent labourmarket indicators: aggregate (business sector) employment rates, sectoralemployment rates and rates o unemployment. All three indicators stem

rom the OECD, but the sectoral rates merit some elaboration. Theserates are generated or each sector as total employment divided by theworking age population. Three sectors have been singled out or theirmarked di erences in skill levels: manu acturing, trade and hotels, andpersonal and social services. Using in ormation on sectoral productivitylevels, workers in manu acturing (ISIC3) 13 belong to the highly quali-fed part o the work orce, whereas workers in trade and hotels (ISIC6)and personal and social services (ISIC9) exhibit markedly lower skillinputs. 14

Time-series resultsA frst step in analysing the tax employment link is to look at individualtime series or selected countries. For the sake o simplicity, the analysiswill ocus on the three role models in comparative wel are state analysis:Germany, Sweden and the US. These countries have been used be ore

or comparative purposes given their well-known institutional di erences(Blundell and MaCurdy, 1999; Esping-Andersen, 1990). Table 3.1 showsthree models or the time series o rate o private sector employment. Themodel includes a lagged dependent variable and two independent variablesto control or di erences in the evolution o the labour markets: unemploy-ment and wages. More importantly, the models contain the three major


59/165


Table 3.1 Time-series results or three countries

Business sector employment to working-age population

GER SWE US

Intercept 40.573*** 38.682*** 6.921(8.109) (5.925) (12.385)

Lagged dependent 0.489*** 0.397*** 0.806***(0.100) (0.092) (0.209)

Unemployment 2 0.853*** 2 0.885*** 2 0.939***(0.066) (0.104) (0.067)

Lag unemployment 0.310* 2 0.002 0.714***(0.139) (0.175) (0.179)

Wage 0.135 2 0.012 0.043(0.086) (0.014) (0.189)

Lag wage 0.073 0.021 2 0.062(0.097) (0.015) (0.193)

Income taxes 0.107 0.037 0.113(0.101) (0.044) (0.094)

Lag income taxes 0.056 0.154** 0.034(0.101) (0.056) (0.105)

Payroll taxes 0.064 2 0.092 0.332(0.247) (0.054) (0.408)

Lag payroll taxes 2 0.811*** 2 0.158*** 0.507(0.251) (0.051) (0.407)Indirect taxes 2 0.197 2 0.481*** 2 0.146*

(0.178) (0.087) (0.079)Lag indirect taxes 2 0.279 0.083 0.011

(0.181) (0.081) (0.458)Adj. R2 1.000 0.998 0.998Nobs. 27 27 27F -Test restricted versus ull (6, 15) 4.5*** 8.75*** 1.25DW-h 2 1.16 2 1.235 2 2.558***

ADF-Test (Intercept, Trend)2

2.7712

3.399***2

4.101***(1,1) (, ) (1, )

Notes:Levels o signifcance: *< 0.10, **< 0.05, ***< 0.01.F -Test compares models with and without tax regressors.DW-h: Durbin-Watson h-statistic or autocorrelation in AR-models. ADF-Test: Augmented Dickey-Fuller test or nonstationarity. Parentheses indicate whereinclusion o trend or intercept was necessary.


60/165


orms o taxation. All exogenous variables enter in contemporaneouslevels and frst temporal lags to allow or some endogeneity.

Table 3.1 shows pronounced di erences in the tax coe cients. The taxvariables exhibit a considerable variation across countries, although theoverall e ect in all three regressions is rather limited. In the case o the US,only indirect taxation has a weakly negative impact on private employment.Taken together, the tax variables do not seem to drive private employmentin the US as theF -test statistics show.15 Sweden is an opposite case to thato the US. Taxation adds to the explanatory power o the model. Moreinterestingly, coe cients or payroll and indirect taxation are negative,whereas income taxation enters with a positive coe cient. For indirecttaxation, it is the contemporaneous level which matters in Sweden, whereasit is the lagged level in the case o payroll taxation. Why income taxationleaves a positive sign is di cult to tell. Since it is the lag o income taxesthat matters, it may be more than pure short-term endogeneity created byeconomic growth. There are good arguments or the case that tax-fnancedpublic goods may enhance growth and employment (Garrett, 1998; Boix,1998), but, as mentioned be ore, our results do not allow us the possibilityto indulge in the question o whether public spending is actually produc-tive. For the purpose o the argument ollowed here, it is sa e to say thatincome taxation seems to be less harm ul than other taxes. In Germany,

fnally, taxation shapes labour market outcomes, though the e ect is pre-dominantly derived rom payroll taxation. I the level o taxation o theprevious period increases by 1 per cent, private sector employment dropsby nearly the same amount. Germany is also an example o countries thathave shi ted the tax mix rom income and payroll to indirect taxation with

ew labour market benefts.The control variables in Table 3.1 show that labour market reactions

di er between the three countries. The test statistics show that autocor-relation and at least or the German case nonstationarity still distort

the results, a act that is in line with fndings o labour economists (Franz,1993: 363). A theoretical explanation o the empirical distortions isthe high correlation between the overall tax burden and indirect taxes.Supplementary time-series analysis (not shown) suggests that the overalltax burden is positively associated with employment only in the Swedishcase but it is weaker than or the three di erent orms o taxation. Hence,although there is some danger o con using the impact o tax mixes withthe impact o overall excessive tax burdens, this danger should not beoverstated.

All in all, the results have important consequences or the subsequentdiscussion. First, country-specifc e ects are crucial in understanding thelink between taxation and employment. Taxation alone cannot resolve


61/165


this problem, even though tax orms di er in their impact on nationallabour markets. Second, country-specifc e ects oster the suspicion thatlabour market regimes di er between these three countries, and that thesedi erences might explain whether taxation matters or not. As expectedby Daveri and Tabellini (2000), tax e ects are presumably smaller in theatomistic and exible US labour market and higher in continental andnorthern European countries. This is, o course, no direct proo or theimpact o labour market institutions on the tax employment link, but itdoes oster the case or this.

Consecutive cross-sectionsWe have seen that the tax-employment relationship di ers over countries,

but does it also di er over time? For this purpose we assume that all coun-tries behave similarly and analyse the cross-temporal evolution or thepooled sample. Table 3.2 shows consecutive cross-section regressions or18 OECD countries over six points in time beginning in 1971 and endingin 1996. I begin with a look at control variables. Di erences in unemploy-ment are signifcant in the explanation o private sector employment. Thisepitomizes the severe employment problems o some countries. It is alsoimportant to understand why the usually positive long-term co-evolution

Table 3.2 Cross-section results or six periods

Business sector employment to working-age population

1971 1976 1981 1986 1991 1996

Intercept 63.11*** 35.38** 56.75*** 51.25*** 56.04*** 50.27***(10.12) (15.31) (10.99) (9.45) (8.49) (10.73)

Unemployment 3.10*** 2.03*** 1.62** 1.31*** 1.42*** 0.82**(0.55) (0.54) (0.32) (0.27) (0.27) (0.28)

Wage 0.29 0.69* 0.36 0.49*** 0.46*** 0.54**(0.24) (0.34) (0.21) (0.15) (0.14) (0.19)

Income taxes 0.51 0.53 0.53* 0.47** 0.21 0.26(0.32) (0.32) (0.25) (0.21) (0.20) (0.21)

Payroll taxes 0.99*** 0.66* 0.64** 0.73*** 0.72*** 0.61***(0.27) (0.31) (0.21) (0.14) (0.17) (0.18)

Indirect taxes 0.09 0.16 0.04 0.09 0.57 0.67**(0.26) (0.34) (0.30) (0.21) (0.28) (0.32)

Adj. R2 0.75 0.23 0.77 0.91 0.93 0.89Nobs. 18 18 18 18 18 18F -Test 10.04** 3.88** 7.38*** 22.48*** 30.19*** 18.73***

Note: Levels o signifcance: *< 0.10, **< 0.05, ***< 0.01


62/165


o wage and employment is only visible or some years. In 1996 higherwages even seemed to reduce employment rather than increase with it.Income taxation is not o signifcant importance or employment levels.As ar as the other two sources o taxation are concerned, a temporalpattern seems to be suggestive: throughout most o the period investi-gated, payroll taxes exert a negative impact on private employment; inthe 1990s, indirect taxes also show a negative impact on employment orthe frst time.

Nevertheless, the evidence is not su cient or corroborating the hypoth-esis that there are strong cross-temporal di erences. One reason is thelimited robustness o the results. Finland and Japan are important outliersbecause the model underestimates their employment per ormance. Inaddition, Australia and New Zealand are signifcantly below the averageo the model. Additional variables do little to improve the situation.Wage inequality as measured by the OECD (1996), or instance, does nothelp greatly in accounting or these outliers. It is interesting to note thatantipodean wel are states per orm worse than the average, although bothAustralia and New Zealand rely on income taxation as the most importantsource o unding or their wel are states. Clearly, this deviation cannot beexplained by tax mixes. New Zealand and Australia share high levels o targeting (Korpi and Palme, 1998) as well as a comparatively low degree

o progressiveness in income taxation. In New Zealand, in particular, per-sonal income tax does not exempt low-wage workers (Gangho , 2006: 9).This could mean that the implications o income taxes or labour marketsin these countries are more akin to those o payroll taxes in other countries(see Messere, 1993).

Several cross-sectional tests have been per ormed to control or theoverall tax burden. This could explain the deviant position o Japan which is the country with the lowest overall tax burden in the sample butnot that o other outliers. Finally, regressions with relative tax shares, i.e.

individual tax orms as a percentage o total taxation, reveal equal or evenstronger results or payroll and indirect taxes. It seems plausible, there ore,to assume that the tax mix matters even once controlled or the overalltax burden.

All things considered, the regressions show a role or the tax mix. It isobvious that econometric problems and sample selection loom large in thefndings. Country e ects are plainly visible once time-series are compared.This fnding is reasonable given the discussion o di erent theoreticalmodels o labour markets. Taxes matter di erently in di erent labourmarkets, and their in uence over the persistence o unemployment is alsodi erent. Hence, I have to be very cautious when I proceed to the presenta-tion o results or pooled data.


63/165


64/165


sensitivity analyses show that the three tax orms maintain signifcance even

i in uential cases such as Ireland are excluded. It is important to stress thatthe two equations presented here measure frst-round e ects only. Little issaid about indirect e ects via lower GDP growth or changes in the capital labour ratio.17All things considered, the idea that payroll and indirect taxesmatter more than income taxes is justifed on the grounds o these models.Income taxes exert only a moderate in uence on both employment andunemployment. Payroll and indirect taxes matter much more drastically orlabour market per ormance at least in the short-term perspective.

Pooling sectoral dataThe last question o interest in this section is to show whether there aresectoral di erences or the impact o taxes on employment. A question o

Table 3.3 Pooled results or employment and unemployment

Business Sector Employment(First Di erence)

Unemployment(First Di erence)

Intercept 2.986*** 2.673 ***(0.917) (0.576)

GDP growth 0.700*** 0.231(0.165) (0.178)

Wage 0.025*** 0.044(0.007) (0.054)

Income taxes 0.069 0.076(0.152) (0.127)

Payroll taxes 0.511* 0.470**(0.313) (0.266)Indirect taxes 0.574*** 0.534***

(0.202) (0.221)Adj. R2 0.48 0.31Nobs. 85 85Fixed E ects Time, Country TimeF -Test (no FE) 2.87*** 4.35***Condition Index 8.79 12.699AR(1), N(0,1) 1.32 0.37

Notes:Levels o signifcance: *< 0.1, **< 0.05, ***< 0.01.Observations are frst di erences o fve-year averages; 17 countries, fve periods.Coe cients o Fixed E ects (FE) not shown.Panel-corrected standard errors in parentheses.Condition index or multicollinearity.AR(1), N(0,1): asymptotic test or serial correlation in residuals.


65/165


major interest posed in the previous section dealt with di erences betweenhighly productive sectors and those with low levels o productivity pro-ducing (non-tradable) goods and services. I start with the two sectors

manu acturing (ISIC3) and trade, restaurants and hotels (ISIC6). Again,the underlying assumption is that productivity levels are considerablyhigher in ISIC3. As in the previous case, the data have been trans ormed

rom annual observations to frst di erences in fve-year averaged data.The exogenous variables are the same with the exception o wages whichare now industry-specifc and operationalized as total compensation o (private sector) employees in the sector relative to GDP.

Table 3.4 shows the result o pooled regressions or both ISIC3 andISIC6. Whereas fxed time e ects were necessary in the case o manu-

acturing employment, the equation with service sector employment asthe dependent does not include any fxed e ects. The table shows thatemployment in both sectors is driven by growth, whereas wages only

Table 3.4 Pooled results or sectoral data

Business Sector Employment Level

Employment in ISIC3(First Di erence)

Employment in ISIC6(First Di erence)

Intercept 2.002*** 0.070(0.775) (0.255)

GDP growth 0.559** 0.239***(0.261) (0.088)

Wage 0.311 0.468**(0.264) (0.246)

Income taxes 0.061 0.018(0.117) (0.053)

Payroll taxes 0.260 0.145***(0.179) (0.059)

Indirect taxes 0.124 0.127(0.325) (0.143)

Adj. R2 0.64 0.45Nobs. 56 56FE Time noneF -Test (no FE) 2.407*Condition index 10.34 6.82AR(1), N(0,1) 1.49 0.82

Notes:Levels o signifcance: *< 0.1, **< 0.05, ***< 0.01.14 countries, our periods (fve-year averages).


66/165


67/165


and that the insurance component, symbolized by the relationship to ben-efts B , is highest or payroll taxes, ollowed by income and indirect taxes.The table shows a stylized rank ordering o the three tax orms in termso employment e ects. For example, the competitive model (CM) holdsthat the tax base o indirect taxes (Z ) is best or employment, ollowedby income (T ) and payroll (S ) taxes. In general, the CM seems to endorseindirect taxation, whereas the wage-bargaining model puts income taxes inthe most avourable light. E ciency models produce results that are most

di cult to summarize and lead to no clear winner.Taken together, three points merit repetition. First, in a second-bestworld progressivity is not a bad thing. The reason is that it decreases wagepressure, especially i one assumes a wage-bargaining model. One must becare ul with its normative implications, since the employment e ects arepositive only under specifc theoretical and empirical assumptions, butthese are not unrealistic. Next, in a second-best world a broad base maynot always be a good thing. Only in the case o the competitive model it isclear that general consumption taxes should be less harm ul. Exempting

sensitive sectors such as the low-skilled could theoretically improve labourmarket per ormance in wage-bargaining and e ciency-wage models.Third, even rom an insurance perspective contribution-fnancing is notnecessarily pre erable or the low-wage sector. Insurance via redistributionis a more viable option or poor people.

The empirical section has tested the relationship o tax orms on employ-ment outcomes. There is evidence that payroll and indirect taxation havemuch more visible and detrimental e ects on taxation than those o incometaxation. The cross-country results are not in allible, but relativelyrobust. This pattern depends on country di erences much more than ontemporal evolutions. And most importantly, this pattern is more obviousin the so-called low-wage sector, as illustrated by ISIC6. On basis o this

Table 3.5 A synopsis o theoretic results

Tax Base Progression Insurance

Z (x) . T (x) . S (x)

T 9(x) . S 9(x) . Z 9(x)

S ~ B . T ~ B . Z ~ B

CM Z f T f S Z f S f T S f T f Z WBT T f Z f S T f S f Z ?EWT ? Z f S f T S f T f Z

Note: Rank ordering symbol f shows tax orm with best labour market outcome.Inequality symbol. shows tax orm with a bigger base, a higher progressivity or moreinsurance.


68/165


69/165


14. For a detailed descriptive analysis see Kemmerling (2003).15. For all three regressions we run a model without the tax variables and compared it to

the model which includes them. TheF -test shows whether there is a systematic di erencebetween the two.

16. As stated in Chapter 2, aggregate tax ratios to be analysed with caution (OECD, 1999b:28). In particular, since tax revenues are expressed as proportions o GDP, they areprone to problems o reverse causality. A decrease (increase) in employment leads to adecrease (increase) in revenues and will automatically induce a perverse sign or regres-sions with employment or unemployment as the dependent variable.

17. In an accompanying paper (Kemmerling, 2005), I also dealt with the problem o cross-temporal instability o the tax coe cients. The hunch was that this problem probablyarises due to deepening international economic integration. The results were in line withthis claim.

18. For a discussion o the role o international economic integration see Wood (1994) orBlanchard and Wol ers (1999).


70/165


71/165


72/165

59

4. Political economy applied to taxmixes

In Chapter 2 we saw that the relative importance o income taxes romcountry to country, and that there is also a switch away rom income taxesin most countries. In Chapter 3 I argued that this trend against progressiv-

ity was not to the beneft o employment. There is evidence that the switchaway rom income taxation hurt employees with low wages more thanothers. I policy-makers are pure social planners, they should have reversedthis trend. The question arises when governments have the capacity tore orm their tax systems and why they want to do this. The prime task o this chapter lies in delineating the political motives or labour taxation inmore detail by paying attention to the key political con icts and

Date post:	14-Apr-2018
Category:	Documents
Upload:	zhapull
View:	230 times
Download:	0 times

Taxing the Working Poor

Documents