The Growth of Government A Public Choice Perspective DENNIS C. MUELLER*

has been given to the question of the growthof M[UCH ATTENTION in recent That has and

years. government government growndramatically,cannot be questioned.

grown,

I. The Facts Total government expenditure in the United States in 1984 asa perof national centage gross product (GNP) was 34.4 percent, up from 23 percentin 1949 and 10percentin 1929 (see Table 1). Moreover,this growth is neither confined to this century nor to the United States. Federal governmentexpendituresas a percentageof national income were only 1.4 percent of nationalincome in 1799. They rose to double that figure by the end of the nineteenth century, but were still only 3 percent of GNP in 1929. Startingin the 1930s, however,federal expenditurestook off, risingsevenfold as a percentageof GNP over the next 50 years (Table 1). Table 2 presents comparabledata for the United States and other Organizationfor Economic Cooperation and Development (OECD) countries. Although spanninga much shorter timeperiod, these data reveal across-the-board increases in the relative sizeof governmentin The OECD smallest country. governmentsector, in Switzerland, every accounts for 30percent of gross domesticproduct(GDP). In contrast, governmentexpendituresin Denmark and the Netherlands absorbed more than 60percentof GDP in 1982. Nor is thegrowthof government *Mr. Mueller is Professor ofEconomics atthe Universityof Maryland,College Park;he is a graduateof ColoradoCollege and PrincetonUniversity.This paperwas writtenwhile the author was aVisitingScholarin the FiscalAffairs Departmentof the Fund. He thanksPeter Coughlin,WallaceOtes, and parSeminaron March10-20, 1986,for ticipantsof the IMF Institute'sPricing Policy helpful discussions of thetopic. 115

International Monetary Fund is collaborating with JSTOR to digitize, preserve, and extend access to Staff Papers - International Monetary Fund ® www.jstor.org

116

DENNIS C. MUELLER

Table 1. GovernmentExpenditurein Relation toNationalIncome and GrossNationalProduct in theUnitedStates,1799-1985 Federal, Total Total State, and Federal Federal Total Total Local Federal Federal Expendi- Federal, Expendi- Expenditures tures State, Expendi- Expendi- tures (As percenttures tures (As (In and Local (In millions (As percent- percent- millions Expenditures age of of current national age of age of of current (As percentdollars) income) GNP) GNP) dollars) age of GNP) Year (1) (2) (3) (4) (5) (6) 10 1.4 1799 10 1.1 1809 21 2.4 1819 15 1.6 1829 27 1.6 1839 42 1.7 1849 66 1.5 1859 316 4.6 1869 5.0 3.7 267 3.2 1879 309 2.9 2.6 1889 563 3.4 1899 694 2.3 1909 16.7 1919 12,402 3.0 1929 3.0 10.0 10,300 3,100 11.7 9.7 1939 19.4 17,600 8,800 16.2 15.0 23.0 1949 38,800 59,300 1959 92,100 18.9 26.8 131,000 19.4 30.4 1969 183,600 286,800 20.8 31.1 1979 503,500 750,800 23.3 1,258,100 34.4 1984 851,800 1985a 959,100 Sources:Figuresfor columns 2 and 3 are from Kendrick (1955, pp.10-12; for column3, Kendrickreportsdata fromKuznetz).Figuresfor columns4, 5, and 6 are from United States, EconomicReportof the President(1985 and 1986, Tables B-1, B-72, and B-74). a Preliminary figure. outside of the United States limited to the post-World War II period. Table 3 presents data for three European countries and the United States that indicate a considerably smaller government sector before World War I than after it and a considerably smaller government sector after World War I than after World War II. The figures in Tables 1, 2, and 3 suggest three questions. What caused the increase in the relative size of government over the past two centuries? What caused the growth of government to accelerate after World

117

THE GROWTH OF GOVERNMENT

Table 2. General GovernmentExpenditureand Revenue in Relation to Gross Domestic Productat Current Prices in OECD Countries (In percent) 1960

1982

Expenditure Revenue Expenditure Revenue Country 22.1 25.4 36.3 Australia 34.2 32.1 31.0 50.3 Austria 46.7 30.3 27.5 56.6 45.5 Belgium 28.9 Canada 26.0 56.5 45.4 24.8 27.3 60.7 50.7 Denmark 26.7 30.0 41.3 Finland 39.7 34.6 34.9 50.7 France 46.9 49.4 45.3 Germany,Fed. Rep. of 32.5 35.1 17.4a 21.1 37.04a Greece 31.3 28.2 36.4 34.4 36.0 Iceland (1960-80) 28.0 24.8 57.1 Ireland(1960-81) 42.3 30.1 28.8 53.7 41.5 Italy 20.7 34.2 18.3 30.2 Japan 30.5 32.5 54.3 51.5 Luxembourg(1960-80) 33.7 63.7 Netherlands 33.9 55.8 29.9 33.1 48.8 52.8 Norway 17.0 17.6 42.7 33.2 Portugal (1960-81) 18.8 34.1 18.8 30.6 Spain (1964-81) 31.1 32.2 67.3 Sweden 59.7 Switzerland

17.2a

23.3

Turkey (1962) United Kingdom United States

19.1 18.0 32.6 30.3 27.6 27.3 37.6 32.0

Mean (unweighted)

26.3 27.7 47.0

Coefficientof variation 0.24 0.21 0.22

30.0a

33.2

47.4

43.7 41.9 0.21

Source:Saundersand Klau (1985, p. 29). Current disbursements only. in Relation Table 3. TaxLevels in theMajorWestern Countries to NationalIncome, 1900-01 to 1924-25 (In percent) 1924-25 1913-14 1900-01 Country 20.01 14.11 14.96 France 29.21 10.51 Germany,Fed. Rep. of 7.99 24.77 11.29 9.99 United Kingdom 11.06 6.68 7.76 United States Source:Seligman(1925, Chapter1), as reportedin Tanzi(1986b, p. 5).

118

DENNIS C. MUELLER

WarII? Whatexplainsthe large disparitiesin levels andratesof growth in governmentamong countries? Public choice, the economic analysisof political institutions,would seem to be the naturalanalyticaltool for answeringthese questions,and in recentyearsit hasfrequentlybeen usedfor this task. Areviewof these efforts follows.

II. Explanationsfor the Size and Growthof Government The same explanationsthat have been given for why government exists should, logically, explainwhy it attainsa givensize in onecountry and not in another, or why it starts to grow at a more rapid rate at a particulartime. Thus, in reviewingthe hypothesizedcauses forthe size and growthof government,one is essentiallyalso reviewingthe explanations for the existenceof government.If each explanationis represented and rates ofgrowth as a variableor a variableset, then differences in size in size must beexplainedby differencesin these variables. Government as Provider of Public Goods and Eliminator of Externalities The traditionalexplanationfor why governmentsexist is that they exist to provide public goodsand to eliminate or alleviate externalities. Let us assume that this is theonly functiongovernments perform. Each citizencan then bepositedto have a demand for the public goods,which is a functionof the individual'sincome, the relativeprice of public to private goods, and perhapsother taste variables.If it is assumed that voting takes place usingmajorityrule, that citizens votedirectlyon the governmentexpenditureissue, and that theonly issue to be decided is the levelof governmentexpenditures,then one canapplyBlack's(1958) median votertheoremand writegovernmentexpendituresas a function of the characteristics of themedianvoter.1LettingX be a compositeof privategoods and G the compositeof publicgoods (with Pxand Pgbeing their respective prices),Ymthe income of the median voter, and Z a vectorof tasteparameters,then onecan write agovernment expenditure equationin logarithmsfor the median voter: + 'yIn Z + xL. InG = a + a In Pg+ 3 In Ym

(1)

See Barrand Davis(1966), Davis and Haines(1966), Borcherding and Deacon (1972), Bergstromand Goodman(1973), and Deacon (1977a, 1977b).

THE GROWTH OF GOVERNMENT

119

An explanation for the relative growth of government can be obtained from equation (1) if any of the following conditions are met: * the demand for public goods is inelastic (-1 < a), and Pghas risen relative to Px * the demand for public goods is elastic (-1 > a), and Pghas fallen relative to Px * because Ymhas been increasing over time, for changes in Ymto explain growing G relative to X, P must be greater than unity * some taste variable could change in the appropriate way, given the sign of '.2 Let us start with the last possibility. The very definitions of public goods and externalities connote geographic proximity. The smoke from a factory harms more individuals in a densely populated community than when the population is thinly dispersed around the factory. A park is easier to reach and probably of more utility in a densely populated community than in a rural area. Increasing urbanization has occurred throughout the last century in every developed country and has been taking place for well over a century in most. Urbanization or population density is an obvious choice for a Z variable with a predicted positive sign on its y. It is surprising, therefore, how little empirical support for this hypothesis one finds.3 No other "taste" variable has garnered both compelling a priori and empirical support. Nor can one account for much of the relative increase in government size from increasing personal income. Although some estimates of the income elasticity of demand for G((p) are greater than unity,4 a greater number are less than unity, and very few indeed are significantly greater than unity.5 2For discussionof these possibilitiesrelative to the growth of government issue, see Borcherding(1977a), Buchanan(1977), and Bennett and Johnson (1980, pp. 59-67). 3See Borcherding (1977a)and Deacon (1977b), but for a critique of this literature seeOates (1985a). Most work inestimatingequation (1)has been at the local governmentalunit level, and many problems in public goods and externalitiesmay be resolvedat higherlevels of governmentalaggregation.But Mueller andMurrell(1985) did not find apositiverelationshipbetweengovernment expendituresand urbanizationacrosscountries. 4Deacon (1977b) has noted that in most studiespark and recreationexpenditures appearas income-elastic. 5There is good reasonto believethatexistingestimates ofthe incomeelasticity of demand for G, based on state and local cross-sectionaldata, are biased downward.Most studies assumethat the cost of providing government services is the same acrosscommunities.But a givenlevelof safety maybe providedmore cheaplyin a wealthycommunitythan in a poor one. Thus theprice of safety is lower inwealthycommunitiesand, given that theprice elasticityof this service

120

DENNIS C. MUELLER

local governExistingstudiesall estimateP usingdata from state and ment jurisdictions.6Most redistributiontakesplace at the nationallevel, however, and redistributionhas been one of the most rapidlygrowing componentsof federal expenditures.Estimates of13based onstate and local governmentdata may not be reasonable approximationsof the incomeelasticityof redistributionexpendituresat the national level.But estimatesof the incomeelasticityof charitycontributionsalso tend to lie below unity, suggestingthat thisadjustmentalso would notaccount for the growthof government(Clotfelter(1985, Chapter 2)). The remainingcandidate for explaining government growthis the price elasticityof demand.Most estimatesof a suggestthat it is significantly greaterthan -1 and thus imply a relativegrowthin government, if there has been a relative increase in its price. Baumol (1967) has argued that we might expect a relative increase in the price of government-provided"goods," given that many of them (education, police protection) are services. Because productivityincreases come typicallyembodied largelyfromtechnological change,and this in turn is in capital equipment,there is less potentialfor productivityadvancesin service sectors such asthe government. Althoughthe argumenthas intuitiveplausibility,it is not clear how far it can be pushed. The militaryservicesare quite capital-intensive today researchand developand spend vast sums on productivity-enhancing ment. Similarly,computers, xerography,and other innovationshave increasesin many white-collarjobs. Thus itis not brought productivity apparenta priori that productivityincreasesin governmentcould not keep pace with those in the private sector, at least with those in the privateservicesector. But it appearsthatthey have not.A fair consensus is less than unity, wealthy communities willconsume less, other things being communiequal. With thepriceof governmentservices heldconstant across all ties, this wealth-priceeffect is shifted to the incomeelasticity,biasingit downward(Hamilton (1983)).SchwabandZampelli (1986)observed ajumpin p from near zero to unity when this income-pricerelationshipwas properlyestimated. But in terms ofaccountingfor the long-rungrowthof government,this adjustment merelyshifts some of theexplanationof governmentgrowth,usingequation (1), from the price term to the income term. The Hamilton-SchwabZampellicritiqueimpliesthat the growthof income should, other thingsbeing equal, bring down the cost of providinggovernmentservices, thus partially offsettingthe "Baumol effect"on price. The total effect ofchangesin income on expendituresmeasuredby Schwaband Zampelliwas roughlyzero. 6 MuellerandMurrell (1986)estimatedgovernmentsize relativeto GDP at the national level.Although alwayspositiveand oftensignificant,the coefficients on income in theirequationswere too small toprovidemuch of anexplanationof the growthof government.

THE GROWTHOF GOVERNMENT

121

that suggeststhat govexists amongstudies of government productivity ernmentproductivitylags privatesector productivityand may in factbe zero or negative.7As Buchanan (1977, pp. 8-9) has noted, lagging productivityin the governmentsector may be more symptomaticof the "problem"of government growththan the cause of it. Be that as it may, direct testing and measurement ofthe "Baumol effect" are nonexistent.Pommerehneand Schneider(1982) offered an indirecttest by comparingtheirpredictedlevels ofgovernmentoutputin each Swissmunicipality, assuming zero productivity growth in the public with actual levels. a correlation between sector, They interpreted high their predictedexpendituresand actualexpendituresas supportfor the Baumol effect. If Pommerehneand Schneider'sresults are assumed to confirm the existence of a Baumol effect, how much of the growthin government does that effect explain?8Some parts of the governmentbudget (for example, puretransfers,interest payments)are difficult tothink of as "goods" whoseprice rises relativeto private goods.The budget component for which Baumol's effect seems most appropriateis perhaps what the OECD characterizes as "final consumption"-that is, the goods and servicesactuallyabsorbedby government.Finalconsumption expenditures for the OECD countries from 1960 through 1983 are presentedin Table4. If it is assumed that the priceof governmentgoods and services roserelative to the price of private goods by2 percent a year, governmentserviceswould be61.6 percentmoreexpensiverelative to privategoods in 1983than they were in 1960. If theprice elasticityof demand for governmentservices were -0.5,9 then the Baumol effect would imply a 12 percent relative increase in final consumptionexpenditures.This figure comes close to the 14 percentincreaseactually observedfor the United States between 1960 and 1983, but it falls conshort of the of 50 increase siderably average percentobservedacross all of the countries listedin Table4. Thus, at best, the Baumoleffect seems capableof explainingabout a quarterof the increasein final consumption expendituresfor the averageOECD country. 7See in particularFuchs (1968), Gollop and Jorgenson (1980), Ross and Burkhead(1974, Chapter 6),and the discussion inPommerehneand Schneider (1982, pp. 312-13). 8 It is interestingto note thatPommerehne andSchneiderestimatedthe Baumol effect forthe OECD countrywith the smallestgrowthin governmentsince WorldWar II. 9This figure seems reasonable from the studies surveyed by Borcherding (1977a, p. 49).

122

DENNIS C. MUELLER

Table4. Total Final Consumption Expenditure Government of in Relationto GDP, 1960-83 (In percent) 1971 1960 1975 Country 13.6 19.2 20.0 21.0 Canada 13.0 13.4 14.4 France Germany,Fed. Rep. of 13.4 16.9 20.5 20.0 15.5 15.4 19.5 12.8 Italy 10.1 10.2 8.0 8.0 Japan 16.4 17.8 21.8 22.0 United Kingdom 19.1 16.9 18.5 United States Averageof the above

13.4 15.6 17.3

Australia Austria Belgium Denmark Finland Greece Iceland Ireland Luxembourg Netherlands New Zealand Norway Portugal Spain Sweden Switzerland Turkey

9.4 15.4 17.6 12.5 17.2 13.0 14.8 12.4 14.1 16.4 17.7 13.3 21.3 24.6 27.2 11.9 15.2 17.1 11.7 12.5 15.2 10.0 11.1 12.3 8.5 12.5 15.2 18.6 14.9 17.3 11.7 9.8 17.4 12.8 16.0 10.7 12.9 14.8 17.9 19.3 19.5 12.9 10.5 13.5 15.0 7.4 9.2 8.6 22.5 15.8 23.8 12.6 8.8 10.9 10.5 13.4 12.6 10.8

Averageof the above

11.3 14.3

1983 16.3

19.3 18.3 18.7 19.4 18.8 20.2 17.7 17.0 14.6 12.3 28.5 13.5

16.2 17.8

14.7 16.7 11.9 Overallaverage Source:OECD, as reportedin Tanzi(1986a, p. 17).

17.9

Government as Redistributor of Income and Wealth The government giveth and it taketh away. Several writers have criticized the view that government exists to provide public goods and alleviate externalities, arguing that this is essentially a normative description of government-a theory of what government ought to do-not a description of what it actually does. These writers argue that a positive theory of government must analyze the

THE GROWTH OF GOVERNMENT

123

redistributivenature of government activity.Aranson and Ordeshook (1981) pressedthe point most forcefully, emphasizingthat all government expenditureshave a redistributivecomponent.Roads mustbe built in this location or that. The contracts to construct the roads mustbe given to one set of firms, to the loss of all others. As Aranson and Ordeshookview it, to understandwhat governmentis and why it grows, one must analyzeits redistributiveactivities. Meltzer and Richard(1978, 1981, 1983) have presentedperhapsthe simplest and yet most elegant publicchoice analysisof the growth of government.Theirmodel presumesthat all governmentactivityconsists in redistribution.Thisredistribution occurs by means ofper capitalumpof financed from a earned sum grants r, proportionaltax of t levied on all income. If y is mean per capitaincome, a balancedgovernmentbudget implies r=ly.

(2)

An individual'sutility depends on his consumption,c, and leisure, 1. Letting n be the fraction of total time worked, one has the identities I =1 - n (3) c=(1 - t)y +r.

(4) and Richard Meltzer assumed that income depends on an ability or productivityfactorx, which is randomlydistributed across thepopulation. Given the hours oneworks, n,one's incomeis higherthe higheris one's x factor: y =nx.

(5)

Given t and r, an individual'sonly choice is how much to work, n. Maximizing U(c, ) with respectto n, given equations (3)-(5),one gets, as a first-ordercondition, Uc(1 - t)x = UI

(6)

=(1 - t).

(7)

or

The marginalrate of substitutionbetween leisureand consumptionis of an individual's time.From equatedto the net-of-taxmarginal product equation (7) one can obtain the number of hours an individual works. For the specific case of a Stone-Geary utilityfunction, U = ln (c + y) + a In(l + X), one obtains for optimaln

124

DENNIS C. MUELLER

(1- t)(1 + )x - a(r + y)8 (1- t)(1 + a)x The denominator of equation (8) must be positive, but with small enoughx the numerator can benegative. Obviouslyn cannot be negative, thus there is acritical level ofability,x0, at whichoptimaln = 0, at or below whichan individualchooses not to work.Lettingn = 0, one can see from equation (8) that a (r + 1)

(9) ) (1 -t(1 Althoughr and tare exogenousfromthe point of view of the individual, they are endogenousto the political system.Substitutingequation (8) backinto the individual'sutilityfunctiondemonstrates thatthe individual'sutilityultimatelydependson r and t.Whenchoosingr and t, the rational voter considers this and takes into account the relationship betweenr andt given byequation(2). Now ay/lt < 0. Meanincomefalls as the tax rate rises because ofthe negative incentiveeffects of higher taxes on effort.10 Thusr is afunctionof t, risingat adiminishingrate until = and then falling (see Figure 1). Voters who work have -dy/dt ylt indifference curvessuch asU' and U2(U2> U1), since positively sloped taxes lower and increased subsidiesraise it.Voterswho do higher utility not work do not have their utilities affected by changes in t. Their indifference curvesare horizontalstraightlines such asU3 andU4, with U4> U3. Each rational voter recognizes that r =yt constitutes the opportunityset in choosingt (or r). Each voter choosesthatt - r combination along the r = yt curve that maximizes his or herutility. Voters who do not workall choose thetothat maximizes the lump-sumtransfer. t than to.If all votershave thesame The voterwith x > x0 favors a lower utility function and differonly in their abilityfactors, x, voters with a higherx havesteeperutilityfunctionsand favor alowert. The voters are in essence confronted with a one-dimensionalchoice, with t uniquely definingr. The median voter theorem can beapplied, and the preferences of the median voter are decisive. If U1 and U2 are indifference curves for the median voter, then tm - r is the optimal tax-subsidy

combination. Meltzer andRichardsuggestedthat one explanationfor the secular growthin governmentin this countryand aroundthe worldover thepast two centuries hasbeen the expansionof suffrage.Those added to the 0Note that as t rises more individuals choose not to work: axo/lt = a ( + a)/(1 + X)(1 - t)2> 0.

125

THE GROWTHOF GOVERNMENT

U4 U3

r=yt

t,

t

t

Figure 1. OptimalChoiceof t voting rolls are more often than not those with incomes (ability levels) below the median. Thus the median voter changes to someone with a flatter indifference map in Figure 1, with tangency further to the right. A decline in the income of the median voter relative to the mean has the same effect. Increased inequality of income as well as increased suffrage were thus given by Meltzer and Richard as the primary causes for the growth of government, with some empirical support for their hypothesis presented. Peltzman (1980) has also presented an explanation for the growth of government that depends on the shape of the distribution of income. Peltzman did not, however, make use of the median voter theorem in developing his argument. Rather, he envisaged a form of representative government in which candidates compete for votes by promising to redistribute income toward those voters or groups of voters that agree to join the candidate's coalition of supporters. Peltzman reasoned that the more equal was the distribution of income among the potential supporters of a candidate, the more bargaining strength they would have. Thus the candidate must promise a greater amount of redistribution the

126

DENNIS C. MUELLER

more equal is the initialdistribution ofincome amongvoters. Peltzman pointedto the spreadof educationas animportantfactorincreasingthe equalityof pretransferincomesand thusleadingto a growthin the size of government.Peltzman'shypothesisdependson increasingequalityof income to drive thegrowthof government,whereas Meltzerand RichBut Peltzman, too,found empirical ard'srests on increasing inequality. data he examined. in the his for hypothesis support Both the Meltzer-Richardand Peltzmanpapersdiscussed therole of governmentas if governmentwere to engage exclusivelyin redistribution. Aranson and Ordeshook (1981) and Brunner(1978) also placed primaryemphasison government'sredistributiveactivities. Butif redistributionis the primaryactivityof government,then some additional logicalargumentsare missingto explainthe growthin governmentto the sizes now observed in different countries. Alternatively,government activityis not exclusivelyredistributive. Governmenthas grown to far greater size than is necessary justto achieveredistribution.If onegroupor a coalitionof groupscanmake use of the democraticmachineryof governmentto achievea greatershare of the pie, then onewould thinkthat thegroupor coalitionoughtto be able to do so in such a way as not to use up such a large fraction of thepie in bringingabout theredistribution.The numberof programsand people making up governmentseems much larger than necessary justto achieve redistribution(see, for example, Table 4). A second set of logicaldifficultiesbesets Meltzer andRichard's useof the median voter theorem to explain the growthof government.This theorempostulatesa single-dimensionalissue space in which issues are decided by a directvote of the individual citizens."Outsideof Switzerland, no countrymakes muchuse of direct (as opposed to representative) democracy,and even in Switzerlandit is not employedin all cantons, much lessat the nationallevel. Nor arepoliticalagendaslimitedto single-dimensionaltax-subsidyor expenditureissues. Now it is common in economicsto assumethat the outcomesof a theoremhold in the real world evenif the assumptionsunderlyingthe theoremare only roughly satisfied. But usuallywhen this is done there is no alternative theorem whose assumptionsbetter fit the conditionsobservedin the real world. Here this is not the case. Theoremsdo exist about what happenswhen issue spaces are multidimensionaland representativedemocracyrather than direct democracyis used. Manyof these theoremssuggestthat no variantsexist, but theassumptionsunderlyingthese are so 1 Multidimensional extreme that it is highly improbablethat any political body voting on actual issueswould eversatisfythem; see Plott (1967), Sen (1970, pp.37-55), Enelow and Hinich (1984, Chapters3 and 4).

THE GROWTH OF GOVERNMENT

127

equilibriumoutcome may result. Giventhese alternativetheorems,it is not clearhow muchweight one can give to the very specific predictions model. that emerge from the Meltzer-Richard An additionalconceptualdifficultysurroundsMeltzer andRichard's relianceon the extensionof suffrageas an explanationof the growthof productivityand income government.The addition of individuals with below median changes the identity of the median voter, making the formermedian voter no longer decisive and worse off thanbefore the extensionof suffrage.Thus an extension ofsuffragethat wouldresultin a new median voterof lower income would never beapprovedunder majorityrule.12 If the sizeof governmentis endogenousto a political theorem isapplicaprocessthat usesmajorityrule and the median voter ble, then the extensionof suffragecannot be endogenousto the same politicalsystem. This difficultycan be circumventedby assumingthat changesin suffrageare governed bysome other form ofvotingrule, such as aconstitutional amendment. Butthis way out of the logical problemraisesother difficulties. Constitutionalchanges require even more support than a simple majority.How can thesechangescome about, given their redistributionalimplicationsin a world of rational, self-interested voters? Moreover,other constitutionalchangeswith importantredistributional to the U.S. Constiimplicationshave occurred,such as the amendment tution allowing Congressto levy an incometax. If othervotingrulesand perhapsother motivationsexplain these actions, how can one be sure that these other factorsdo not also affect the sizeof government? Both Meltzer and Richard and Peltzman'3assumed that all redistribution isfrom rich topoor. But this characterization of the redistribution thatgovernmentinducesdoes not fit the facts. Some redistribution is in the direction of the rich,'4and much redistributionis difficultto 12 Of course an expansionof suffragethat added voterswith above-median income to the list of voters would also changethe identityof the median voter. Such an expansionshould also beopposed bythe medianvoter andall voters to with lower incomes-if thehigher-incomeindividualswho would be added the voter rollswere already payingtaxes-since the new voters wouldfavor lower taxes andtransfers,makingthe previousmajorityworse off. Butthe additionof higher-incomecitizensto the tax base-say, after the annexationof a wealthy eventhoughsuch community-mightbe favoredby a previouswinning majority, additionwouldchangethe identityof the medianvoter, since the expansionof the tax base mightmore thanmake up for the loss in capacityto chose the tax subsidyrates (see Mueller(1979, pp. 134-45)). 13Peltzman(1980) backedawayfrom this assumptionat the end of hispaper 285-87). (pp. 14 of special FratianniandSpinelli(1982)have notedthe increasing importance programsto help businessin their discussionof the growthof governmentin Italy.

128

DENNIS C. MUELLER

or poor. Indeed, if allgovernmentactivitycan categorizein terms of rich be characterizedas some form ofredistribution,the mostsalientfeature of thisprocessof redistributionis probablyits lackof a single-directional flow. 5 The multidimensional charactergovernment of redistributionmakes it difficultto rationalize allgovernment activity aspurelyredistributionally motivated.If all government programs simplytake fromone groupand if to and all citizens participateat both ends of the redisgive another, from theprocess?Whydo notthe citizens who tributionalprocess, gains simply abolish the governmentand save the tremendous deadweight There must either be some clear losses from zero-sum redistribution? gainersfrom theredistributionalprocess,who arein apositionto sustain and enlarge their gains, or all governmentactivityis not purely redistributionalin character.If the formerpossibility explainsthe growthof government,who are those gainingfrom governmentand how do they achieve theirgoalswithin the rulesof a democraticprocess?If some, not insignificant proportionof government activity is not purely redistributional,but say is to provide publicgoods, then one again has a logical problemin explaininggovernment growthas a result of redistributionalstruggles. Once it is admitted that a large componentof government expenditure is to provide public goods, then all redistributionalobjectivescan be achievedsimplyby changingthe taxshares of individuals orgroupsof individuals.16 One typicallydoes not have to a to or spend money on, give money to, group give that group greater command overprivategoods. Despite these criticisms,it is difficultto suppressthe impressionthat an importantcomponentof the explanationfor the growthof government lies in government'sredistributionalactivities, so substantial has been the growthin the transfercomponentof governmentbudgets, as Table 5 shows (see also the discussion in Tanzi (1986a)). But these argumentsmake clear that the hypotheses put forward so far, which attemptto explain the growthof governmentin simple redistributional terms, are inadequate. Some additionalelements are needed to complete the story. Two villains often mentioned as instrumentalin the growthof governmentare interestgroupsand bureaucrats. 5Aransonand Ordeshook(1981)emphasizedthe multidimensional character of the redistributionalprocess, as did Brunner(1978) to some extent. 16See Muellerand Murrell (1985). Of course those groupsthat pay no taxes to finance thepublicgoods portionof the budgetcan be subsidizedfurtheronly by an expenditureor transferprogram,but notenoughgroupssuch asthis exist in most countries. Onemightobject to accountfor currentgovernment activity that tax cuts cannot always be designed to benefit specific groups,but the numberof tax loopholes and the complexityof tax-loophole legislationbelies this point.

129

THE GROWTHOF GOVERNMENT

Table 5. GovernmentExpendituresin Relation toGDP in the United States and FifteenDevelopedCountries(DCs), 1953-74 (In percent) Expenditure Category Total government United States Averageof 15 DCs SD of 15 DCs CV of 15 DCs

1953/54

1958/59

1963/64

1968/69 1973/74

27.5 27.0 28.0 31.1 32.2 29.9 31.7 28.9 35.8 39.4 4.1 4.3 5.9 4.8 7.2 14.1 14.2 15.0 16.6 18.3

Defense 12.25 9.9 United States 4.05 3.3 3.24 Averageof 15 DCs SD of 15 DCs 2.46 1.55 1.32 1.13 CV of 15 DCs 60.7 46.5 40.7

8.45 8.95 5.7 2.83 2.51 39.9

Transfersa 5.5 6.7 United States of 14 DCs 11.9 Average 4.3 SD of 14 DCs CV of 14 DCs 36.4 32.5

7.5 12.9 13.8 4.2 4.3 4.9 31.0

Nontransfer,nondefense 9.2 United States Averageof 15 DCs 12.64 SD of 15 DCs 2.89 CV of 15 DCs 22.9 18.5

13.4 10.9 12.05 14.7 13.54 16.61 2.11 3.27 2.32 19.7 14.4

1.03 41.1

8.7 11.0 16.2 18.8 5.9 30.3 31.6 15.5 18.37 3.47 18.9

Sources:Peltzman(1980), as reportedin North and Wallis (1982). All sample data arefrom national accounts of OECD countries;U.S. data are fromUnited States, EconomicReport ofthe President. Note: Samplecountries areAustralia,Austria,Belgium,Canada,Denmark, Finland,France, the FederalRepublicof Germany, Italy, Japan,the Netherlands, Norway, Sweden, Switzerland,and the United Kingdom. "SD of 15 countries"is the standard deviationfor 15(14) country sample;"CV" is the coefficientof variation. aTransferpaymentswere not brokendown separatelyfor Switzerland.

Interest Groups and the Growth of Government The pioneering publicchoice analysisof the questionof government

size might be regarded as Tullock's (1959) classic discussion of majority

rule. Tullockpresentedan examplein which acommunityof 100farmers votes on proposals to repair access roads, which benefit only a few farmers each.Using a majorityrule, a coalition of 51 farmersis predicted, with a political outcome in which the only roads repairedare those that service the 51farmersin the winningcoalition.Because these 51 farmerspay only 51 percentof the costsof repairingtheirroads, they

130

DENNIS C. MUELLER

vote to have themmaintained at ahigherlevel of repairthanthey would if they had to coverthe full costs.Thus, majorityrule might be said to lead to a level of governmentexpendituresthat is excessive, relative to the Paretooptimallevel thatwould occurunder theunanimityrule, in one of two senses.First, more isspent repairingthose roads thanwould be spent under theunanimityrule. Second, if the unanimityrulewere in use, there would be no incentive to have the government (thatis, the communityto whichthe 100farmersbelong)repairthe roads at all.Each smallgroupof farmers couldagreeamongthemselves torepairtheir own roads. Therepairof access roads would not be apublicissue at all inthe of 100. community WhereasTullock'sexampleof road repair nicelyillustrates howgovernmentmay become too large under majorityrule, it also illustrates some of the troublesome questions raised earlier. Ifa coalition of 51 farmers canimpose taxes on their neighborswithoutthe neighborsreceivingany benefits, why do not the 51simplytake the money as a cash transfer andrepairthe roadsthemselves atthe optimallevel, rather than make suboptimallylarge road repairsthroughthe government? The logrollingunder majorityrule that Tullockdescribesresembles the kind of pork-barrellogrollingunder majorityrule that one often associates withpressuresfrominterestgroups. Virtually everystudythat discusses thegrowthof governmentmentionsinterestgroupsas a possible cause and then goes on to discusssome other cause forthe growth of government, leavingthe role of interestgroupsin limbo. The questions of whether, how,andwhy interestgroupscontribute tothe growth of governmentremainlargelyunanswered. If one begins to ask these questions, the first issue one confronts is whether interestgroups pursueredistributive favors from government, or the provisionof public goodsand alleviationof externalities. A moment's reflectionbringsto mindthe names ofinterestgroupswith each objective. A farmers'cooperative seekingprice supportswouldseem to belong to the category of a redistribution-orientedinterest group, a naturalists' clubwantingto save the whales is inthe categoryof a club seeking a publicgood. Thus in terms of the overtobjectivesof interest groupsit wouldseem possible, if such groupsdo play an importantrole in explaining government growth,that interestgroups expandeitherthe redistributiveor the public good expenditurecomponents,or both. Becker (1983) has developed a model of the influence of interest groups(or, as he callsthem, pressuregroups)that isrelevant tothe issue at hand. At first glance it appearsthat Becker is analyzingthe purely redistributivegainsof interestgroups.He states that"the basicassumption of the analysisis that taxes, subsidies,regulations,and other polit-

THE GROWTH OF GOVERNMENT

131

ical instruments are used to raise thewelfareof moreinfluentialpressure He assumes asingle good,income, to be distribgroups"(pp. 373-74). uted among the interest groups, with some groups getting a subsidyS and otherspayinga tax R. The summation ofthe subsidiesreceivedover all interestgroupsis less than the taxes paid, however, because of the transaction costs ofcollecting and distributingsubsidies andthe deadincomesare taxed and weightlosses fromadverseincentive effects when subsidiesgranted. to increase(reduce)its subsidies(taxes). Each group applies pressure With diminishingreturnsto applyingpressureassumed,an equilibrium is obtained in which the pressureto increase subsidiesjust equals the opposing pressureto reduce taxes. Foreach groupthe marginalcost of applyingadditionalpressurejust equalsthe marginalgain from reduced taxes or increasedsubsidies. Several, not totally obvious propositions follow from this analysis. One is that "competitionamong pressure groups favors efficient methodsof taxation" (Becker (1983, p. 386)). Those groupsthat can becheaplysubsidizedor are expensiveto tax do better. In particular,"politicallysuccessfulgroupstend tobe small relative to the sizeof the groupstaxedto pay their subsidies"(Becker(1983, p. 385)). AlthoughBecker'sanalysisis entirelycouchedin terms oftaxes, subsidies, and regulations-as if all government activitywere of a redistributivenature-one of the implicationsof the analysisis that thegovernmentdoes provide publicgoods and alleviate externalities whenever the collectivegains from these activities exceedthe transactioncosts of bringingthem about. "Politicalpolicies that raise efficiency are more likely to be adoptedthan policies that lowerefficiency" (Becker (1983, p. 384)). This propositionhas broadimplicationsfor the types of activities that are subsidizedand taxed and for the types of interest groups helped and harmed.If education haspositive spillovereffects, teachers' for their "indusunions will have more successwinning public support union. the a Moreover, than, supportis more likely say, plumbers' try" to take the form of more fundsfor more teachers thanof higher wages for existingteachers.Factor ownersin theliquorandcigaretteindustries in the are morelikely to sufferfrom excise taxesthan are factor owners because of the externalities perceivednegative plumbinggoods industry from drinkingalcohol and smokingtobacco. Becker'spaper gives insightinto howinterestgroupactivityand government activity might be linked. It also demonstrates that the expendituresand taxes that interestgroups bringabout have more than public merelyredistributivecharacteristics.Groupswhose interests have good or externalityattributesare morelikelyto be successfulthanthose

132

DENNIS C. MUELLER

seeking pure redistribution.Groups whose productiveactivities have negativeexternalities are morelikely to be taxed. North and Wallis (1982) have linked the growth of governmentto interest group activity. They drew a parallel between the growth of governmentand thegrowthof white-collar andmanagerialemployment in the privatesector. Bothwere seen as a responseto the greatertransactioncosts fromorganizinga marketeconomywith increasingspecialization. "Growing specializationalso created a host of new interest groups"(p. 340). The demands thesegroups presson governmentare not simply for a redistributivehandout but are to alleviate the transactioncosts these groupsbearwithinan increasinglyspecializedsociety. Thus, as with Becker's analysis, the influence of interest groups on governmentactivityis seen as havingboth an efficiency-enhancingdimension as well as a redistributive dimension.North and Wallis presentedthe data in Table5, taken fromPeltzman(1980), to illustrate that nondefense, nontransferexpendituresof governmenthave grown fasterthan totalgovernmenthas grownsince World War II, and almost as fast as transfers. The only directevidencethat interestgroupsaffectthe sizeof government presented to date is by Mueller and Murrell(1985, 1986). They described apolitical processin whichparties supplyinterestgroupswith favorsin exchangefor the interestgroups' support.When these favors take the form ofgoods targetedto specificinterestgroups,but with some spilloversfor other groups, governmentgrows larger. The numberof organizedinterestgroupsin a countrywas shown to have apositiveand significanteffect onthe relative size ofthe governmentsectorin a crosssectional sample of OECD countries for theyear 1970. The analysesof Becker andMuellerand Murrellare essentiallystatic, describingan equilibriumin which interest-grouppressuresare in balance, an equilibriumin which governmentis biggerthan itwould be in the absenceof the influenceof interestgroups. Theydo not describe a process in whichinterest-groupinfluence leads togrowinggovernment size. Olson (1982) has discussed the conditionsfavoringthe growthof interest groups, however, and Murrell(1984) has presented evidence consistent with Olson'shypotheses concerningthe causes of interestgroup formation. The stable economic and political environmentin Westerndevelopedcountries since World War II will have facilitated the to Olson'sthesis, and this growthin growthin interestgroups, according turn may help to explain the relatively poor macroeconomicperformance of some countries inthe 1970s and 1980s. If the number of effective interestgroupsin developedcountrieshas grownsince World WarII, then their growthcould help to explain the relativegrowthof

THE GROWTH OF GOVERNMENT

133

government. Government growth and macroeconomic inefficiency would, in turn, be tied together. Unfortunately,however,the possible interrelationshipsof interest groups, government growth,and macroeconomic performanceremainlargely unexplored.17 Bureaucracy and the Growth of Government Governmentprogramsdo not come into existence merely because some interest group wants them and the legislatureauthorizesthem. They must be "manufactured."More often thannot, the supplierof a programis partof the governmentitself, a governmentbureau.Government may grownot only becauseincreasingexpendituresare demanded by citizens, interest groups, or legislators, but also because they are demanded by the bureaucracysupplying governmentprograms.The governmentbureaucracies are an independentforce, whichpossiblymay lead to increasing government size. To understandthe problem,considerFigure2. Let G be a measurable outputfrom apublicprogram-say, the numberof studentssuccessfully output,somehow completinga yearof school work. The demand for this revealedthroughthe political process,is the lineDD'. Let the minimum cost of supplyingthis outputbe scheduleSS'. The optimalquantityof G to purchasefrom the point of view of the communityis Go, where the marginalbenefit to the communityof another studentsuccessfullycompleting a year of schoolingjust equals the marginalcost. The minimum total costof supplyingthe optimal outputis Po Go.But, if members ofthe school-systembureaucracydesire a greater budget expenditurethan PoGo, it may occur. Why?Because a bureaucracydoes notsimplyoffer up a supplyschedule such asSS' to the legislatureand ask thelegislatureto pick a point on it. Instead,it offers up a total budgetB = PBGBthat may exceed the optimal budget from the point of view of the demanders because GB> Go, or PB> PO,or both. How can the bureaucracyget the legisof legislaturelatureto "buy"this higher budget?Two characteristics the work in of the favor bureaucracyin achieving bureaucracy relationship this end. First, the legislatureand the bureaucracyare typicallyin a bilateral needs thelegisbureaucracy obviously monopolysituation. Whereas the laturein that thelegislaturemustapproveall expendituremeasuresif the 17Mueller and Murrell(1985, 1986)made allowance for interest groupsand governmentsize both being endogenousvariables.

134

DENNIS C. MUELLER

D

St

Po

S

Go

D'

G

Figure2. Demand and Supply ofa Bureau'sOutput

bureaucracyis to be sustained,the legislaturealso, at least in the short run, is in equal need of the bureaucracy.If the bureaucracyrefuses to supply any output, allows a significantdeterioration ofquality, or the like, it may be membersof the legislaturethat bearthe hostilereaction of the citizenryat the next election, not members of thebureaucracy.If BO= POGois the optimalbudgetfromthe point of view of thelegislature andBBthe optimalbudgetfromthe perspectiveof the bureaucracy,then by analogywith other bilateralmonopoly situations one canprobably expect a budget B*, with Bo < B* < BB, to emerge from the bargaining

process. The bureaucracyis aided in its bargainingwith the legislaturein that the legislaturetypicallyhas at best a vagueidea of what SS' is-or often even of what G is, for thatmatter. Indeed, it is often preciselybecause the good cannot be supplied in well-defined, divisible unitsthat the governmentis in the business of supplyingit. The outputs of a police department(crimeprevention),nationaldefense department (security), and a healthdepartmentare very difficult to measure.Even with education there are definitionalproblems. Althoughone can certainlycount the numberof students whoare in school andcomplete a school year, determiningwhetherthey have "successfully"completedthe year-that

THE GROWTH OFGOVERNMENT

135

is, how much they have learned during the year-is another matter. Indeed, one typicallymust rely on members ofthe bureaucracyitself, the experts,to supplythis information. These same bureaucratic experts are also often theonly ones who can define the cost of obtaininga given level of output, however defined. Here the monopoly nature of the bureauhelps it to achieve itsgoals because thelegislaturehas no other supplier of the good or service with which to compare costs. If the bureauclaims it cansupplyG1at a total cost ofB1, and G2at total cost of B2, it is often difficultfor the legislatureto countereffectivelythat it instead demandsG1at total cost B3< B . That the bureaucrats will wish to obtain a Bgreaterthan Bo seems almostobvious. Evenif the bureausuppliesGo, it seems natural that it wouldpreferan effective unitpricefor supplyingGothat isgreaterthan PO.This extra revenue could be used to offer higher salaries, more leisure (because of a largerstaff), more perquisites (paidtravel to conventions), and a whole host ofamenitiesthat mightmake abureaucrat's life on andoff the job more pleasant.But the powerof the bureaucracy to obtain these benefits should not be exaggerated. Salaryincreases are of visible exercises bureaucratic very power;traveland otherperquisites can often be easily monitored.A wise legislatureshould be capableof exercisingsome control over increases inP at a given G. One way sometimes used tojustify a largersalaryis to increaseGthat is, to expand the bureau's output-and then to demand higher salariesthat allowfor the expandeddemandsplacedon the bureaucracy. Niskanen (1971, p. 38) postulated that a bureaucrat's "salary, perare all quisitesof the office, publicreputation, power [and] patronage" of to the size the bureau. Niskanen this related uses postulate positively the to analyzethe consequencesof assumingthat bureaucrats maximize size of their budgets.Not surprisingly,the modelimplies largerbudgets than are desiredby the legislativedemanders. Niskanen'sanalysishas become the theoreticalunderpinningfor animportant partof the literature on the growthof government. The model of the budget-maximizingbureaucrat has acertain resonance with modelsof the corporationthat assumethat managersmaximize the corporation'ssize, its growth in size, or other size-related variables such as white-collar staff (Baumol (1959), Marris (1964), Williamson(1964)). The behavioralunderpinningand empirical support for these modelscan to someextent be cited insupportof the postulate bureaucrat. But onemustnot be too quickto of the budget-maximizing generalize. The managerof a companywith $10 billion in sales may be able to a justifyto the boardof directors and stockholderslarger salarythan he

136

DENNIS C. MUELLER

could if the company had sales of $1 billion, and company size and managerial compensationare positivelycorrelated.But the head of a bureau with abudget of $10 billion does not necessarilyget paid more than that ofa bureau with abudgetof $1 billion. Salaries acrossgovernment bureaucraciestend to be much more uniform than are salaries across companies. Moreover, the top officers in bureausare typically yearsat most.Thus political appointees,who stay at the bureau for four if size of the size and the even bureau, salarywere positively expanding not to benefit would be related, directly the bureaucrat who likely broughtabout the increase. If the growthof bureaus benefits thetop members of the bureaucracy,it must in general be from the nonpecuniarydimensions of a bureaucrat's rewardsthat accompanythe growthin a bureau's size. Even at middlelevels, salarylevels do not differ muchacrossbureaus. Undersecretariesearnthe sameregardlessof whichdepartmentthey are in. But the chances for promotion in a rapidly growing bureau are thanin a shrinkingone. Thus,middle-levelbureaucrats certainly greater do have a financial incentiveto encouragethe rapid expansionof their bureausbecause itincreases the likelihood of their promotionto a higher rank. Career bureaucrats are alsolikely to be with the bureau long enough to benefit directlyfrom the expansion,unlike their short-term superiors. Although this analysis provides a rationale for the promotion of growthin size by middle-level careerbureaucrats,it greatlycomplicates to fulfill theirgoals to the the story of why these individuals are allowed at thetop of the bureaudo not benefit loss of society. If the bureaucrats from the growth in bureau size, why do they not curtailgrowth?Are middle-levelbureaucratsable to deceiveboth thelegislativeoverseersof the bureau and their superiorswithin the bureau about the truemagnitudes of P and G ?18

Bureaucratsand interestgroupsstand equallyhigh on all lists of the causes of the growth of government, but empirical support for the is sparse. Perhaps the best evidence in bureaucracy-size relationship favorof the hypothesisis thatindicatingthat governmentbureaucracies do havehigherunit coststhanprivatefirmswhentheysupplymeasurable outputs,such as tonsof garbagecollected. Borcherdingdescribes this as "the BureaucraticRule of Two... removalof an activityfrom the priIf unit costs riseby vate sectorwill doubleits unit costs ofproduction."19 8For further critical discussion of the bureaucracy-size-of-government hypothesis, see Musgrave(1981, pp. 91-95). '9Borcherding (1977a, p. 62). For a summary of the evidence, see Orzechowski's(1977) paper in the samevolume (Borcherding(1977b)).

THE GROWTH OF GOVERNMENT

137

this much when directcomparisonswith privatesector alternatives are possible, how muchmore are they inflated when thebureaucracyknows it cannotbe subjectedto a comparisonwith privatemarket alternatives? Romer and Rosenthal(1978, 1979, 1982) have developedand tested a level a model in which bureaucrats can force voters to choosehigher of government expenditures than the level voters most under the prefer of in set institutional which the choice is made. rules Consider particular for the medianvoter Figure3. Let U1, U2, and U3 be indifference curves in a communitydefined over unitsof private goodsX consumedby the voter and publiclyprovide G. Let BB' be the budget constraintline facing the voter, given his or her tax share of thecosts of providingG. The medianvoter'smost preferredquantityof G, givenhis tax share and budgetconstraintBB', is then Go. If the quantityof G for the community were determinedby majorityrule with all levels ofG as possible outcomes, Go would be the chosen level of the public good for the community.

Private goods B

ul

U2 \U'

goods

BudgetModel Figure3. Government Size in a Bureaucrat-Determined

138

DENNIS C. MUELLER

Suppose,however,that thecommunitydoes notget to choose from all possiblelevels of G, but is simply givena choicebetween GI, a level of expendituresproposedby the bureau thatsuppliesG, and G2, a reversion level set by the constitutionto take effect when thebureau'sproposed level fails to achieve a majorityof the votes in a referendum. Given thechoice betweenG1and G2, the median voter prefersG . Ifthe reversion level of expendituresis below the most preferredlevel of expendituresof the median voter, the budget setter can induce the medianvoter to vote for a level of expendituresabove his or herideal point if the budget setter does not choose a level too much above the ideal point (for example, above G4). The kind ofinstitutionalsettingjust described issufficientlyrare that it is unlikelythat much of thegrowthof governmentcan be explained by a directapplicationof Romer andRosenthal'smodel. But theevidence they presentfor school districtsin Oregonis importantbecause it does imply that the school supervisorsand superintendentswho make the budgetproposalsare cognizantof the powerthey have to induce larger in power.If other bureaucrats budgets,and that the bureaucrats use this that bureaucratic other contextsare similarlymotivated,the hypothesis power increasesgovernmentsize is a possibly significantfactor in explainingthe size of government. Both Niskanen's (1971) model and that of Romer and Rosenthal (1978, 1979, 1982) are static. These models explain why government mightbe largerthanthe legislaturewouldpreferif it knewthe unit costs of the outputs it thought it was buying, and why the level of outputs might be largerthan the medianvoter's mostpreferredquantity.They do not directly explainwhy government grows. Indirectly,however, they perhapsdo offer an explanation.The bureaucracy'sability to expand the budget beyondthe amount thelegislature or citizens demanddependsin part on its abilityto misrepresent the trueprices and quantitiesof publiclyprovided goods.The abilityto misrepresentis likelyto dependin turnon the size andcomplexityof the budget itself. The biggerthe bureaucracyis, the more difficult it is for outsidersto monitor itsactivity,and themore insiders there are who are workingto increase the sizeof the bureaucracy.Thus thegrowthof the bureaucracyis likely to depend on its absolute size. To see this relationshiplet us define G, as the amount of publicly providedgoods the citizens or legislature trulydemand. Let B, be the total size of the budget. B, is greaterthan Gtto the extentto whichthe bureaucracyis capable of forcing a greater expenditureof resources towardthe bureaucracythan is demanded;that is, B, = at Gt,

att

> 1.

(10)

THE GROWTH OF GOVERNMENT

139

Now let = eaBt

(11)

and let the amount of publicly provided goods demanded grow at a constant rate nequal to, say, the growthin nationalincome: Gt = cen'.

(12)

Then Bt = ceaent.

The growthin the budget, g, is then g = lnB, - InB,_ = a(B, - B,-1) + n.

(13)

(14)

The growth rate of the budget both exceeds the growth in national income, n, and increases withthe absolute difference betweenthis period's and last period's budget.Other functional forms for oa,will yield other relationshipsbetweeng and B,; so long as a, increases withbudget size, however, the growthin the size of the budget can be expected to increase with its absolute size. Equation (13) is broadly consistent with the pattern of growth of pasttwo governmentexpenditureobservedin the United States over the centuries: slow butsteady initial growth, gradually shifting into more rapidrates ofgrowth.But this samepatterncould beexpectedfrom any process that resulted in an S-shaped growth path for government. S-shaped growthpaths for organizationsare sufficientlycommon that one should not claimtoo much fora hypothesisthat bureaucratic power increasessize of governmentmerelybecause thehypothesisis consistent with at least the initial segment of the S, and so too are the data. Governmentgrowthcannotoutpacenational incomeindefinitely,and so the subsequentslowdown in growth characteristicof the S must also come. Indeed, the conservativegovernmentsthat in recent years have taken office in the United States, the United Kingdom, the Federal Republicof Germany,and several otherdevelopedcountriesmay have alreadyinitiatedthe slowdown. This consideration raises a more general point. Given the steady past two growthin governmentboth since World War II and over the centuries,any variableexhibitinga steadygrowthover the same period, or for that matterany variable with a stable trend factor, is likely to be with Tullock correlated (1983) has made a government growth. highly similarpoint in critiqueof Meltzer and Richard's(1983) empiricalsupport for their hypothesis.A key variablein their model is the ratio of median to mean income. This ratio has been virtuallyconstant since

140

DENNIS C. MUELLER

WorldWar II, yet it "explains"a significantfraction ofthe growthof government. Testinghypotheses by using time-seriesdata that have strong trend componentsis a familiarproblem.That it is presentin the literature on the growthof governmentis unfortunate ifunsurprising.The existence of the problem reinforces the importanceof those studies that try to explainlevels of governmentexpenditureby using cross-sectionaldata, even thoughthese studiestypicallydo rely on static rather than dynamic models of governmentsize. Fiscal Illusion The hypothesisthat bureaucraticpowerincreasesthe size of government presumesthat the bureaucracycan deceive thelegislatureabout the true costs ofsupplyingdifferent levelsof output. The fiscal illusion hypothesis presumesthat the legislaturecan deceive the citizens about the truesize of government.Tanzi(1980) has tracedthe argumentback to Mill and also cites Paretoas a source for thehypothesis.But it is to Puviani (1903) that credit mustgo for emphasizingthe importanceof fiscalillusion to apositive theoryof government(see also the discussion in Buchanan(1967, pp. 126-43)). The fiscalillusion explanationfor governmentsize assumes that citibill. Tobring zens measurethe size ofgovernment bythe size of their tax about an increase in governmentsize, for which the citizens are not willingto pay voluntarily,the legislative-executiveentities must increase the citizens'tax burdenin such away that the citizensare unaware that they are payingmore in taxes, or be willingto pay the price of citizen displeasureat the next election. If tax burdenscan be disguisedin this way, citizens havethe illusion thatgovernmentis smaller thanit actually is, and governmentcan grow beyondthe levels citizensprefer. Althoughthispropositionseems reasonableenough,to developit into a model to be used for explainingthe size andgrowthof governmentone must make somespecificassumptionsabout thekindsof tax burdensthat can bedisguised.Mill (1970, pp.220-21) felt thatdirect taxes were more visible and, by implication,that excessive governmentgrowth would have torely on indirecttaxes. But the citizensof Bostonhad no illusions of the British taxon tea two centuries about the burden ago, and one can like bank collection income of that taxes, argue employers'withholding of propertytaxes with mortgagepayments,makesthese formsof direct taxation lessvisible than sometypes of indirecttaxation,such asliquor and cigarettetaxes. The issueof what sourcesof revenueare less visible

THE GROWTH OF GOVERNMENT

141

to citizens, as well as the magnitudeof any fiscal illusioncaused, must be regardedas largely empirical. The empiricalliteratureon fiscal illusionis too large to be reviewed here. Fortunately,Oates (1985b)has recentlycompleteda comprehensive review of it. He breaks the fiscal illusionargumentinto five subhypotheses:(1) a taxburden ismore difficult tojudgethe morecomplex is the tax structure;(2) renters are less able to judge their share of propertytaxes in the communitythan arehomeowners;(3) built-in tax increasesbecauseof the progressivityof the taxstructure are less clearly elastic taxstructuresmore perceivedthan arelegislated changes, making conduciveto government growththan are inelastic structures;(4) the implicit future tax burdens inherentin the issuance of debt are more difficult to evaluate than are equivalentcurrenttaxes; (5) citizens donot treat lump-sumcash subsidies to their governmentas being as much theirs asthey would a cashsubsidyto themselves(the "flypaper" effect). Each of these hypotheses implies a relationshipbetween the size or growth of governmentand the relevant fiscal illusionvariable. Oates carefully examined the evidence in support of each, and concluded ".. . that although all five cases entail plausible illusion hypotheses,

none of them havevery compelling empiricalsupport"(Oates (1985b, p. 26)). Fiscalillusion, like theother hypothesesreviewedin this paper, does not by itself provide a persuasiveexplanationfor the growth of government. The lackof strongempiricalsupportfor thefiscal illusionhypothesis, despiteits intuitiveappeal, maybe due tothe rathervague wayin which fiscal illusion has been defined and modeled inthe literature. Forexample, it is not clear from the literature whetherfiscal illusion isa kind of short-runmyopia on the part of voters that allows for temporaryincreasesin expenditures,or a kind ofpermanent astigmatism indefinitely obscuringthe true size of government.The latter is obviouslya much strongerhypothesis.The taxrevoltsin Europe and the United States in the 1970s,andthe deficit revolt ofthe 1980s,symbolizedso dramatically in the United States, sugby the Gramm-Rudman-Hollings legislation that fiscal illusion not gest may permanentlyimpairvoters'vision. Evensee the and rise up to chain Leviathan. tually, they may light From a public choice perspective,the most seriousdeficiencyin the literature on fiscalillusion isthe gap between thepotentialexistence of fiscal illusion onthe part of voters andits consequencesfor the size of government.Who isit withinthe governmentthattakesadvantageof the fiscal illusion, bureaucratsor elected politicians?Is it perhapsgroups outside of governmentwho gain-perhaps interest groups?If it is insiders,why do they not simplyincrease theirsalariesratherthanexpand

142

DENNIS C. MUELLER

personneland programs?Althoughfiscal illusionexplains why government could grow larger than citizens prefer, it does not provide an explanationfor why it would. Such anexplanationcan be foundonly in the other theories of governmentgrowth examined here. To account fiscal illusion must be combined with completelyfor government growth, some other hypothesis that can supply the driving force behind the growthof government.Some suggestionsare presentedin the concluding section. (For further critique of the fiscal illusion hypothesis, see Musgrave(1981, pp. 98-104) and Oates (1985a).) III. Conclusions The five possible explanationsof governmentsize reviewedin this paperstem fromtwo quite differentconceptualizationsof the state. The first three hypotheses (governmentas provider of public goods and eliminatorof externalities,governmentas redistributor ofincome and wealth,andinterestgroupsas inducersof governmentgrowth)are essentially drawnfrom a classicaltheory of the democraticstate (Pateman (1970)). Ultimateauthoritylies withthe citizens.The state existsto carry out "the will of thepeople." State policies are reflectionsof the preferences of individualvoters. In thepublicchoice literature thestate often appearsas simply a voting rule that transformsindividualpreferences into political outcomes. Most of the classic works on public choicefrom Arrow (1951; revised edition1963), Downs (1957), Black (1958), and Buchanan and Tullock (1962) up through Groves and Ledyard (1977)-are based on this citizen-over-stateview of the polity. The fourthand fifthhypothesesreviewedhere (bureaucraticpoweras fiscal illusionas enhancerof government inducerof government growth, growth)place the state above the citizens. It is the preferencesof the state, or of theindividualsin the government,that aredecisive.Citizens and political institutionsconstitute at most (loose) constraintsagainst whichpoliticalleadersand bureaucrats pursuetheirown personalinterests. This state-rules-citizenview of politics underliesPuviani's(1903) work and characterizes that of Niskanen (1971) and Brennan and Buchanan(1980). If either of these two conceptionsof the state isfully accurate,20 then the other must berejected-and so too the set of hypothesesassociated with it in this paper. But both viewsmight be correct to somedegree. Government officials and bureaucrats may have discretionarypower to 2 Tanzi(1980)has discussed both of these conceptionsof the state-as well as a third, the paternalisticstate-in the context of the fiscal illusionissue.

THE GROWTH OF GOVERNMENT

143

advancetheir own interestsat the citizens'expense to some degree, but citizens'preferences,as registered through existingpoliticalinstitutions, mayalso constitutea consequentialconstraint. Ifso, then allfive hypotheses may help to explainthe size and growthof government. Only one of the studies reviewed here-the paper by Pommerehne and Schneider (1982)-attempts to explain the size of government througha modelthat incorporatesboth viewsof thestate. Theseauthors first estimatedthe demandfor governmentfor 48 Swissmunicipalities, which operate under direct (as opposed to representative) democracy. from thisequationto simulate They then used the estimated coefficients what the levels of governmentexpenditurewould be in the 62 Swiss municipalitiesthat haverepresentativedemocracy. Theyfind that allof the individualspendingcategoriesare underestimatedfromthe parameter estimatesbased on the direct democracies. Therepresentativedemocraciesspent 28 percentmore than one wouldpredicton the basis of estimatedoverthe direct democracies. the expenditure equation The use of a representativeform of government changesthe nature ofthe political outcomes substantially,making government considerablylarger than it would beif citizensdirectlydeterminedoutcomes. Moreover,in those Swissmunicipalitiesin whichrepresentativedemocracyexists, the size of governmentis smaller if the citizens have the right to call a referendumand therebyreverse agovernmentdecision. These results of Pommerehneand Schneidersuggestratherstronglythat the existenceof between the citizens andpolitical a layer of representative government would outcomesexpandsthe size of thepublicsectorconsiderably. They appearto supportthe state-over-citizenview of government,and Pommerehne and Schneider (1982, pp. 319-22) interprettheir results as indicatingthe importanceof "the supply side of local services." This inferencemay be warranted,but it is alsopossible that the existence of representativedemocracyfacilitatesthe attainment ofprivategains by interestgroups. Both Peltzman(1980) and Mueller and Murrell(1985, 1986) have seen the growthof governmentas a by-productof the competition for votes between candidates and parties. Thus government growth (or size) in these models is dependent on the representative nature of the democratic process, although the models assume that are citizens'preferences,as channeledby interestgroup representation, the drivingforce behindgovernment programs. Regardlessof which of theseinterpretationsis correct, Pommerehne and Schneider's results forcefullydemonstrate whatmay be regardedas the single, most importantmessage publicchoice has to teach-the rules of the game do affectthe outcomes of thegame. Institutions matter.In on politicaloutcomes, Switzerland,the more direct is citizen influence

144

DENNIS C. MUELLER

of government.Among the developedcountries, the smaller is the scale citizens of Switzerland are able to exercise control overgovernment more effectivelythan anywhereelse. Only Switzerland makes much use of directdemocracyand the referendum,and ithas thestrongestfederalistsystemin theworld.It alsohas thesmallestpublicsectoramongthe developed countries (see Table 2). The results of Pommerehne and Schneidersuggest that these facts are related. Pommerehneand Schneider'sstudy,as well as the others reviewedfor to that exists the hypothesis government provide publicgoods and to eliminate externalities(see SectionII, the firstsubsection),explainsthe relative size of governmentacross political jurisdictions.It also sheds light on the third of the three questionsposed in the opening section: whatexplainsthe largecross-countrydisparitiesin the levelsand ratesof governmentgrowth?Of course, any staticmodel ofgovernmentsize can each variable become a dynamicmodel ofgovernment growth by writing as a percentagechangeratherthan astaticlevel. Amongthe explanatory variablesin the cross-sectional modelsof governmentsize, increasesin price and interest group strength,and perhapsincome, appear as the most promisingcandidatesto explain the growth of government,although it is doubtful that changes in these variables canexplain all of governmentgrowth. Meltzer and Richard(1978, 1981, 1983) and Peltzman(1980) have addressedthe issue of thegrowthof government directly.The key variable in theiranalysesis the skewness of the income distribution,the ratio of median to mean income.This variablechangesso slowly over time that it isvery difficultto believe that it hasbeen the drivingforce behind around the world.The expansionof suffrage,which government growth Meltzer and Richard emphasized, also does not seem to be a likely cause, particularlyin Europe. More promisingas anexplanationof the growthof governmentis the hypothesisof bureaucraticpower. Bureaucraticpower stems frombuof absolutesize, which reaucraticsize. Growth in size is then a function leads to the predictionthat governmentgrows exponentially.The bureaucraticpowerhypothesismightalso explainthe ratcheteffect in government growth, to which attention was first drawnby Peacock and Wiseman(1961). The expansionof governmentbureaucracyin response to an external shock such as a war increases the bureaucracy'sdiscretionarypower, allowingit to achieve a permanentincrease in size. The citizen'srole in a representativedemocracyis morepassivethan in a direct democracy,and Pommerehne and Schneider's (1982) work suggeststhat even this difference leadsto a significant fillipto government size. Today'scitizen, confrontedby expandedand morecomplex

THE GROWTH OF GOVERNMENT

145

government structures at the local, state, and federal levels, must feel that he is more a passive spectator of the democratic process, as he watches a campaign commercial on television, than did the citizen of 150 years ago. How much of the growth of government in the intervening years can be explained by a slackening of the reins of government in citizens' hands, and how much is a reflection of the preferences of citizens transmitted through the political process, remains, alas, a somewhat open question. REFERENCES Aranson,Peter H., and Peter C.Ordeshook,"Regulation,Redistribution,and Public Choice," Public Choice (Dordrecht), Vol. 37 (No. 1, 1981), pp. 69-100. Arrow, KennethJ., Social Choiceand IndividualValues,2nd ed. (New York: Wiley, 1963). Barr, James L., and Otto A. Davis, "An ElementaryPoliticaland Economic Theory of the Expendituresof Local Governments,"SouthernEconomic Journal(ChapelHill, NorthCarolina),Vol. 33 (October1966),pp. 149-65. Baumol, WilliamJ., BusinessBehavior, Value andGrowth(New York: Macmillan, 1959). , "Macroeconomics ofUnbalanced Growth:The Anatomy of Urban Crisis,"AmericanEconomicReview(Nashville,Tennessee),Vol. 57 (June 1967), pp. 415-26. Becker, GaryC., "A Theoryof CompetitionAmong PressureGroupsfor Political Influence," QuarterlyJournal of Economics (New York), Vol. 98 (August 1983),pp. 371-400. Bennett, JamesT., and Manuel H.Johnson, ThePoliticalEconomyof Federal GovernmentGrowth,1959-1978(CollegeStation:Texas A & MUniversity Press, 1980). Bergstrom,TheodoreC., and Robert P.Goodman,"PrivateDemands forPublic Goods," AmericanEconomic Review(Nashville, Tennessee), Vol. 63 (June 1973), pp. 280-96. Black, Duncan, The Theoryof Committees andElections(Cambridge:CambridgeUniversityPress, 1958). Borcherding,Thomas E. (1977a), "The Sources of Growth of Public Expendituresin theUnited States, 1902-70,"in Budgetsand Bureaucrats: The Sourcesof GovernmentGrowth,ed. by ThomasE. Borcherding (Durham, North Carolina: DukeUniversity Press),pp. 45-70. , ed. (1977b), Budgets and Bureaucrats: TheSources of Government Growth(Durham,North Carolina:Duke UniversityPress). , and Robert T.Deacon, "The Demand forthe Servicesof Non-Federal Governments,"AmericanEconomic Review(Nashville, Tennessee), Vol. 62 (December 1972), pp. 891-901.

146

DENNIS C. MUELLER

Brennan, Geoffrey, and James M. Buchanan, The Powerto Tax: Analytical Foundationsof a Fiscal Constitution(Cambridge: Cambridge University Press, 1980). Brunner, Karl, "Reflectionson the Political Economy of Government: The PersistentGrowth of Government,"SchweizerischeZeitschrift farVolkswirtschaftund Statistik(Bern), Vol. 114 (September1978), pp. 649-80. Buchanan,JamesM., Public Financein aDemocraticProcess: Fiscal Institutions and theIndividual Choice Press, (ChapelHill: Universityof North Carolina 1967). , "Why Does Government Grow?," in Budgetsand Bureaucrats: The Sourcesof GovernmentGrowth,ed. by ThomasE. Borcherding(Durham, North Carolina:Duke UniversityPress, 1977), pp. 3-18.

, and GordonTullock, Calculusof Consent(Ann Arbor: Universityof MichiganPress, 1962). Clotfelter,CharlesT., Federal TaxPolicyand CharitableGiving(Chicago:Universityof ChicagoPress, 1985). Davis, Otto A., and G. H. Haines, Jr., "A PoliticalApproachto a Theory of Public Expenditures:The Case of Municipalities,"National Tax Journal (Columbus,Ohio), Vol. 19 (September1966), pp. 259-75. Deacon, Robert T. (1977a), "Private Choiceand Collective Outcomes:Evidence from Public SectorDemand Analysis,"National TaxJournal(Columbus,Ohio), Vol. 30 (December), pp. 371-86. (1977b), "Reviewof the Literatureon the Demandfor PublicServices," in NationalConferenceon NonmetropolitanCommunityServicesResearch, Paperspreparedfor U.S. Senate, Committeeon Agriculture,Nutritionand Forestry,95th Congress,1st session, July 12, 1977 (Washington:Government PrintingOffice), pp. 207-30. Downs, Anthony,An EconomicTheory ofDemocracy (NewYork:Harperand Row, 1957). Enelow, JamesM., and Melvin J.Hinich, TheSpatialTheory of Voting(Cambridge:CambridgeUniversityPress, 1984). Fratianni,Michele, and FrancoSpinelli, "The Growthof Governmentin Italy: Evidencefrom 1861to 1979,"Public Choice(Dordrecht),Vol. 39 (No. 2, 1982), pp. 221-43. Fuchs,VictorR., The ServiceEconomy (NewYork:ColumbiaUniversityPress, 1968). Gollop, Frank M., and Dale W. Jorgenson, "U.S. ProductivityGrowth by Industries,1947-73," in New Developmentsin ProductivityMeasurement andAnalysis,ed. by JohnW. Kendrickand BeatriceN. Vaccara(Chicago: Universityof ChicagoPress, 1980), pp. 17-124. Allocation ofPublic Goods: Groves,Theodore,andJohnO. Ledyard, "Optimal A Solutionto the 'Free Rider'Problem,"Econometrica(Chicago),Vol. 45 (May 1977), pp.783-809. Hamilton,Bruce W., "The FlypaperEffect and OtherAnomalies,"Journalof Public Economics(Amsterdam),Vol. 22 (December 1983), pp. 347-61.

OF GOVERNMENT THEGROWTH

147

Kendrick,M. Slade, A Centuryand a Half of FederalExpenditures(New York: NationalBureauof EconomicResearch, 1955). Marris, Robin,The EconomicTheory of'Managerial' Capitalism (London:Macmillan, 1964). Meltzer, Allan H., and Scott F. Richard, "Why Government Grows(and Grows) in a Democracy,"Public Interest(New York), Vol. 52 (Summer 1978), pp. 111-18. , "A RationalTheory of the Size of Government,"Journalof Political Economy (Chicago),Vol. 89 (October 1981), pp. 914-27. , "Testsof a RationalTheoryof the Size of Government,"Public Choice (Dordrecht),Vol. 41 (No. 3, 1983), pp. 403-18. Mill, John Stuart,Principlesof PoliticalEconomy (Harmondsworth, England: PenguinBooks, 1970). Mueller, Dennis C., Public Choice(Cambridge:CambridgeUniversityPress, 1979). , and Peter Murrell, "Interest Groups and the Political Economy of GovernmentSize," in PublicExpenditureand GovernmentGrowth,ed. by FrancescoForte and Alan Peacock(Oxford:Basil Blackwell, 1985), pp. 13-36. , "InterestGroups and the Size of Government,"Public Choice (Dordrecht), Vol. 48 (No. 2, 1986), pp. 125-45. Murrell, Peter, "An Examination of the FactorsAffecting the Formation of InterestGroupsin OECD Countries,"Public Choice(Dordrecht),Vol. 43 (No. 2, 1984), pp. 151-71. Musgrave,RichardA., "Leviathan Cometh-Or Does He?," in Taxand ExpenditureLimitations,ed. by Helen F. Ladd and T. Nicolaus Tideman (Washington:Urban Institute, 1981), pp. 77-120. Government(ChiNiskanen, WilliamA., Jr., Bureaucracyand Representative cago: Aldine-Atherton,1971). North, Douglass C., and John Joseph Wallis, "American GovernmentExpenditures:A HistoricalPerspective,"AmericanEconomicReview,Papers and Proceedings oftheNinety-Fourth AnnualMeetingof theAmericanEconomic Association(Nashville, Tennessee),Vol. 72(May1982), pp.336-40. Oates, WallaceE. (1985a), On the Measurement of Congestionin theProvision of Local PublicGoods, Working Paper85-11 (College Park:Universityof Maryland, Departmentof Economics,October). (1985b), On the Nature andMeasurementof Fiscal Illusion: ASurvey, Working Paper85-13 (College Park:Universityof Maryland,Department of Economics,November). Olson, Mancur,The Riseand Declineof Nations:EconomicGrowth,Stagflation, and Social Rigidities(New Haven, Connecticut: YaleUniversity Press, 1982). Orzechowski,William,"Economic Models of Bureaucracy:Survey,Extensions, and Evidence," in Budgetsand Bureaucrats:The Sourcesof Government Growth,ed. by Thomas E. Borcherding(Durham,North Carolina: Duke UniversityPress, 1977), pp. 229-59.

148

DENNISC. MUELLER

Pateman,Carole,Participationand DemocraticTheory(Cambridge: Cambridge UniversityPress, 1970). Peacock,Alan T., and JackWiseman,The Growthof PublicExpenditurein the UnitedKingdom(Princeton,New Jersey:PrincetonUniversityPress,1961). Peltzman,Sam, "The Growthof Government,"Journalof Law andEconomics (Chicago),Vol. 23 (October 1980), pp. 209-87. Plott, CharlesR., "A Notion ofEquilibriumand ItsPossibilityUnder Majority Rule," American EconomicReview(Nashville, Tennessee),Vol. 57 (September 1967), pp. 787-806. Pommerehne,WernerW., and FriedrichSchneider,"UnbalancedGrowthBetween Public and Private Sectors:An EmpiricalExamination,"in Public Finance and Public Employment,ed. by Robert H. Haveman (Detroit, Michigan:WayneState UniversityPress, 1982), pp. 309-26. Puviani, Amilcare, Teoria dell'llusioneFinanziaria,1st ed. (Palermo: 1903; Milano:IstitutoEditorialeInterazionale, 1973). Romer, Thomas,and HowardRosenthal,"PoliticalResourceAllocation,ControlledAgendas, and the StatusQuo," Public Choice(Dordrecht),Vol. 33 (Winter1978), pp. 27-43. , "BureaucratsVersus Voters: On the Political Economy of Resource Allocation by Direct Democracy,"QuarterlyJournalof Economics(New York), Vol. 93 (November1979), pp. 563-87. ,"Median Voters or Budget Maximizers:Evidence from School ExpenditureReferenda,"EconomicInquiry(LongBeach, California),Vol. 20 (October 1982), pp. 556-78. Ross, JohnP., andJesse Burkhead,Productivityin the LocalGovernmentSector (Lexington,Massachusetts:LexingtonBooks, 1974). Saunders,Peter, and FriedrichKlau, The Roleof the PublicSector:Causesand Consequencesof the Growthof Government,OECDEconomicStudies,No. 4 (Paris:Organizationfor EconomicCooperationandDevelopment,1985). Schwab, Robert M., and Ernest M. Zampelli, "Disentanglingthe Demand Services:The Case Functionfromthe ProductionFunctionfor Local Public of PublicSafety"(unpublished;CollegePark:Universityof Maryland,Departmentof Economics, 1986). Sen, AmartyaK., CollectiveChoiceand SocialWelfare(San Francisco:HoldenDay, 1970). Seligman,EdwinR.A., Studiesin PublicFinance(New York:Macmillan,1925). Tanzi, Vito, "Toward a Positive Theory of Public Sector Behavior: An Interpretationof Some ItalianContributions"(unpublished;Washington: InternationalMonetaryFund, 1980). (1986a), "PublicExpenditureand Public Debt: An Internationaland HistoricalPerspective,"in PublicExpenditure:TheKeyIssues,ed. by John Bristow and Declan McDonagh (Dublin: Institute of Public Administration), pp. 6-41. (1986b),"The Growthof PublicExpenditurein IndustrialCountries:An Internationaland HistoricalPerspective"(unpublished;Washington:International Monetary Fund).

THE GROWTH OF GOVERNMENT

149

Tullock, Gordon,"Problemsof MajorityVoting,"Journalof PoliticalEconomy (Chicago),Vol. 67 (December 1959),pp. 571-79. , "FurtherTestsof a RationalTheoryof the Size ofGovernment,"Public Choice(Dordrecht),Vol. 41 (No. 3, 1983), pp. 419-21. United States, EconomicReport ofthe President,Transmitted to the Congress February1985 (Washington:GovernmentPrintingOffice, 1985). to the CongressFebruary , EconomicReportof the President,Transmitted 1986 (Washington:GovernmentPrintingOffice, 1986). Williamson,Oliver E., The Economicsof DiscretionaryBehavior(Englewood Cliffs, New Jersey:Prentice-Hall,1964).