TAXATION AND ECONOMIC GROWTH pdf

More formally, we candecompose the growth rate of theeconomy’s output into its differentcomponents:where the real GDP growth rate in country i is denoted y iand the netinvestment rate ex

TAXATION AND ECONOMIC GROWTH

ERIC ENGEN * &

JONATHAN SKINNER **

Abstract - Tax reforms are sometimes

touted as having strong macroeconomic growth effects Using three approaches,

we consider the impact of a major tax reform—a 5 percentage point cut in marginal tax rates—on long-term growth rates The first approach is to examine the historical record of the U.S.

economy to evaluate whether tax cuts have been associated with economic growth The second is to consider the evidence on taxation and growth for a large sample of countries And finally,

we use evidence from microlevel studies

of labor supply, investment demand, and productivity growth Our results suggest modest effects, on the order of 0.2 to 0.3 percentage point differences

in growth rates in response to a major tax reform Nevertheless, even such small effects can have a large cumula- tive impact on living standards.

INTRODUCTION

By now, a presidential campaign isincomplete without at least oneproposal for tax reform Recent propos-als suggested that by reducing marginal

tax rates, or by replacing the currentfederal income tax with a consumption-type tax, the United States can experi-ence increased work effort, saving, andinvestment, resulting in faster economicgrowth For example, Steve Forbesvaulted briefly into the political limelightbased almost solely on his advocacy of aflat tax which cut nearly every person’stax bill, but which was supposed tobalance the budget by stimulatingeconomic growth The Kemp Commis-sion suggested that its general principlesfor tax reform would almost double U.S.economic growth rates over the nextfive to ten years.1 Most recently,presidential candidate Robert Doleproposed a 15 percent across-the-boardincome tax cut coupled with a halving

of the tax on capital gains, with apredicted increase in gross domesticproduct (GDP) growth rates from about2.5 to 3.5 percentage points

Others have questioned whether taxreform would have such beneficialeffects on economic growth.2 If tax cutsfail to produce the projected boost ineconomic growth, tax revenues coulddecline, putting upward pressure on thedeficit, worsening levels of nationalsaving, and leading to laggard economicgrowth in the future At this stage,however, there is little agreement about

* Federal Reserve Board, Washington, D.C 20551.

** Department of Economics, Dartmouth College, Hanover, NH

03755, and NBER, Cambridge, MA 02138.

Mozambique because its (per capita)capital stock is so much larger and moretechnologically advanced and itsworkers have more skills, or humancapital The growth rate of economicoutput therefore will depend on thegrowth rate of these resources—physical capital and human capital—aswell as changes in the underlyingproductivity of these general inputs inthe economy More formally, we candecompose the growth rate of theeconomy’s output into its differentcomponents:

where the real GDP growth rate in

country i is denoted y iand the netinvestment rate (expressed as a fraction

of GDP), equivalently the change over

time in the capital stock, is given by k i .The percentage growth rate in theeffective labor force over time is written

m i, while the variable µi measures theeconomy’s overall productivity growth.There are two other relevant variables inequation 1, which are the coefficientsmeasuring the marginal productivity ofcapital, αi, and the output elasticity oflabor, βi.3 For example, if there were aone percentage point increase in thegrowth rate of the (skill-adjusted) laborforce and β were equal to 0.75, theimplied increase in the economic growthrate would be 0.75 percentage point.Alternatively, if the investment rate were

to rise by one percentage point and α

were 0.10, the growth rate of outputwould rise by 0.10 percentage point.This theoretical framework allows us tocatalog the five ways that taxes mightaffect output growth, corresponding toeach of the variables on the right-handside of equation 1 First, higher taxes

whether a major tax reform wouldprovide an economic boon to the UnitedStates or impede economic growth

In this paper, we reexamine the ship between economic growth andtaxation in light of the accumulatedeconomic evidence, both from theUnited States and other countries Whilemany economists would agree with theproposition that “high taxes are bad foreconomic growth,” we show that thisproposition is not necessarily obvious,either in theory or in the data However,

relation-we find that the evidence is consistentwith lower taxes having modest positiveeffects on economic growth While suchgrowth effects are highly unlikely toallow tax cuts to pay for themselves,they can contribute to substantialdifferences in the level of economicactivity and living standards, particularlyover the long term

SHOULD WE EXPECT TAXES TOAFFECT GROWTH? A THEORETICAL

Before jumping into the morass ofempirical evidence, it is useful to firstask the question: How does tax policyaffect economic growth? By discourag-ing new investment and entrepreneurialincentives? By distorting investmentdecisions because the tax code makessome forms of investment moreprofitable than others? Or by discourag-ing work effort and workers’ acquisition

of skills? These questions are oftenaddressed in an accounting frameworkfirst developed by Solow (1956) In this

approach, the output, y, of an economy,

typically measured by GDP, is mined by its economic resources—the

deter-size and skill of its workforce, m, and

the size and technological productivity

of its capital stock, k Thus, a country

like the United States might be expected

to have a greater per capita output than

2

1

2

can discourage the investment rate, or

the net growth in the capital stock (k i inequation 1 above), through highstatutory tax rates on corporate andindividual income, high effective capitalgains tax rates, and low depreciationallowances Second, taxes may attenu-

ate labor supply growth m i by

discour-aging labor force participation or hours

of work, or by distorting occupationalchoice or the acquisition of education,skills, and training Third, tax policy hasthe potential to discourage productivitygrowth µ by attenuating research anddevelopment (R&D) and the develop-ment of venture capital for “high-tech”

industries, activities whose spillovereffects can potentially enhance theproductivity of existing labor and capital

Fourth, tax policy can also influence themarginal productivity of capital bydistorting investment from heavily taxedsectors into more lightly taxed sectorswith lower overall productivity

(Harberger, 1962, 1966) And fifth,heavy taxation on labor supply candistort the efficient use of human capital

by discouraging workers from ment in sectors with high social produc-tivity but a heavy tax burden In otherwords, highly taxed countries mayexperience lower values of α and β,which will tend to retard economicgrowth, holding constant investmentrates in both human and physical capital(Engen and Skinner, 1992) We showthis graphically in Figure 1, whichfocuses on a fixed level of the capital

employ-stock K, shown by the width of the

horizontal axis (A similar analysis holdsfor labor market distortions.) Supposethat the income tax on the corporatesector, as well as subsidies to non-corporate owner-occupied housing,distort the allocation of the capital stock

between the corporate (c) and corporate (nc) sectors (In other coun-

non-tries, the distortion may arise betweensectors which escape taxation such as

the underground economy or scale agriculture, versus the manufactur-ing sector which is easily taxed orheavily regulated.) The line denoted

small-MP(c) is the value of the marginal

product of capital in the corporate

sector, while MP(nc) denotes the value

of the marginal pro-duct in thenoncorporate sector Without any taxdistortion, the profit-maximizing and

most efficient point is C; the mar-ginal

productivity of capital is equalized in

both sectors and the economy-wide return on capital is R*, as shown by thedotted line (The allocation of the total

capital stock, K, is Q* units of capital in

the noncorporate sector and K-Q* units

in the corporate sector.) With a tax of

AB on corporate capital only, there is a

distortion in the allocation of capital;

capital flows from the corporate to thenoncorporate sector, so the new

allocation is Q units of capital in the noncorporate sector and K-Q units of

capital in the corporate sector The net

loss in output is given by ABC, the

traditional Harberger welfare losstriangle Under some plausible restric-tions, the average rate of return for the

entire capital stock, R, will correspond to

the rate of return on new investment,given in equation 1 by α.4 Hence, adistortionary tax on capital (or on labor)will be reflected in lower overall rates of

return on new investment (from R* to

R), leading to laggard growth rates.

We have outlined five possible nisms by which taxes can affect eco-nomic growth Therefore, it mightappear that taxes should play a centralrole in determining long-term growth

mecha-However, the conventional Solow growth

model implies that taxes should have no impact on long-term growth rates In

part, this result occurs by assumption,since productivity growth µ is assumed

to be fixed and unaffected by tax policy

But this paradoxical result holds also

because of a distinction betweenchanges in the level of GDP and changes ingrowth rates of GDP For example,suppose that, in the year 2000, a “taxand spend” president is elected in theUnited States and tax rates are increased

by ten percentage points across theboard (Ignore the effects of the extragovernment spending on the economy.)The extra tax distortion reduces laborsupply and investment, causing a

sudden decline in short-term growth

rates But once the U.S economy hadadjusted to the harsh new tax regime, itwould revert back to its original growthpath, albeit at a lower absolute levelthan it would have been in the absence

of the tax hikes (In the Solow model,both investment and labor supplygrowth revert back to their original ratesdetermined by long-term populationgrowth.) In other words, the simpleSolow model implies that tax policy,however distortionary, has no impact on

long-term economic growth rates, even

if it does reduce the level of economic

output in the long-term

So then how can taxation affect outputgrowth rates? We focus on two possiblemechanisms The first is that when the

structure of taxes changes, short-term

output growth rates would be expected

to change as well along a possiblylengthy transition path to the newsteady state If one believes that theDole or the Forbes tax reform wouldexpand output by five percentage pointsand it takes ten years to make thetransition to the new steady state,growth rates will be higher, on average,

by about 0.5 percentage points duringthis period before settling back down totheir long-run values.5 Ten years is along-term horizon for presidentialcandidates but is still the short-term inthe Solow model And these short-termeffects are clearly important, since theyresult in a permanent increase in GDP

The second possibility arises within thecontext of the new class of “endog-enous growth” models (e.g., Romer,1986; Lucas, 1990) In these models,the stable growth rate of the Solowmodel, stapled down by technology andworkforce productivity growth, isreplaced by steady-state growth rateswhich can differ, persistently, because oftax and expenditure policies pursued bythe government (e.g., King and Rebelo,1990) The endogenous growthframework emphasizes factors such as

“spillover” effects and “learning bydoing,” by which firm-specific decisions

to invest in capital or in R&D, orindividual investments in human capital,can yield positive external effects(e.g.,on µ ) that benefit the rest of the

economy In these models, taxes can

then have long-term, persistent effects

on output growth However, thequestion still remains: what is themagnitude of these tax effects oneconomic growth?

A number of recent theoretical studieshave used endogenous growth models

to simulate the effects of a fundamentaltax reform on economic growth.6 All ofthese studies conclude that reducing thedistorting effects of the current taxstructure would permanently increaseeconomic growth Unfortunately, themagnitude of the increase in economicgrowth is highly sensitive to certainassumptions embodied in the economicmodels used in these studies, with littleempirical guidance or consensus aboutkey parameter values Consequently,these studies reached substantiallydifferent conclusions concerning themagnitude of the boost in growth rates

At one extreme, Lucas (1990) calculatedthat a revenue-neutral change thateliminated all capital income taxes whileraising labor income taxes wouldincrease growth rates negligibly At theother extreme, Jones, Manuelli, and

Rossi (1993) calculated that eliminatingall distorting taxes would raise averageannual growth rates by a whopping four

to eight percentage points.7 (An

“across-the-board” reduction indistortionary tax rates in these models,rather than complete elimination ofdistortionary taxes, would be expected

to have a smaller positive effect oneconomic growth.) Most recently, thesimulation model in Mendoza, Razin,and Tesar (1994) suggests relativelymodest differences in economic growth

of roughly 0.25 percentage pointsannually as the consequence of a 10percentage point change in tax rates.These simulation models of endogenousgrowth fail to provide a comfortablerange of plausible effects of taxes ongrowth and thus tend to raise morequestions than they answer Moreover,they are likely to miss many relevantcharacteristics of the U.S tax system Nomacroeconomic model allows for thepossibility of a firm undertakingfinancial restructuring to reduce taxableincome, or of timing issues in deferredtaxes, or the possibility of tax evasion.8Often the simulation analysis is per-formed in terms of a single flat-rate tax

in the context of a (single) tive agent model Ultimately, one needs

representa-to consider the empirical record representa-to makeinformed judgments about whether taxpolicy exerts a strong influence oneconomic growth

Below, we take three separate proaches to judge the empirical record.First, we take a quick look at the U.S.historical record to see if there is aneasily discernible link between changes

ap-in U.S tax policy and changes ap-ineconomic growth across time Second,

we consider whether differences ingrowth rates across countries can beattributed, at least partially, to variation

in tax policy Third and finally, we survey

the microlevel studies of how taxesaffect specific subsectors of theeconomy and build up from thesemicrolevel studies to make inferencesabout aggregate tax effects.

AN INFORMAL LOOK AT TAXES AND U.S

ECONOMIC GROWTHAnecdotal stories about the U.S taxcode can sometimes have a largerimpact on the policy debate than astack of statistical studies The KempCommission (NCR, 1996), for example,highlighted the complaint of onefrustrated businessman:

As an entrepreneur, I experience first handthe horrors of our tax system It has growninto a monstrous predator that kills in-centives, swallows time, and chokes thehopes and dreams of many We haveabandoned several job-creating businessconcepts due to the tax complexities thatwould arise

While this testimony is suggestive thatthe tax system adversely affects incen-tives, it is not entirely clear whether theentrepreneur is concerned about the taxrate per se or the complexity of the taxsystem more generally And we are notsure what fraction of entrepreneurs are

of like mind, or how much investment isaffected adversely by the tax code Forexample, surveys from a few decadesago indicate that typical businesspeopledid not view taxes as an impediment tobusiness decisions; in one study con-ducted in Britain in the early 1960s, not

a single executive out of the sample of

181 replied that they abandoned theintroduction of a new plant or equip-ment during the past seven yearsbecause of tax changes (Corner andWilliams, 1965).9 More recent surveystudies suggest a larger impact oftaxation on the discount rates used toevaluate private investment projects(Poterba and Summers, 1995); even

among these tax-savvy Fortune 1000executives, 36 percent reported that acorporate tax cut from 34 to 25 percentwould not make them more likely toengage in investment projects.10

A slightly more rigorous approach is tolook at the historical evidence fromtime-series changes in taxation andoutput growth The Kemp Commission’sreport (NCR, 1996) relied on time-seriescomparisons to argue that the patternsare self-evident:

America has experienced three periods ofvery strong economic growth in this cen-tury: the 1920s, the 1960s, and the1980s Each of these growth spurts co-incided with a period of reductions inmarginal tax rates In the eight years fol-lowing the Harding–Coolidge tax cuts,the American economy grew by morethan five percent per year Following theKennedy tax cuts in the early 1960s, theeconomy grew by nearly five percent peryear In the seven years following the

1981 Reagan tax cuts, the economy grew

by nearly four percent per year while realfederal revenues rose by 26 percent.This approach does not try to performthe “growth accounting” exercisedetailed in the theoretical section, butasks simply whether there are discern-ible differences in GDP growth followingtax cuts We consider the latter two taxreforms in Figure 2, which shows realGDP growth rates (both total and percapita) in the United States between

1959 and 1994 in the bottom panel,with the relevant tax series graphed inthe upper two panels.11 To smooth outyear-to-year volatility in GDP growthrates, we present three-year movingaverages of GDP growth rates in thebottom panel of Figure 2, both foraggregate growth rates and for percapita growth rates The two economicexpansions noted above during the1960s and the 1980s are apparent, as

FIGURE 2. Average Tax Rates, Marginal Tax Rates, and GDP Growth in the United States, 1959–95

are the other expansions followingrecessions (shown by the shadedregions) The general slowdown ineconomic growth over the last threedecades can be seen also.

Moving to the top panel of Figure 2, wenext consider the ratio of tax revenue toGDP—a commonly used measure of theaverage tax burden The top line showsU.S federal government revenue(measured on a National Income andProduct Accounts (NIPA) basis) as apercentage of GDP The lower line isstate and local government tax revenuemeasured on a NIPA basis as a percent-age of GDP Since 1959, the averagefederal tax rate has risen by about twopercentage points, but has generallyhovered around 20 percent of GDP; theaverage individual income tax rate hasremained relatively constant, whilegrowth in social insurance taxes havebeen mostly offset by the decline incorporate and excise taxes State andlocal government average tax burdenshave risen by about three percentagepoints over the last three decades

The Kennedy–Johnson tax cuts in 1964

resulted in a small decline in the averagetax rate Real GDP growth averaged arobust 4.8 percent over the subsequent

1964 to 1969 period However, theextent to which this growth was caused

by the tax cuts is unclear, as GDPgrowth had averaged over five percent

in the two years prior to 1964.

The Reagan tax cuts also lowered theaverage tax rate, and real GDP growthaveraged a healthy 3.9 percent from

1983 to 1989, significantly above thepreceding period from 1980 to 1982that was dominated by recession.12 But

it is a difficult task to sort out whetherthe strong growth during the 1980s wasthe consequence of supply-side effects

of lowering marginal tax rates, tional Keynesian aggregate demandeffects fueled by tax cuts and expandingdefense expenditures, or a recovery thatwould have occurred without the taxchange.13 Indeed, Feldstein andElmendorf (1989) suggest a somewhatdifferent cause for the 1980s expansion:expansionary monetary policy combinedwith a strong dollar and active businessinvestment

tradi-Over the longer term, since 1959, boththe average federal tax rate and theaverage state-local tax rate have risen—

by about two and three percentagepoints, respectively At the same time,average growth rates in real GDP havedeclined, from 4.4 percent during the1960s to only 2.4 percent in 1986–95.These coincident trends over the lastthree and a half decades are consistentwith the hypothesis that higher taxeshave stunted economic growth Beforearriving at conclusions about taxationand growth from this single observation(which does not account for otherfactors that were also changing over thistime period), we note that the averagetax rate series is unlikely to reflect the

marginal tax distortion, which economic

theory suggests is more important inaffecting economic growth throughhouseholds’ and firms’ choices ofsaving, investment, and employment.The middle panel of Figure 2 shows themarginal individual income tax ratesrelevant for households at the 75th,50th, and 25th percentiles of theincome distribution in each year(Hakkio, Rush, and Schmidt, 1996).14From 1960 to the early 1980s, marginaltax rates at the 75th percentile grewwhile marginal tax rates at the 25thpercentile declined slightly There wassome reduction in output growthcoincident with the increase in the

upper-middle class marginal tax rates.

However, GDP growth rates continued

to fall over the past decade even as themarginal tax rates for both upper- andlower-income households declined.15 Inother words, the time-series correlationbetween marginal tax rates and growthrates yields a decidedly mixed picture;

some decades were correlated positively,and others negatively

Finally, we correct the first sentence ofthe quotation from the Kemp Commis-sion above The most rapid growth rates

in this century were, in fact, during theperiod 1940–45, when output grew at12.5 percentage points annually Duringthis same period, the federal tax systemexpanded dramatically, with medianmarginal tax rates rising from 3.6percent in 1940 to 25 percent in 1945

Yet it would be ludicrous to claim onthat basis that higher taxes have apositive effect on output growth, giventhe obvious confounding events duringthis period Nevertheless, highlightingthe period 1940–45 is useful for twopurposes The first is that it illustratesthe risks of trying to discern incentiveeffects of taxation using short-termtime-series data This is a point rein-forced by the experience of Sweden’stax reform, when the economy fell into

a recession just after a tax reformtrimming marginal tax rates substantially(Agell, Englund, and Sodersten, 1996)

And second, it suggests that one shouldlook most carefully at GDP growth ratesbefore and after the early 1940s whenthe federal income tax experienced itsmajor expansion Stokey and Rebelo(1995) looked for this break in long-term output growth rates and wereunable to find any significant difference

On the other hand, given the majordisruptions in economic activity occur-ring during the 20th century, it may beasking too much of the data to detect

what might be very small differences ingrowth rates, on the order of 0.5 per-centage points, caused by thedistortionary effects of taxation

More formal econometric methods mayhold greater promise for uncovering thepure effects of taxation on economicgrowth, because that type of analysisattempts to control for other factorsthat affect output independently of taxpolicy The problem is that time-seriesanalysis is best suited for detectingshort-term effects of changes in taxpolicy on output growth, which, asnoted above, may reflect Keynesianexpansionary effects of deficit spending

or other unmeasured factors associatedwith tax cuts In addition, figuring out

which characteristics of a particular tax

reform—changes in top marginal taxrates, depreciation allowances, taxprogressivity, tax rates on capital gains—caused changes in growth rates isparticularly problematic in aggregatetime-series analysis For these reasons,

we turn our attention next to country studies

cross-TAX POLICY AND GROWTH: THE COUNTRY EVIDENCE

CROSS-An alternative empirical approach is todraw on the experience of differentcountries to investigate how tax policyaffects economic growth Countrieshave very different philosophies abouttaxation and very different methods ofcollecting their revenue During the pastseveral decades, some countries haveincreased taxation quite dramatically,while, in other countries, tax rates haveremained roughly the same Somecountries incorporated value-addedtaxation in the 1960s (e.g., France andBritain), while others shifted away fromcorporate taxation (the United States).The advantage of using such cross-

country comparisons is that we can usemany countries with different taxstructures and GDP growth rates to testfor correlation (and, one hopes, causa-tion) between tax policy and growth.

In general, studies of taxation usingcross-country data suggest that highertaxes have a negative impact on outputgrowth, although these results are notalways robust to the tax measure used

Using reduced-form cross-sectionregressions, Koester and Kormendi(1989) estimated that the marginal taxrate—conditional on fixed average taxrates—has an independent, negativeeffect on output growth rates Skinner(1988) used data from African countries

to conclude that income, corporate, andimport taxation led to greater reductions

in output growth than average exportand sales taxation Dowrick (1992) alsofound a strong negative effect ofpersonal income taxation, but no impact

of corporate taxes, on output growth in

a sample of Organisation for EconomicCo-operation and Development (OECD)countries between 1960 and 1985

Easterly and Rebelo (1993) found somemeasures of the tax distortion (such as

an imputed measure of marginal taxrates) to be correlated negatively withoutput growth, although other mea-sures of the tax distortion were insignifi-cant in the growth equations

Most empirical studies of taxation andgrowth are “reduced form” estimates inthat they specify a linear model ofoutput growth rates, with tax rates,labor resource growth, and investmentrates on the right-hand side of theequation However, taxes do notnecessarily enter the growth accountingframework in equation 1 in a linearfashion We explored this possibility inEngen and Skinner (1992), where theprimary growth effect of tax distortions

on production is hypothesized to

depress the economy-wide return oncapital, α, and on labor, β (as inequation 1 and Figure 1) Using cross-country data for 1970–85, Engen andSkinner found that an increase of 2.5percentage points in the average taxburden (total taxes divided by GDP) ispredicted to reduce long-term outputgrowth rates by 0.18 percentage points,holding constant the supply of invest-ment and labor

A recent McKinsey (1996) study points

to the potential importance of theintersectoral allocation of capital Thestudy observed that Japan and Germanyboth had much higher rates of invest-ment But because U.S investmentappeared to be allocated to moreprofitable (i.e., higher productivity)sectors, the net increment to theeffective capital stock, and hence tonational income, was considerablygreater in the United States, despite thelower investment rate Similarly, Kingand Fullerton (1984), in their study oftax systems in the United Kingdom,Sweden, West Germany, and the UnitedStates, found a strong negative correla-tion between economic growth and theintersectoral variability in investment taxrates.16

Of course, nearly any tax will tend todistort economic behavior along somemargin, so the objective of a well-designed tax system is to avoid highlydistortionary taxes and raise revenuefrom the less distortionary ones There issome evidence that how a countrycollects taxes matters for economicgrowth Figure 3, reproduced fromMendoza, Milesi-Ferretti, and Asea(1996), shows the correlation amongthe OECD countries between incometaxes and economic growth (panels Aand B) and consumption taxes andeconomic growth (panel C), over theperiod from 1965 to 1991 These scatter

plots, largely confirmed in regressionanalysis, suggest that income taxation ismore harmful to growth than broad-based consumption taxes.

It is useful to consider the growtheffects of a major tax reform using thesecross-country regression estimates

Suppose that marginal tax rates are cut

by a uniform five percentage points andaverage tax rates are cut by 2.5 percent

of GDP, leading to a (static) revenue loss

of $185 billion annually This cal tax reform was chosen because it is

hypotheti-on the outer fringe of politically feasibletax reform, losing more than twice asmuch revenue as the tax proposalsupported by presidential candidateRobert Dole Were such a plan enacted,the tax-to-GDP ratio would revert tolevels last seen in 1958 As noted above,the estimated coefficient from Engenand Skinner (1992) that ignores possiblechanges in the supply of capital andlabor implies an increase in long-term

growth rates of 0.18 percentage points.Including estimates of the responsive-ness of investment to the marginal taxrate from Mendoza, Milesi-Ferretti, andAsea (1996) suggests that this hypo-thetical tax reduction would increaseinvestment by 1.35 percent, boostingthe predicted growth rate effect of thetax cut to 0.32 percentage pointsannually.17

SANDTRAPS IN CROSS-COUNTRYECONOMETRIC ANALYSIS

To this point, we have been taking theresults of the cross-country econometricstudies at face value Any empiricalstudy must be treated with somecaution; but, in many of the studiescited above, particularly the cross-country studies, one must be particularlycareful in the interpretation of thecoefficients (Levine and Renelt, 1992;Slemrod, 1995) We consider justfour of these potential problems below

FIGURE 3B. Growth and the Labor Income Tax, OECD Countries

First, studies of taxation and growthmay find negative growth effectsresulting from taxation, but it is moredifficult to measure the potentialbenefits of the spending financed by therevenue collected The combined impact

of distortionary taxes and beneficialgovernment expenditures may yield anet improvement in the workings of theprivate sector economy (e.g., Barro,

1990, 1991a,b) An example of thedeleterious effects caused by theabsence of government spending comes

from the World Development Report

(World Bank, 1988, p.144):

According to the Nigerian Industrial velopment Bank (NIDB), frequent poweroutages and fluctuations in voltage af-fect almost every industrial enterprise

De-in the country To avoid productionlosses as well as damage to machineryand equipment, firms invest in genera-tors One large textile manufacturingenterprise estimates the depreciated capi-tal value of its electricity supply invest-ment as $400 per worker Typically, asmuch as 20 percent of the initial capitalinvestment for new plants financed by theNIDB is spent on electric generators andboreholes

That is, when the government of Nigeria

did not provide the necessary electricity

supply, private firms were forced togenerate electricity on their own, andpresumably at much higher cost Clearly,

a tax in Nigeria earmarked for (new)government expenditures on improvingthe electrical system would be likely toenhance economic growth even if thetaxes distorted economic activity Theproblem is that taxes are not necessarilyearmarked to those expenditures mostconducive to economic growth, eitherbecause of political “inefficiencies” orbecause of redistributional policies thatmay yield benefits for society but willnot be reflected in robust GDP growthrates (Atkinson, 1995).18 Thus, one

must be careful in interpreting thecoefficients on tax and output growthstudies to remember that these esti-mates reflect just one part—the costs—

of a combined tax and expendituresystem

Second, one should be very wary of thedata, particularly from developingcountries with large agricultural orinformal sectors where the measure-ment of income is difficult indeed.19Even in developed countries, it is wellknown that GDP measures suffer frombiases and mismeasurement of produc-tivity in service sectors, for example.20Measuring “the” effective tax rate iseven more difficult, given the widevariety of tax distortions, methodsfor measuring them, and variationacross countries in administrativepractices

Third, there are real difficulties withreverse causation; one does not knowwhether regression coefficients reflectthe impact of investment on GDPgrowth rates, for example, or thereverse influence of GDP growth rates

on investment, or both effects bined (Blomstrom, Lipsey, and Zejan,1996) Sometimes these biases creep inbecause of the way the regressionvariables are constructed Suppose onewanted to estimate an explicitly short-term relationship between the change inthe tax burden, typically measured asthe ratio of tax revenue to GDP, and thepercentage growth rate in GDP Anypositive measurement error (or short-term shock) in GDP will shift GDPgrowth rates up but also tend to shiftthe tax-to-GDP ratio down, therebyintroducing a spurious negative bias inthe estimated coefficient.21 One can try

com-to avoid such bias by introducing asexplanatory variables the percentagegrowth rate in the level of taxation, or

of government expenditures, rather