tax reform of the century—the swedish experiment · 2019. 4. 11. · slogan “tax reform of the...

SYMPOSIUM: WHAT CAN TAX REFORM DELIVER?

643

TAX REFORM OF THECENTURY—THESWEDISH EXPERIMENTJONAS AGELL, * PETER ENGLUND, *

& JAN SÖDERSTEN *

Abstract - What can changes in taxstructure accomplish? The Swedish taxreform of 1991 is the most far-reachingreform in any industrialized country inthe postwar period. It represents athorough application of a strategy ofrate cuts cum base broadening, and ithas affected a myriad of economicincentives in a more or less substantialway. This paper reviews the lessons froma major evaluation effort, sponsored bythe Swedish government and involving alarge number of researchers.

INTRODUCTION

Sweden might be best known as thehome of film director Ingemar Bergmanand—for better or for worse—as theprototype welfare state. What might beless well known is that Sweden recentlyimplemented the most far-reaching taxreform in any western industrializedcountry. Although Sweden was alatecomer to the bandwagon of world-wide tax reforms of the 1980s, with theU.S. Tax Reform Act of 1986 (TRA 86) asa celebrated example, the architects of

the Swedish tax reform of 1991 (TR 91)applied the strategy of rate cuts cumbase broadening in an unusuallythorough manner. Under the catchyslogan “tax reform of the century,”marginal income taxes were dramaticallylowered and various tax shelterseliminated. According to prereformestimates, the rate cuts entailed arevenue loss on the order of six percentof gross domestic product (GDP).Measured in this way, TRA 86 standsout as a relatively modest endeavor,with a projected revenue loss of one totwo percent of GDP due to rate cuts.

It goes without saying that sharplyreduced income tax rates represent asensitive political issue, particularly so ina country where tax policy used tocenter on the idea that a steeplyprogressive income tax is an efficientway of transferring resources from therich to the poor. To understand whyTR 91 gathered wide political support,both external and internal consider-ations are important. The integration ofworld capital markets during the 1980simplied that it became more difficult totax capital income at rates that differedvery much from those applicableelsewhere. Also, the major changes inthe tax structures of a number of other

*Department of Economics, Uppsala University, S-751 20

Uppsala, Sweden.

NATIONAL TAX JOURNAL VOL. XLIX NO. 4

644

countries provided an impulse. However,it would be wrong to view TR 91 as amechanical response to these develop-ments. After all, Ronald Reagan’s taxpolicy is not a very natural source ofinspiration for Sweden’s Social Demo-crats. While many of the arguments ofthe proponents of TR 91 bear a strikingresemblance to those put forth in thecontext of TRA 86,1 their origin shouldbe traced primarily to the domesticdebate.

As early as 1978, Nobel LaureateGunnar Myrdal complained that highmarginal tax rates had turned Swedeninto a “nation of wanglers.” Accordingto Myrdal (1978), the progressiveincome tax had created such strongincentives for high-income individuals toexploit various tax avoidance schemes(including outright tax fraud) that theSwedish tax system no longer redistrib-uted income. Myrdal’s view carriedparticular weight since he belonged tothe political left. Parties more to theright, and quite a few economists, hadfor long warned that the income taxcreated large disincentive effects. ButMyrdal seemed to suggest that alsothose in favor of egalitarian outcomes,and concerned less about efficiency, hadreason to reconsider the role of theprogressive income tax. AlthoughMyrdal offered no hard facts to substan-tiate his claim, the perception that therich could avoid their fair share of thetax burden was probably instrumental insoftening the Social Democrat’s tradi-tional resistance to proposals involvinglower marginal tax rates.

But there were also other influentialarguments. Since the late 1970s, wheninflation reached double-digit levels,there was widespread concern that thetax system promoted the wrong kind ofinvestments. Investments in noncor-porate assets, and housing in particular,

were given preferential tax treatment.Many Swedish economists argued, asdid many of their colleagues in theUnited States, that the uneven playingfield created substantial efficiencylosses. The nonuniform treatment of thereturns on different assets also createdconsiderable scope for a number ofstraightforward tax arbitrage operations,more often than not involving purchasesof low-taxed assets with borrowedmoney. This spelled bad news for thetax collector—for several years, house-holds claimed tax deductions to such anextent that the net revenue from taxeson personal capital income was nega-tive.

The corporate income tax attractedsimilar criticism. A variety of tax allow-ances, and a high statutory tax rate,were once part of a deliberate strategyof stimulating firms to plow back profitsinto their businesses. The idea was thatlarge and expanding firms were goodfor growth. During the 1980s, there wasa shift in emphasis. According to thenew view, the corporate tax breaks werean obstacle to efficient capital alloca-tion. The high rates of profit retentionrequired to take advantage of thevarious tax allowances created a capitallock-in effect, which prevented neces-sary structural readjustments. Finally,toward the mid-1980s, an overheatedeconomy and concerns about an acutelabor shortage seemed to give addedweight to the argument that there wasan urgent need to strengthen laborsupply incentives.

TR 91 was presented as a way ofremedying all of these problems in onegiant stroke. According to its propo-nents, the reform would avoid theclassical goal conflict between efficiencyand income distribution. In spite ofdrastic marginal tax cuts, high-incomeearners were not supposed to gain


645

15 percent was to be financed byelimination of loopholes and preferentialrules for taxing earned income. Enlargedtax bases due to a generally moreefficient economy would offset roughlyfive percent of the revenue loss. Thereform hence brought a major realloca-tion of the total tax burden away fromearned income to consumption and toindividual capital income (including thereturn to housing).

A noteworthy feature of TR 91 was themove away from the principle of globalincome taxation toward a dual incometax, by introducing separate tax sched-ules for earned income and capitalincome.3 The new taxation of earnedincome meant that almost 85 percent ofthe income earners would pay only localincome tax. In 1991, the countrywideaverage of the local income tax was 31percent. A national income tax of 20percent was imposed for incomesexceeding 185,000 kronor (equivalent toU.S. $33,500, at the 1991 exchangerate), which meant that the top marginaltax rate on earned income was set at 51percent. TR 91 implied that the marginalrate would be reduced by between 24and 27 percentage points for largegroups of full-time employees (Figure 1).

The new proportional capital incometax was set at 30 percent, and levied ondividends, interest income, and bothlong- and short-term realized nominalcapital gains. As before, interest on allkinds of debt would be fully deductible.A stated purpose of the separate capitalincome tax was to reduce the value ofinterest deductions and to limit thescope for tax avoidance in variousforms.

Although the initial ambition of TR 91was to levy a uniform VAT (of anunchanged 23 percent) on all commer-cial turnover of goods and services,

relative to other groups. As a result of agenerally more efficient economy, allstrata in society should gain. Moreover,total tax revenue was not supposed todecrease—the Swedish tax take shouldremain the highest in the world. Therevenue gains from broader tax baseswere to make up for the losses fromlower tax rates.

What can changes in tax structureaccomplish? Did the proverbial freelunch materialize? This paper reviewsthe lessons from a major evaluationeffort, commissioned by the Swedishgovernment and involving a largenumber of foreign and Swedishresearchers.2 The next section gives abrief outline of TR 91. We proceed bydiscussing the considerable difficulty inevaluating a tax reform in the midst of avery sharp recession. We review theevidence on behavioral responses, andwe seek to identify areas where TR 91mattered the most. We also present anassessment of the overall effects onefficiency and equity. Our concludingremarks draw some lessons for taxreform more generally.

WHAT TR 91 IMPLIED

Prereform estimates presented by theMinistry of Finance indicated that therate cuts for the personal income taxalone, together with additional outlaysfor housing and child allowances tocushion the distributional effects of thereform, would reduce revenues by anamount equivalent to between six andseven percent of GDP. Nearly 40 percentof this loss was to be recouped througha new system of taxing capital income.A broadening of the value added tax(VAT) (of 23 percent) to include goodsand services previously exempted, orgranted lower rates, would yieldadditional revenue on the order of 30percent of the budget loss, while almost


646

several areas have remained tax exempt.These include various cultural and socialservices and housing rents. Housingcosts have risen as a result of the taxreform, however, partly because the VATwas broadened to include real estatemaintenance, heating, and electricity.Owner-occupiers were further affectedby the inclusion of expenditures onhousing investment in the new VAT.Before TR 91, the tax was 13 percent.

The changes in the corporate incometax were no less dramatic. The statutorytax rate was reduced from 57 to 30percent. In order to maintain anunchanged level of revenue from thecorporate tax, the rate reduction wascombined with substantial broadeningof the base. The possibility to under-value inventories for tax purposes waseliminated, and the investment fundssystem, introduced in the mid 1950s asthe main tool of a countercyclical fiscalpolicy, was discontinued.

Even a casual comparison revealsobvious similarities between TRA 86 andTR 91. Both reforms were far reachingwith the same intent of reducing variousbehavioral distortions from the taxsystems. The approach to tax reformwas similar, combining substantial taxcuts with a broadening of the base.TR 91 was much larger in scope, both interms of revenue effects from the ratecuts and by covering a wider range oftax instruments. Obviously, thesedifferences partly reflect the vastdifference between Sweden and theUnited States in the financing require-ments for the public sector. As aproportion of GDP, the total tax yieldwas almost 56 percent in Sweden in1989, compared to 30 percent for theUnited States. However, the relativeimportance of different taxes, summa-rized in Table 1, is broadly similar. Inboth countries, there is a heavy relianceon receipts from the personal incometax and from Social Security contribu-

FIGURE 1. Marginal Tax Rate 1989–91 at Different Levels of Tax Assessed Income; 1 SEK = 0.181 U.S. Dollars (1991)


647

tions. Taxes on goods and services aremore important in Sweden, whereascorporate and property taxes contributea larger share of tax revenue in theUnited States. The reallocation of thetotal tax burden brought by TR 91, awayfrom personal income to consumption,is clearly visible.

A NATURAL EXPERIMENT?

A major reshuffling of the tax structuremay seem like that rare opportunity tosharpen the estimates of the behavioralelasticities that would interest a publicfinance economist. However, theimplementation of TR 91 was accompa-nied by the most severe economicdownturn since the 1930s. Between1991 and 1993 GDP fell by more thanfive percent, open (i.e., excluding thoseenrolled in various labor marketprograms) unemployment rose from lessthan two percent to more than eightpercent, asset prices tumbled, andresidential construction activity came toa virtual standstill. At the same time,income inequality increased and thegovernment deficit reached recordheights. In this environment, it is clearlyhard to sort out cause and effect withmuch precision.4 Macroeconomic timeseries for the years surrounding TR 91

contain very little of the kind of inde-pendent information needed to discrimi-nate between alternative hypothesesabout behavioral responses.Microeconomic panel data, allowing theanalyst to control for individual variationover time, may seem like a safer bet, butthere is still an obvious risk that themacroeconomic noise pollutes themicroeconometric analysis.

The Swedish tax evaluation effortincludes several contributions which tryto make do with data sets including pre-and postreform observations. But themacroeconomic turmoil has led anumber of researchers to confineattention to the information that can begathered from an ex ante evaluation ofTR 91. These evaluations have twoingredients in common. The first is anassessment—based on previous findingsreported in the literature or on newfindings based on prereform datasets—of the reasonable value of somerelevant behavioral parameter. Thesecond ingredient is a careful assess-ment of the change in the incentivestructure implied by TR 91. This is a farfrom trivial exercise, since TR 91contained several provisions withcounteracting influences on the relevantincentive margin. Combining these

TABLE 1SOURCES OF TAX REVENUE IN SWEDEN AND THE UNITED STATES

Share of Total Receipts (Percent)

Sweden U.S.

1989 1991 1991

Taxes on personal incomes 39.3 34.2 34.9(including capital gains)

Taxes on corporate incomes 3.8 3.1 7.3Social Security contributions 26.7 28.6 29.8Payroll taxes 2.5 3.0 —Property taxes 3.3 4.1 11.2Taxes on goods and services 24.2 26.9 16.8Miscellaneous taxes 0.1 0.1 0.0Total receipts (percent) 100.0 100.0 100.0Share of taxes in GDP (percent) 55.6 52.7 29.5GDP (Billion U.S. 1991 dollars) — 237 5,611

Source: Revenue Statistics of OECD Member Countries 1965–94 (OECD, Paris, 1995).

Revenue Source


648

ingredients, the ex ante evaluationproduces a set of predicted responses,which serve the useful purpose ofscrutinizing the political arguments. Dorecent research and refined analysis ofthe incentive structure produce the kindof behavioral responses promised by thearchitects of TR 91?

An additional complication stems fromthe fact that the sharp drop in theactivity level cannot be treated asindependent of TR 91. Toward the endof the 1980s, the Swedish economyshowed many signs of overheating.5

Due to a high demand pressure, wagecosts increased substantially, which ledto eroded competitiveness in worldmarkets. At the same time, and in thewake of financial deregulation, theindebtedness of households and firmsreached very high levels. This left theSwedish economy in a vulnerableposition when the internationaleconomy slowed down in the early1990s. While the international recessionmay explain part of the rise in unem-ployment, a more important reason forthe severity of the recession is probablythat macroeconomic policy was firmlydevoted to nonaccommodation. As hadbeen the case previously in many otherWestern European countries, fightinginflation and defending an overvaluedexchange rate now became the toppriority for Swedish economicpolicymaking. As a result, manufactur-ing output fell sharply.

These developments, certainly unrelatedto TR 91, explain a major part of thesevere economic recession. But TR 91may help to explain why the slump inthe exporting sectors spread to thesheltered parts of the economy. Bylowering the value of household interestdeductions, TR 91 increased real after-tax borrowing costs and stimulatedindebted households to sell off assets.

Also, higher borrowing costs, inconjunction with certain tax hikes aimedspecifically at the housing sector,contributed to the collapse of theconstruction sector. Thesecontractionary impulses amplified thedownturn and contributed to a generalweakening of aggregate demand.6

James Tobin once cautioned that it takesa heap of Harberger triangles to fill anOkun gap. Whatever its long-termmerits, it is easy to conclude that thetiming of TR 91 was unfortunate. Whilepart of this can be blamed on bad luck,TR 91 was designed to be implementedin a situation with fair macroeconomicwind. The official inquiry which precededTR 91 contains a wealth of material, butthere was hardly any discussion of themacroeconomic aspects.

BEHAVIORAL RESPONSES

As suggested by Slemrod (1992), theevidence about the responses to themajor U.S. tax reforms of the 1980s canbe interpreted as an indication of abehavioral hierarchy. The most respon-sive decisions, at the top of the hierar-chy, are those involving the timing oftransactions, followed by a variety offinancial, accounting, and evasionresponses. The least responsive deci-sions, at the bottom, concern the “real”ones, including labor supply, savings,and investment. Let us review theSwedish evidence.

Tax Avoidance

A characteristic feature of the pre-TR 91tax system was the highly nonuniformtreatment of income from capital. Thereturns to different assets were taxed atvastly differing rates, and the tax paidon a given income could vary systemati-cally depending on the identity of thetaxpayer, the kind of income concerned,and when it was reported for taxation.


649

The steeply progressive tax schedule, incombination with the treatment ofchildren as separate taxpayers, meantthat the tax burden could be reducedconsiderably by shifting capital incomefrom parents in high tax brackets tochildren with little or no earned income.Likewise, differences among corporatefirms in availability and possibilities totake advantage of various nondebt taxshields were extensively exploited. Onseveral occasions, schemes to avoid taxwere set up as joint ventures betweentaxpaying private corporations and taxexempt institutions in the public sector,including several Swedish cities (theoperations included sale and leasebackof icebreakers, municipal sewagesystems, hospital equipment, etc.).

In Sweden as elsewhere, the strategy ofthe tax planners was to claim deductibleexpenses against fully taxable incomeand report income in forms grantedpreferential tax treatment. The conces-sions to specific forms of householdsavings in the late 1970s and early1980s provide a simple illustration ofthis. Savings in special bank accountswere offered a tax-free return plus aninitial tax credit, which effectivelyimplied a negative tax. However,nothing prevented households frompaying for the contributions to thescheme by borrowed money anddeducting the interest against earnedincome. The special bank account couldhence easily be transformed into amoney machine. Private pension plans,which received the equivalent ofconsumption tax treatment, providedsimilar opportunities for reaping thedouble benefits of tax exemption andfull interest deductibility.

As in other countries, much effort wasdevoted to transforming corporatesource income into low-taxed capitalgains. Complicated schemes were set

up, often involving the use of newfinancial instruments, where the taxlegislation was either unclear orpreliminary in nature. The capital gainstax provided additional opportunities fortax avoidance. Before TR 91, short-termgains on shares held for less than twoyears were fully taxed, while on long-term gains, the tax rate was 40 percentof the income tax rate. The holding-period distinction thus implied that ashort-term loss of one krona could beused to offset the tax on a long-termgain, two and a half times as large.

The decades preceding TR 91 witnesseda continuing battle between the taxplanners and the tax authorities. Newand increasingly complicated rules wereset up to combat tax avoidance, and atleast to some extent the added com-plexity created new and unforeseenopportunities to avoid tax. Rather thancontinuing along the same path, TR 91cut the Gordian knot by focusing on thecircumstances that had opened up thetax planning, that is, the asymmetric taxrules and the high tax rates. Familytransactions set up to exploit differencesin the marginal tax rates betweenparents and children were renderedmeaningless, since individual capitalincome was taxed at a proportional ratewith no exemption. The uniform tax oncapital income also meant that the taxcould be levied at the source.7 Transac-tions among firms driven by differencesin taxpaying status were less rewardingbecause of the lowered statutory taxrate, while the broadening of the taxbase reduced the differences in availabil-ity of tax shields between companies.The elimination of the holding-perioddistinction for capital gains meant afurther blow to tax planning.

The balance between taxation of laborincome and corporate source capitalincome was much discussed in the tax


650

reform process and during the followingyears. TR 91 retained a difference ofsome 15 percentage points between thetotal (including payroll taxes) marginalrate on earned income and the totalcorporate and personal tax on profits.Hence, for owners of corporate firmsactive as managers, there remained aclear incentive to reclassify labor incomeas corporate profits. While a thorough-going attack against tax avoidancewould seem to require that this incen-tive be eliminated, the tax legislators doin fact face a difficult dilemma posed bythe increased openness of the Swedisheconomy. Following the deregulation offinancial markets and the elimination ofcurrency controls at the end of the1980s, Swedish households have bothlegal and easy access to internationalportfolio investments. Though taxationof individual income from capital followsthe residence principle, Swedish taxauthorities in practice have littleopportunity to enforce taxes on incomeearned abroad. The prospect of intro-ducing a functioning reporting systemfrom foreign banks and/or a system ofwithholding taxes on dividends andinterest payments seems bleak indeed.If, on the other hand, the tax legislatorswould attempt to keep domesticinvestors at home by more lenienttaxation, the difficulties of defendingthe line of demarcation against laborincome would magnify.

Consumption

Private consumption is normallyconsidered to be one of the least volatileof all macroeconomic variables. InSweden, however, consumption fell byabout five percent between 1991 and1993.8 Can changes in tax structureexplain the consumption bust? TR 91may have had a negative impact onconsumption via three channels: (1)intertemporal substitution in response to

an increase in the real after-tax interestrate,9 (2) a downward revision ofexpected future labor income, and (3)wealth effects due to reevaluations inasset markets, in particular, in themarkets for residential real estate. Ofthese channels, only (3) seems to holdsome promise in explaining the con-sumption bust.

With a large intertemporal elasticity ofsubstitution, periods of high expectedinterest rates should coincide with rapidconsumption growth, and periods of lowinterest rates with stagnant consumption(e.g., Hall, (1988)). However, althoughthere are reasons to be cautious aboutthe information that can be drawn fromrepresentative agent models andaggregate data, empirical studies inSweden and elsewhere suggest that theintertemporal elasticity of substitution isclose to zero.10 Moreover, some basicaspects of the data are hard to reconcilewith a story of intertemporal substitu-tion. Between 1986 and 1989, whenconsumption growth was brisk, theaverage real after-tax interest rate wasnegative. Between 1991 and 1993,when consumption growth was nega-tive, the average real after-tax interestrate was exceptionally high. Ifintertemporal substitution is to charac-terize the data, consumption growthshould follow the opposite pattern.

When households make a downwardrevision of their forecasts of future laborincome, consumption should fall.However, while there are good reasonsto believe that permanent labor incomefell during the Swedish consumptionbust,11 TR 91 is not likely an importantfactor. Indeed, most assessments of theefficiency effects of TR 91—discussedbelow—suggest that it lowered excessburdens due to tax wedges in the labormarket. If anything, this effect shouldincrease permanent income.


651

Capitalization effects in asset markets,whether due to TR 91 or somethingelse, may certainly affect consumption.The consumption boom of the late1980s was associated with an increasein real estate prices, and the bustperiod with decreasing prices. Ofcourse, correlation is not the same ascausation. But recent macroeconometricwork indicates that variables such as realhousing prices and windfalls in thehousing market seem to have power inexplaining aggregate Swedish consump-tion behavior.12 Moreover, TR 91 led—as discussed below—to a dramaticincrease in the rental cost of housingand contributed to the sharp fall inhouse prices after 1991. Although thesehousing market adjustments are unlikelyto explain a very large part of theconsumption bust, they seem like themost important mechanism for anadverse consumption response toTR 91.

Consumption Pattern and AssetComposition

While TR 91 may have had a smalleffect on the aggregate consumptionlevel, it mattered more for the composi-tion of consumption and savings. Themove toward a broader base for theVAT implied substantial tax hikes forsome previously favored consumptioncategories. Available data and a feweconometric studies suggest that therewas a strong negative demand responsefor some of these categories, includinghotel and restaurant services anddomestic tourism.13 These responsesserve as a healthy reminder of the factthat one of the guiding principles ofTR 91—the purported superiority of asystem with uniform tax rates on goodsand services—has no obvious connec-tion with the tax structure implied bythe “inverse elasticity rule” of models ofoptimal commodity taxation.

Our prior was that household assetchoice—at the top of the responsehierarchy—was an area where TR 91should matter. Microeconometricevidence suggested that the old taxsystem created strong tax clienteleeffects, in the sense that householdstended to specialize in assets accordingto their marginal tax rate.14

Macroeconometric evidence suggestedthat the large shifts from financial tononfinancial savings outlets (e.g.,consumer durables and housinginvestment) during the decades preced-ing TR 91 were highly correlated withafter-tax returns.15

TR 91 removed many of the asymme-tries on the asset side of households’balance sheets. The 30 percent tax rateon personal capital income mitigatedthe tax disadvantage of bank savingsand reduced the tax premium forinvestments in durables and real estate.There were also important conse-quences for the treatment of liabilities.In the early 1980s, nominal interestexpenses were fully deductible againstoften quite high marginal income taxrates. This created a strong incentive toinflate balance sheets by purchasingassets with borrowed money. In 1980,an individual with the marginal tax rateof an average blue-collar worker paid areal after-tax interest rate of minusseven percent. In 1991, when nominalinterest expenses became deductible atthe new flat tax rate of 30 percent andinflation was lower, the same individualpaid a real after-tax interest rate of plusseven percent.

On balance, TR 91 gave householdsstrong incentives to shift from real tofinancial savings outlets and to shrinkbalance sheets by selling off assets andamortizing debt. Households seem tohave adjusted accordingly. In the early1990s, net investments in tangibles and


652

durables turned negative, and house-hold financial savings increased dramati-cally. Between 1988 and 1992, netlending as a share of disposable incomeincreased by an astonishing 13 percent-age points, while the nonfinancialsavings ratio decreased by more than 8percentage points.16 Although the rapidfall in inflation and the macroeconomiccrisis played a role, calculations reportedin Agell, Berg, and Edin (1995) indicatethat about one-third of the increase inthe net lending rate can be attributed toTR 91.

Housing

Valuing and taxing the services renderedby owner-occupied housing consti-tute a classical difficulty in implement-ing a comprehensive income tax. InSweden, this has been handled byimputing a measure of the implicitincome from an owner-occupied homeat one to two percent of market valueand adding this imputed income on topof other taxable income. With nominalinterest income being fully taxable andinterest payments deductible, thisintroduced substantial asymmetries inthe tax treatment of housing and othersectors of the economy, and within thehousing sector itself, especially with thehigh inflation and nominal interest ratesof the 1970s and 1980s. Owner-occupied housing was tax advantagedrelative to other forms of householdinvestment, user costs of owner-occupied housing varied across house-holds according to marginal tax rates,and owner-occupied housing had ageneral tax advantage17 over rentalhousing. Apart from the provisions ofthe income tax code, housing was alsofavored through government-guaran-teed interest-subsidized loans, a lowerVAT rate for building and construction,and income-dependent housingallowances.

As a reflection of the progressivity of thetax schedule, with widely differingmarginal tax rates, after-tax interestpayments varied a lot across house-holds. Before TR 91, this resulted inrental costs of owner-occupied housingranging from two percent in the uppermarginal tax bracket up to six percent atlow marginal tax rates, according tocalculations discussed in Englund,Hendershott, and Turner (1995). Withthe new flat 30 percent tax schedule forall forms of capital income, after-taxinterest costs became the same for allhouseholds. Owner-occupied housingwas no longer cheaper for high-incomegroups. But TR 91 also affected housingcosts through increased VAT andreduced interest subsidies. The com-bined impact of TR 91 was to funda-mentally transform the price structure inthe housing market. Rental costsincreased substantially and becameessentially uniform across households.

Based on microsimulations, we predictthat these cost changes should lead toan aggregate demand decrease on theorder of 15 percent. Since the stock ofhousing capital adjusts so sluggishly, theimmediate response to the reformshould be on prices and new construc-tion. Simulations with a perfect fore-sight model in the vein of Poterba(1984) suggest a drop in prices ofaround 10 to 15 percent upon an-nouncement of TR 91 and a sharp fall innew construction.18 These predictions offalling prices and a virtual standstill ofnew construction have been borne outby developments after the implementa-tion of TR 91. The reform year of 1991marks the peak of house prices and ofthe rate of new construction. Construc-tion virtually ceased after 1992, andnominal house prices fell by a total of19 percent from the peak in the thirdquarter of 1991 to the trough in thefirst quarter of 1993.19 The total price


653

fall from peak to trough is well in linewith expectations, although here aselsewhere it is hard to isolate the effectsof TR 91 from those of the severerecession. The timing, however, issomewhat puzzling in view of the factthat the reform could reasonably havebeen expected by 1989 and wascertainly known when the reform billwas passed by Parliament in June 1990.The fact that it took two years until thiswas reflected in house prices casts somedoubt on the rationality of pricing in theSwedish housing market.

When house prices started to fall,however, they did so quite rapidly. Infact, this is the first episode in moderntimes with falling nominal prices. Thisraises two issues. First, the capital lossesof homeowners gave rise to a poten-tially sizable redistribution of incomeboth vertically, since the averagehomeowner is higher up the incomedistribution than the average renter, andhorizontally from homeowner to renterat the same income level. Second,falling house prices dug deep holes inthe net wealth of many homeowners,even to the point of creating negativenet equity. This most likely created atemporary lock-in effect, ashomeowners short of equity could notput up the down payment necessary fora new house. Indeed, the transactionsvolume in the secondary marketdecreased substantially.

Labor Supply

In the public debate surrounding thereform, the potential impact on laborsupply played a major role, althoughmany economists cautioned againstoverly optimistic supply projections. Thedevelopment of the wedge betweenlabor costs to the employer (includingwage taxes) and the after-tax remunera-tion of blue- and white-collar employees

(accounting for income taxes, as well asVAT and other indirect taxes) is depictedin Figure 2. We see that TR 91 marks asharp break of a long-term trend, whereby the end of the 1970s the marginaltake-home pay of the average blue-collar worker had fallen to less than 30percent of pretax labor costs. In relativeterms, the increase from 1989 to 1991was 23 percent for the blue-collarworker and as much as 76 percent forthe upper-tier white-collar worker.

Although these numbers clearlydemonstrate the magnitude of thereform, one should not be misled intogeneralizing to all groups in society. Infact, looking across a representativesample of all individuals, it appears thataround a quarter saw an increase in thetax wedge.20 The explanation is that alarge fraction of Swedish wage earnerswork less than full-time at relativelymodest marginal-tax rates that were notmuch affected by the reform, and thatmany households in this category areentitled to income-dependent housingallowances that were increased as partof the reform. A similar pattern with amixture of increased and decreasedmarginal tax rates holds for TRA 86(Hausman and Poterba, 1987). In fact,as is emphasized by Auerbach andSlemrod (1997, forthcoming), it lies inthe nature of a reform aimed atmaintaining the degree of redistributionthat it is very difficult to lower marginaltaxes for everybody.

Various labor supply studies, conductedbefore TR 91 and as part of the evalua-tion effort, tend to confirm the “elastic-ity pessimism” underlying Slemrod’sresponse hierarchy, with labor supply atthe bottom. A representative estimateof the compensated wage elasticity ofhours worked among Swedish prime-age males is on the order of magnitudeof 0.1,21 but the estimates are so


654

imprecise that the predicted hoursresponse to the reform of the represen-tative white-collar worker of Figure 2has a typical confidence interval rangingfrom 1.5 to 15 percent. Unfortunately,there is a paucity of ex post studies ofthe impact of the reform, but the panelstudy of Klevmarken (1996) finds thatchanges in marginal wages between1985 and 1992 were associated withstatistically significant increases of thesupply of hours worked, both for menand for women.22

Although a very large amount ofresearch has focused on hours worked,it is possible that other margins of laborsupply response may prove moreimportant in the longer run. It isnoteworthy that TR 91 implied a largechange in the incentive to undertakeinvestment in education. TR 91 loweredtax rates at higher (postinvestment)income levels, but left tax rates more or

less unchanged at lower (preinvestment)income levels. Edin and Holmlund(1995) calculate internal rates of returnto investing in a four-year university pro-gram. In the early 1980s, the return wasfour to five percent before tax and oneto three percent after tax, implyingeffective tax rates between 40 and 90percent depending on the year chosenas a basis for the calculations. After thereform, the effective tax rate fell to 25percent. The impact of this reduction isnot confined to formal education butapplies more broadly to the choicebetween careers with different earningsprofiles.

The Corporate Response

The new corporate tax rules meant anoteworthy departure from the previouslong-standing policy of stimulatingbusiness investment in fixed capitalthrough a combination of a high

FIGURE 2. The Marginal Take-Home Pay per Dollar of Employer Costs, of Blue- and White-Collar Workers, 1952–93

Note: Marginal take-home pay is calculated as (1-marginal tax rate)/[(1 + wage taxes)(1 + indirect taxes)]

Source: DuRietz (1994)


655

statutory tax rate and generous allow-ances to investing firms. The tax ratewas cut almost in half, and, to keep thetax payments of the corporate sectorroughly constant, the base of the taxwas considerably broadened. Many ofthe innovative incentive provisions thathad set the Swedish tax system apartwere eliminated, notably the investmentfunds system.23

Though there was a widespread viewamong policymakers that cutting thestatutory tax rate in half would itselfgreatly improve investment incentives,estimates using conventional methodol-ogy indicate that the corporate taxreform had little effect on the cost ofcapital. The base broadening largelyoffset the effects of the tax cut. How-ever, TR 91 somewhat reduced theprevious strong incentives to use debtrather than equity as a source offunds.24

An important complication in evaluatingthe effects of the corporate tax reform isthat Swedish companies to a largeextent both paid corporate income taxand abstained from fully using thegenerous tax allowances. A detailedstudy (Forsling, 1996) indicates that,over the period 1979–88, the averagerate of utilization of tax allowances(deductions for depreciation, contribu-tions to investment funds, and under-valuation of inventories) among tax-paying firms was a mere 72 percent.Only one out of five firms used themaximum allowed by the tax code. Anincrease in the rate of utilization from72 to 76 percent would have beensufficient to completely eliminate all taxpayments. Conventional estimates ofthe cost of capital or effective tax rates,assuming full utilization of existing taxallowances, may therefore give amisleading picture of the incentiveeffects of the corporate tax.

Recent research (Kanniainen andSödersten, 1994) has attributed thisrather odd tax behavior to the uniformreporting convention used in Sweden(and several other OECD countries).Firms can distribute cash dividends onlyto the extent of their after-tax profits,taking account of fiscal depreciation,contributions to investment funds, etc.Hence, corporate civil law imposes adividend constraint on using taxallowances, and, in practice, thisconstraint seems to have been moretight than the upper limits set by the taxcode. When tax allowances on existingassets have not been fully used, anadditional investment project will notaffect total tax payments, that is, at themargin, the corporate tax rate is zero.Put differently, the corporate income taxis effectively turned into a tax ondistributions or a cash flow tax with noimpact on the cost of capital.25 Themechanism involved here is similar tothat analyzed in the literature aboutdividend taxation.26 To the extent thatpaying dividends is the only way to getcash to the shareholders (share repur-chases are disallowed in Sweden), thefirm is in a “trapped equity” regimewhere the corporate tax is capitalized inshare prices.

The possibility that large groups ofSwedish firms effectively faced a zeromarginal corporate tax rate makes itunclear to what extent the old taxsystem actually did offer an advantageto debt finance. Even the direction ofchange in the incentives for borrowingbrought by TR 91 is unclear, as the basebroadening would be expected to raisethe rate of utilization of the still remain-ing tax shields.27 Firms experiencing aswitch from being taxed on their cashflows to being subject to a regularincome tax would find the value ofinterest deductions increase at themargin, despite the sharp cut in the


656

statutory tax rate. Moreover, given thatthe financial markets in Sweden arehighly integrated with the internationalmarkets for debt and equity, it seemsunlikely that the switch to the dualpersonal income tax with a flat rate 30percent tax on personal capital incomeswould be of much importance forcorporate financial decisions. The after-tax costs of funds of the large Swedishfirms are more likely determined by theoperations of international portfolioinvestors, say, United States pensionfunds, than by the savings and portfoliodecisions of Swedish households.

Auerbach, Hassett, and Södersten(1995) focus on the effects on businessfixed investment. A model of equipmentinvestment is estimated in order todetermine which of several potentialregimes best described investmentbehavior before the reform. Eventhough the regression results do notsettle the issue, evidence on the use oftax allowances and investment fundsgenerally supports the view that the pre-reform corporate tax system hadessentially no effect on investment. Thechange in the user cost of capital due tothe reform is found to be very small andswamped in recent years by the impactof the rise in real interest rates anddecline in profitability. The authorsconclude, “with some confidence,” thatthe effects of TR 91 itself (as opposed tocontemporaneous macroeconomicfactors) on investment are likely to havebeen minor.

INCOME DISTRIBUTION

In the political process of selling TR 91to various interest groups and theelectorate in general, a key element wasthe claim that the reform would bedistributionally neutral. This wasinterpreted in a bookkeeping sense tomean an unchanged relation between

an exogenously given distribution ofpretax factor income and the distribu-tion of income after taxes and allow-ances. Against this background, it wasnatural that a mechanical evaluation ofthe distributional impact along theselines was one element of the evaluationeffort, although it comes natural for aneconomist to point to the limitations ofsuch an exercise.

During the 1980s, a growing number ofcritics came to doubt whether theSwedish tax system achieved much interms of redistribution. There were threeingredients to the critique. First, variousloopholes and tax arbitrage activitiescreated a wedge between “true”income and taxable income. Second,substitution between market labor andleisure and household productioncreated a discrepancy between taxableincome (which only derives from marketactivities) and potential income (whichalso includes the value of leisure andhome production). Third, taxation basedon yearly income redistributes incomeacross different phases in an individual´slife cycle, i.e., from more toward lessproductive ages, but it is less clear howmuch redistribution actually is achievedacross households with different lifetimeincomes.28

The recent study by Björklund, Palme,and Svensson (1995), however, giveslittle support to the critics of the oldsystem. According to this analysis, theold Swedish tax system certainlyachieved a substantial amount ofredistribution of yearly incomes.29 Theamount of redistribution is not muchaffected by going from actual income toa measure of full income. Also, whenthe authors follow a panel of individualsover the period 1974–91 and take thesum of discounted income over thisperiod as a measure of lifetime income,the conclusion is that the tax system


657

redistributes almost as much in terms oflifetime income as in terms of yearlyincome. Of course, this only sayssomething about the tax system as itevolved over this 18-year period andgives no conclusive evidence on thelong-run redistributive properties of thesystem of the late 1980s. Nevertheless,it leads us to be somewhat skepticalabout some of the more popular viewsof the old tax system as being void ofany redistributive effect.

Björklund, Palmer, and Svensson (1995)also examine differences in pre- andpost-tax Gini coefficients under the oldand new tax rules. Looking at theaggregate amount of redistributionacross all groups of households, thedifferences appear minuscule; i.e., atleast in the short term (up to 1992),TR 91 seems to have lived up to thepromise of “neutrality“ with respect tothe income distribution. This conclusion,however, conceals important differencesin the structure of the tax and subsidysystem. Broadly speaking, there are threecounteracting differences between theold and new tax systems. First, thetaxation of earned income is clearly lessredistributive under TR 91. Second, withthe new flat tax on capital income, taxpayments become more proportional toactual capital income than under the oldsystem, where it was possible to avoidcapital taxes altogether through taxarbitrage. Since capital income isconcentrated at the top of the incomedistribution, this tends to make the newsystem more redistributive than the oldone. Third, child and housing allowancesplay a larger role after TR 91 than before.Since they largely redistribute incomefrom households without children tofamilies with children, TR 91 represents ashift of emphasis toward more redistribu-tion across various phases of the lifecycle rather than between householdswith different lifetime incomes.

By simply comparing Gini coefficientsbefore and after taxes, one takes anunduly narrow view on income distribu-tion. In particular, one takes the pretaxdistribution of factor income forgranted, thereby glossing over thestrong assumptions about tax incidenceimplicit in such an excercise. Whereas itmay be reasonable for a small openeconomy like the Swedish one toassume that the pretax return to capitalis determined in international marketsand unaffected by Swedish tax policy,the assumptions about the incidence onwages merit more attention than theyare commonly given by income distribu-tion analysts. In fact, there has been arecent trend in Sweden, as elsewhere,toward more inequality of factorincomes, and one may ask if thisdevelopment has been induced by taxreform to any extent. This shoulddepend on the relative supply responsesof high- and low-skilled workers and onthe degree of complementarity inproduction between different types oflabor. However, given the generallysmall labor supply responses, it appearsunlikely that the changing wagestructure has primarily been induced bychanges in tax structure. In the longerrun, the strengthened incentives toinvest in human capital should be moreimportant for the wage distribution. It ishard, though, to have a very definiteopinion on the implications for thedistributional analysis of TR 91.30

EFFICIENCY — A DOUBLE DIVIDEND?

A main argument of the proponents ofTR 91 was that economic efficiencyshould improve. For many politicians,efficiency was here interpreted as asynonym for various easy-to-observeresponses, such as increased laborsupply and higher savings. In somequarters, there was also a hope thatTR 91 should deliver an easily detectable


658

growth bonus. For an economist,however, efficiency is defined in termsof not directly observable areas betweencompensated demand and supplycurves, and there is no simple relationbetween the implied tax distortions andthe magnitude of behavioral responses.Much the same goes for economicgrowth. According to the standardneoclassical growth model, a badlydesigned tax system may createimportant negative level effects, withoutthe long-term rate of growth beingaffected. As we will see shortly, TR 91even contained provisions which tendedto reduce registered GDP growth but toincrease consumption opportunities.

From an excess burden perspective, themost important aspect of TR 91 is that itimplied a shift in the tax burden fromhighly taxed labor income to lightlytaxed housing capital. As a conse-quence, the relative prices of leisure andhousing consumption increased.Although there is room for disagree-ment on the exact magnitudes of somekey behavioral elasticities, there is littlereason to question the soundness of thisstrategy. Table 2 shows the results froman attempt at quantifying the marginalefficiency cost due to labor supplydistortions of the Swedish tax system.31

The compensated elasticities of labor

supply are chosen to reflect the range offindings in recent empirical studies.32

The marginal tax wedges (differencesbetween productivity and real take-home pay) for some categories ofemployees, before and after TR 91,account for income, payroll, and indirecttaxes.

A key point is that a small behavioralelasticity matters if the marginal taxwedge is sufficiently high. The logicbehind the marginal welfare cost perdollar of tax revenue implies that thereis a region of tax wedges at which theefficiency cost starts to increase rapidly.At some marginal tax wedge, anadditional tax hike creates additionalexcess burdens but no extra tax rev-enue. Although the table brings out thesensitivity of results to alternativebehavioral assumptions, one can hardlyrule out the possibility that prereformtax wedges were close to that level. Thepoint estimates of labor supply elastici-ties, not to mention the confidenceintervals, reported in recent studies areconsistent with the view that the pre-TR 91 tax system had marked negativeincentive effects.

Of course, microeconomic estimatessuggesting high marginal welfare costsof taxation do not necessarily imply that

TABLE 2MARGINAL EXCESS BURDEN PER KRONA OF TAX REVENUE (IN PERCENT)

Compensated Labor Supply Elasticity

0.05 0.11 0.25

Marginal tax wedge (in percent):

62 (average blue-collar worker, 1991) 8.2 19.0 54.763 (average earned income, 1991) 8.6 20.1 59.070.5 (average blue-collar worker, 1988) 12.7 31.6 121.871.5 (average white-collar worker, 1991) 13.4 33.973 (average earned income, 1988) 14.6 37.879 (average white-collar worker, 1988) 22.0 65.3 2,280.085.5 (average senior white-collar worker, 1988) 41.0 192.5 —

Tax rate which maximizes tax revenue 94.5 89.5 79.5

Source: Agell, Englund, and Södersten (1995).

175.0139.1


659

tax cuts are in order. When there is abinding revenue constraint, lower taxwedges on labor income make senseonly if other taxes can be raised in a lessdistortionary way. The remarkableaspect of the Swedish situation was thatthere was scope for a “double divi-dend.” Higher taxes on housinggenerated a substantial part of therevenue required to finance the tax cuton labor income, but they also reducedintersectoral investment distortions. Dueto the generous tax and subsidy rules,discussed above, housing investmentwas given a considerable advantageover investments in other sectors.Available estimates indicate (for realisticvalues of the debt-equity mix, inflation,etc.) that new investments in owner-occupied housing could reap a netmarginal subsidy—the (financial) cost ofcapital for the prospective homeownerwas well below the real rate of interest.TR 91 did much to promote a lessinefficient allocation of investmentresources.

The case of housing illustrates that theshort-term response of an aggregateproduction measure, such as GDP, mayprovide a poor indicator of the welfareeffects of tax reform. Before TR 91,household savings was channeled intoan activity where the marginal produc-tivity of capital was considerably belowthe opportunity cost, which in an openeconomy can be approximated by theworld real rate of interest. At the sametime, the housing sector gave a substan-tial contribution to Swedish GDP. In anyyear, housing investment, valued fromthe production side, added to aggregateinvestments. TR 91 gave householdsstrong incentives to redirect theirsavings to other uses, including netpurchases of foreign assets.33 Duringthe adjustment phase (when thehousing sector shrinks and the netforeign asset position improves), GDP

growth tends to slacken. In spite of thisnegative production effect, aggregateconsumption possibilities tend toincrease—every krona’s worth of savingstransferred from the housing sector tointernational asset markets implies thatSwedish national income increases withthe difference between the world realrate of interest and the marginalproductivity of housing capital.

Undoubtedly, TR 91 has affectedeconomic efficiency along a number ofmargins in addition to those justdiscussed. The reduced progressivity ofthe income tax has enhanced educa-tional incentives. The new corporateincome tax brought about a moreuniform treatment of investmentprojects within the corporate sector. Thebroader base for the VAT implied highertax wedges on the “white” consumerservice sector, which competes with do-it-yourself activities and servicesproduced in the underground economy.However, in these cases, there is scantevidence on the behavioral response.

Finally, an important objective of TR 91was to simplify the tax code, and thereis reason to believe that TR 91 did muchto reduce the transaction complexity ofthe tax system. TR 91 made it much lessprofitable to invest resources in a varietyof tax avoidance activities. Surveyevidence indicates that households’ timespent on tax compliance declinedsubstantially in the years after TR 91.34

There is also evidence suggesting thatthe tax authorities got an easier work-load.

Concluding remarks

What is the verdict on TR 91 five yearsafter its implementation? Have we seenthe behavioral responses that the reformarchitects expected? Did the reformcontribute to a more efficient economy?


660

Although we should keep our fingerscrossed in view of the severe crisis in theSwedish economy in the early 1990s,we concur with Auerbach and Slemrod(1997, forthcoming), who argued intheir survey of TRA 86 that there hasbeen a hierarchy of responses. Anumber of financial activities related totax planning were rendered meaninglessand have been virtually wiped out. Wehave also seen large and expectedeffects on portfolio composition ingeneral, with a shrinking of both sidesof private sector balance sheets.Households have been induced to shiftout of owner-occupied housing,resulting in falling house prices and astandstill of new construction. In thelonger run, this will result in a moreefficient allocation of the capital stock.

At the other end of the spectrum, majorreal activities such as labor supply andsavings appear quite insensitive, at leastas far as can be inferred from short-runbehavior. Since real activities can beexpected to be more important from awelfare perspective, it might be conjec-tured that the reform only has made asmall contribution to increase theefficiency of the Swedish economy, likeAuerbach and Slemrod (1997, forthcom-ing) conclude for TRA86. While such aconclusion may be warranted at U.S. taxrates, it is hardly correct when startingout from marginal tax rates of 70 percentand more. At such levels—close to thetop of the Laffer curve—the marginalexcess burden is highly nonlinear in taxrates, implying that accounting forheterogeneity and uncertainty about thecorrect elasticity values is very important;the expected aggregate excess burden ismuch larger than the marginal excessburden for the average taxpayer evalu-ated at point estimates of elasticities.

Although a standard efficiency calcula-tion—comparing hypothetical equilibria

before and after the reform—unam-biguously shows that TR 91 wasefficiency improving, these benefitswere not without costs. In fact, thereform may be viewed as an investmentwith quite visible short-run costs thathave to be weighed against less visible,and perhaps also less certain, long-runbenefits. The short-run costs of TR 91were primarily of two types. First, byshifting savings out of real assets suchas housing and consumer durables intofinancial assets, the reform implied areduction of effective demand. Since thereform was implemented in a recessionwhen output was arguably demanddetermined, we conclude that it led to afurther deepening of the recession withaccompanying short-term productionlosses. Second, the rapid implementa-tion of the reform led to sizable capitallosses in the housing sector, with anensuing arbitrary horizontal redistribu-tion across households. These observa-tions serve as a reminder of the pointemphasized, e.g., by Feldstein (1976),that one has to distinguish between taxreform and de novo tax design. While acomparison of the pre- and post-1991tax systems comes out in favor of thepost-1991 system, an evaluation of thereform has to weigh the long-runbenefits against the short-term costs.

The costs of the reform were notunavoidable. The government took arather careless attitude toward thetransition problem. A more gradualphasein of the reform certainly wouldhave dampened the short-run costsconsiderably without giving up any ofthe long-run benefits. While this wouldbe a rather obvious recommendation toa benevolent dictator, things are morecomplex in a parliamentary democracywhere one cannot tie the hands offuture governments. Indeed, bringingabout political consensus about areform as far reaching as TR 91 involved


661

a rather delicate balance of the gainsand losses of different interest groups insociety. If parts of the reform wouldhave been implemented gradually orwith a lag, they would also have beenmore susceptible to future politicalpressures.

Even a ”tax reform of the century”implemented with such force as TR 91did not stay unaffected for very long.Already after three years, in 1994, onecould make a list of some 75 taxchanges involving minor and majordeviations from the original reform.Such a count of changes obviously is acheap argument; after all, the worldchanges continuously, and one wouldhope that the tax law adapts. Neverthe-less, some of the changes representreversals of the guiding principles ofTR 91. One example is the numerouschanges in different VAT rates. Anotheris the increase of the top marginal taxrate on earned income from 50 to 55percent. The latter change was pre-sented as a temporary measure and partof a package to cut the growing budgetdeficits after 1992. While these deficitsare largely attributable to the recession,they partly reflect an underfinancing ofthe reform by two to three percent ofGDP. It is not surprising that it waseasier to sell an underfunded reformpackage, where all groups could beportrayed as winners, than a fullyfinanced reform, where some groupswould appear as losers, at least in ashort-run accounting sense.

In what direction should we expect theSwedish tax system to evolve in thefuture? The 1980s was a decade of taxreforms aimed at a more uniform taxstructure, conspicuously ignoring muchof the development in public financesince the 1970s, emphasizing the role ofdifferentiated taxation in fundinggovernment expenditure. In the 1990s,

the pendulum of tax reform discussionhas swung in the opposite direction,with a renewed emphasis on differentia-tion, e.g., for environmental reasons oras a means of fighting unemployment.When these factors now are beingconsidered in Sweden, future reformscan build on a tax structure that hasfewer counterproductive asymmetriesthan the pre-TR 91 system.

The Swedish public sector ranks amongthe largest in the world. As long as thisremains, Swedish tax rates will inevitablybe high. Our discussion suggests that,especially with a high aggregate taxrate, the structure of taxation matters.TR 91 has reduced the aggregate excessburden. However, the ever growingintegration with international capitaland labor markets will undoubtedly putincreasing pressure on the Swedish taxsystem in the future.

ENDNOTES

We are grateful to Joel Slemrod for his commentson a first draft of this article.

1 See Auerbach and Slemrod (1997) for an extensivediscussion.

2 The evaluation project is summarized (in Swedish)in Agell, Englund, and Södersten (1995) (anEnglish version will be published by MacmillanPress.) The 1995 Autumn issue of the SwedishEconomic Policy Review includes a selection of thebackground papers. A complete list of backgroundpapers is available from the National Institute ofEconomic Research, Box 3116, 103 62 Stockholm,Sweden.

3 See Sørensen (1994) for further discussion.4 From the point of view of evaluation research, the

timing of TRA 86 was more fortunate. Asdiscussed by Auerbach and Slemrod (1997,forthcoming), the U.S. macroeconomy was wellbehaved in the years surrounding TRA 86.

5 For an overview of the Swedish macroeconomicexperience, see Calmfors (1993).

6 However, compared with the internationalrecession and the nonaccommodative exchangerate policy, TR 91 was most likely a less decisivefactor. A back-of-the-envelope calculation of Agell,Englund, and Södersten (1995) suggests thatTR 91 may explain about one-fifth of the sharpdecline in Swedish GDP between 1991 and 1993.


662

7 The Swedish Central Securities Depository handlesboth domestic source taxation and the withholdingtax on dividends paid to foreign shareholders.

8 For a discussion of consumption behavior inSweden, see Berg (1994).

9 A complication in assessing the effects of TR 91 onintertemporal incentives stems from the fact thatTR 91 contained provisions with counteractingeffects on the real after-tax interest rate. The lowerstatutory tax rate on capital income strengthenedincentives, but some of the base-broadeningingredients, such as the abolishment of thefavorable treatment of long-term capital gains,worked in the opposite direction. On balance,however, it seems reasonable to conclude thatTR 91 implied lower effective marginal tax rates fora majority of households. This can be contrasted tothe case of TRA 86, where the conclusion seems tobe that the impact on savings incentives was of anambiguous sign (Auerbach and Slemrod, 1997).

10 For Swedish evidence on the intertemporalelasticity of substitution, see Campbell andMankiw (1991) and Agell, Berg, and Edin (1995).

11 A more likely factor in a drop in permanent laborincome is the sharp increase in unemploymentafter 1991. While higher unemployment mayinduce consumers to revise their expectations offuture labor income, it may also boost precaution-ary savings. When consumers are prudent, moreuncertain earnings prospects can have strongnegative effects on current consumption (e.g.,Caballero, 1991). Although we believe that part ofthe increase in unemployment can be explained bythe short-term contractionary impact of TR 91, themajor impulses should, as already discussed, besought elsewhere.

12 See Agell, Berg, and Edin (1995).13 See, e.g., Hultkrantz (1995).14 See, e.g., Agell and Edin (1990).15 For an analysis of household asset choice, before

and after TR 91, see Agell, Berg, and Edin (1995).16 According to the National Accounts of Statistics

Sweden, household net lending is defined as thesum of borrowing and lending in the creditmarket, net purchases of corporate shares,individual insurance savings, and savings in otherinterest bearing assets.

17 The latter advantage, however, was offset byspecial interest subsidies at high rates to recent-vintage rental housing.

18 See Åsberg and Åsbrink (1994) for a presentationof the model including a discussion of thesensitivity of the results to demand and supplyelasticities.

19 These effects are qualitatively similar to, but muchlarger in magnitude than, those of TRA 86according to the discussion in Poterba (1990).

20 This fraction applies to a comparison between1985 and 1992 for a sample of households fromthe HUS panel (Klevmarken et al.,1995). It

accounts for the broadening of the VAT coverageand other factors affecting the real wage.

21 See studies by Blomquist and Hansson-Brusewitz(1990), Ackum Agell, and Meghir (1995), andAronsson and Palme (1995).

22 The estimated effects are quite large. Women whoexperience a tax decrease of more than 20percentage points are estimated to increase yearlylabor supply by 400 hours, i.e., by roughly 30 per-cent of average work hours. Correspondingestimates for men are 200 hours and 10 percent.

23 See Taylor (1982) and Södersten (1989) for adiscussion of the investment funds system.

24 See Södersten (1993).25 For further explanation of the Kanniainen–Södersten

theory, see also Sørensen (1995) in this journal.26 Cf. Auerbach (1979).27 Though data are still scant, preliminary observa-

tions seem to confirm this assumption.28 The two latter lines of criticism received some

support from a study by Hansson and Norrman(1986), based on a cross section of households.

29 Measuring pretax income by the sum of taxablelabor income and an imputed measure of capitalincome (net wealth times a three percent realinterest rate) and using the square-root equiva-lence scale to adjust for differences in householdcomposition, the Gini coefficent based on yearlypretax income was around 0.25 during the 1980s,a quite low number by international standards.After taxes and housing and child allowances, thecoefficient was around 0.20, i.e., a sizableredistribution.

30 It should also be pointed out that the calculationsreported above are confined to personal incometaxes and disregard the incidence of the corporateincome tax. However, as is seen from Table 1,corporate tax revenues did not change much as aresult of TR 91, and assumptions about corporatetax incidence should not be crucial.

31 See Agell, Englund, and Södersten (1995). Thenumbers derive from a partial equilibrium model,where an individual maximizes a single-periodconstant-elasticity-of-substitution utility function inleisure and goods consumption, subject to a linearincome tax system, and where there is noavoidance type of response that creates adiscrepancy between true and taxable income. Tomake life easy, it is assumed that all incrementaltax revenue is returned as a lump sum (all incomeeffects are eliminated). The marginal welfare costper unit of tax revenue is defined as ∆ΕΒ/∆Τ,where ∆ΕΒ is the change in excess burden and ∆Τthe change in tax revenue.

32 The highest elasticity, 0.25, is of relevance forwomen’s labor supply. The intermediate elasticity,0.11, is typical for many studies of the labor supplyof married prime-age males. The smallest elasticity,0.05, is taken from a recent study on the laborsupply of blue-collar workers.


663

33 The drastic increase in household financial savingsin the early 1990s was accompanied by a rapidimprovement of the current account.

34 See Malmer and Persson (1994).

REFERENCES

Ackum Agell, Susanne, and Costas Meghir.“Male Labour Supply in Sweden: Are IncentivesImportant?” Swedish Economic Policy Review 2No. 2 (Fall, 1995): 391–418.

Agell, Jonas, and Per-Anders Edin. “MarginalTaxes and the Asset Portfolios of SwedishHouseholds.” Scandinavian Journal ofEconomics 92 No. 1 (January, 1990): 47–64.

Agell, Jonas, Lennart Berg, and Per-AndersEdin . “The Swedish Boom to Bust Cycle: TaxReform, Consumption and Asset Structure.”Swedish Economic Policy Review 2 No. 2 (Fall,1995): 271–314.

Agell, Jonas, Peter Englund, and JanSödersten. Svensk skattepolitik i teori ochpraktik. 1991 års skattereform (Swedish TaxPolicy in Theory and Practice. The 1991 TaxReform). Stockholm: Fritzes, 1995.

Aronsson, Thomas, and Mårten Palme. “ADecade of Tax and Benefit Reforms in Sweden—Effects on Labour Supply, Welfare and Inequal-ity.” Tax Reform Evaluation Report No. 3.Stockholm: National Institute of EconomicResearch, 1995.

Åsberg, Per, and Stefan Åsbrink.“Capitalisation Effects in the Market for Owner-Occupied Housing: A Dynamic Approach.” TaxReform Evaluation Report No. 2. Stockholm:National Institute of Economic Research,1994.

Auerbach, Alan. “Wealth Maximization andthe Cost of Capital.” Quarterly Journal ofEconomics 93 No. 3 (August, 1979): 433–46.

Auerbach, Alan, Kevin Hassett, and JanSödersten. “Taxation and Corporate Invest-ment: The Impact of the 1991 Swedish TaxReform.” Swedish Economic Policy Review 2No. 2 (Fall, 1995): 361–83.

Auerbach, Alan, and Joel Slemrod. “TheEconomic Effects of the Tax Reform Act of1986.” Journal of Economic Literature 35 No. 1(forthcoming, March, 1997).

Berg, Lennart. “Household Savings and Debts:The Experience of the Nordic Countries.” OxfordReview of Economic Policy 10 No. 2 (Summer,1994): 42–53.

Björklund, Anders, Mårten Palme, andIngemar Svensson. “Tax Reforms and IncomeDistribution: An Assessment Using Different

Income Concepts.” Swedish Economic PolicyReview 2 No. 2 (Fall, 1995): 229–66.

Blomquist, Sören, and Urban Hansson-Brusewitz. “The Effect of Taxes on Male andFemale Labor Supply in Sweden.” Journal ofHuman Resources 25 No. 3 (1990): 317–57.

Caballero, Ricardo. “Earnings Uncertainty andAggregate Wealth Accumulation.” AmericanEconomic Review 81 No. 4 (September, 1991):859–71.

Calmfors, Lars. “Lessons from the Macroeco-nomic Experience of Sweden.” EuropeanJournal of Political Economy 9 No. 1 (March,1993): 25–72.

Campbell, John, and Gregory Mankiw. “TheResponse of Consumption to Income: A Cross-Country Investigation.” European EconomicReview 35 No. 4 (May, 1991): 723–67.

DuRietz, Gunnar. Välfärdsstatens finansiering(Funding the Welfare State). Stockholm: CityUniversity Press, 1994.

Edin, Per-Anders, and Bertil Holmlund. “TheSwedish Wage Structure: The Rise and Fall ofSolidarity Wage Policy in Sweden.” In Differ-ences and Changes in Wage Structure, edited byRichard Freeman and Lawrence Katz. Chicago:University of Chicago Press, 1995.

Englund, Peter, Patric Hendershott, andBengt Turner. “The Tax Reform and theHousing Market.” Swedish Economic PolicyReview 2 No. 2 (Fall, 1995): 319–56.

Feldstein, Martin. “On the Theory of TaxReform.” Journal of Public Economics 6 No. 1-2(July–August, 1976): 77–104.

Forsling, Gunnar. “Utilization of Tax Allow-ances: A Survey of Swedish Corporate Firms,1979–93.” Tax Reform Evaluation Report No. 22.Stockholm: National Institute of EconomicResearch, 1996.

Hall, Robert. “Intertemporal Substitution inConsumption.” Journal of Political Economy 96No. 2 (April, 1988): 339–57.

Hansson, Ingemar, and Erik Norrman.“Fördelningseffekter av inkomstskatt ochutgiftskatt” (Distributional Effects of Income Taxand Expenditure Tax). In Utgiftsskatt. Teknik ocheffekter (Expenditure Taxation. Techniques andEffects). SOU 1986: 40. Stockholm: LiberFörlag, 1986.

Hausman, Jerry, and James Poterba.“Household Behavior and the Tax Reform Act of1986.” Journal of Economic Perspectives 1 No.1 (Summer, 1987): 101–19.

Hultkrantz, Lars. “On Determinants of SwedishRecreational Domestic and Outbound Travel.”


664

Tax Reform Evaluation Report No. 7. Stockholm:National Institute of Economic Research, 1995.

Kanniainen, Vesa, and Jan Södersten. “Costsof Monitoring and Corporate Taxation.” Journalof Public Economics 55 No. 2 (1994): 307–21.

Klevmarken, Anders. “Did the Tax CutsIncrease Hours of Work? A Pre-Post Analysis ofSwedish Panel Data.” Unpublished manuscript,Department of Economics, Uppsala University,1996.

Klevmarken, Anders (with Irene Andersson,Peter Brose, Erik Grönqvist, Paul Olovsson,and Marianne Stoltenberg-Hansen). “LaborSupply Responses to Swedish Tax Reforms 1985–1992.” Tax Reform Evaluation Report No. 11.Stockholm: National Institute of EconomicResearch, 1995.

Malmer, Håkan, and Annika Persson.“Skattereformens effekter på skattesystemetsdriftskostnader, skatteplanering och skattefusk”(The Effects of the Tax Reform on ComplianceCosts, Tax Planning, and Tax Fraud). InÅrhundradets skattereform (Tax Reform of theCentury), edited by Håkan Malmer, AnnikaPersson, and Åke Tengblad, 5–364. Stockholm:Fritzes, 1994.

Myrdal, Gunnar. “Dags för ett nyttskattesystem!” (Time for a New Tax System!)Ekonomisk Debatt 6 (November, 1978): 493–506.

Poterba, James. “Tax Subsidies to Owner-Occupied Housing.” Quarterly Journal of

Economics 99 No. 4 (November, 1984): 729–52.

Poterba, James. “Taxation and HousingMarkets: Preliminary Evidence on the Effects ofthe Recent Tax Reform.” In Do Taxes Matter?:The Impact of the Tax Reform Act of 1986,edited by Joel Slemrod, 141–60. Cambridge,MA: MIT Press, 1990.

Slemrod, Joel. “Do Taxes Matter?: Lessonsfrom the 1980s.” American Economic Review82 No. 2 (May, 1992): 250–6.

Södersten, Jan. “The Investment Funds SystemReconsidered.” Scandinavian Journal ofEconomics 91 No. 4 (1989): 671–87.

Södersten, Jan. “Sweden.” In Tax Reform andthe Cost of Capital. An International Compari-son, edited by Dale Jorgenson and Ralph Landau,270–99. Washington, D.C.: The BrookingsInstitution, 1993.

Sørensen, Peter Birch. “From the GlobalIncome Tax to the Dual Income Tax: Recent TaxReforms in the Nordic Countries.” Inter-national Tax and Public Finance 1 No. 1 (May,1994): 57–79.

Sørensen, Peter Birch. “Changing Views ofthe Corporate Income Tax.” National Tax Journal48 No. 2 (June, 1995): 279–94.

Taylor, John. “The Swedish Investment FundsSystem as a Stabilization Policy Rule.” BrookingsPapers on Economic Activity 13 No. 1 (1982):57–99.

tax reform of the century—the swedish experiment · 2019. 4. 11. · slogan “tax reform of the...

Documents