ELECTORAL SYSTEMS AND PUBLIC SPENDINGiversen/PDFfiles/Melesi-FerettiPerottiR… · “greater...

ELECTORAL SYSTEMS AND PUBLIC SPENDING

GIAN MARIA MILESI-FERRETTI

ROBERTO PEROTTI

MASSIMO ROSTAGNO

We study the effects of electoral institutions on the size and composition ofpublic expenditure in OECD and Latin American countries We emphasize thedistinction between purchases of goods and services which are easier to targetgeographically and transfers which are easier to target across social groups Wepresent a theoretical model in which voters anticipating government policymak-ing under different electoral systems have an incentive to elect representativesmore prone to transfer (public good) spending in proportional (majoritarian)systems The model also predicts higher total primary spending in proportional(majoritarian) systems when the share of transfer spending is high (low) Afterdening rigorous measures of proportionality to be used in the empirical investi-gation we nd considerable support for our predictions

I INTRODUCTION

In modern democracies elected representatives making de-cisions on scal policy face a basic trade-off between allegiance toa social constituency and allegiance to a geographic constituencyElected ofcials represent a specic district but also typicallyadvance the interests of specic social groups that spread acrossmany districts or the whole nation This trade-off is relevant toscal policymaking because it parallels the distinction betweenthe two main types of government spending transfers and pur-chases of goods and services The former are mostly targeted togroups of individuals with certain social characteristics such asthe unemployed and the elderly the scope for targeting themgeographically is therefore limited The latter (which we will callpublic goods) instead are typically targeted along geographicallines

For useful comments and suggestions we thank Alberto Alesina V VChari Torsten Persson Guido Tabellini Jose Tavares two anonymous refereesand participants at seminars at Ente Einaudi in Rome New York University YaleUniversity and the University of California Los Angeles We are grateful toArend Lijphart for generously sharing his data Ernesto Talvi helped us withelectoral results from Uruguay Jo Buelens provided much needed claricationson the Belgian electoral system and Asbjorn Fidjestol on the Norwegian systemJeffrey Frieden Mark Hallerberg Ron Rogowski Ernesto Stein and Paul Zackgave us useful bibliographical suggestions Manzoor Gill and Nada Mora providedexcellent research assistance The views expressed in this paper do not necessar-ily reect those of the International Monetary Fund or the European CentralBank Roberto Perotti is grateful to the National Science Foundation for nancialsupport

copy 2002 by the President and Fellows of Harvard College and the Massachusetts Institute ofTechnologyThe Quarterly Journal of Economics May 2002

609

In this paper we study how the electoral system shapes thistrade-off and the incentives of elected ofcials to allocate reve-nues to the two types of spending We show that proportionalsystems are more geared to spending on transfers while majori-tarian systems are more prone to public good spending Thusloosely speaking our model captures the widespread notion thatldquoproportional systems allow representation of a greater variety ofinterestsrdquo (a frequent claim by advocates of this system) whileldquomajoritarian systems are more grounded in local interestsrdquo

The model is based on the logic of strategic delegation ofChari Jones and Marimon [1997] and Besley and Coate [1999]extended to a framework with two types of government spendingand with different electoral systems In a majoritarian systemeach district elects one representative If the distribution of dif-ferent social groups is similar across districts all representativeswill belong to the same social group Hence all elected represen-tatives derive utility from the same type of transfers but eachderives utility from a different public good It follows that electorswill have an incentive to vote for individuals with stronger pref-erences for public goods relative to transfers in order to biasgovernment expenditure on public goods toward their district Inequilibrium the result is just high expenditure on public goods

In a proportional system each district elects more than onerepresentative Hence more than one social group will be repre-sented in Parliament in contrast to the majoritarian systemeach representative now derives utility from a different type oftransfer Individuals have an incentive to vote for representativeswith stronger preference for transfers in order to bias the spend-ing decisions of the government toward their own type of trans-fers In equilibrium high spending on transfers is the result Themodel then predicts that spending on transfers is higher in pro-portional systems and spending on public goods is higher inmajoritarian systems Total government spending is higher inproportional systems if the median voter values relatively littlethe public good and relatively highly private consumption andtransfers lower in the opposite case

One virtue of our model is that it captures common and webelieve plausible views both of the properties of different elec-toral systems and of different types of government spendingOther recent positive models of electoral systemsmdashsuch as Pers-son and Tabellini [1999a 2000] and Lizzeri and Persico [2001]mdashexploit the differences between alternative types of government

610 QUARTERLY JOURNAL OF ECONOMICS

spending but they emphasize a different breakdown that be-tween a ldquouniversalrdquo expenditure and a more targetable one Ourmodel does not have a universal type of spending but two typesof goods with different targeting characteristics this is whatgenerates the dichotomy between ldquoallegiance to social constitu-enciesrdquo and ldquoallegiance to geographic constituenciesrdquo and thatbetween ldquogreater variety of interestsrdquo in proportional systems andldquogreater importance of local interestsrdquo in majoritarian systems1

In the empirical part of this paper we construct rigorouslydened measures of the degree of proportionality of electoralsystems in OECD and Latin American countries and use them toexplore the reduced-form relationship between electoral systemsand government spending In both cross-section and panel regres-sions we nd considerable support for the predictions of ourmodel for OECD countries and weaker results for Latin AmericaIn particular we document the existence of a strong and veryrobust positive relationship between the degree of proportionalityof the electoral system and the size of transfer spending amongOECD countries

The plan of the paper is as follows The next section presentsthe model Sections III and IV solve it in the majoritarian andproportional systems respectively Sections V and VI discuss howto operationalize voting systems in view of an empirical test of themodel Section VII presents the cross-sectional evidence SectionVIII the panel evidence Section IX discusses further the relation-ship to the recent literature both theoretical and empirical Sec-tion X concludes Some of the more technical passages in thesolution of the model are presented in Appendix 1 details on theconstruction of the electoral variables are given in Appendix 2Appendix 3 presents the data

1 We discuss these models more in detail in Section IX an importantdifference with our model is that they allow for binding promises by candidateswhich we rule out We also rule out strategic voting Austen-Smith and Banks[1988] and Baron and Diermeier [2001] show in different contexts how voters canbehave strategically in their electoral decision internalizing the expected coali-tion bargaining that will lead to policy formation after the election Strategicvoting can lead electors to pick a party whose preferences or policy platform aremore distant from theirs than another partyrsquos Key issues are how the right topropose a government coalition is attributed (typically depending on vote shares)and what is each partyrsquos utility out of the status quo outcome in case an agreementis not reached As we will see these issues do not arise in our model because theright to form a government is attributed randomly and all representatives whorefuse an offer to take part in the government receive the same utility

611ELECTORAL SYSTEMS AND PUBLIC SPENDING

II THE MODEL

II1 Population and the Fiscal System

The country is populated by a continuum of individuals withtotal mass 1 The population is divided into three groups A Band C with size A B and C respectively These sizes can bedifferent but no group can include more than 50 percent or lessthan 25 percent of the population2 The country is composed ofthree geographic regions A region can be thought of as the basicsubnational unit of the country hence government spendingcannot be targeted more nely than a region

There are two types of government spending transfers andpurchases of goods and services or ldquopublic goodsrdquo Typically thegovernment xes the eligibility criteria for a specic transfer andall citizens who meet the criteria are then eligible for that trans-fer regardless of their region of residence For instance old agepensions are paid to all national residents above a certain agewho have paid enough contributions and unemployment benetsare paid to all unemployed individuals with a work history Incontrast spending on goods and services is local in nature Thegovernment can always decide to build a school or to hire morepolicemen in a city and not in another it is a matter of policy howevenly distributed these expenditures are on the national territory3

Of course the distinction is not always precise Certain goodsor services purchased by the government are available virtuallyto the whole population (for instance a plane in a state-ownedairline company) But it is rarer for transfers to households pro-vided by the central government to be explicitly localized legis-lation usually does not bar citizens from a certain transfer onlybecause of where they live4 Thus we believe that by and largethe distinction we have made is conceptually and empiricallysound

We capture this difference between the two types of govern-

2 As we will see this condition ensures that all three groups are representedin a proportional system

3 Some public goodsmdashsuch as defensemdashclearly have a nationwide external-ity However expenditure on these goods is still localized a military base can bebuilt in a specic state Ceteris paribus residents of a state prefer to have themilitary base in their own state than in another state

4 Of course transfers can end up being more concentrated in certain areasbecause of the demographic or labor market characteristics of these areas thusFlorida receives more old-age pension expenditure per person than most otherstates and high-unemployment areas receive a larger share of unemployment-related transfers


ment spending in a simple way Because of some different under-lying characteristics individuals in the three groups differ in thetypes of transfers they are entitled to an individual in group jbenets from the transfer sj but not from the transfers specic tothe other groups In contrast individuals in region k derive utilityfrom public good spending in region k gk and not from the publicgoods specic to the other regions

All individuals have the same productivity which we nor-malize at 1 The utility of individual i of group j in region k is

(1) U ijk 5 (1 2 t) i is ji(1 2 i)gk

1 2 i

where t is the proportional tax rate and (1 2 t) is therefore anindividualrsquos posttax income Thus individuals have Cobb-Doug-las preferences over public goods and private income and Cobb-Douglas preferences over the breakdown of disposable incomeinto primary income and transfer income Within each group theparameters i and i are distributed uniformly over the intervals[ L H ] and [ L H ] respectively with L L $ 0 and H H 15

II2 The Electoral System and Government Formation

The values of taxes t transfers s j and public goods gk aredecided by elected representatives We describe rst how repre-sentatives are elected and then how their preferences are aggre-gated to deliver the policy outcomes [ts jgk]

The rst stage corresponds to the electoral system We x thenumber of representatives at three and characterize an electoralsystem by the number of electoral districts At one extreme eachof the three regions is a separate electoral district electing onerepresentative we call this the majoritarian system At theother extreme the whole country makes up a single electoraldistrict electing three representatives we call this the propor-tional system

Representatives of group j from different regions all deriveutility from the same transfer sj but from different public goodsWhen a district comprises more than one region as in the pro-portional system it is not necessary for our purposes to specify

5 The assumption of a uniform distribution ensures the existence of a non-cycling majority when individuals vote on the two issues contemporaneously Theassumption could be relaxed if we assumed that voting on the two issues issequentialmdashsee Appendix 1


fully the process by which public good spending within the dis-trict is then allocated to the regions within the district we simplyassume that whatever process is in place the equilibrium out-come is a uniform allocation of total public good spending amongthe regions6 Our model focuses on the allocation to districts atthe central government level in turn in deciding its allocationthe central government takes into account that it will be dividedequally among the regions in the district

The second stage describes how governments are formed andhow their decisions are taken The literature has provided a largenumber of possibilities here In our setup one simple way toformalize government formation is the following One of the threerepresentatives (the Prime Minister) is randomly selected to forma government He makes an offer to join the government toanother representative subject to the constraint that the govern-ment maximizes the joint utility of its members If the secondrepresentative accepts the government is formed and policy isdecided by maximizing the joint utility function of the two repre-sentatives If the second representative does not accept thePrime Minister makes an offer to the third representative If healso refuses then no spending on transfers and public goods isauthorized and all representatives receive a status-quo utility ofzero It should be clear that (i) it is not in the interest of the PrimeMinister to offer more than one representative to join the govern-ment (ii) it is in the interest of a representative to accept an offerby the Prime Minister7

Besides its simplicity this formulation of the political processhas the important virtue that it separates the electoral systemfrom the formation of government that is electoral systems differonly in the way representatives are elected and not in the waygovernments are assumed to be formed

6 The within-district allocation can be interpreted as a subnational voting orbargaining process that takes as given the total amount of public goods allocatedto that district by the central government All results would go through even if wetreated the subnational governments symmetrically to the national governmentnamely the subnational government at the district level is formed by the repre-sentative of a randomly chosen subregion who then invites the representativefrom another subregion to join the government

7 The assumption that the government has to maximize the joint utility ofits members reduces the importance of agenda control As an alternative we couldassume that two of the three representatives are selected at random to form agovernment If they agree the government is formed and policy is decided bymaximizing a joint utility function If they do not then no spending on eithertransfers or public goods can be authorized and all representatives receive astatus quo utility of 0


III MAJORITARIAN SYSTEM

In this system each of the three districts elects one represen-tative We solve the model backward starting with the policieschosen by the government

III1 The Policy Formation in the Government

In our main analysis we shall assume for simplicity that onegroup is larger than the other two8 Because all districts have thesame composition the three representatives will all belong to thelargest group Assume that group B is such a group thus thegovernment will be composed of two B-individuals elected in twodifferent districts Denote by k1 and k2 the districts where thetwo members of the government have been elected and let anasterisk denote an elected individual Taking logs the govern-ment will maximize the joint utility

(2) VM(k1 k2) 5 ( k1k1

1 k2k2

) log (1 2 t) 1 ( k1(1 2 k1

)

1 k2(1 2 k2

)) log sB 1 (1 2 k1) log gk1 1 (1 2 k2

) log gk2

where the superscript M denotes a majoritarian system and k i

and k irepresent the preferences of the individuals elected in

district ki These representatives (and their constituencies) wantdifferent public goods but both derive utility only from the trans-fer sB Maximization of the above objective function is subject tothe government budget constraint

(3) t 5 BsB 1 gk11 gk2

where t is the proportional tax rate and the aggregate income ofthe economy is 1 To understand the above expression recall thatthe per capita transfer is sB only individuals of type B (a fraction

B of the population) receive it there is no reason for the tworepresentatives in the government to vote for a positive transferthat benets the other two groups

Let s j = js j denote total spending on transfers to group j lets g and t denote the shares in GDP of transfers expenditure onthe public goods and total expenditure ie s j= A

Cjs j g

8 The assumption that a group larger than the others exists considerablysimplies the algebra in the majoritarian case but in no way is it essential to ourargument All our results would go through if we assumed that all groups have thesame size one-thirdmdashsee the discussion at the end of this section


k= 13 gk and t s + g It is straightforward to show that the

government policies that maximize (2) are

(4)

tM(k1k2) 52 2 ( k1

k1 1 k2k2

)

2

sBM(k1k2) 5

k1(1 2 k1

) 1 k2(1 2 k2

)

2 sA

M(k1k2) 5 0 sCM(k1k2) 5 0

gk1

M(k1k2) 51 2 k1

2 gk2

M(k1k2) 51 2 k2

2 gk3

M(k1k2) 5 0

gM(k1k2) ordm gk1

M(k1k2) 1 gk2

M(k1k2) 52 2 k1 2 k2

2

where tM (k1 k2) indicates the equilibrium value of total primaryspending in the majoritarian system when the government isformed by representatives from districts k1 and k2 and similarlyfor the other scal policy variables Similar results obtain in thecase that all groups have the same size9

III2 The Choice of the Representatives

In the rst stage each group selects simultaneously by ma-jority voting its own representative among its members so thatthe space of possible candidates spans the rectangle with length[ L H ] and height [ L H ] In Appendix 1 we show that theindividual with median values of the parameters and is thedecisive voter in each group despite the fact that the issue spaceis bi-dimensional The median voter of group B in region k1maximizes with respect to k1

and k1the utility function10

9 When all groups are the same size the election result is random In thiscase we have two possible outcomes for government formation The rst occurs ifthe government is formed by two representatives belonging to the same socialgroup In this case the policy choice is analogous to the one in equation (4) aboveThe second outcome occurs if the government coalition is formed between tworepresentatives belonging to different social groups (this will always occur whenthree candidates belonging to different parties are elected) In this case theoptimal policy choice will be analogous to the one of equation (4) for taxation andpublic good spending (with the subscripts k1 k2 now indicating representativesfrom different social groups as well as regions) for transfer spending the totalis unchanged but its composition now reects the preferences of the two so-cial groups in government with sk1

M (k1 k2) = ( k1(1 2 k1)) 2 and sk2M (k1 k2) =

( k2(1 2 k2))2 10 For simplicity we omit the social group and region subscripts from the

utility parameters


(5) E(VmBk1

M ) 5 Or= 2

3

[ m m log (1 2 tM(k1kr))

1 m(1 2 m) log sBM(k1kr) 1 (1 2 m) log gk1

M(k1kr)]

where tM (k1 kr) sBM (k1 kr) and gk1

M (k1 kr) are given by (4)Taking rst-order conditions and imposing symmetry betweenthe two districts we obtain the and preferred by the medianvoter in a majoritarian system

(6) M 5m

2 2 m M 5 m

Hence the median voter wants a representative with the medianvalue of but a value of below the median The logic is similarto that of Besley and Coate [1999] except that there are two typesof public expenditures In a majoritarian system all representa-tives and members of the government benet from the sametransfer but from different public goods Hence the median voterin district k tries to bias the decision of the government towardhis own public good by electing an individual with preference forhigh spending on public goods relative to transfers In equilib-rium the result is just high spending on the two public goods thatget funded

Substituting (6) into (4) one nally gets

tM 5 1 2m m

2 2 m

(7) sM 5m(1 2 m)

2 2 m

gM 52(1 2 m)

2 2 m

Again similar results obtain when all groups have the samesize11

11 In this case the solution for the values of and chosen by the medianvoter for each party representative in the district turns out to be a weightedaverage of the values in equation (6) and of those that occur under a proportionalsystem (see equation (12) below) The weights are equal to the respective proba-bilities that a government will be formed by two representatives of the same partyor by representatives of two different parties Results are available from theauthors upon request


IV PROPORTIONAL SYSTEM

Because each group has more than 25 percent but less than50 percent of total population in this system a representativefrom each group is elected

IV1 The Policy Formation in the Government

Suppose that the government is formed by representatives ofgroup j1 and j2 who maximize joint utility12

(8) VP( j1 j2) 5 ( j1 j1 1 j2 j2) log (1 2 t) 1 j1(1 2 j1

) log sj1

1 j2(1 2 j2

) log sj21 (2 2 j1

2 j2) log (g3)

where j1and j1

are the two utility parameters of the represen-tative from group j1 and similarly for j2

and j2 The maximi-zation of the above objective function is subject to the governmentbudget constraint

(9) t 5 j1sj1 1 j2s j2 1 g

It is straightforward to show that the solutions to this problemare

(10)

tP( j1 j2) 52 2 ( j1 j1 1 j2 j2)

2

s j1P( j1 j2) 5

j1(1 2 j1)

2 s j2

P( j1 j2) 5j2(1 2 j2)

2 s j3

P( j1 j2) 5 0

sP( j1 j2) ordm s j1P 1 sj2

P 5j1(1 2 j1) 1 j2(1 2 j2)

2

gP( j1 j2) ordm g j1P( j1 j2) 1 gj2

P( j1 j2) 1 g j3P( j1 j2) 5

2 2 j1 2 j22

For given and total spending on each of the two instru-ments is the same as under a majoritarian system (see equation(4)) however the optimal choices of and by the medianvoters will now be different as we show next

12 Because in equilibrium public good spending is divided equally within thedistrict individuals are indifferent to the region of origin of a representative theonly relevant characteristic is to which social group the representative belongs


IV2 The Choice of the Representatives

The median voter of group j1 maximizes with respect to j1

and j1 the utility function13

(11) E(Vj1m

P )= Or= 2

3

[ m m log (1 2 tP( j1 jr))+ m(1 2 m) log sj1

P ( j1 jr)

+ (1 2 m) log gP( j1 jr)]

where the values of tP sj2

P and gP are given by (10) Takingrst-order conditions and imposing symmetry between the twogroups the values of and preferred by the median voter in theproportional system are

(12) P 5m

2 2 m P 5

m(2 2 m)

1 1 m(1 2 m)

Hence the median voter selects a candidate with higher andlower than the median This pattern is exactly the opposite thatunder a majoritarian system The reason is intuitive in a propor-tional system spending on public goods is uniform across regionsbut each member of the government benets from a different typeof transfer Hence the median voter tries to bias the decision ofthe government toward his own transfer by electing an individualwith a preference for high spending on transfers relative to publicgoods In equilibrium the result is just high spending on the twotypes of transfers that get funded Using (12) in (10) one nallygets

tP 51 1 m(1 2 2 m)

1 1 m(1 2 m)

(13) sP 52 m(1 2 m)

1 1 m(1 2 m)

gP 51 2 m

1 1 m(1 2 m)

13 Again when this does not create any ambiguity we omit the region andthe group subscripts


IV3 Predictions

Comparing equations (6) and (13) it is easy to see that

sP sM

(14) gP gM

tP tMlaquo m

1(2 2 m)

Thus the model delivers three predictions First suppose thatone compares the outcome across electoral systems holding con-stant the median voterrsquos preference parameters then we havejust shown that

1 spending on transfers is higher in a proportional systemand

2 spending on goods and services is higher in a majoritariansystem

Now consider two countries with different values of m andm In the rst transfers are larger than public good spending

under both electoral systems from (7) and (13) this implies thatm gt 2(3 2 m ) In the second country public good spending is

larger under both electoral systems implying that m lt 1(3 22 m ) For m lt 1 2(3 2 m ) gt 1(2 2 m ) hence from (14) inthe rst country total spending is larger in a proportional systemtP gt tM Conversely because 1(3 2 2 m ) lt 1(2 2 m ) againfrom (14) in the second country total spending is lower in aproportional system tM gt tP Hence we have the third predictionof the model

3 Total government spending is higher in a proportionalsystem if transfer spending is large relative to public goodspending regardless of the electoral system converselytotal government spending is higher in a majoritariansystem if transfer spending is low relative to public goodspending regardless of the electoral system

Note that these results also hold when we compare the twoelectoral systems holding constant the number of parties in Par-liament or in government For example when all social groupshave the same size we can have three parties in Parliament evenin a majoritarian system However while for given preferences ofthe representatives the collective choices concerning public spend-ing would be the same as in a proportional system it is still thecase that the representatives chosen under a majoritarian systemhave a stronger preference for public goodsrsquo spending In fact


when selecting candidates voters internalize the higher likeli-hood of a conict of interest in government between public spend-ing priorities than between transfer priorities

V OPERATIONALIZING VOTING SYSTEMS

To test these hypotheses we need quantitative measures ofthe degree of proportionality of a system The key variable weconstruct is the share of electoral votes that guarantees a party aparliamentary seat in an electoral district of average size Thisvariable formally dened in subsection V2 is denoted by UMS(upper marginal share) Clearly the more proportional is a sys-tem the easier it is for small parties to gain political representa-tion and hence proportionality is declining in the UMS

Constructing such measure is not a straightforward matterhowever because real-life electoral systems are invariably morecomplicated than the stylized systems of the model Three fea-tures of actual electoral systems need to be taken into accountwhen constructing measures of proportionality the number oftiers used to allocate seats the presence or absence of legalthresholds to bar smaller parties from entering parliament andthe method used for translating seats into votes

V1 Electoral Tiers

Electoral systems in general have one or two tiers In two-tiersystems a certain portion of parliamentary seats are allocated ina second tier comprising fewer larger districts each encompass-ing several rst-tier districts This second tier typically serves thefunction of increasing the degree of proportionality in the elec-toral system We will use T1 and T2 as shortcuts for ldquorst tierrdquoand ldquosecond tierrdquo

Let Sik denotes the number of seats attributed to parties onthe basis of votes in district k of tier i14 or its district size Si =

k Sik is the total number of seats attributed to parties on thebasis of votes in tier i or the tier size Di is the number ofdistricts in tier i Si SiD i is the average district size of tier i15

14 The subscripts i and k in this section have a different meaning from thesame letters in the presentation of the model (Sections IIndashIV)

15 In some electoral systems (see subsection V4) S1k can be an upper boundto the number of seats attributed in a T1 district some seats could be left unlledtransferred to T2 and attributed there in these cases the number of seatsattributed in T2 depends on the number of seats effectively attributed in T1


As is standard in the literature whenever there are two cham-bers all of these magnitudes refer to the lower chamber

V2 Voting Method

A voting method describes how party votes are converted intoseats in each district of an electoral system We distinguish be-tween three voting methods

1 MajorityPlurality All the Sik district seats are attributedto those parties that win an absolute majority of votes(Majority systems) or just more votes than other candi-dateslists (Plurality systems) In our sample Majoritymethods are represented by the Two-Round method(France) and the Alternative Vote method (Australia)16

Plurality methods are represented by the First-Past-the-Post method (United Kingdom and several others)

2 Highest Average The share of votes obtained by eachparty in each district is divided sequentially by a set ofdivisors 1 2 3 in the case of the drsquoHondt formula and135 in the case of the St Lague formula Each of theSik highest quotients entitles the party that obtains themto a seat17

3 Largest Remainder In each district k of tier i rst a quotais calculated dened as 1Sik in the Hare formula1(Sik + 1) in the Droop formula and 1(Sik + 2) in theImperiali formula Then each party is allocated as manyseats as full quotas it has obtained The seats left unlledafter this allocation can be transferred to a higher tier ifit exists or attributed in the same district to the parties orcandidates with the largest remainders18

16 In the Australian system voters rank candidates on the ballot Any candi-date with an absolute majority is elected If no candidate reaches an absolute major-ity the candidate with the lowest number of rst preferences is eliminated and hisvotes reassigned to all other candidates according to the ranking indicated on hisballots The process is repeated until a candidate reaches an absolute majority

17 As an example consider a four-seat district with three parties (A B andC) 100 electors and votes obtained by each party VA = 50 VB = 30 VC = 20Under a drsquoHondt formula the quotients obtained by A are 50 502 and 503 byB 30 302 and 303 by C 20 202 and 203 The three highest quotients are 5030 and 502 Hence A gets two seats B one seat and C gets no seats

18 Two methods the Single Transferable Vote (used in Ireland) and theSingle Nontransferable Vote (used in Japan until 1995) are difcult to t in ourclassication In one important respect (the calculation of upper quotasmdashseebelow) they behave like Largest Remainder methods hence we will classify themas such In the Irish STV method voters rank candidates Any candidate whoserst preferences reach a full Droop quota is elected and his votes above the quotaare redistributed to the remaining candidates following the second preferences he


V3 Legal Thresholds

Legal thresholds serve the purpose of limiting access to Par-liament to parties receiving ldquosmallrdquo shares of votes Formally thelegal threshold of tier i THRi is the minimum share of nationalvotes set by the electoral law that a party must obtain in orderto be eligible for a seat in the tier19

V4 Dening Proportionality of Electoral Systems

We now proceed to construct our measure of proportionalityin three stages We rst determine the share of votes that guar-antees a seat in the average district of each tier abstracting fromelectoral thresholds We then take electoral thresholds into ac-count Finally we determine in which tier the ldquomarginal seatrdquo isallocated20

We start by dening the share of votes that even under themost unfavorable distribution of votes among parties in the dis-trict still guarantees a seat to the party that obtains it

DEFINITION (UPPER QUOTA QI) The upper quota of district k in tieri Qi(Sik) is the share of district votes that would guaranteea party its rst seat in that district if there were no legalthreshold

Note that in general the upper quota depends only on thedistrict size not on the number of parties in the district21 We

has received If no candidates reach a full quota the candidate with the lowestnumber of rst preferences is eliminated The process continues until all Sik seatsare lled In the Japanese SNTV method voters express a single vote theSik most voted candidates are elected

19 Thus THR2 = 05 would state that in order to obtain at least one seatin T2 a party must win at least 5 percent of the national votes If the samethreshold also applies to T1 a party not meeting this threshold must relinquishany T1 seat that it might have obtained Otherwise the threshold is binding onlyfor the allocation of T2 seats Note that if a legal threshold does not exist in tieri this is equivalent to assuming that THRi = 0 Sometimes the electoral lawsstate requirements on the vote shares of a party in a district rather than a tierThese local requirements can be translated into a share of national votes andtherefore into a legal threshold using the procedure described in Appendix 2

20 To implement our measures of proportionality we assume that the dis-tribution of votes among parties is the same in all T1 districts In other words bythis assumption a partyrsquos district share of votes in all T1 districts is equal to itsnational share An obvious shortcoming of this assumption is that it does not dealwell with regional parties However incorporating regional parties would requirea detailed knowledge of actual results election by election and district by districtwhich we do not have

21 In a few cases (Hare and St Lague when the number of parties is belowthe district magnitude) the formula for the upper quota depends on the number of


provide formulas for the upper quota for each voting methodin the Appendix

By taking into account the electoral threshold we cannow dene the share of votes that ensures a party electoralrepresentation in a district of average size in tier i

DEFINITION (UPPER MARGINAL SHARE OF VOTES OF TIER I UMSI) Theupper marginal share of votes of tier i UMS i is the share ofnational votes that guarantees a seat to a party in the districtof average size of that tier UMSi = max(Q i(Si)THRi)

For multitier systems we also need to establish in whichtier the ldquomarginalrdquo seat is allocated that tier is the decisivetier for the purpose of determining the degree ofproportionality

DEFINITION (DECISIVE TIER) The decisive tier of an electoral systemis the tier with the lower Upper Marginal Share of votes

In other words the decisive tier is the tier where it iseasier for a small party to gain representation We can nowdene our proportionality variable

DEFINITION (UPPER MARGINAL SHARE OF VOTES OF AN ELECTORAL SYS-TEM UMS) The upper marginal share of votes of an electoralsystem UMS is the share of national votes that guaranteesa seat in the district of average size of the decisive tier UMS= min(UMS1 UMS2)

V5 Implementing the Denition

We now show how this denition can be applied in practice tothe different types of electoral systems The case of one-tier elec-toral systems is straightforward UMS is just the larger betweenthe upper quota in the average-size district and the electoralthreshold in the only tier that is UMS = max(Q1(S1)THR1)

Constructing UMS in two-tier systems is more complicatedIn order to determine the share of national votes that ensures aseat in the system as a whole we need to address two issues rsthow is the total number of second-tier seats (S2) determined andsecond which votes are used to allocate such seats Table Isummarizes the two-tier systems in our sample

parties In these cases we have collected data on the number of parties in eachelection


With regard to the number of second-tier seats in Remain-der Seats systems (denoted by RS) S2 is variable T1 seats areattributed to parties using a Largest Remainder method theremainder seats after this allocation are lled in T2 In Adjust-ment Seats systems (denoted by AS) S2 (and therefore S1) isxed Because AS systems must attribute a xed number of seatsin T1 they generally use a Highest Average method to attributeT1 seats (the exception in our sample being Ecuador which usesthe Hare formula in T1)

With regard to the votes used to allocate second-tier seats inRemainder Votes systems (RV) seats in T2 district k are at-tributed using those votes not used in the allocation of seats in allthe T1 districts included in k These remainder votes are trans-ferred to k pooled22 and used to attribute seats there typicallyusing a Highest Average method

In Superdistrict Votes systems (SV) all votes cast in a T2districtmdashnot just the remainder votesmdashare used to allocate seatsThis can be done in two ways In Parallel SV systems (SV-P)such as Greece T2 seats are attributed independently of T1 seatsIn this case voters usually cast a separate ballot for each tier theallocation of T1 seats is based on T1 votes and the allocation of T2seats is based on T2 votes In Mixed SV systems (SV-M) seats ineach T2 district are attributed to parties after taking into accountthe seats already attributed to candidates in the T1 districts thatmake up the T2 district in question Specically seats in T2 areattributed in order to achieve an overall distribution of seats toparties as close as possible to the distribution that would obtain

22 The exception is Norway since 1990 which attributes the eight T2 seatsto the lists with the highest quotas among all the quotas unused in the distribu-tion of seats in T1

TABLE ITWO-TIER SYSTEMS

SV RV

AS Denmark Germany Greece Italy 1994ndash1995Sweden 1971ndash1995 Ecuador GuatemalaMexico

Norway 1990ndash1995

RS Belgium Venezuela Austria Italy 1960ndash1995

In bold the parallel systems The electoral systems of Belgium Greece and Venezuela have peculiaritiesthat require some interpretation A detailed description of these systems is available upon request


if all seats were attributed based on the T2 votes method andnumber of districts Effectively then if there are enough seats inT2 the seats attributed to candidates in T1 serve only to deter-mine the names of as many representatives but the distributionof seats to parties is determined wholly in T2 Since we areinterested in the distribution of seats to parties for our purposesthe system effectively works as if S2 = Sto t this is what we willuse in our empirical application23

We can now turn to the determination of the degree of pro-portionality of two-tier systems In SV systems we can deter-mine UMS2 by simply applying its denition UMS2 =max(Q2(S2)THR2)24 The purpose of the second tier is to in-crease the proportionality of the system hence in all these sys-tems UMS2 lt UMS1 and T2 is the decisive tier UMS = UMS2

In two-tier RV systems the lowest share of national votesthat still guarantees a seat to a party in the average district of T2occurs when the party does not win any T1 seat transfers all itsvotes as remainders to T2 and these votes are enough to guar-antee a seat in the average district there Thus as in SV systemsUMS = UMS2 However now it is no longer true that the T2shares of votes are the same as the national shares of votes onlythe remainder votes from T1 are used to allocate T2 seats andsmall parties (for instance all those that did not obtain any seatin T1) have a much larger share of remainder votes than of allnational votes Appendix 2 shows how UMS2 can be calculated inthese cases

VI MEASURES OF PROPORTIONALITY

Typically in the literature (eg Taagepera and Shugart[1989] and Lijphart [1994]) proportionality is captured by a mea-sure of district size rather than by a measure of vote share likeUMS25 Thus in this section we rst convert UMS into a mea-sure of average district size then we present two more measures

23 Obviously for the rst tier in these systems to be completely irrelevant asufcient number of seats must be attributed in the second tier In practice thisis the case for all SV-M systems in our sample

24 Although the formula is the same recall that S2 = Sto t in ASSV-Msystems while in ASSV-P systems S2 lt Sto t

25 Taagepera and Shugart [1989] calculated average effective district mag-nitudes for a number of OECD countries in the seventies and mideighties Ourmeasure differs from theirs because it is based on the rigorously dened notion ofUpper Marginal Share of votes which incorporates the different types of two-tiersystems and legal thresholds


of proportionality some variants of which have often been used inthe literature These measures try to formalize how easy it is forsmaller parties to gain representation in Parliament26

VI1 Average Standardized District Magnitude (SM)

As made clear in the previous section UMS is ceteris pari-bus inversely related to the average district size of the decisivetier For instance in a single-tier electoral system with no thresh-olds which uses the drsquoHondt voting method UMS is equal to theupper quota in the ldquoaveragerdquo district UMS = Q1(S1) = 1(1 +S1) One can invert the above relation and back out a measure ofaverage district magnitude from data on UMS In doing sohowever it is important to partial out the voting method used ineach system27 We use the drsquoHondt formula which does notdepend on the number of parties and therefore provides a one-to-one mapping with UMS By applying the inverse of the drsquoHondtformula to UMS we thus obtain the Average Standardized Dis-trict Magnitude or SM More formally

DEFINITION (AVERAGE STANDARDIZED DISTRICT MAGNITUDE SM) Con-sider the electoral system possibly with two tiers and alegal threshold and consider the electoral system with onetier no legal threshold and the drsquoHondt formula The aver-age standardized district magnitude of system SM is theaverage district size of system in which a party with thesame UMS of would be guaranteed a seat

Hence SM = S = 1UMS 2 1

26 In constructing these measures of proportionality we ignore those tierselecting less than 5 percent of the total assembly size In our sample this excludesthe second tier in the Norwegian ASRV system in 1991ndash1995 which elects 8 outof 165 representatives and the fourth parallel tier in the Greek ASSV systemwhich elects 12 out of 300 representatives

27 Suppose that two one-tier systems have the same value of UMS of 01but the rst uses the St Lague formula with six parties the second the drsquoHondtformula Inverting the formulas for the upper quotas in Appendix 2 subsection 2the average district magnitude would be 85 in the rst system and 9 in the secondsystem Hence it is important to convert the UMS of the system in the averagedistrict size of a standard system with a xed electoral formula In the workingpaper version of this paper [Milesi-Ferretti Perotti and Rostagno 2001] we alsoconsider an alternative measure of proportionality which uses the voting methodof the decisive tier of the electoral system to invert the formula for the upperquota In practice this measure is very strongly correlated with SM


VI2 Average District Magnitude (AM)

A second measure of average effective district magnitudecaptures the notion of how large is the district where the ldquoaver-agerdquo representative is formally elected It is dened as follows

DEFINITION (AVERAGE DISTRICT MAGNITUDE AM) The average dis-trict magnitude of an electoral system is the weighted aver-age of the average district sizes of the two tiers with weightsequal to the proportion of all representatives elected in thetwo tiers

Thus this variable is measured simply by AM = (S1Sto t)S1 + (S2Sto t)S2 Note that for the purposes of computingthis variable S1 and S2 represent the number of representativeseffectively elected in each tier

VI3 The Average Deviation from Proportionality (RAE)

The two variables described so far are ex ante measures ofproportionality being based on institutional characteristics Wealso use one ex post measure based on voting outcomes electionby election This variable was originally dened by Rae [1967] asfollows

DEFINITION (AVERAGE DEVIATION FROM PROPORTIONALITY RAE) TheAverage Deviation from Proportionality (RAE) is the averageof the deviations (in absolute values) of the share of seats ofeach party from its share of votes RAE = p = 1

P u sp 2 vp u where p indexes a party sp is the share of seats and vp is theshare of votes obtained by party p

Thus this variable measures deviations of the shares of seatsin Parliament from the share of votes obtained by parties in eachelection28

VII CROSS-SECTIONAL REGRESSIONS

VII1 The Data

Our sample consists of twenty OECD and twenty LatinAmerican countries listed in Appendix 3 The data consist of the

28 This variable is not independent of the number of parties and their sizeit tends to give a small degree of disproportionality in systems with many smallparties An alternative measure proposed by Gallagher [1991] is dened as thesquare root of the sum of squares of the deviations between seat and votepercentages Thus it gives more weight to large deviations In practice it is highlycorrelated with RAE


political variables described in the previous section the level andcomposition of public spending as well as a number of othercontrol variables described in Appendix 3 The scal data includetotal primary government spending government transfers tohouseholds and expenditure on public goods all by the generalgovernment Transfers are dened as the sum of social securitypayments and other transfers to families plus subsidies torms29 Public goods are dened as the sum of current and capitalspending on goods and services ie the sum of governmentconsumption and of capital spending30 The so-called ldquopork-bar-relrdquo expenditure like building a bridge or hiring civil servants ina certain locality to please one ownrsquos constituency falls mostlyunder one of the two components of our denition of public goodsgovernment consumption and investment

Data for the OECD sample start in 1960 with the exceptionof Greece Portugal and Spain whose data start in the mid-seventiesThe time span for Latin American data is more limitedmdashwe have information on electoral variables for the early ninetiesIn our empirical investigation we rst use the combined OECDand Latin American samples for cross-sectional estimates (basedon averages of all variables over the four-year period 1991ndash1994or the closest available periods31) and then we exploit the time-series dimension of the OECD sample to run panel regressions

Table II displays the cross-sectional 1991ndash1994 averagestandard deviation minima and maxima for each variable for thewhole sample as well as for OECD and the Latin Americancountries separately To make the reading of the empirical resultseasier we will dene all electoral variables as direct measures ofproportionality To do so in the case of RAE we take the negativeof the variable as originally dened although we keep the origi-nal name

A few points are worth noticing because they will play a rolein interpreting our results Latin American countries have onaverage more proportional systems They also have much

29 For most Latin American countries we do not have enough information toseparate social transfers to families from subsidies to rms

30 The residual term in total primary expenditure is property income paid(net of interest payments) which is also largely not subject to political control ateach moment in time In the OECD sample the residual term also includessubsidies to rms

31 For some Latin American countries we use a different period wheneverthe electoral law changes during the 1991ndash1994 period in order to encompass onlyone electoral law The details of the time periods used for Latin American coun-tries are in Appendix 3 subsection 1


TA

BL

EII

SU

MM

AR

YS

TA

TIS

TIC

S

All

Mea

nA

llS

DA

llM

inA

llM

axO

EC

DM

ean

OE

CD

SD

OE

CD

Min

OE

CD

Max

LA

Mea

nL

AS

DL

AM

inL

AM

ax

Avg

dis

tric

tm

agn

(A

M)

161

307

115

015

323

115

017

329

91

120

Sta

nd

dist

rict

mag

n(

SM

)23

639

71

180

820

133

114

83

272

461

180

8R

AE

34

40

326

12

41

90

36

83

76

50

326

1P

rim

ary

expe

ndG

DP

326

151

103

666

454

94

327

666

198

55

103

298

Tra

nfe

rsG

DP

134

95

14

316

213

63

115

316

55

36

14

146

Pub

lic

good

sG

DP

181

56

71

299

219

42

1629

914

44

17

123

5G

DP

per

capi

ta84

0055

0012

0018

100

1320

028

0068

0018

100

3500

1800

1200

7200

Pop

uls

hare

over

659

75

22

919

514

42

113

195

52

22

911

8E

ffn

umb

ofpa

rtie

s3

51

71

38

43

71

61

98

43

31

71

38

4E

TH

NIC

(fra

cti

ndex

)20

119

81

7522

122

11

7518

117

51

159

9O

PE

N(o

penn

ess)

305

173

74

961

301

141

87

629

309

204

74

961

Cro

ss-s

ecti

onal

aver

ages

sta

nda

rdde

viat

ions

min

imum

and

max

imu

mva

lues

for

all

the

vari

able

sfo

rth

efu

llsa

mpl

eth

eO

EC

Dsa

mpl

ean

dth

eL

atin

Am

eric

ansa

mpl

eS

eeth

eA

ppen

dix

for

vari

able

srsquom

nem

onic

sE

ach

cros

s-se

ctio

nalo

bser

vati

onis

the

1991

ndash199

4(o

rcl

oses

tav

aila

ble

peri

od)a

vera

geof

the

vari

able

for

the

cou

ntry

Num

ber

ofob

serv

atio

ns

40fo

rth

ew

hol

esa

mpl

e(3

5fo

rR

AE

)20

for

the

OE

CD

sam

ple

20fo

rth

eL

atin

Am

eric

ansa

mpl

e(1

5fo

rR

AE

)


smaller governments average primary expenditure is 198 per-cent of GDP less than half than in OECD countries indeed thelargest Latin American government is smaller than the smallestOECD government The largest difference between the twogroups of countries is in transfers on average their ratio to GDPis four times higher in OECD countries in contrast the shares ofpublic goods are much closer 219 percent versus 139 percent InLatin America public good spending is much larger than trans-fers but the two items are virtually equal in OECD countries32

Table III displays the average values over 1991ndash1994 coun-try by country of the electoral variables used in our estimationTable IV displays the cross-country correlations among the aver-ages of LogAM (the log of Average District Magnitude) LogSM(the log of Average Standardized District Magnitude) MAJ (a

32 Indeed if one subtracts military spending from public good spendingtransfer spending becomes larger than public good spending The size of militaryspending is to a large extent dictated by international commitments and itsgeographic targeting may be constrained by strategic considerations

TABLE IIIAVERAGE VALUES OF ELECTORAL VARIABLES

AM SM RAE ENPP AM SM RAE ENPP

Australia 10 10 2 51 24 Argentina 107 107 30Austria 194 380 2 10 30 Bolivia 144 149 2 22 33Belgium 84 236 2 11 84 Brazil 178 178 2 18 84Canada 1 1 2 68 23 Chile 20 20 2 18 50Switzerland 78 78 2 09 63 Colombia 49 49 31Germany 108 190 2 20 32 Costa Rica 81 88 2 18 22Denmark 153 490 2 10 44 Dominican Rep 300 390 2 27 29Spain 67 67 2 18 29 Ecuador 44 122 64Finland 133 133 2 14 52 El Salvador 43 43 2 20 30France 1 1 2 52 30 Guatemala 101 290 2 34 47United Kingdom 1 1 2 50 22 Honduras 67 79 2 15 20Greece 57 104 2 19 23 Jamaica 1 1 2 261 13Ireland 40 40 2 15 32 Mexico 806 657 2 12 23Italy 172 293 2 12 55 Nicaragua 100 123 2 09 21Japan 39 39 2 17 34 Panama 17 17 14Netherlands 1500 1483 2 03 42 Paraguay 44 44 2 11 25Norway 87 86 2 12 42 Peru 120 120 39Portugal 105 105 2 37 23 Trin and Tob 1 1 2 81 21Sweden 141 240 2 10 41 Uruguay 52 52 2 03 33United States 10 10 2 47 19 Venezuela 79 1808 2 09 34

AM Average district magnitude SM Average standardized district magnitude RAE (minus) index ofdeviations from proportionality ENPP effective number of political parties in Parliament See Section VI andAppendix 3 for further details on the denition of variables


dummy variable taking the value of one in majoritarian systemsand zero otherwise from Persson and Tabellini [1999a]) andLogMAGN (the log of the measure of effective district magnitudeby Taagepera and Shugart [1989] and Lijphart [1994] which weextended to 1995 using the same methodology) The high corre-lations between LogSM LogAM and LogMAGN and betweenRAE and MAJ are noteworthy Table A1 available from theauthors at httpwwwiueitPersonalPerotti presents the de-tails country by country of each electoral system

VII2 Basic Specication

Our basic cross-section specication is

(15) G i 5 c 1 cOECD 1 X i 1 POP65i 1 LogGDPPCi 1 i

TABLE IVPROPORTIONALITY MEASURES CROSS-COUNTRY CORRELATIONS (1991ndash1994)

Log SM Log AM RAE MAJ Log MAGN

Panel A Whole sample

Log SM 1Log AM 092 1RAE 052 055 1MAJ 2 018 2 017 2 058 1

Panel B OECD

Log SM 1Log AM 097 1RAE 085 084 1MAJ 2 082 2 082 2 092 1Log MAGN 095 095 087 2 087 1

Panel C Latin America


Measures of proportionality calculated as average during the period 1991ndash1994 AM Average districtmagnitude SM Average standardized district magnitude RAE (minus) index of deviations from propor-tionality MAJ dummy variable taking the value of one if the electoral system is majoritarian and zerootherwise MAGN effective district magnitude See Section VI and Appendix 3 for more details on thedenition of variables The variables SM and AM are available for all countries in the sample The variableMAGN is available for OECD countries only The variables RAE and MAJ are available for all OECDcountries and fteen and eighteen Latin American countries respectively


where i indexes a country G is the average percent share in GDPof either total primary expenditure EXP transfers TRAN orpublic goods PGOOD X is one of the three electoral systemvariables that we have introduced before the rst two LogAMand LogSM are the logs of the Average District Magnitude and ofthe Average Standardized District Magnitude respectively33

The third electoral variable is RAE POP65 the share of popula-tion over 65 is a potentially important determinant of the size oftransfer expenditure34 and LogGDPPC the PPP adjusted percapita income of the country in logs of thousands of dollarscaptures possible Wagner Law-type effects In addition to theregression constant c we also include an OECD dummy vari-able cO E C D to allow for the large difference in the averagegovernment spendingGDP ratio between OECD and Latin Ameri-can countries

Our hypotheses imply that an increase in the value of any ofthe electoral variables should be associated with a higher share oftransfers in GDPmdasha positive value of when G = TRANmdashandwith a lower share of public goods in GDPmdasha negative value ofwhen G = PGOOD (see Results 1 and 2 in subsection IV3) Inaddition the effect on total spending depends on the initial shareof transfers in total primary spending and more fundamentallyon the underlying preferences of the median voter We havedocumented above the vastly different shares of transfers in GDPand in government spending between Latin American and OECDcountries If we interpret these differences as reecting at leastin part different patterns of distribution of preferences over scalpolicy between these two groups of countries then when G i =EXP we should expect a positive value of in OECD countrieswhich have a large share of transfers in primary spending and anegative value of in Latin American countries all of which havean extremely low share of transfers (see Result 3 in subsectionIV3)

33 We use logs because increasing the average district magnitude by onerepresentative has a very different effect on government spending when the initialdistrict magnitude is 1 than when it is 50

34 One could argue that POP65 is an endogenous variable as higher spend-ing on certain types of expenditure (notably health) can increase life expectancyWe cannot think of a good instrument in our sample but we note that thecorrelation of POP65 with the electoral variable is quite low in fact when wereestimate our regressions dropping POP65 the estimated coefcients of theelectoral variables are largely unchanged and if anything larger


VII3 Basic Results

In columns (1)ndash(3) of Table V the dependent variable is thetotal primary spendingGDP ratio EXP The estimated coef-cients of the three electoral variables are positive but none issignicant at conventional levels This result is consistent withour theoretical model where a more proportional system canhave an ambiguous impact on total primary government spend-ing depending on the relative strength of its effects on transferand public good spending In fact we will see shortly that thisresult is the combination of two very different patterns in LatinAmerica and in OECD countries

Our model predicts that more proportional systems should beassociated with higher spending on transfers Columns (4)ndash(6)display the same regressions as columns (1)ndash(3) but with theshare of transfers in GDP as the dependent variable Now all theestimated coefcients of the electoral variables are signicant atthe 5 percent level To assess the economic signicance considerthe change in the dependent variable associated with a change inthe electoral variable equal to its range (reported in the third tolast row of the table) The range in the transferGDP ratio variesbetween about 69 (in the LogSM regressions of column (5)) and98 percentage points of GDP (in the RAE regression of column(6)) Besides electoral variables the only other signicant vari-able in the regressions is POP65 which has the expected positivecoefcient Still the explanatory power of these regressions isquite large with adjusted R2s always at or above 8

The reader may question whether we are simply capturing adichotomy between majoritarian and proportional systemsrather than a systematic relation between the degree of propor-tionality and transfers Figure I provides a clear negative answerto this question It shows that for OECD countries the positiverelationship between LogSM and transfers survives (and actuallybecomes stronger) even if one excludes the countries with a ma-jorityplurality system in the sample Australia France UnitedKingdom Canada and the United States A plot of the residualsof a regression of transfers on log GDP per capita and share ofpopulation above 65 (not included for reasons of space) conveysthe same message In Latin America by contrast there is no bivari-ate relation between proportionality and transfers (Figure II)

By comparing the coefcients of the electoral variables ineach total spending regressions with the same coefcient in the


TA

BL

EV

PR

IMA

RY

SP

EN

DIN

G

TR

AN

SF

ER

SP

UB

LIC

GO

OD

SA

ND

EL

EC

TO

RA

LS

YS

TE

MS

(FU

LL

SA

MP

LE)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

105

170

20

60(L

ogA

M)

(11

0)(3

49)

(1

12)

Log

stan

ddi

stm

agn

074

132

20

48(L

ogS

M)

(08

9)(3

06)

(1

03)

RA

E0

320

382

002

(10

5)(2

21)

(0

13)

Pop

sha

reov

er65

113

114

130

125

127

142

20

122

012

20

10(1

86)

(1

87)

(1

95)

(4

04)

(3

96)

(3

81)

(0

34)

(03

5)(0

27)

Log

GD

Ppe

rca

pita

116

083

054

219

169

145

21

642

146

21

72(0

37)

(02

6)(0

16)

(13

7)(1

04)

(07

8)(0

92)

(08

3)(0

93)

OE

CD

dum

my

134

913

82

115

51

201

752

021

108

310

64

106

5(2

10)

(2

14)

(1

57)

(03

7)(0

52)

(00

5)(2

98)

(2

92)

(2

60)

C

onst

ant

275

574

108

62

217

92

174

62

117

529

32

278

328

87

(01

1)(0

23)

(04

2)(1

73)

(1

36)

(08

1)(2

09)

(2

01)

(20

0)

Ran

ge5

253

848

308

516

869

763

022

490

55O

bser

vati

ons

4040

3540

4035

4040

35R

20

760

750

730

840

830

800

440

430

38

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s


corresponding transfer regression it is clear that one shouldexpect a negative but small coefcient of the electoral variable inthe public good regressions This is in fact what we nd in all

FIGURE ITransfers and Standardized District Magnitude OECD Countries

FIGURE IITransfers and Standardized District Magnitude Latin American Countries


public good regressions (columns (6)ndash(9)) the electoral variableshave negative coefcients but never reach statistical signicance

However Table V hides a substantial difference between thetwo subsamples Table VI displays the same regressions as TableV for the subsample of OECD countries In the primary spendingregressions of columns (1)ndash(3) the coefcients of the electoralvariables are three to ve times larger in the OECD subsamplethan in the whole sample and are now signicant at the 10percent level (except for RAE in column (3)) Even stronger re-sults hold for the transfer regressions in columns (4)ndash(6) now thecoefcients of the electoral variables are typically twice as big asin the corresponding columns of Table V and all signicant atleast at the 2 percent level The effect of electoral variables ontransfers explains nearly all of their effect on total spending as aresult we nd small and statistically insignicant negative ef-fects on public good spending (columns (7)ndash(9))

Qualitatively the results for Latin America displayed inTable VII are almost the mirror image of those for OECD coun-tries although they are statistically less strong (we do not displayresults with RAE because in the Latin American sample we haveonly fteen observations on this variable) The effect of electoralvariables on total spending is now negative although statisticallyinsignicant (columns (1)ndash(3)) This is the result of almost noeffect on transfers (columns (4)ndash(6)) and a large negative effect onpublic good spending (columns (7)ndash(9)) although with high p-values between 15 and 20

All the point estimates in Tables V to VII are consistent withthe predictions of subsection IV3 Consistent with Results 1 and2 more proportional systems always have higher transfers andlower public good spending ceteris paribus Consistent with Re-sult 3 more proportional systems are associated with higher totalprimary spending in OECD countries which have high transferspending and with lower primary spending in Latin Americancountries which have low transfers regardless of the electoralsystem

These results complement those of a related literature thathas studied the difference in Latin American and OECD scalpolicy As documented in Gavin and Perotti [1997] in LatinAmerica most of the scal policy response to cyclical variations inthe economy and to external shocks occurs on public good spend-ing in OECD countries on transfer spending This paper showsthat the (cross-country) response of scal policy to electoral insti-


TA

BL

EV

IP

RIM

AR

YS

PE

ND

ING

TR

AN

SF

ER

S

PU

BL

ICG

OO

DS

AN

DE

LE

CT

OR

AL

SY

ST

EM

S(O

EC

DC

OU

NT

RIE

S)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

305

303

20

34(L

ogA

M)

(21

0)

(40

3)

(04

1)L

ogst

and

dist

mag

n2

562

632

028

(Log

SM

)(1

89)

(3

64)

(0

37)

RA

E1

411

622

064

(12

9)(2

56)

(1

14)

Pop

sha

reov

er65

137

115

143

119

095

117

022

025

040

(14

9)(1

17)

(14

0)(2

50)

(1

80)

(1

99)

(0

43)

(04

5)(0

75)

Log

GD

Ppe

rca

pita

171

516

39

157

67

737

056

594

484

574

09(2

36)

(2

23)

(2

03)

(20

5)

(18

0)

(14

8)(1

08)

(11

1)(1

03)

OE

CD

214

210

213

154

212

092

274

42

264

42

254

04

223

17

224

34

224

04

(20

2)

(18

5)

(16

2)(2

04)

(1

70)

(12

6)(0

58)

(06

1)(0

63)

Ran

ge15

28

128

09

1115

18

131

310

45

169

141

413

Obs

erva

tion

s20

2020

2020

2020

2020

R2

034

031

024

061

057

044

20

82

08

20

1

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s


tutions also follows the same pattern it affects mostly public goodspending in Latin America and mostly transfer spending inOECD countries

For Latin America results are statistically much weaker Wehave two candidate explanations for this differencemdash besides theobvious one that our theory ts Latin America less well thanOECD countries First measurement error both the budget vari-ables and the electoral variables are measured less precisely inLatin American countries Second Latin America and its scalpolicy are subject to larger and more frequent shocks than OECDcountries (see eg Gavin et al [1996]) hence it is likely that therole of electoral systems in shaping scal outcomes will be harderto detect in Latin America

VII4 Robustness

Because of the small sample size our benchmark specica-tion is necessarily very parsimonious Several variables that wehave omitted could conceivably be correlated both with the elec-toral systems and with scal outcomes In Table A2 available

TABLE VIIPRIMARY SPENDING TRANSFERS AND ELECTORAL SYSTEMS

(LATIN AMERICAN COUNTRIES)

(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD

Log avg distr magn 2 091 036 2 099(Log AM) (086) (067) (131)

Log stand distr magn 2 108 014 2 096(Log SM) (113) (028) (140)

Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)

Log GDP per capita 2 145 2 077 150 144 2 242 2 183(051) (027) (104) (096) (120) (089)

Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)

Range 438 561 172 071 472 499Observations 20 20 20 20 20 20R2 2 04 000 040 038 009 010

Dependent variables EXP share of total primary spending in GDP (columns (1)ndash(2)) TRAN share oftransfers in GDP (columns (3)ndash(4)) PGOOD share of spending on public goods in GDP (columns (5)ndash(6))averages 1991ndash1994 or closest available period See Appendix 3 for the denition of all variables Estimationby ordinary least squares (t-statistics are in parentheses) () signicant at the 10 percent (5 percent) levelRange is range of variation of dependent variable associated with range of variation of electoral variableholding constant all other variables


from the authors at httpwwwiueitPersonalPerotti web sitewe display the estimated coefcients of the electoral variableswhen we add these omitted variables one at a time in thespecication estimated so far we also display the estimated co-efcient of the added variable In this section we briey discussthe main results to conserve space we focus on the most relevantdependent variables in the two groups of countries transferspending in OECD countries and spending on public goods inLatin American countries Also we present results for LogSMonly results with the other variables are similar

Conventional wisdom has it that proportional systems tendto be associated with a larger number of parties in Parliamentand therefore with larger coalition governments In turn empiri-cally larger coalitions tend to be associated with more expendi-ture particularly on transfers (see Perotti and Kontopoulos[1999]) To address this issue we add the log of the averageeffective number of parliamentary parties LogENPP to the listof independent variables In OECD countries the estimated coef-cient of LogENPP is positive albeit insignicant more impor-tantly the coefcient of LogSM remains signicant at the 5percent level In Latin America the estimated coefcient ofLogSM practically does not change35 Very similar results obtainwhen we use for OECD countries the number of parties in thecoalition from Perotti and Kontopoulos We conclude that theelectoral system has an effect on scal outcomes independent ofits effects on the degree of party fractionalization both in electionsand in Parliamentary representation

In an inuential paper Rodrik [1998] argues that more opensocieties spend more on government transfers as insuranceagainst external shocks When we include openness in our regres-sions its coefcient is always entirely insignicant in both theOECD and Latin American samples the coefcients of the elec-toral variables are largely unaffected The same happens if weinteract the openness variable with the volatility of the terms oftrade as suggested by Rodrik36

One could also argue that more ethnically or linguistically

35 In a study of the effects of electoral institutions on scal performance inLatin America Stein Talvi and Grisanti [1999] nd that the only ldquopoliticalrdquovariable correlated with the size of total spending (inclusive of interest payments)is the effective number of parties

36 These results are potentially important in view of the considerable debatethat Rodrik [1998] has generated See for instance Alesina and Wacziarg [19981999]


ldquofragmentedrdquo countries might have more proportional systems toensure the representation of all minorities ethnolinguistic frac-tionalization might also independently affect the provision ofpublic goods for instance as a means to ensure the consensus ofdifferent groups as in Alesina and Spolaore [1997] When weinclude ethnolinguistic fractionalization in our regressions itscoefcient is insignicant and the coefcient of the electoralvariable is not affected37

Another potentially relevant omitted variable is ideologyCountries with larger district magnitudes tend to be Nordic coun-tries with a Social-Democratic tradition or Southern Europeancountries that have had leftist governments for long periods oftime Thus when we add an ideology variable IDEOL (not avail-able for Latin American countries) which takes values rangingfrom 1 in more rightist governments to 5 in the more leftist ones(see Appendix 3 for a more detailed description) Its coefcienthas a t-statistic very close to 0 while the estimated coefcient ofthe electoral variable is virtually unchanged

Particularly in small samples one has to worry about therobustness of the results to possible outliers In the OECD groupwhen we exclude one country at a time the p-value on theestimated coefcient of the electoral variables in the transferregression never falls below 0538 In Latin America the initialresults are statistically less strong to begin with a similar exer-cise conrms that no individual country is responsible for theseweaker results

One could also worry that the 1991ndash1994 period might not berepresentative for instance because it was a period of widespreadscal consolidations in many countries We do not have a choicefor Latin American countries but for OECD countries we have

37 Alesina Baqir and Easterly [1999] nd that in ethnically fragmentedU S cities the provision of public goods is lower while Ordeshook and Shvestova[1994] present cross-country evidence on the impact of ethnic fractionalization onthe effective number of parties under different electoral systems

38 In principle two countries could have a nontrivial inuence on the re-sults the Netherlands with a very large average district magnitude (its propor-tional system consists of only one district electing 150 representatives) andFrance the only non-English speaking country with a majority system but at thesame time with a very high share of transfers in GDP close to 30 percent Indeedwhen we reestimate all our regressions dropping one country at a time wetypically nd that the exclusion of France causes the estimated coefcients ofelectoral variables and their t-statistics to increase dramatically the exclusion ofthe Netherlands causes them to fall somewhat although they always remainsignicant at the 5 percent level When we exclude both countries point estimatesand t-statistics rise considerably relative to the benchmark regressions of Table VI


annual data over the whole 1960 ndash1995 period When we reesti-mate the regressions of Table VI using cross-sectional data basedon 1960 ndash1995 averages we nd even stronger results withlarger and more signicant coefcients in the total spending andtransfer regressions

VIII TIME-SERIES EVIDENCE INTERACTIVE EFFECTS

In the OECD sample we can exploit the time dimension of thescal data to investigate the following question do electoral in-stitutions affect the response of scal variables to shocks Thespirit of the model suggests that the increase in governmentspending in response to a given shock should be higher in a moreproportional system39 Except for RAE the electoral measures weuse in the previous regressions display limited or no time varia-tion in the OECD sample40 hence it does not make sense to runpure xed effects panel regressions because the xed effects andthe electoral variables would be highly collinear We use twodifferent methodologies designed to estimate the responses tocommon shocks and to country-specic shocks respectively

VIII1 Common Shocks

To study the response to common shocks we borrow themethodology of Blanchard and Wolfers [2000] we estimate re-gressions of the type

(16) Git 5 c i 1 d t(1 1 Xi) 1 POP65it 1 LogGDPPCit 1 it

where i indexes the country and t the period c i is a country-specic intercept d t is a time dummy and the other variableshave been dened previously Thus in this specication the timedummies capture the common shocks and their effects depend onthe value of the electoral variable This dependence is captured bythe coefcient

Without a richer dynamic specication (which would be dif-cult to estimate given the short time series) it would not be

39 Strictly speaking this question is outside the model However it couldeasily be incorporated in a version of the model where the current policy acts asthe status quo

40 There is more time series variation in the Latin American sample due tothe frequent changes in electoral laws engendered by new administrations How-ever it is extremely difcult to reconstruct precisely electoral laws further backthan the early nineties and in many countries meaningful democratic electionsbegan only in the mid- or late eighties


reasonable to interpret this regression as capturing the effects ofelectoral institutions on year-to-year changes in scal policyhence all our variables are averages over a ve-year period andthe time index t refers to ve-year periods beginning in 1960 orthe earliest available year

Table VIII reports estimates of equation (16) (by nonlinearleast squares) on the sample of twenty OECD countries with thespendingGDP ratio (columns (1)ndash(3)) and the transferGDP ratio(columns (4)ndash(6)) as the dependent variables41 In the case ofRAE which has meaningful time series variation the regressionalso includes the variable by itself The rst row of Table VIIIdisplays the effects of the ldquopurerdquo time effects ie the increase inthe spendingGDP ratio that would have been experienced by a

41 The presentation of results also follows Blanchard and Wolfers [2000]

TABLE VIIICOMMON SHOCKS AND ELECTORAL SYSTEMS

PANEL REGRESSIONS INDUSTRIAL COUNTRIES

Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN

Time 1491 1402 1429 1319 1278 1240Log AM p time dummy 010 013

(208) (493)Log SM p time dummy 007 011

(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443

( 2 117) ( 2 104) ( 2 108) ( 2 294) (2 268) ( 2 253)Min 082 086 068 078 080 064Max 134 122 112 143 132 114Range 052 036 044 065 052 050Observations 112 112 112 112 112 112R2 092 092 091 096 096 096

Dependent variable share of total primary spending in GDP (columns (1)ndash(3)) and share of transfers inGDP (columns (4)ndash(6)) ve-year averages from 1960 (earliest available year) to 1995 See Appendix 3 for thedenition of all variables Estimation by nonlinear least squares (t-statistics are in parentheses) ()indicates statistical signicance at the 10 percent (5 percent) level Range is range of variation of dependentvariable associated with range of variation of electoral variable holding constant all other variables(minimum and maximum values also reported)


country with the average value of the electoral variable Thus thepure time effect is equal to the difference between the estimatedtime effects in 1990 ndash1995 (the last one) and 1960 ndash1964 The nextrow reports the estimate of with its t-statistic The rows labeledldquoMinrdquo and ldquoMaxrdquo answer the following question suppose that acommon shock causes the spendingGDP ratio to increase by 1percentage point in the country with average values of the elec-toral variable by how much does the spendingGDP ratio in-crease in the countries with the lowest and highest values of theelectoral variable The answer is given by 1 + Xmin and 1 +

Xmax respectively The row labeled ldquoRangerdquo displays the differ-ence between the two

Table VIII provides some support for the notion that inresponse to common shocks primary spending increases by morein more proportional systems than in other systems and muchstronger support for the notion that transfers increase by more inmore proportional systems In the primary spending regressionsall the estimates of are positive and signicant in the case ofLogAM (at the 5 percent level) and RAE (at the 10 percent level)According to the point estimates the range of variation in thespendingGDP ratio in response to a shock that causes the sameratio to increase by 1 percentage point in the ldquoaveragerdquo country isbetween 36 and 52 percentage points of GDP

In the transfers regressions (columns (4)ndash(6)) the coefcientsof the electoral variables are very close to those in the spendingregressions and the t-statistics are much highermdashthe p-valuesare all below 001 The implied range of variation in the transferGDP ratio in response to the ldquoaveragerdquo shock is even larger from5 to 65 percentage points of GDP Note also that the ldquopurerdquo timeeffect (rst row) is similar across all columns

VIII2 Country-Specic Shocks

We now allow electoral institutions to interact with country-specic shocks To do so we need a macroeconomic shock whoseeffects on spending are a priori clear An increase in unemploy-ment will cause most types of government spending to increaserelative to GDP both because of the working of automatic stabi-lizers and because of the discretionary response by the govern-ment Thus we estimate regressions of the type

(17)

Git 5 c i 1 d t 1 U it 1 U itX i 1 POP65it 1 LogGDPPCit 1 it


Uit is the unemployment rate in country i in period t Thuscaptures the interaction of a shock to unemployment with theelectoral system if a country with a more proportional systemresponds to a shock to unemployment by increasing spending ortransfers more is positive As before because RAE displaysmeaningful time series variation we also include the variable byitself in addition to its interaction with U

Table IX displays the results The estimated coefcient of Uis indeed always positive and signicant over a ve-year horizonin a majority system (where LogAM and LogSM are equal to 0) anincrease in unemployment by 1 percentage point is associatedwith an increase in the total primary spendingGDP ratio bybetween 42 and 48 percentage points and an increase in thetransfersGDP ratio by between 27 and 31 percentage points

The estimates of the interactive term coefcient are al-ways positive but they are signicant at the 5 percent level onlywhen transfersGDP is the dependent variable (columns (4)ndash(6))

TABLE IXCOUNTRY-SPECIFIC SHOCKS AND ELECTORAL SYSTEMS


Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN

Log AM p unempl rate 014 015(160) (315)

Log SM p unempl rate 009 011(125) (280)

RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)

Log GDP per capita 459 446 394 269 262 217(190) (184) (162) (204) (197) (160)

Range 072 047 046 077 057 053Observations 112 112 112 112 112 112R2 090 090 090 094 094 094



The implied economic signicance is also considerable from therow labeled ldquorangerdquo a 1 percentage point shock to unemploymentcauses the transfersGDP ratio to increase by between 53 and 77percentage points more in the most proportional electoral systemthan in the least proportional one

Both types of panel regressions are robust to outliers Wereestimated all regressions in Tables VIII and IX dropping onecountry at a time and both France and the Netherlands at thesame time The estimates of change only minimally and so dotheir p-values

IX RELATION WITH THE LITERATURE

We now discuss the relationship between our contributionand the existing literature in terms of both theory and empiricsIn Persson and Tabellini [1999a 2000 2001] two candidatesmake binding promises on the provision of a ldquouniversalrdquo type ofspending and on a ldquotargetablerdquo one (which is both district- andgroup-specic) The driving force is uncertainty by the policymak-ers over the distribution of votersrsquo preferences and therefore overthe identity of the median voter In a majoritarian system thecandidates compete for ldquoswingrdquo districts by directing the tar-getable instrument toward a narrower constituency identied intheir model with the ldquomiddle classrdquo Hence majoritarian systemshave higher spending on the more targetable instrument and alower provision of the universal public good By focusing on votersin a limited number of districts politicians in a majoritariansystem fail to internalize the overall distortions induced by tax-ation decisions thus also leading to a larger government

Thus in terms of the composition of spending these modelspredict that the universal type of spending will be higher inproportional systems while the targetable type of spending willbe higher in majoritarian systems in terms of the level of spend-ing the prediction is that total government expenditure will behigher in majoritarian systems In contrast our model predictsthat expenditure on transfers will be higher in proportional sys-tems while expenditure on purchases of goods and services willbe higher in majoritarian systems the effect of the electoralsystem on total government spending depends on the share of


transfers in total spendingmdashand more deeply the underlyingdistribution of voters preferences42

In estimating these models Persson and Tabellini mostlycapture the properties of the electoral system via a ldquomajoritarianrdquodummy variable43 In Persson and Tabellini [1999a] public goodsare the universal expenditure and transfers the targetable one Inthe empirical application the universal public good is denedmore specically as expenditure on order and safety healthtransportation and education in a cross section of 50 countriesthere is evidence that expenditure on these universal public goodsis indeed higher in proportional systems and that total govern-ment expenditure is higher in majoritarian systems In Perssonand Tabellini [2000 2001] welfare transfers are instead the uni-versal expenditure and local public goods are the targetable oneIn the empirical application they nd based on a panel of 60countries that majoritarian systems tend to have lower overallexpenditure and especially lower welfare spending44

Another difference with our approach is that our theory andour empirical results underscore the importance of disaggregat-ing the sample the relation between degree of proportionalityand size of transfers is positive and very robust among OECDcountries but not so for Latin American countries

X CONCLUDING REMARKS

In this paper we have studied the effects of electoral institu-tions on the size and composition of public expenditure in OECDand Latin American countries We have emphasized the distinc-tion between purchases of goods and services which are easier to

42 In Lizzeri and Persico [2001] two candidates make binding promises onthe level of spending on two types of government expenditure a universal publicgood and transfers that can be targeted geographically or by groups Even whenthe public good is more valuable to individuals in a majoritarian system wherethe spoils of ofce go to the winner on average one could have lower provision ofthe universal public good

43 For some regressions Persson and Tabellini [1999a] also use the inverseof the average district magnitude from Cox [1997] However the construction ofthis variable does not take into account the specic features of two-tier systemsthat we highlighted in Sections V and VI

44 However the main focus of these papers is on the dichotomy betweenpresidential and parliamentary systems in this case Persson and Tabelliniconsistently nd that presidential systems tend to have lower government expen-diture In an interesting study at a more disaggregated level Baqir [2001] ndsthat in a cross section of U S cities larger district councils are associated withlarger city governments and that spending is higher when councils are elected ldquoatlargerdquo than when they are elected by city district


target geographically and transfers which are easier to targetacross social groups We presented a theoretical model in whichvoters anticipating government policymaking under differentelectoral systems have an incentive to elect representatives moreprone to transfer (public good) spending in proportional (majori-tarian) systems The model also predicts higher total primaryspending in proportional (majoritarian) systems when the shareof transfer spending is high (low)

To test our predictions we have dened and constructedrigorous measures of proportionality that take into account theexistence of different voting methods multiple electoral tiers andelectoral thresholds In the empirical investigation we havefound strong support for our predictions in OECD countries andmuch weaker evidence for Latin American countries Interest-ingly the positive relation between transfers and the degree ofproportionality in OECD countries holds even within propor-tional systems highlighting the importance of constructing mea-sures of proportionality that go beyond the majoritarianpropor-tional dichotomy In future research it would be interesting toexamine the relation between electoral systems and the distribu-tion of expenditure within countries since existing models (in-cluding ours) have strong implications in this regard

APPENDIX 1

Given the assumption of uniform distribution of preferences( ) over the rectangle with vertices L H L H the voter withmedian preferences over and is the decisive voter in eachelectoral system This result would still hold if we relaxed theassumption about uniform distribution of preferences as long asthese are nondegenerate and bounded and allowed the represen-tative to be selected by a sequential vote on and regardless ofthe order of voting

Consider the majoritarian system rst and suppose initiallythat the sequencing on votes is rst on and then on Considerthe problem solved by individual i of group B in district k1 (forbrevity we will omit the subscript B to indicate the group whenthis does not create any ambiguity) Starting from the secondstage let k1

denote the value of that has prevailed by majorityvoting in the rst stage The problem of individual i in district k1is to nd the optimal value of for a representative who alreadyhas a value of equal to k1

Denote this optimal value of from


the perspective of individual i by ik1 This is found by maximiz-

ing with respect to k1expression (5) (with the index i replacing

the median voter index m) subject to tM sBM and g1

M being givenby suitable modications of expressions (4) and taking as given

k1 k2

k3 k2

and k3 It is easy to see that one obtains

ik1= ik1

Thus in the second stage an individual with medianvalue of prevails in each district

In the rst stage the optimal value of for a representativefrom the point of view of individual i in district k1 ik1

is foundby maximizing the same expression above with respect to ik1

with k1

= m Again it is easy to see that one obtains ik1=

ik1(2 2 ik1

) Hence the individual with median value of isdecisive in the rst stage

A similar reasoning applies when voting rst on then on and in the proportional system

APPENDIX 2

As we mentioned in the text in two-tier electoral systemsthere is a distinction between the upper bound on the number ofrepresentatives that can be elected in a district and the actualnumber of representatives that can be elected We denote with anasterisk actual sizes thus Sik denotes the actual district size ofdistrict k in tier i and analogous denitions hold for Si and Sinote that for T2 districts S2k = S2k always ie they are alwaysconditional on T1 results Clearly S2 = 0 for one-tier systems

In Remainder Seats systems S2 is variable in principle allor no seats could be attributed in T1 (S1 = Sto t) but typicallyonly some are (S1 S1) and the remainder seats are lled in T2(S2 = S2 = Sto t 2 S1) In Adjustment Seats systems S2 is xedhence S1 = S1 and S2 = S2 = Sto t 2 S1

21 Local Requirements

A rst type of local requirement sets a minimum share ofvotes that a party must obtain in a given district in order toparticipate in the allocation of seats in that district this is equiva-lent to a legal threshold of the same size applying to that tier Asecond type of local requirement denes a fraction h of an upperquota that a party must win in at least one district in T1 in orderto have access to the allocation of seats in T2 To translate thisrequirement into a legal threshold note that the binding con-straint is meeting the requirement in the largest district of T1


Hence the legal threshold in T2 is THR2 = hQ1 (S1max) We haveassembled information on the maximum district size in T1 andwe use it in constructing THR2 whenever the local requirement isof this type

22 Upper Quotas

We now provide the formulas for the upper quota in thedifferent voting methods and formulas In all cases we will referto ldquoparty Ardquo as the party whose share is equal to the upper quotaWe will assume that there are P parties in each district

MajorityPlurality Methods

FPTP clearly Q i(Sik) = 5Alternative Vote Q i(Sik) = 5Two-Round Q i(Sik) = 5

Largest Remainder Methods

In LR methods in which ldquoremaindersrdquo are transferred to anupper tier the formulas for the upper quotas are clearly the sameas the quotas themselves any party needs to reach the full quotato be assured of a seat The exception is the LR Imperiali methodwhere the number of quotas can exceed the number of seatsHenceHare Q i(Sik) = 1Sik

Droop Qi(Sik) = 1(1 + Sik)45

Imperiali Qi(Sik) = 1(1 + Sik)In LR methods in which remainder seats are attributed

within the same electoral tier the formulas for the Droop andImperiali quotas are unchanged For the Hare quota insteadthey depend on the relation between the number of seats in thedistrict and the number of parties P competing for the seats (seeLijphart and Gibberd [1977] and Gallagher [1991]) If Sik lt Pthe upper quota is 1(1 + Sik) as for the other methods If Sik $P the upper quota is (P 2 1)PSik for the Hare formula

45 In the Japanese SNTV method the share that guarantees a candidateelection without resorting to any second-round allocation of votes is dened by lawas the Droop quota hence Qi(Sik) = 1(1 + Sik) For the purpose of calculatingupper quotas we consider the Irish STV method as equivalent to a Droop formulaIn fact the Irish system establishes that the Droop quota guarantees election fora candidate We assume that this is also the quota for a party Hence Q i(Sik) =1(1 + Sik)


Highest Average Methods

drsquoHondt The least favorable distribution of votes for party Ais when another party (say party P) gets all the remaining votesParty A still gets one seat if party P gets a vote share VP which isjust short of VASik Thus when the vote share of the two partiesare divided by 12 Sik 2 1 party P gets a seat each timeThe last quotient VP Sik is (innitesimally) smaller than VA hence party A gets the last representatives Hence Qi(Sik) isdened by the conditions

VA 1 VP 5 1 VA 5 VPS ik Qi(S ik) 5 VA

which gives Qi(Sik) = 1(1 + Sik) This formula is valid irrespec-tive of the relation between number of parties and number ofseats

Modied St Lague We present the conditions for the Modi-ed St Lague formula since this is always used instead of thepure St Lague formula Under Modied St Lague the partyshares are divided by 1435 instead than by 135 In thiscase we need to distinguish between the case in which the num-ber of parties is larger than the number of seats (Sik lt P) and theopposite case In the former case the worst distribution of votesfor party A obtains when Sik other parties get a share of voteswhich is the same as the share of party A This implies that thevote share to ensure election is implicitly given by VA 14 = (1 2VA )14Sik which in turn implies Q i(Sik) = 1(1 + Sik) Ifinstead the number of parties is ldquointermediaterdquo (Sik $ P $Sik 2 + 1) then Q i(Sik) = 14(16 Sik 2 02P + 16) [Gal-lagher 1992] Finally if the number of parties is ldquosmallrdquo (P lt(Sik 2 + 1) then Q i(Sik) = 14(2Sik 2 P + 24) [Lijphart andGibberd 1977]

23 Calculation of Upper Marginal Share in Two-Tier RVSystems

In this Appendix we show the calculation for the case when aLargest Remainder method is used in T1 In our sample onlyNorway after 1990 uses a Highest Average method to attributeT1 seats The eight T2 seats are attributed to the T1 district listswith the eight highest averages after the averages that earned aseat Unlike in a Largest Remainder method in a Highest Aver-age method a seat can be attributed in T1 even to a party thatgets less than the upper quota Therefore it is more difcult to


approximate the share of votes that go to T2 as remainders andto calculate UMS2 However because the number allocated in T2seats is less than 5 percent of the assembly size by the rule we setout above we do not consider T2 in this case In any case becauseT2 is so small we know that UMS2 must be very close to UMS1This leaves Austria and Italy 1960 ndash1993 to represent two-tier RVsystems Seats are attributed in a T1 district only if the upperquota is met46 hence on average each seat attributed in T1 usesa fraction of district votes equal to Q1(S1) (assuming as it isalways the case in practice that this is larger than THR1 seesubsection V3) and a fraction of national votes equal to Q1(S1)D1 47 A total of S1 seats are attributed in T1 this way thusleaving a fraction 1 2 S1Q1(S1)D1 of national votes to be usedfor the allocation of the remaining S2 T2 seats48 Assuming thatall T2 districts have the same size and ignoring for the momentlegal thresholds to be guaranteed a seat in a T2 district a partyneeds a share Q2(S2) of all the average T2 district votes hencethe same fraction of all T2 votes hence a fraction Q2(S2)(1 2S1Q1(S1)D1) of national votes UMS2 can then be computed asUMS2 = max (Q2(S2)(1 2 S1Q1(S1)D1) THR2)

APPENDIX 3

31 List of Countries

OECD countries Australia Austria Belgium Canada Den-mark Finland France Germany Greece Ireland Italy JapanNetherlands Norway Portugal Spain Sweden SwitzerlandUnited Kingdom United States

Latin American countries (all variables are taken as averagesover the 1991ndash1994 period unless otherwise indicated) ArgentinaBolivia (1993ndash1994) Brazil (1991ndash1992) Chile Colombia (1992ndash1993) Costa Rica Dominican Republic Ecuador El Salvador Gua-

46 Recall that when a Largest Remainder method is used in T1 districts ofRV systems Q1(S1k) is the actual share of district votes used for each seatallocated in district k of T1

47 This is clearly an approximation because we replace the average ofQ1(S1 k) with Q1(S1)

48 We do not have information on S1 for each election however Taageperaand Shugart [1989] do provide information on the average value of S1 where theaverage is taken across elections in a given country In the empirical implemen-tation of our variables we use this average value as a proxy for the actual valueof S1 Throughout these calculations we assume that there are no abstentions orinvalid ballots


temala Honduras Jamaica (1993ndash1994) Mexico Nicaragua (1992ndash1994) Panama (1991ndash1993) Paraguay (1993ndash1994) Peru (1990ndash1991) Trinidad and Tobago (1992ndash1994) Uruguay (1991ndash1992)Venezuela

32 Fiscal Variables

OECD Countries

All scal variables are from the OECD Economic OutlookDatabase and refer to the general government

TRAN Transfers to households Dened as SSPG + TRPGusing the mnemonics of the Economic Outlook Data-base SSPG Social security benets to householdsTRPG Other transfers to households

PGOOD Public goods dened as the sum of governmentconsump-tion and government investment net of depreciationCGW + CGNW + CAPEXP CGW Government con-sumption wages CGNW Government consumption ex-cluding wages CAPEXP government investment net ofdepreciation plus net capital transfers paid

EXP Total primary government expenditure dened asTRAN + PGOOD + residual item where residual itemTSUB + YPEPG 2 GNINTP TSUB Subsidies to rmsYPEPG Property income paid by government GNINTPNet interest payments by government

Latin American Countries

For Argentina Bolivia Brazil Chile Colombia Costa RicaEcuador Mexico Panama Paraguay Peru Uruguay VenezuelaGavin and Perotti [1997] based on IMF data World Bank dataand national data For Dominican Republic El Salvador Guate-mala Honduras Jamaica Nicaragua Trinidad and Tobago IMFdata

All denitions are consistent and follow the IMF Govern-ment Financial Statistics classication which in turn is for largeaggregates virtually identical to the OECD Economic Outlookclassication


33 Political Variables

Log AM Logarithm of average district magnitude AM is de-ned as the weighted average of the average districtsizes of the two electoral tiers with weights equal tothe proportion of all representatives elected in thetwo tiers (see Section VI for details) Sources forOECD countries authorsrsquo calculations based on Ta-agepera and Shugart [1989] Lijphart [1994] andnational sources for Latin American countries Inter-Parliamentary Union (various years) Political data-base of the Americas (httpwwwgeorgetownedupdbaenglishhtml) and national sources (mostlyconstitutions and electoral laws in place at the timeswe take our cross-sectional observations)

Log SM Logarithm of Standardized Average District Magni-tude SM is dened as 1UMS 2 1 where UMS is theminimum share of national votes which guarantees aParliamentary seat to a party (see Section VI for de-tails) Sources for OECD countries authorsrsquo calcula-tions based on Taagepera and Shugart [1989] Lijphart[1994] and national sources for Latin American coun-tries Inter-Parliamentary Union (various years) Politi-cal database of the Americas and national sources

LogENPP Logarithm of the effective number of Parliamentaryparties ENPP is dened as psp

21 where sp is the shareof seats of party p Source for OECD countries Lijphart[1994] and unpublished data from Lijphart extended tothe early 1990s using national sources for Latin Ameri-can countries Inter-Parliamentary Union (variousyears) Political database of the Americas and nationalsources

IDEOL Ideological conguration of government The variabletakes values from 1 (dominant right-wing party) to 5(dominant left-wing party) Source Woldendorp Ke-man and Budge [1993] and updates from Perotti andKontopoulos [1999] (available only for OECD countries)


MAJ Dummy variable taking the value of one if the elec-toral system is majoritarian and zero otherwiseSource Persson and Tabellini [1999a]

MAGN Effective district magnitude Source Taagepera andShugart [1989] Lijphart [1994] and authorsrsquo exten-sions (until 1995)

34 Other Variables

ETHNIC Index of ethnolinguistic fractionalization for 1960It measures the probability that two randomly se-lected people from a given country will not belong tothe same ethnolinguistic group Source Atlas Nar-odov Mira [1964] as reported in Easterly and Le-vine [1997]

LogGDPPC Log of real GDP per capita in thousands of 1985international dollars Source Summers and Heston[1991] and Penn World Tables 56 update updatedfor the years after 1992 using World Bank data

OPEN Ratio of exports of goods and services plus importsof goods and services over two times GDP SourceWorld Bank World Development Indicators andPenn World Tables 56 update

POP65 Ratio of population above 65 to total populationSource World Bank World Development Indicators

U Unemployment rate Source OECD Economic Out-look Database

INTERNATIONAL MONETARY FUND AND CENTRE FOR ECONOMIC POLICY RESEARCH

EUROPEAN UNIVERSITY INSTITUTE AND CENTRE FOR ECONOMIC POLICY RESEARCH

EUROPEAN CENTRAL BANK

REFERENCES

Alesina Alberto Reza Baqir and William Easterly ldquoPublic Goods and EthnicDivisionsrdquo Quarterly Journal of Economics CXIV (1999) 1243ndash1284

Alesina Alberto and Enrico Spolaore ldquoOn the Number and Size of NationsrdquoQuarterly Journal of Economics CXII (1997) 1027ndash1056


Alesina Alberto and Romain Wacziarg ldquoOpenness Country Size and Govern-mentrdquo Journal of Public Economics LXIX (1998) 305ndash321

Alesina Alberto and Romain Wacziarg ldquoIs Europe Going Too Farrdquo Carnegie-Rochester Conference Series on Public Policy LI (1999) 1ndash42

Austen-Smith David and Jeffrey Banks ldquoElections Coalitions and LegislativeOutcomesrdquo American Political Science Review LXXXII (1988) 405ndash422

Baqir Reza ldquoDistricting and Government Overspendingrdquo IMF Working Paper0196 August 2001 Journal of Political Economy forthcoming

Baron David P and Daniel Diermeier ldquoElections Governments and Parlia-ments in Proportional Representation Systemsrdquo Quarterly Journal of Eco-nomics CXVI (2001) 933ndash968

Besley Timothy and Stephen Coate ldquoCentralized versus Decentralized Provisionof Local Public Goods A Political Economy Analysisrdquo NBER Working PaperNo 7084 1999

Blanchard Olivier Jean and Justin Wolfers ldquoThe Role of Shocks and Institutionsin the Rise of European Unemployment The Aggregate Evidencerdquo EconomicJournal CX (2000) C1ndash33

Chari V V Larry E Jones and Ramon Marimon ldquoOn the Economics of Split-Ticket Voting in Representative Democraciesrdquo American Economic ReviewLXXXVII (1997) 957ndash976

Cox Gary W Making Votes Count (New York NY Cambridge University Press1997)

Easterly William and Ross Levine ldquoAfricarsquos Growth Tragedy Policies and EthnicDivisionsrdquo Quarterly Journal of Economics CXII (1997) 1203ndash1250

Gallagher Michael ldquoProportionality Disproportionality and Electoral SystemsrdquoElectoral Studies X (1991) 33ndash51

Gavin Michael Ricardo Hausmann Roberto Perotti and Ernesto Talvi ldquoManag-ing Fiscal Policy in Latin America and the Caribbean Volatility Procyclical-ity and Limited Creditworthinessrdquo IADB Working Paper No 326 1996

Gavin Michael and Roberto Perotti ldquoFiscal Policy in Latin Americardquo NBERMacroeconomics Annual Ben Bernanke and Julio Rotemberg eds (Cam-bridge MA MIT Press 1997)

Inter-Parliamentary Union Chronicle of Parliamentary Elections and Develop-ments Geneva Switzerland (various years)

Lijphart Arend Electoral Systems and Party Systems (Oxford UK Oxford Uni-versity Press 1994)

Lizzeri Alessandro and Nicola Persico ldquoThe Provision of Public Goods underAlternative Electoral Incentivesrdquo American Economic Review XCI (2001)225ndash239

Milesi-Ferretti Gian Maria and Roberto Perotti with Massimo Rostagno ldquoElec-toral Rules and Public Spendingrdquo Centre for Economic Policy Research Dis-cussion Paper No 2742 March 2001

Ordeshook Peter C and Olga V Shvestova ldquoEthnic Heterogeneity DistrictMagnitude and the Number of Partiesrdquo American Journal of Political Sci-ence XXXVIII (1994) 100ndash123

Perotti Roberto and Yianos Kontopoulos ldquoFragmented Fiscal Policyrdquo mimeoColumbia University (1999) Journal of Public Economics forthcoming

Persson Torsten and Guido Tabellini ldquoThe Size and Scope of GovernmentComparative Politics with Rational Politiciansrdquo European Economic ReviewXLIII (1999a) 699ndash735

Persson Torsten and Guido Tabellini ldquoPolitical Economics and Public FinancerdquoNBER Working Paper No 7097 1999b in Handbook of Public EconomicsVol 3 Alan Auerbach and Martin Feldstein eds (Amsterdam The Nether-lands Elsevier Science Publishers) forthcoming

Persson Torsten and Guido Tabellini Political Economics (Cambridge MA MITPress 2000)

Persson Torsten and Guido Tabellini ldquoPolitical Institutions and Policy Out-comes What Are the Stylized Factsrdquo CEPR Discussion Paper No 2872 June2001

Rae Douglas W The Political Consequences of Electoral Laws (New Haven CTYale University Press 1967)


Rodrik Dani ldquoWhy Do More Open Economies Have Bigger Governmentsrdquo Jour-nal of Political Economy CVI (1998) 997ndash1032

Stein Ernesto Ernesto Talvi and Alejandro Grisanti ldquoInstitutional Arrange-ments and Fiscal Performance The Latin American Experiencerdquo in FiscalInstitutions and Fiscal Performance James Poterba and Jurgen von Hageneds (Chicago IL University of Chicago Press for NBER 1999)

Summers Robert and Alan Heston ldquoPenn World Tables Mark 5 An ExpandedSet of International Comparisonsrdquo Quarterly Journal of Economics CVI(1991) 327ndash368

Taagepera Rein and Matthew Soberg Shugart Seats and Votes (New Haven CTYale University Press 1989)

Woldendorp J H Keman and I Budge ldquoA Compilation of Political Data onIndustrialized Parliamentary Democraciesrdquo European Journal of PoliticalResearch XXV (1993) 1ndash120


In this paper we study how the electoral system shapes thistrade-off and the incentives of elected ofcials to allocate reve-nues to the two types of spending We show that proportionalsystems are more geared to spending on transfers while majori-tarian systems are more prone to public good spending Thusloosely speaking our model captures the widespread notion thatldquoproportional systems allow representation of a greater variety ofinterestsrdquo (a frequent claim by advocates of this system) whileldquomajoritarian systems are more grounded in local interestsrdquo

The model is based on the logic of strategic delegation ofChari Jones and Marimon [1997] and Besley and Coate [1999]extended to a framework with two types of government spendingand with different electoral systems In a majoritarian systemeach district elects one representative If the distribution of dif-ferent social groups is similar across districts all representativeswill belong to the same social group Hence all elected represen-tatives derive utility from the same type of transfers but eachderives utility from a different public good It follows that electorswill have an incentive to vote for individuals with stronger pref-erences for public goods relative to transfers in order to biasgovernment expenditure on public goods toward their district Inequilibrium the result is just high expenditure on public goods

In a proportional system each district elects more than onerepresentative Hence more than one social group will be repre-sented in Parliament in contrast to the majoritarian systemeach representative now derives utility from a different type oftransfer Individuals have an incentive to vote for representativeswith stronger preference for transfers in order to bias the spend-ing decisions of the government toward their own type of trans-fers In equilibrium high spending on transfers is the result Themodel then predicts that spending on transfers is higher in pro-portional systems and spending on public goods is higher inmajoritarian systems Total government spending is higher inproportional systems if the median voter values relatively littlethe public good and relatively highly private consumption andtransfers lower in the opposite case

One virtue of our model is that it captures common and webelieve plausible views both of the properties of different elec-toral systems and of different types of government spendingOther recent positive models of electoral systemsmdashsuch as Pers-son and Tabellini [1999a 2000] and Lizzeri and Persico [2001]mdashexploit the differences between alternative types of government







II THE MODEL












(1) U ijk 5 (1 2 t) i is ji(1 2 i)gk

1 2 i


















(2) VM(k1 k2) 5 ( k1k1

1 k2k2

) log (1 2 t) 1 ( k1(1 2 k1

)

1 k2(1 2 k2

)) log sB 1 (1 2 k1) log gk1 1 (1 2 k2

) log gk2




(3) t 5 BsB 1 gk11 gk2




Cjs j g





(4)

tM(k1k2) 52 2 ( k1

k1 1 k2k2

)

2

sBM(k1k2) 5

k1(1 2 k1

) 1 k2(1 2 k2

)

2 sA

M(k1k2) 5 0 sCM(k1k2) 5 0

gk1

M(k1k2) 51 2 k1

2 gk2

M(k1k2) 51 2 k2

2 gk3

M(k1k2) 5 0

gM(k1k2) ordm gk1

M(k1k2) 1 gk2

M(k1k2) 52 2 k1 2 k2

2






M (k1 k2) = ( k1(1 2 k1)) 2 and sk2M (k1 k2) =


utility parameters


(5) E(VmBk1

M ) 5 Or= 2

3



M(k1kr)]



(6) M 5m

2 2 m M 5 m



tM 5 1 2m m

2 2 m

(7) sM 5m(1 2 m)

2 2 m

gM 52(1 2 m)

2 2 m








(8) VP( j1 j2) 5 ( j1 j1 1 j2 j2) log (1 2 t) 1 j1(1 2 j1

) log sj1

1 j2(1 2 j2

) log sj21 (2 2 j1

2 j2) log (g3)

where j1and j1



(9) t 5 j1sj1 1 j2s j2 1 g


(10)

tP( j1 j2) 52 2 ( j1 j1 1 j2 j2)

2

s j1P( j1 j2) 5

j1(1 2 j1)

2 s j2

P( j1 j2) 5j2(1 2 j2)

2 s j3

P( j1 j2) 5 0


P 5j1(1 2 j1) 1 j2(1 2 j2)

2


P( j1 j2) 1 g j3P( j1 j2) 5

2 2 j1 2 j22







(11) E(Vj1m

P )= Or= 2

3


P ( j1 jr)

+ (1 2 m) log gP( j1 jr)]



(12) P 5m

2 2 m P 5

m(2 2 m)

1 1 m(1 2 m)


tP 51 1 m(1 2 2 m)

1 1 m(1 2 m)

(13) sP 52 m(1 2 m)

1 1 m(1 2 m)

gP 51 2 m

1 1 m(1 2 m)



IV3 Predictions


sP sM

(14) gP gM

tP tMlaquo m

1(2 2 m)














V1 Electoral Tiers








V2 Voting Method










V3 Legal Thresholds





























SV RV

































VII1 The Data












TA

BL

EII

SU

MM

AR

YS

TA

TIS

TIC

S

All

Mea

nA

llS

DA

llM

inA

llM

axO

EC

DM

ean

OE

CD

SD

OE

CD

Min

OE

CD

Max

LA

Mea

nL

AS

DL

AM

inL

AM

ax

Avg

dis

tric

tm

agn

(A

M)

161

307

115

015

323

115

017

329

91

120

Sta

nd

dist

rict

mag

n(

SM

)23

639

71

180

820

133

114

83

272

461

180

8R

AE

34

40

326

12

41

90

36

83

76

50

326

1P

rim

ary

expe

ndG

DP

326

151

103

666

454

94

327

666

198

55

103

298

Tra

nfe

rsG

DP

134

95

14

316

213

63

115

316

55

36

14

146

Pub

lic

good

sG

DP

181

56

71

299

219

42

1629

914

44

17

123

5G

DP

per

capi

ta84

0055

0012

0018

100

1320

028

0068

0018

100

3500

1800

1200

7200

Pop

uls

hare

over

659

75

22

919

514

42

113

195

52

22

911

8E

ffn

umb

ofpa

rtie

s3

51

71

38

43

71

61

98

43

31

71

38

4E

TH

NIC

(fra

cti

ndex

)20

119

81

7522

122

11

7518

117

51

159

9O

PE

N(o

penn

ess)

305

173

74

961

301

141

87

629

309

204

74

961

Cro

ss-s

ecti

onal

aver

ages

sta

nda

rdde

viat

ions

min

imum

and

max

imu

mva

lues

for

all

the

vari

able

sfo

rth

efu

llsa

mpl

eth

eO

EC

Dsa

mpl

ean

dth

eL

atin

Am

eric

ansa

mpl

eS

eeth

eA

ppen

dix

for

vari

able

srsquom

nem

onic

sE

ach

cros

s-se

ctio

nalo

bser

vati

onis

the

1991

ndash199

4(o

rcl

oses

tav

aila

ble

peri

od)a

vera

geof

the

vari

able

for

the

cou

ntry

Num

ber

ofob

serv

atio

ns

40fo

rth

ew

hol

esa

mpl

e(3

5fo

rR

AE

)20

for

the

OE

CD

sam

ple

20fo

rth

eL

atin

Am

eric

ansa

mpl

e(1

5fo

rR

AE

)


















Panel B OECD












VII3 Basic Results






TA

BL

EV

PR

IMA

RY

SP

EN

DIN

G

TR

AN

SF

ER

SP

UB

LIC

GO

OD

SA

ND

EL

EC

TO

RA

LS

YS

TE

MS

(FU

LL

SA

MP

LE)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

105

170

20

60(L

ogA

M)

(11

0)(3

49)

(1

12)

Log

stan

ddi

stm

agn

074

132

20

48(L

ogS

M)

(08

9)(3

06)

(1

03)

RA

E0

320

382

002

(10

5)(2

21)

(0

13)

Pop

sha

reov

er65

113

114

130

125

127

142

20

122

012

20

10(1

86)

(1

87)

(1

95)

(4

04)

(3

96)

(3

81)

(0

34)

(03

5)(0

27)

Log

GD

Ppe

rca

pita

116

083

054

219

169

145

21

642

146

21

72(0

37)

(02

6)(0

16)

(13

7)(1

04)

(07

8)(0

92)

(08

3)(0

93)

OE

CD

dum

my

134

913

82

115

51

201

752

021

108

310

64

106

5(2

10)

(2

14)

(1

57)

(03

7)(0

52)

(00

5)(2

98)

(2

92)

(2

60)

C

onst

ant

275

574

108

62

217

92

174

62

117

529

32

278

328

87

(01

1)(0

23)

(04

2)(1

73)

(1

36)

(08

1)(2

09)

(2

01)

(20

0)

Ran

ge5

253

848

308

516

869

763

022

490

55O

bser

vati

ons

4040

3540

4035

4040

35R

20

760

750

730

840

830

800

440

430

38

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s












TA

BL

EV

IP

RIM

AR

YS

PE

ND

ING

TR

AN

SF

ER

S

PU

BL

ICG

OO

DS

AN

DE

LE

CT

OR

AL

SY

ST

EM

S(O

EC

DC

OU

NT

RIE

S)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

305

303

20

34(L

ogA

M)

(21

0)

(40

3)

(04

1)L

ogst

and

dist

mag

n2

562

632

028

(Log

SM

)(1

89)

(3

64)

(0

37)

RA

E1

411

622

064

(12

9)(2

56)

(1

14)

Pop

sha

reov

er65

137

115

143

119

095

117

022

025

040

(14

9)(1

17)

(14

0)(2

50)

(1

80)

(1

99)

(0

43)

(04

5)(0

75)

Log

GD

Ppe

rca

pita

171

516

39

157

67

737

056

594

484

574

09(2

36)

(2

23)

(2

03)

(20

5)

(18

0)

(14

8)(1

08)

(11

1)(1

03)

OE

CD

214

210

213

154

212

092

274

42

264

42

254

04

223

17

224

34

224

04

(20

2)

(18

5)

(16

2)(2

04)

(1

70)

(12

6)(0

58)

(06

1)(0

63)

Ran

ge15

28

128

09

1115

18

131

310

45

169

141

413

Obs

erva

tion

s20

2020

2020

2020

2020

R2

034

031

024

061

057

044

20

82

08

20

1

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s




VII4 Robustness




(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD



Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)


Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)





















VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES








































II THE MODEL












(1) U ijk 5 (1 2 t) i is ji(1 2 i)gk

1 2 i


















(2) VM(k1 k2) 5 ( k1k1

1 k2k2

) log (1 2 t) 1 ( k1(1 2 k1

)

1 k2(1 2 k2

)) log sB 1 (1 2 k1) log gk1 1 (1 2 k2

) log gk2




(3) t 5 BsB 1 gk11 gk2




Cjs j g





(4)

tM(k1k2) 52 2 ( k1

k1 1 k2k2

)

2

sBM(k1k2) 5

k1(1 2 k1

) 1 k2(1 2 k2

)

2 sA

M(k1k2) 5 0 sCM(k1k2) 5 0

gk1

M(k1k2) 51 2 k1

2 gk2

M(k1k2) 51 2 k2

2 gk3

M(k1k2) 5 0

gM(k1k2) ordm gk1

M(k1k2) 1 gk2

M(k1k2) 52 2 k1 2 k2

2






M (k1 k2) = ( k1(1 2 k1)) 2 and sk2M (k1 k2) =


utility parameters


(5) E(VmBk1

M ) 5 Or= 2

3



M(k1kr)]



(6) M 5m

2 2 m M 5 m



tM 5 1 2m m

2 2 m

(7) sM 5m(1 2 m)

2 2 m

gM 52(1 2 m)

2 2 m








(8) VP( j1 j2) 5 ( j1 j1 1 j2 j2) log (1 2 t) 1 j1(1 2 j1

) log sj1

1 j2(1 2 j2

) log sj21 (2 2 j1

2 j2) log (g3)

where j1and j1



(9) t 5 j1sj1 1 j2s j2 1 g


(10)

tP( j1 j2) 52 2 ( j1 j1 1 j2 j2)

2

s j1P( j1 j2) 5

j1(1 2 j1)

2 s j2

P( j1 j2) 5j2(1 2 j2)

2 s j3

P( j1 j2) 5 0


P 5j1(1 2 j1) 1 j2(1 2 j2)

2


P( j1 j2) 1 g j3P( j1 j2) 5

2 2 j1 2 j22







(11) E(Vj1m

P )= Or= 2

3


P ( j1 jr)

+ (1 2 m) log gP( j1 jr)]



(12) P 5m

2 2 m P 5

m(2 2 m)

1 1 m(1 2 m)


tP 51 1 m(1 2 2 m)

1 1 m(1 2 m)

(13) sP 52 m(1 2 m)

1 1 m(1 2 m)

gP 51 2 m

1 1 m(1 2 m)



IV3 Predictions


sP sM

(14) gP gM

tP tMlaquo m

1(2 2 m)














V1 Electoral Tiers








V2 Voting Method










V3 Legal Thresholds





























SV RV

































VII1 The Data












TA

BL

EII

SU

MM

AR

YS

TA

TIS

TIC

S

All

Mea

nA

llS

DA

llM

inA

llM

axO

EC

DM

ean

OE

CD

SD

OE

CD

Min

OE

CD

Max

LA

Mea

nL

AS

DL

AM

inL

AM

ax

Avg

dis

tric

tm

agn

(A

M)

161

307

115

015

323

115

017

329

91

120

Sta

nd

dist

rict

mag

n(

SM

)23

639

71

180

820

133

114

83

272

461

180

8R

AE

34

40

326

12

41

90

36

83

76

50

326

1P

rim

ary

expe

ndG

DP

326

151

103

666

454

94

327

666

198

55

103

298

Tra

nfe

rsG

DP

134

95

14

316

213

63

115

316

55

36

14

146

Pub

lic

good

sG

DP

181

56

71

299

219

42

1629

914

44

17

123

5G

DP

per

capi

ta84

0055

0012

0018

100

1320

028

0068

0018

100

3500

1800

1200

7200

Pop

uls

hare

over

659

75

22

919

514

42

113

195

52

22

911

8E

ffn

umb

ofpa

rtie

s3

51

71

38

43

71

61

98

43

31

71

38

4E

TH

NIC

(fra

cti

ndex

)20

119

81

7522

122

11

7518

117

51

159

9O

PE

N(o

penn

ess)

305

173

74

961

301

141

87

629

309

204

74

961

Cro

ss-s

ecti

onal

aver

ages

sta

nda

rdde

viat

ions

min

imum

and

max

imu

mva

lues

for

all

the

vari

able

sfo

rth

efu

llsa

mpl

eth

eO

EC

Dsa

mpl

ean

dth

eL

atin

Am

eric

ansa

mpl

eS

eeth

eA

ppen

dix

for

vari

able

srsquom

nem

onic

sE

ach

cros

s-se

ctio

nalo

bser

vati

onis

the

1991

ndash199

4(o

rcl

oses

tav

aila

ble

peri

od)a

vera

geof

the

vari

able

for

the

cou

ntry

Num

ber

ofob

serv

atio

ns

40fo

rth

ew

hol

esa

mpl

e(3

5fo

rR

AE

)20

for

the

OE

CD

sam

ple

20fo

rth

eL

atin

Am

eric

ansa

mpl

e(1

5fo

rR

AE

)


















Panel B OECD












VII3 Basic Results






TA

BL

EV

PR

IMA

RY

SP

EN

DIN

G

TR

AN

SF

ER

SP

UB

LIC

GO

OD

SA

ND

EL

EC

TO

RA

LS

YS

TE

MS

(FU

LL

SA

MP

LE)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

105

170

20

60(L

ogA

M)

(11

0)(3

49)

(1

12)

Log

stan

ddi

stm

agn

074

132

20

48(L

ogS

M)

(08

9)(3

06)

(1

03)

RA

E0

320

382

002

(10

5)(2

21)

(0

13)

Pop

sha

reov

er65

113

114

130

125

127

142

20

122

012

20

10(1

86)

(1

87)

(1

95)

(4

04)

(3

96)

(3

81)

(0

34)

(03

5)(0

27)

Log

GD

Ppe

rca

pita

116

083

054

219

169

145

21

642

146

21

72(0

37)

(02

6)(0

16)

(13

7)(1

04)

(07

8)(0

92)

(08

3)(0

93)

OE

CD

dum

my

134

913

82

115

51

201

752

021

108

310

64

106

5(2

10)

(2

14)

(1

57)

(03

7)(0

52)

(00

5)(2

98)

(2

92)

(2

60)

C

onst

ant

275

574

108

62

217

92

174

62

117

529

32

278

328

87

(01

1)(0

23)

(04

2)(1

73)

(1

36)

(08

1)(2

09)

(2

01)

(20

0)

Ran

ge5

253

848

308

516

869

763

022

490

55O

bser

vati

ons

4040

3540

4035

4040

35R

20

760

750

730

840

830

800

440

430

38

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s












TA

BL

EV

IP

RIM

AR

YS

PE

ND

ING

TR

AN

SF

ER

S

PU

BL

ICG

OO

DS

AN

DE

LE

CT

OR

AL

SY

ST

EM

S(O

EC

DC

OU

NT

RIE

S)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

305

303

20

34(L

ogA

M)

(21

0)

(40

3)

(04

1)L

ogst

and

dist

mag

n2

562

632

028

(Log

SM

)(1

89)

(3

64)

(0

37)

RA

E1

411

622

064

(12

9)(2

56)

(1

14)

Pop

sha

reov

er65

137

115

143

119

095

117

022

025

040

(14

9)(1

17)

(14

0)(2

50)

(1

80)

(1

99)

(0

43)

(04

5)(0

75)

Log

GD

Ppe

rca

pita

171

516

39

157

67

737

056

594

484

574

09(2

36)

(2

23)

(2

03)

(20

5)

(18

0)

(14

8)(1

08)

(11

1)(1

03)

OE

CD

214

210

213

154

212

092

274

42

264

42

254

04

223

17

224

34

224

04

(20

2)

(18

5)

(16

2)(2

04)

(1

70)

(12

6)(0

58)

(06

1)(0

63)

Ran

ge15

28

128

09

1115

18

131

310

45

169

141

413

Obs

erva

tion

s20

2020

2020

2020

2020

R2

034

031

024

061

057

044

20

82

08

20

1

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s




VII4 Robustness




(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD



Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)


Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)





















VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES





































(1) U ijk 5 (1 2 t) i is ji(1 2 i)gk

1 2 i


















(2) VM(k1 k2) 5 ( k1k1

1 k2k2

) log (1 2 t) 1 ( k1(1 2 k1

)

1 k2(1 2 k2

)) log sB 1 (1 2 k1) log gk1 1 (1 2 k2

) log gk2




(3) t 5 BsB 1 gk11 gk2




Cjs j g





(4)

tM(k1k2) 52 2 ( k1

k1 1 k2k2

)

2

sBM(k1k2) 5

k1(1 2 k1

) 1 k2(1 2 k2

)

2 sA

M(k1k2) 5 0 sCM(k1k2) 5 0

gk1

M(k1k2) 51 2 k1

2 gk2

M(k1k2) 51 2 k2

2 gk3

M(k1k2) 5 0

gM(k1k2) ordm gk1

M(k1k2) 1 gk2

M(k1k2) 52 2 k1 2 k2

2






M (k1 k2) = ( k1(1 2 k1)) 2 and sk2M (k1 k2) =


utility parameters


(5) E(VmBk1

M ) 5 Or= 2

3



M(k1kr)]



(6) M 5m

2 2 m M 5 m



tM 5 1 2m m

2 2 m

(7) sM 5m(1 2 m)

2 2 m

gM 52(1 2 m)

2 2 m








(8) VP( j1 j2) 5 ( j1 j1 1 j2 j2) log (1 2 t) 1 j1(1 2 j1

) log sj1

1 j2(1 2 j2

) log sj21 (2 2 j1

2 j2) log (g3)

where j1and j1



(9) t 5 j1sj1 1 j2s j2 1 g


(10)

tP( j1 j2) 52 2 ( j1 j1 1 j2 j2)

2

s j1P( j1 j2) 5

j1(1 2 j1)

2 s j2

P( j1 j2) 5j2(1 2 j2)

2 s j3

P( j1 j2) 5 0


P 5j1(1 2 j1) 1 j2(1 2 j2)

2


P( j1 j2) 1 g j3P( j1 j2) 5

2 2 j1 2 j22







(11) E(Vj1m

P )= Or= 2

3


P ( j1 jr)

+ (1 2 m) log gP( j1 jr)]



(12) P 5m

2 2 m P 5

m(2 2 m)

1 1 m(1 2 m)


tP 51 1 m(1 2 2 m)

1 1 m(1 2 m)

(13) sP 52 m(1 2 m)

1 1 m(1 2 m)

gP 51 2 m

1 1 m(1 2 m)



IV3 Predictions


sP sM

(14) gP gM

tP tMlaquo m

1(2 2 m)














V1 Electoral Tiers








V2 Voting Method










V3 Legal Thresholds





























SV RV

































VII1 The Data












TA

BL

EII

SU

MM

AR

YS

TA

TIS

TIC

S

All

Mea

nA

llS

DA

llM

inA

llM

axO

EC

DM

ean

OE

CD

SD

OE

CD

Min

OE

CD

Max

LA

Mea

nL

AS

DL

AM

inL

AM

ax

Avg

dis

tric

tm

agn

(A

M)

161

307

115

015

323

115

017

329

91

120

Sta

nd

dist

rict

mag

n(

SM

)23

639

71

180

820

133

114

83

272

461

180

8R

AE

34

40

326

12

41

90

36

83

76

50

326

1P

rim

ary

expe

ndG

DP

326

151

103

666

454

94

327

666

198

55

103

298

Tra

nfe

rsG

DP

134

95

14

316

213

63

115

316

55

36

14

146

Pub

lic

good

sG

DP

181

56

71

299

219

42

1629

914

44

17

123

5G

DP

per

capi

ta84

0055

0012

0018

100

1320

028

0068

0018

100

3500

1800

1200

7200

Pop

uls

hare

over

659

75

22

919

514

42

113

195

52

22

911

8E

ffn

umb

ofpa

rtie

s3

51

71

38

43

71

61

98

43

31

71

38

4E

TH

NIC

(fra

cti

ndex

)20

119

81

7522

122

11

7518

117

51

159

9O

PE

N(o

penn

ess)

305

173

74

961

301

141

87

629

309

204

74

961

Cro

ss-s

ecti

onal

aver

ages

sta

nda

rdde

viat

ions

min

imum

and

max

imu

mva

lues

for

all

the

vari

able

sfo

rth

efu

llsa

mpl

eth

eO

EC

Dsa

mpl

ean

dth

eL

atin

Am

eric

ansa

mpl

eS

eeth

eA

ppen

dix

for

vari

able

srsquom

nem

onic

sE

ach

cros

s-se

ctio

nalo

bser

vati

onis

the

1991

ndash199

4(o

rcl

oses

tav

aila

ble

peri

od)a

vera

geof

the

vari

able

for

the

cou

ntry

Num

ber

ofob

serv

atio

ns

40fo

rth

ew

hol

esa

mpl

e(3

5fo

rR

AE

)20

for

the

OE

CD

sam

ple

20fo

rth

eL

atin

Am

eric

ansa

mpl

e(1

5fo

rR

AE

)


















Panel B OECD












VII3 Basic Results






TA

BL

EV

PR

IMA

RY

SP

EN

DIN

G

TR

AN

SF

ER

SP

UB

LIC

GO

OD

SA

ND

EL

EC

TO

RA

LS

YS

TE

MS

(FU

LL

SA

MP

LE)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

105

170

20

60(L

ogA

M)

(11

0)(3

49)

(1

12)

Log

stan

ddi

stm

agn

074

132

20

48(L

ogS

M)

(08

9)(3

06)

(1

03)

RA

E0

320

382

002

(10

5)(2

21)

(0

13)

Pop

sha

reov

er65

113

114

130

125

127

142

20

122

012

20

10(1

86)

(1

87)

(1

95)

(4

04)

(3

96)

(3

81)

(0

34)

(03

5)(0

27)

Log

GD

Ppe

rca

pita

116

083

054

219

169

145

21

642

146

21

72(0

37)

(02

6)(0

16)

(13

7)(1

04)

(07

8)(0

92)

(08

3)(0

93)

OE

CD

dum

my

134

913

82

115

51

201

752

021

108

310

64

106

5(2

10)

(2

14)

(1

57)

(03

7)(0

52)

(00

5)(2

98)

(2

92)

(2

60)

C

onst

ant

275

574

108

62

217

92

174

62

117

529

32

278

328

87

(01

1)(0

23)

(04

2)(1

73)

(1

36)

(08

1)(2

09)

(2

01)

(20

0)

Ran

ge5

253

848

308

516

869

763

022

490

55O

bser

vati

ons

4040

3540

4035

4040

35R

20

760

750

730

840

830

800

440

430

38

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s












TA

BL

EV

IP

RIM

AR

YS

PE

ND

ING

TR

AN

SF

ER

S

PU

BL

ICG

OO

DS

AN

DE

LE

CT

OR

AL

SY

ST

EM

S(O

EC

DC

OU

NT

RIE

S)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

305

303

20

34(L

ogA

M)

(21

0)

(40

3)

(04

1)L

ogst

and

dist

mag

n2

562

632

028

(Log

SM

)(1

89)

(3

64)

(0

37)

RA

E1

411

622

064

(12

9)(2

56)

(1

14)

Pop

sha

reov

er65

137

115

143

119

095

117

022

025

040

(14

9)(1

17)

(14

0)(2

50)

(1

80)

(1

99)

(0

43)

(04

5)(0

75)

Log

GD

Ppe

rca

pita

171

516

39

157

67

737

056

594

484

574

09(2

36)

(2

23)

(2

03)

(20

5)

(18

0)

(14

8)(1

08)

(11

1)(1

03)

OE

CD

214

210

213

154

212

092

274

42

264

42

254

04

223

17

224

34

224

04

(20

2)

(18

5)

(16

2)(2

04)

(1

70)

(12

6)(0

58)

(06

1)(0

63)

Ran

ge15

28

128

09

1115

18

131

310

45

169

141

413

Obs

erva

tion

s20

2020

2020

2020

2020

R2

034

031

024

061

057

044

20

82

08

20

1

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s




VII4 Robustness




(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD



Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)


Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)





















VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES













































(2) VM(k1 k2) 5 ( k1k1

1 k2k2

) log (1 2 t) 1 ( k1(1 2 k1

)

1 k2(1 2 k2

)) log sB 1 (1 2 k1) log gk1 1 (1 2 k2

) log gk2




(3) t 5 BsB 1 gk11 gk2




Cjs j g





(4)

tM(k1k2) 52 2 ( k1

k1 1 k2k2

)

2

sBM(k1k2) 5

k1(1 2 k1

) 1 k2(1 2 k2

)

2 sA

M(k1k2) 5 0 sCM(k1k2) 5 0

gk1

M(k1k2) 51 2 k1

2 gk2

M(k1k2) 51 2 k2

2 gk3

M(k1k2) 5 0

gM(k1k2) ordm gk1

M(k1k2) 1 gk2

M(k1k2) 52 2 k1 2 k2

2






M (k1 k2) = ( k1(1 2 k1)) 2 and sk2M (k1 k2) =


utility parameters


(5) E(VmBk1

M ) 5 Or= 2

3



M(k1kr)]



(6) M 5m

2 2 m M 5 m



tM 5 1 2m m

2 2 m

(7) sM 5m(1 2 m)

2 2 m

gM 52(1 2 m)

2 2 m








(8) VP( j1 j2) 5 ( j1 j1 1 j2 j2) log (1 2 t) 1 j1(1 2 j1

) log sj1

1 j2(1 2 j2

) log sj21 (2 2 j1

2 j2) log (g3)

where j1and j1



(9) t 5 j1sj1 1 j2s j2 1 g


(10)

tP( j1 j2) 52 2 ( j1 j1 1 j2 j2)

2

s j1P( j1 j2) 5

j1(1 2 j1)

2 s j2

P( j1 j2) 5j2(1 2 j2)

2 s j3

P( j1 j2) 5 0


P 5j1(1 2 j1) 1 j2(1 2 j2)

2


P( j1 j2) 1 g j3P( j1 j2) 5

2 2 j1 2 j22







(11) E(Vj1m

P )= Or= 2

3


P ( j1 jr)

+ (1 2 m) log gP( j1 jr)]



(12) P 5m

2 2 m P 5

m(2 2 m)

1 1 m(1 2 m)


tP 51 1 m(1 2 2 m)

1 1 m(1 2 m)

(13) sP 52 m(1 2 m)

1 1 m(1 2 m)

gP 51 2 m

1 1 m(1 2 m)



IV3 Predictions


sP sM

(14) gP gM

tP tMlaquo m

1(2 2 m)














V1 Electoral Tiers








V2 Voting Method










V3 Legal Thresholds





























SV RV

































VII1 The Data












TA

BL

EII

SU

MM

AR

YS

TA

TIS

TIC

S

All

Mea

nA

llS

DA

llM

inA

llM

axO

EC

DM

ean

OE

CD

SD

OE

CD

Min

OE

CD

Max

LA

Mea

nL

AS

DL

AM

inL

AM

ax

Avg

dis

tric

tm

agn

(A

M)

161

307

115

015

323

115

017

329

91

120

Sta

nd

dist

rict

mag

n(

SM

)23

639

71

180

820

133

114

83

272

461

180

8R

AE

34

40

326

12

41

90

36

83

76

50

326

1P

rim

ary

expe

ndG

DP

326

151

103

666

454

94

327

666

198

55

103

298

Tra

nfe

rsG

DP

134

95

14

316

213

63

115

316

55

36

14

146

Pub

lic

good

sG

DP

181

56

71

299

219

42

1629

914

44

17

123

5G

DP

per

capi

ta84

0055

0012

0018

100

1320

028

0068

0018

100

3500

1800

1200

7200

Pop

uls

hare

over

659

75

22

919

514

42

113

195

52

22

911

8E

ffn

umb

ofpa

rtie

s3

51

71

38

43

71

61

98

43

31

71

38

4E

TH

NIC

(fra

cti

ndex

)20

119

81

7522

122

11

7518

117

51

159

9O

PE

N(o

penn

ess)

305

173

74

961

301

141

87

629

309

204

74

961

Cro

ss-s

ecti

onal

aver

ages

sta

nda

rdde

viat

ions

min

imum

and

max

imu

mva

lues

for

all

the

vari

able

sfo

rth

efu

llsa

mpl

eth

eO

EC

Dsa

mpl

ean

dth

eL

atin

Am

eric

ansa

mpl

eS

eeth

eA

ppen

dix

for

vari

able

srsquom

nem

onic

sE

ach

cros

s-se

ctio

nalo

bser

vati

onis

the

1991

ndash199

4(o

rcl

oses

tav

aila

ble

peri

od)a

vera

geof

the

vari

able

for

the

cou

ntry

Num

ber

ofob

serv

atio

ns

40fo

rth

ew

hol

esa

mpl

e(3

5fo

rR

AE

)20

for

the

OE

CD

sam

ple

20fo

rth

eL

atin

Am

eric

ansa

mpl

e(1

5fo

rR

AE

)


















Panel B OECD












VII3 Basic Results






TA

BL

EV

PR

IMA

RY

SP

EN

DIN

G

TR

AN

SF

ER

SP

UB

LIC

GO

OD

SA

ND

EL

EC

TO

RA

LS

YS

TE

MS

(FU

LL

SA

MP

LE)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

105

170

20

60(L

ogA

M)

(11

0)(3

49)

(1

12)

Log

stan

ddi

stm

agn

074

132

20

48(L

ogS

M)

(08

9)(3

06)

(1

03)

RA

E0

320

382

002

(10

5)(2

21)

(0

13)

Pop

sha

reov

er65

113

114

130

125

127

142

20

122

012

20

10(1

86)

(1

87)

(1

95)

(4

04)

(3

96)

(3

81)

(0

34)

(03

5)(0

27)

Log

GD

Ppe

rca

pita

116

083

054

219

169

145

21

642

146

21

72(0

37)

(02

6)(0

16)

(13

7)(1

04)

(07

8)(0

92)

(08

3)(0

93)

OE

CD

dum

my

134

913

82

115

51

201

752

021

108

310

64

106

5(2

10)

(2

14)

(1

57)

(03

7)(0

52)

(00

5)(2

98)

(2

92)

(2

60)

C

onst

ant

275

574

108

62

217

92

174

62

117

529

32

278

328

87

(01

1)(0

23)

(04

2)(1

73)

(1

36)

(08

1)(2

09)

(2

01)

(20

0)

Ran

ge5

253

848

308

516

869

763

022

490

55O

bser

vati

ons

4040

3540

4035

4040

35R

20

760

750

730

840

830

800

440

430

38

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s












TA

BL

EV

IP

RIM

AR

YS

PE

ND

ING

TR

AN

SF

ER

S

PU

BL

ICG

OO

DS

AN

DE

LE

CT

OR

AL

SY

ST

EM

S(O

EC

DC

OU

NT

RIE

S)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

305

303

20

34(L

ogA

M)

(21

0)

(40

3)

(04

1)L

ogst

and

dist

mag

n2

562

632

028

(Log

SM

)(1

89)

(3

64)

(0

37)

RA

E1

411

622

064

(12

9)(2

56)

(1

14)

Pop

sha

reov

er65

137

115

143

119

095

117

022

025

040

(14

9)(1

17)

(14

0)(2

50)

(1

80)

(1

99)

(0

43)

(04

5)(0

75)

Log

GD

Ppe

rca

pita

171

516

39

157

67

737

056

594

484

574

09(2

36)

(2

23)

(2

03)

(20

5)

(18

0)

(14

8)(1

08)

(11

1)(1

03)

OE

CD

214

210

213

154

212

092

274

42

264

42

254

04

223

17

224

34

224

04

(20

2)

(18

5)

(16

2)(2

04)

(1

70)

(12

6)(0

58)

(06

1)(0

63)

Ran

ge15

28

128

09

1115

18

131

310

45

169

141

413

Obs

erva

tion

s20

2020

2020

2020

2020

R2

034

031

024

061

057

044

20

82

08

20

1

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s




VII4 Robustness




(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD



Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)


Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)





















VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES





































(4)

tM(k1k2) 52 2 ( k1

k1 1 k2k2

)

2

sBM(k1k2) 5

k1(1 2 k1

) 1 k2(1 2 k2

)

2 sA

M(k1k2) 5 0 sCM(k1k2) 5 0

gk1

M(k1k2) 51 2 k1

2 gk2

M(k1k2) 51 2 k2

2 gk3

M(k1k2) 5 0

gM(k1k2) ordm gk1

M(k1k2) 1 gk2

M(k1k2) 52 2 k1 2 k2

2






M (k1 k2) = ( k1(1 2 k1)) 2 and sk2M (k1 k2) =


utility parameters


(5) E(VmBk1

M ) 5 Or= 2

3



M(k1kr)]



(6) M 5m

2 2 m M 5 m



tM 5 1 2m m

2 2 m

(7) sM 5m(1 2 m)

2 2 m

gM 52(1 2 m)

2 2 m








(8) VP( j1 j2) 5 ( j1 j1 1 j2 j2) log (1 2 t) 1 j1(1 2 j1

) log sj1

1 j2(1 2 j2

) log sj21 (2 2 j1

2 j2) log (g3)

where j1and j1



(9) t 5 j1sj1 1 j2s j2 1 g


(10)

tP( j1 j2) 52 2 ( j1 j1 1 j2 j2)

2

s j1P( j1 j2) 5

j1(1 2 j1)

2 s j2

P( j1 j2) 5j2(1 2 j2)

2 s j3

P( j1 j2) 5 0


P 5j1(1 2 j1) 1 j2(1 2 j2)

2


P( j1 j2) 1 g j3P( j1 j2) 5

2 2 j1 2 j22







(11) E(Vj1m

P )= Or= 2

3


P ( j1 jr)

+ (1 2 m) log gP( j1 jr)]



(12) P 5m

2 2 m P 5

m(2 2 m)

1 1 m(1 2 m)


tP 51 1 m(1 2 2 m)

1 1 m(1 2 m)

(13) sP 52 m(1 2 m)

1 1 m(1 2 m)

gP 51 2 m

1 1 m(1 2 m)



IV3 Predictions


sP sM

(14) gP gM

tP tMlaquo m

1(2 2 m)














V1 Electoral Tiers








V2 Voting Method










V3 Legal Thresholds





























SV RV

































VII1 The Data












TA

BL

EII

SU

MM

AR

YS

TA

TIS

TIC

S

All

Mea

nA

llS

DA

llM

inA

llM

axO

EC

DM

ean

OE

CD

SD

OE

CD

Min

OE

CD

Max

LA

Mea

nL

AS

DL

AM

inL

AM

ax

Avg

dis

tric

tm

agn

(A

M)

161

307

115

015

323

115

017

329

91

120

Sta

nd

dist

rict

mag

n(

SM

)23

639

71

180

820

133

114

83

272

461

180

8R

AE

34

40

326

12

41

90

36

83

76

50

326

1P

rim

ary

expe

ndG

DP

326

151

103

666

454

94

327

666

198

55

103

298

Tra

nfe

rsG

DP

134

95

14

316

213

63

115

316

55

36

14

146

Pub

lic

good

sG

DP

181

56

71

299

219

42

1629

914

44

17

123

5G

DP

per

capi

ta84

0055

0012

0018

100

1320

028

0068

0018

100

3500

1800

1200

7200

Pop

uls

hare

over

659

75

22

919

514

42

113

195

52

22

911

8E

ffn

umb

ofpa

rtie

s3

51

71

38

43

71

61

98

43

31

71

38

4E

TH

NIC

(fra

cti

ndex

)20

119

81

7522

122

11

7518

117

51

159

9O

PE

N(o

penn

ess)

305

173

74

961

301

141

87

629

309

204

74

961

Cro

ss-s

ecti

onal

aver

ages

sta

nda

rdde

viat

ions

min

imum

and

max

imu

mva

lues

for

all

the

vari

able

sfo

rth

efu

llsa

mpl

eth

eO

EC

Dsa

mpl

ean

dth

eL

atin

Am

eric

ansa

mpl

eS

eeth

eA

ppen

dix

for

vari

able

srsquom

nem

onic

sE

ach

cros

s-se

ctio

nalo

bser

vati

onis

the

1991

ndash199

4(o

rcl

oses

tav

aila

ble

peri

od)a

vera

geof

the

vari

able

for

the

cou

ntry

Num

ber

ofob

serv

atio

ns

40fo

rth

ew

hol

esa

mpl

e(3

5fo

rR

AE

)20

for

the

OE

CD

sam

ple

20fo

rth

eL

atin

Am

eric

ansa

mpl

e(1

5fo

rR

AE

)


















Panel B OECD












VII3 Basic Results






TA

BL

EV

PR

IMA

RY

SP

EN

DIN

G

TR

AN

SF

ER

SP

UB

LIC

GO

OD

SA

ND

EL

EC

TO

RA

LS

YS

TE

MS

(FU

LL

SA

MP

LE)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

105

170

20

60(L

ogA

M)

(11

0)(3

49)

(1

12)

Log

stan

ddi

stm

agn

074

132

20

48(L

ogS

M)

(08

9)(3

06)

(1

03)

RA

E0

320

382

002

(10

5)(2

21)

(0

13)

Pop

sha

reov

er65

113

114

130

125

127

142

20

122

012

20

10(1

86)

(1

87)

(1

95)

(4

04)

(3

96)

(3

81)

(0

34)

(03

5)(0

27)

Log

GD

Ppe

rca

pita

116

083

054

219

169

145

21

642

146

21

72(0

37)

(02

6)(0

16)

(13

7)(1

04)

(07

8)(0

92)

(08

3)(0

93)

OE

CD

dum

my

134

913

82

115

51

201

752

021

108

310

64

106

5(2

10)

(2

14)

(1

57)

(03

7)(0

52)

(00

5)(2

98)

(2

92)

(2

60)

C

onst

ant

275

574

108

62

217

92

174

62

117

529

32

278

328

87

(01

1)(0

23)

(04

2)(1

73)

(1

36)

(08

1)(2

09)

(2

01)

(20

0)

Ran

ge5

253

848

308

516

869

763

022

490

55O

bser

vati

ons

4040

3540

4035

4040

35R

20

760

750

730

840

830

800

440

430

38

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s












TA

BL

EV

IP

RIM

AR

YS

PE

ND

ING

TR

AN

SF

ER

S

PU

BL

ICG

OO

DS

AN

DE

LE

CT

OR

AL

SY

ST

EM

S(O

EC

DC

OU

NT

RIE

S)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

305

303

20

34(L

ogA

M)

(21

0)

(40

3)

(04

1)L

ogst

and

dist

mag

n2

562

632

028

(Log

SM

)(1

89)

(3

64)

(0

37)

RA

E1

411

622

064

(12

9)(2

56)

(1

14)

Pop

sha

reov

er65

137

115

143

119

095

117

022

025

040

(14

9)(1

17)

(14

0)(2

50)

(1

80)

(1

99)

(0

43)

(04

5)(0

75)

Log

GD

Ppe

rca

pita

171

516

39

157

67

737

056

594

484

574

09(2

36)

(2

23)

(2

03)

(20

5)

(18

0)

(14

8)(1

08)

(11

1)(1

03)

OE

CD

214

210

213

154

212

092

274

42

264

42

254

04

223

17

224

34

224

04

(20

2)

(18

5)

(16

2)(2

04)

(1

70)

(12

6)(0

58)

(06

1)(0

63)

Ran

ge15

28

128

09

1115

18

131

310

45

169

141

413

Obs

erva

tion

s20

2020

2020

2020

2020

R2

034

031

024

061

057

044

20

82

08

20

1

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s




VII4 Robustness




(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD



Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)


Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)





















VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES



































(5) E(VmBk1

M ) 5 Or= 2

3



M(k1kr)]



(6) M 5m

2 2 m M 5 m



tM 5 1 2m m

2 2 m

(7) sM 5m(1 2 m)

2 2 m

gM 52(1 2 m)

2 2 m








(8) VP( j1 j2) 5 ( j1 j1 1 j2 j2) log (1 2 t) 1 j1(1 2 j1

) log sj1

1 j2(1 2 j2

) log sj21 (2 2 j1

2 j2) log (g3)

where j1and j1



(9) t 5 j1sj1 1 j2s j2 1 g


(10)

tP( j1 j2) 52 2 ( j1 j1 1 j2 j2)

2

s j1P( j1 j2) 5

j1(1 2 j1)

2 s j2

P( j1 j2) 5j2(1 2 j2)

2 s j3

P( j1 j2) 5 0


P 5j1(1 2 j1) 1 j2(1 2 j2)

2


P( j1 j2) 1 g j3P( j1 j2) 5

2 2 j1 2 j22







(11) E(Vj1m

P )= Or= 2

3


P ( j1 jr)

+ (1 2 m) log gP( j1 jr)]



(12) P 5m

2 2 m P 5

m(2 2 m)

1 1 m(1 2 m)


tP 51 1 m(1 2 2 m)

1 1 m(1 2 m)

(13) sP 52 m(1 2 m)

1 1 m(1 2 m)

gP 51 2 m

1 1 m(1 2 m)



IV3 Predictions


sP sM

(14) gP gM

tP tMlaquo m

1(2 2 m)














V1 Electoral Tiers








V2 Voting Method










V3 Legal Thresholds





























SV RV

































VII1 The Data












TA

BL

EII

SU

MM

AR

YS

TA

TIS

TIC

S

All

Mea

nA

llS

DA

llM

inA

llM

axO

EC

DM

ean

OE

CD

SD

OE

CD

Min

OE

CD

Max

LA

Mea

nL

AS

DL

AM

inL

AM

ax

Avg

dis

tric

tm

agn

(A

M)

161

307

115

015

323

115

017

329

91

120

Sta

nd

dist

rict

mag

n(

SM

)23

639

71

180

820

133

114

83

272

461

180

8R

AE

34

40

326

12

41

90

36

83

76

50

326

1P

rim

ary

expe

ndG

DP

326

151

103

666

454

94

327

666

198

55

103

298

Tra

nfe

rsG

DP

134

95

14

316

213

63

115

316

55

36

14

146

Pub

lic

good

sG

DP

181

56

71

299

219

42

1629

914

44

17

123

5G

DP

per

capi

ta84

0055

0012

0018

100

1320

028

0068

0018

100

3500

1800

1200

7200

Pop

uls

hare

over

659

75

22

919

514

42

113

195

52

22

911

8E

ffn

umb

ofpa

rtie

s3

51

71

38

43

71

61

98

43

31

71

38

4E

TH

NIC

(fra

cti

ndex

)20

119

81

7522

122

11

7518

117

51

159

9O

PE

N(o

penn

ess)

305

173

74

961

301

141

87

629

309

204

74

961

Cro

ss-s

ecti

onal

aver

ages

sta

nda

rdde

viat

ions

min

imum

and

max

imu

mva

lues

for

all

the

vari

able

sfo

rth

efu

llsa

mpl

eth

eO

EC

Dsa

mpl

ean

dth

eL

atin

Am

eric

ansa

mpl

eS

eeth

eA

ppen

dix

for

vari

able

srsquom

nem

onic

sE

ach

cros

s-se

ctio

nalo

bser

vati

onis

the

1991

ndash199

4(o

rcl

oses

tav

aila

ble

peri

od)a

vera

geof

the

vari

able

for

the

cou

ntry

Num

ber

ofob

serv

atio

ns

40fo

rth

ew

hol

esa

mpl

e(3

5fo

rR

AE

)20

for

the

OE

CD

sam

ple

20fo

rth

eL

atin

Am

eric

ansa

mpl

e(1

5fo

rR

AE

)


















Panel B OECD












VII3 Basic Results






TA

BL

EV

PR

IMA

RY

SP

EN

DIN

G

TR

AN

SF

ER

SP

UB

LIC

GO

OD

SA

ND

EL

EC

TO

RA

LS

YS

TE

MS

(FU

LL

SA

MP

LE)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

105

170

20

60(L

ogA

M)

(11

0)(3

49)

(1

12)

Log

stan

ddi

stm

agn

074

132

20

48(L

ogS

M)

(08

9)(3

06)

(1

03)

RA

E0

320

382

002

(10

5)(2

21)

(0

13)

Pop

sha

reov

er65

113

114

130

125

127

142

20

122

012

20

10(1

86)

(1

87)

(1

95)

(4

04)

(3

96)

(3

81)

(0

34)

(03

5)(0

27)

Log

GD

Ppe

rca

pita

116

083

054

219

169

145

21

642

146

21

72(0

37)

(02

6)(0

16)

(13

7)(1

04)

(07

8)(0

92)

(08

3)(0

93)

OE

CD

dum

my

134

913

82

115

51

201

752

021

108

310

64

106

5(2

10)

(2

14)

(1

57)

(03

7)(0

52)

(00

5)(2

98)

(2

92)

(2

60)

C

onst

ant

275

574

108

62

217

92

174

62

117

529

32

278

328

87

(01

1)(0

23)

(04

2)(1

73)

(1

36)

(08

1)(2

09)

(2

01)

(20

0)

Ran

ge5

253

848

308

516

869

763

022

490

55O

bser

vati

ons

4040

3540

4035

4040

35R

20

760

750

730

840

830

800

440

430

38

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s












TA

BL

EV

IP

RIM

AR

YS

PE

ND

ING

TR

AN

SF

ER

S

PU

BL

ICG

OO

DS

AN

DE

LE

CT

OR

AL

SY

ST

EM

S(O

EC

DC

OU

NT

RIE

S)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

305

303

20

34(L

ogA

M)

(21

0)

(40

3)

(04

1)L

ogst

and

dist

mag

n2

562

632

028

(Log

SM

)(1

89)

(3

64)

(0

37)

RA

E1

411

622

064

(12

9)(2

56)

(1

14)

Pop

sha

reov

er65

137

115

143

119

095

117

022

025

040

(14

9)(1

17)

(14

0)(2

50)

(1

80)

(1

99)

(0

43)

(04

5)(0

75)

Log

GD

Ppe

rca

pita

171

516

39

157

67

737

056

594

484

574

09(2

36)

(2

23)

(2

03)

(20

5)

(18

0)

(14

8)(1

08)

(11

1)(1

03)

OE

CD

214

210

213

154

212

092

274

42

264

42

254

04

223

17

224

34

224

04

(20

2)

(18

5)

(16

2)(2

04)

(1

70)

(12

6)(0

58)

(06

1)(0

63)

Ran

ge15

28

128

09

1115

18

131

310

45

169

141

413

Obs

erva

tion

s20

2020

2020

2020

2020

R2

034

031

024

061

057

044

20

82

08

20

1

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s




VII4 Robustness




(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD



Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)


Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)





















VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES







































(8) VP( j1 j2) 5 ( j1 j1 1 j2 j2) log (1 2 t) 1 j1(1 2 j1

) log sj1

1 j2(1 2 j2

) log sj21 (2 2 j1

2 j2) log (g3)

where j1and j1



(9) t 5 j1sj1 1 j2s j2 1 g


(10)

tP( j1 j2) 52 2 ( j1 j1 1 j2 j2)

2

s j1P( j1 j2) 5

j1(1 2 j1)

2 s j2

P( j1 j2) 5j2(1 2 j2)

2 s j3

P( j1 j2) 5 0


P 5j1(1 2 j1) 1 j2(1 2 j2)

2


P( j1 j2) 1 g j3P( j1 j2) 5

2 2 j1 2 j22







(11) E(Vj1m

P )= Or= 2

3


P ( j1 jr)

+ (1 2 m) log gP( j1 jr)]



(12) P 5m

2 2 m P 5

m(2 2 m)

1 1 m(1 2 m)


tP 51 1 m(1 2 2 m)

1 1 m(1 2 m)

(13) sP 52 m(1 2 m)

1 1 m(1 2 m)

gP 51 2 m

1 1 m(1 2 m)



IV3 Predictions


sP sM

(14) gP gM

tP tMlaquo m

1(2 2 m)














V1 Electoral Tiers








V2 Voting Method










V3 Legal Thresholds





























SV RV

































VII1 The Data












TA

BL

EII

SU

MM

AR

YS

TA

TIS

TIC

S

All

Mea

nA

llS

DA

llM

inA

llM

axO

EC

DM

ean

OE

CD

SD

OE

CD

Min

OE

CD

Max

LA

Mea

nL

AS

DL

AM

inL

AM

ax

Avg

dis

tric

tm

agn

(A

M)

161

307

115

015

323

115

017

329

91

120

Sta

nd

dist

rict

mag

n(

SM

)23

639

71

180

820

133

114

83

272

461

180

8R

AE

34

40

326

12

41

90

36

83

76

50

326

1P

rim

ary

expe

ndG

DP

326

151

103

666

454

94

327

666

198

55

103

298

Tra

nfe

rsG

DP

134

95

14

316

213

63

115

316

55

36

14

146

Pub

lic

good

sG

DP

181

56

71

299

219

42

1629

914

44

17

123

5G

DP

per

capi

ta84

0055

0012

0018

100

1320

028

0068

0018

100

3500

1800

1200

7200

Pop

uls

hare

over

659

75

22

919

514

42

113

195

52

22

911

8E

ffn

umb

ofpa

rtie

s3

51

71

38

43

71

61

98

43

31

71

38

4E

TH

NIC

(fra

cti

ndex

)20

119

81

7522

122

11

7518

117

51

159

9O

PE

N(o

penn

ess)

305

173

74

961

301

141

87

629

309

204

74

961

Cro

ss-s

ecti

onal

aver

ages

sta

nda

rdde

viat

ions

min

imum

and

max

imu

mva

lues

for

all

the

vari

able

sfo

rth

efu

llsa

mpl

eth

eO

EC

Dsa

mpl

ean

dth

eL

atin

Am

eric

ansa

mpl

eS

eeth

eA

ppen

dix

for

vari

able

srsquom

nem

onic

sE

ach

cros

s-se

ctio

nalo

bser

vati

onis

the

1991

ndash199

4(o

rcl

oses

tav

aila

ble

peri

od)a

vera

geof

the

vari

able

for

the

cou

ntry

Num

ber

ofob

serv

atio

ns

40fo

rth

ew

hol

esa

mpl

e(3

5fo

rR

AE

)20

for

the

OE

CD

sam

ple

20fo

rth

eL

atin

Am

eric

ansa

mpl

e(1

5fo

rR

AE

)


















Panel B OECD












VII3 Basic Results






TA

BL

EV

PR

IMA

RY

SP

EN

DIN

G

TR

AN

SF

ER

SP

UB

LIC

GO

OD

SA

ND

EL

EC

TO

RA

LS

YS

TE

MS

(FU

LL

SA

MP

LE)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

105

170

20

60(L

ogA

M)

(11

0)(3

49)

(1

12)

Log

stan

ddi

stm

agn

074

132

20

48(L

ogS

M)

(08

9)(3

06)

(1

03)

RA

E0

320

382

002

(10

5)(2

21)

(0

13)

Pop

sha

reov

er65

113

114

130

125

127

142

20

122

012

20

10(1

86)

(1

87)

(1

95)

(4

04)

(3

96)

(3

81)

(0

34)

(03

5)(0

27)

Log

GD

Ppe

rca

pita

116

083

054

219

169

145

21

642

146

21

72(0

37)

(02

6)(0

16)

(13

7)(1

04)

(07

8)(0

92)

(08

3)(0

93)

OE

CD

dum

my

134

913

82

115

51

201

752

021

108

310

64

106

5(2

10)

(2

14)

(1

57)

(03

7)(0

52)

(00

5)(2

98)

(2

92)

(2

60)

C

onst

ant

275

574

108

62

217

92

174

62

117

529

32

278

328

87

(01

1)(0

23)

(04

2)(1

73)

(1

36)

(08

1)(2

09)

(2

01)

(20

0)

Ran

ge5

253

848

308

516

869

763

022

490

55O

bser

vati

ons

4040

3540

4035

4040

35R

20

760

750

730

840

830

800

440

430

38

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s












TA

BL

EV

IP

RIM

AR

YS

PE

ND

ING

TR

AN

SF

ER

S

PU

BL

ICG

OO

DS

AN

DE

LE

CT

OR

AL

SY

ST

EM

S(O

EC

DC

OU

NT

RIE

S)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

305

303

20

34(L

ogA

M)

(21

0)

(40

3)

(04

1)L

ogst

and

dist

mag

n2

562

632

028

(Log

SM

)(1

89)

(3

64)

(0

37)

RA

E1

411

622

064

(12

9)(2

56)

(1

14)

Pop

sha

reov

er65

137

115

143

119

095

117

022

025

040

(14

9)(1

17)

(14

0)(2

50)

(1

80)

(1

99)

(0

43)

(04

5)(0

75)

Log

GD

Ppe

rca

pita

171

516

39

157

67

737

056

594

484

574

09(2

36)

(2

23)

(2

03)

(20

5)

(18

0)

(14

8)(1

08)

(11

1)(1

03)

OE

CD

214

210

213

154

212

092

274

42

264

42

254

04

223

17

224

34

224

04

(20

2)

(18

5)

(16

2)(2

04)

(1

70)

(12

6)(0

58)

(06

1)(0

63)

Ran

ge15

28

128

09

1115

18

131

310

45

169

141

413

Obs

erva

tion

s20

2020

2020

2020

2020

R2

034

031

024

061

057

044

20

82

08

20

1

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s




VII4 Robustness




(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD



Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)


Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)





















VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES






































(11) E(Vj1m

P )= Or= 2

3


P ( j1 jr)

+ (1 2 m) log gP( j1 jr)]



(12) P 5m

2 2 m P 5

m(2 2 m)

1 1 m(1 2 m)


tP 51 1 m(1 2 2 m)

1 1 m(1 2 m)

(13) sP 52 m(1 2 m)

1 1 m(1 2 m)

gP 51 2 m

1 1 m(1 2 m)



IV3 Predictions


sP sM

(14) gP gM

tP tMlaquo m

1(2 2 m)














V1 Electoral Tiers








V2 Voting Method










V3 Legal Thresholds





























SV RV

































VII1 The Data












TA

BL

EII

SU

MM

AR

YS

TA

TIS

TIC

S

All

Mea

nA

llS

DA

llM

inA

llM

axO

EC

DM

ean

OE

CD

SD

OE

CD

Min

OE

CD

Max

LA

Mea

nL

AS

DL

AM

inL

AM

ax

Avg

dis

tric

tm

agn

(A

M)

161

307

115

015

323

115

017

329

91

120

Sta

nd

dist

rict

mag

n(

SM

)23

639

71

180

820

133

114

83

272

461

180

8R

AE

34

40

326

12

41

90

36

83

76

50

326

1P

rim

ary

expe

ndG

DP

326

151

103

666

454

94

327

666

198

55

103

298

Tra

nfe

rsG

DP

134

95

14

316

213

63

115

316

55

36

14

146

Pub

lic

good

sG

DP

181

56

71

299

219

42

1629

914

44

17

123

5G

DP

per

capi

ta84

0055

0012

0018

100

1320

028

0068

0018

100

3500

1800

1200

7200

Pop

uls

hare

over

659

75

22

919

514

42

113

195

52

22

911

8E

ffn

umb

ofpa

rtie

s3

51

71

38

43

71

61

98

43

31

71

38

4E

TH

NIC

(fra

cti

ndex

)20

119

81

7522

122

11

7518

117

51

159

9O

PE

N(o

penn

ess)

305

173

74

961

301

141

87

629

309

204

74

961

Cro

ss-s

ecti

onal

aver

ages

sta

nda

rdde

viat

ions

min

imum

and

max

imu

mva

lues

for

all

the

vari

able

sfo

rth

efu

llsa

mpl

eth

eO

EC

Dsa

mpl

ean

dth

eL

atin

Am

eric

ansa

mpl

eS

eeth

eA

ppen

dix

for

vari

able

srsquom

nem

onic

sE

ach

cros

s-se

ctio

nalo

bser

vati

onis

the

1991

ndash199

4(o

rcl

oses

tav

aila

ble

peri

od)a

vera

geof

the

vari

able

for

the

cou

ntry

Num

ber

ofob

serv

atio

ns

40fo

rth

ew

hol

esa

mpl

e(3

5fo

rR

AE

)20

for

the

OE

CD

sam

ple

20fo

rth

eL

atin

Am

eric

ansa

mpl

e(1

5fo

rR

AE

)


















Panel B OECD












VII3 Basic Results






TA

BL

EV

PR

IMA

RY

SP

EN

DIN

G

TR

AN

SF

ER

SP

UB

LIC

GO

OD

SA

ND

EL

EC

TO

RA

LS

YS

TE

MS

(FU

LL

SA

MP

LE)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

105

170

20

60(L

ogA

M)

(11

0)(3

49)

(1

12)

Log

stan

ddi

stm

agn

074

132

20

48(L

ogS

M)

(08

9)(3

06)

(1

03)

RA

E0

320

382

002

(10

5)(2

21)

(0

13)

Pop

sha

reov

er65

113

114

130

125

127

142

20

122

012

20

10(1

86)

(1

87)

(1

95)

(4

04)

(3

96)

(3

81)

(0

34)

(03

5)(0

27)

Log

GD

Ppe

rca

pita

116

083

054

219

169

145

21

642

146

21

72(0

37)

(02

6)(0

16)

(13

7)(1

04)

(07

8)(0

92)

(08

3)(0

93)

OE

CD

dum

my

134

913

82

115

51

201

752

021

108

310

64

106

5(2

10)

(2

14)

(1

57)

(03

7)(0

52)

(00

5)(2

98)

(2

92)

(2

60)

C

onst

ant

275

574

108

62

217

92

174

62

117

529

32

278

328

87

(01

1)(0

23)

(04

2)(1

73)

(1

36)

(08

1)(2

09)

(2

01)

(20

0)

Ran

ge5

253

848

308

516

869

763

022

490

55O

bser

vati

ons

4040

3540

4035

4040

35R

20

760

750

730

840

830

800

440

430

38

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s












TA

BL

EV

IP

RIM

AR

YS

PE

ND

ING

TR

AN

SF

ER

S

PU

BL

ICG

OO

DS

AN

DE

LE

CT

OR

AL

SY

ST

EM

S(O

EC

DC

OU

NT

RIE

S)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

305

303

20

34(L

ogA

M)

(21

0)

(40

3)

(04

1)L

ogst

and

dist

mag

n2

562

632

028

(Log

SM

)(1

89)

(3

64)

(0

37)

RA

E1

411

622

064

(12

9)(2

56)

(1

14)

Pop

sha

reov

er65

137

115

143

119

095

117

022

025

040

(14

9)(1

17)

(14

0)(2

50)

(1

80)

(1

99)

(0

43)

(04

5)(0

75)

Log

GD

Ppe

rca

pita

171

516

39

157

67

737

056

594

484

574

09(2

36)

(2

23)

(2

03)

(20

5)

(18

0)

(14

8)(1

08)

(11

1)(1

03)

OE

CD

214

210

213

154

212

092

274

42

264

42

254

04

223

17

224

34

224

04

(20

2)

(18

5)

(16

2)(2

04)

(1

70)

(12

6)(0

58)

(06

1)(0

63)

Ran

ge15

28

128

09

1115

18

131

310

45

169

141

413

Obs

erva

tion

s20

2020

2020

2020

2020

R2

034

031

024

061

057

044

20

82

08

20

1

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s




VII4 Robustness




(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD



Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)


Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)





















VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES



































IV3 Predictions


sP sM

(14) gP gM

tP tMlaquo m

1(2 2 m)














V1 Electoral Tiers








V2 Voting Method










V3 Legal Thresholds





























SV RV

































VII1 The Data












TA

BL

EII

SU

MM

AR

YS

TA

TIS

TIC

S

All

Mea

nA

llS

DA

llM

inA

llM

axO

EC

DM

ean

OE

CD

SD

OE

CD

Min

OE

CD

Max

LA

Mea

nL

AS

DL

AM

inL

AM

ax

Avg

dis

tric

tm

agn

(A

M)

161

307

115

015

323

115

017

329

91

120

Sta

nd

dist

rict

mag

n(

SM

)23

639

71

180

820

133

114

83

272

461

180

8R

AE

34

40

326

12

41

90

36

83

76

50

326

1P

rim

ary

expe

ndG

DP

326

151

103

666

454

94

327

666

198

55

103

298

Tra

nfe

rsG

DP

134

95

14

316

213

63

115

316

55

36

14

146

Pub

lic

good

sG

DP

181

56

71

299

219

42

1629

914

44

17

123

5G

DP

per

capi

ta84

0055

0012

0018

100

1320

028

0068

0018

100

3500

1800

1200

7200

Pop

uls

hare

over

659

75

22

919

514

42

113

195

52

22

911

8E

ffn

umb

ofpa

rtie

s3

51

71

38

43

71

61

98

43

31

71

38

4E

TH

NIC

(fra

cti

ndex

)20

119

81

7522

122

11

7518

117

51

159

9O

PE

N(o

penn

ess)

305

173

74

961

301

141

87

629

309

204

74

961

Cro

ss-s

ecti

onal

aver

ages

sta

nda

rdde

viat

ions

min

imum

and

max

imu

mva

lues

for

all

the

vari

able

sfo

rth

efu

llsa

mpl

eth

eO

EC

Dsa

mpl

ean

dth

eL

atin

Am

eric

ansa

mpl

eS

eeth

eA

ppen

dix

for

vari

able

srsquom

nem

onic

sE

ach

cros

s-se

ctio

nalo

bser

vati

onis

the

1991

ndash199

4(o

rcl

oses

tav

aila

ble

peri

od)a

vera

geof

the

vari

able

for

the

cou

ntry

Num

ber

ofob

serv

atio

ns

40fo

rth

ew

hol

esa

mpl

e(3

5fo

rR

AE

)20

for

the

OE

CD

sam

ple

20fo

rth

eL

atin

Am

eric

ansa

mpl

e(1

5fo

rR

AE

)


















Panel B OECD












VII3 Basic Results






TA

BL

EV

PR

IMA

RY

SP

EN

DIN

G

TR

AN

SF

ER

SP

UB

LIC

GO

OD

SA

ND

EL

EC

TO

RA

LS

YS

TE

MS

(FU

LL

SA

MP

LE)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

105

170

20

60(L

ogA

M)

(11

0)(3

49)

(1

12)

Log

stan

ddi

stm

agn

074

132

20

48(L

ogS

M)

(08

9)(3

06)

(1

03)

RA

E0

320

382

002

(10

5)(2

21)

(0

13)

Pop

sha

reov

er65

113

114

130

125

127

142

20

122

012

20

10(1

86)

(1

87)

(1

95)

(4

04)

(3

96)

(3

81)

(0

34)

(03

5)(0

27)

Log

GD

Ppe

rca

pita

116

083

054

219

169

145

21

642

146

21

72(0

37)

(02

6)(0

16)

(13

7)(1

04)

(07

8)(0

92)

(08

3)(0

93)

OE

CD

dum

my

134

913

82

115

51

201

752

021

108

310

64

106

5(2

10)

(2

14)

(1

57)

(03

7)(0

52)

(00

5)(2

98)

(2

92)

(2

60)

C

onst

ant

275

574

108

62

217

92

174

62

117

529

32

278

328

87

(01

1)(0

23)

(04

2)(1

73)

(1

36)

(08

1)(2

09)

(2

01)

(20

0)

Ran

ge5

253

848

308

516

869

763

022

490

55O

bser

vati

ons

4040

3540

4035

4040

35R

20

760

750

730

840

830

800

440

430

38

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s












TA

BL

EV

IP

RIM

AR

YS

PE

ND

ING

TR

AN

SF

ER

S

PU

BL

ICG

OO

DS

AN

DE

LE

CT

OR

AL

SY

ST

EM

S(O

EC

DC

OU

NT

RIE

S)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

305

303

20

34(L

ogA

M)

(21

0)

(40

3)

(04

1)L

ogst

and

dist

mag

n2

562

632

028

(Log

SM

)(1

89)

(3

64)

(0

37)

RA

E1

411

622

064

(12

9)(2

56)

(1

14)

Pop

sha

reov

er65

137

115

143

119

095

117

022

025

040

(14

9)(1

17)

(14

0)(2

50)

(1

80)

(1

99)

(0

43)

(04

5)(0

75)

Log

GD

Ppe

rca

pita

171

516

39

157

67

737

056

594

484

574

09(2

36)

(2

23)

(2

03)

(20

5)

(18

0)

(14

8)(1

08)

(11

1)(1

03)

OE

CD

214

210

213

154

212

092

274

42

264

42

254

04

223

17

224

34

224

04

(20

2)

(18

5)

(16

2)(2

04)

(1

70)

(12

6)(0

58)

(06

1)(0

63)

Ran

ge15

28

128

09

1115

18

131

310

45

169

141

413

Obs

erva

tion

s20

2020

2020

2020

2020

R2

034

031

024

061

057

044

20

82

08

20

1

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s




VII4 Robustness




(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD



Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)


Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)





















VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES







































V1 Electoral Tiers








V2 Voting Method










V3 Legal Thresholds





























SV RV

































VII1 The Data












TA

BL

EII

SU

MM

AR

YS

TA

TIS

TIC

S

All

Mea

nA

llS

DA

llM

inA

llM

axO

EC

DM

ean

OE

CD

SD

OE

CD

Min

OE

CD

Max

LA

Mea

nL

AS

DL

AM

inL

AM

ax

Avg

dis

tric

tm

agn

(A

M)

161

307

115

015

323

115

017

329

91

120

Sta

nd

dist

rict

mag

n(

SM

)23

639

71

180

820

133

114

83

272

461

180

8R

AE

34

40

326

12

41

90

36

83

76

50

326

1P

rim

ary

expe

ndG

DP

326

151

103

666

454

94

327

666

198

55

103

298

Tra

nfe

rsG

DP

134

95

14

316

213

63

115

316

55

36

14

146

Pub

lic

good

sG

DP

181

56

71

299

219

42

1629

914

44

17

123

5G

DP

per

capi

ta84

0055

0012

0018

100

1320

028

0068

0018

100

3500

1800

1200

7200

Pop

uls

hare

over

659

75

22

919

514

42

113

195

52

22

911

8E

ffn

umb

ofpa

rtie

s3

51

71

38

43

71

61

98

43

31

71

38

4E

TH

NIC

(fra

cti

ndex

)20

119

81

7522

122

11

7518

117

51

159

9O

PE

N(o

penn

ess)

305

173

74

961

301

141

87

629

309

204

74

961

Cro

ss-s

ecti

onal

aver

ages

sta

nda

rdde

viat

ions

min

imum

and

max

imu

mva

lues

for

all

the

vari

able

sfo

rth

efu

llsa

mpl

eth

eO

EC

Dsa

mpl

ean

dth

eL

atin

Am

eric

ansa

mpl

eS

eeth

eA

ppen

dix

for

vari

able

srsquom

nem

onic

sE

ach

cros

s-se

ctio

nalo

bser

vati

onis

the

1991

ndash199

4(o

rcl

oses

tav

aila

ble

peri

od)a

vera

geof

the

vari

able

for

the

cou

ntry

Num

ber

ofob

serv

atio

ns

40fo

rth

ew

hol

esa

mpl

e(3

5fo

rR

AE

)20

for

the

OE

CD

sam

ple

20fo

rth

eL

atin

Am

eric

ansa

mpl

e(1

5fo

rR

AE

)


















Panel B OECD












VII3 Basic Results






TA

BL

EV

PR

IMA

RY

SP

EN

DIN

G

TR

AN

SF

ER

SP

UB

LIC

GO

OD

SA

ND

EL

EC

TO

RA

LS

YS

TE

MS

(FU

LL

SA

MP

LE)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

105

170

20

60(L

ogA

M)

(11

0)(3

49)

(1

12)

Log

stan

ddi

stm

agn

074

132

20

48(L

ogS

M)

(08

9)(3

06)

(1

03)

RA

E0

320

382

002

(10

5)(2

21)

(0

13)

Pop

sha

reov

er65

113

114

130

125

127

142

20

122

012

20

10(1

86)

(1

87)

(1

95)

(4

04)

(3

96)

(3

81)

(0

34)

(03

5)(0

27)

Log

GD

Ppe

rca

pita

116

083

054

219

169

145

21

642

146

21

72(0

37)

(02

6)(0

16)

(13

7)(1

04)

(07

8)(0

92)

(08

3)(0

93)

OE

CD

dum

my

134

913

82

115

51

201

752

021

108

310

64

106

5(2

10)

(2

14)

(1

57)

(03

7)(0

52)

(00

5)(2

98)

(2

92)

(2

60)

C

onst

ant

275

574

108

62

217

92

174

62

117

529

32

278

328

87

(01

1)(0

23)

(04

2)(1

73)

(1

36)

(08

1)(2

09)

(2

01)

(20

0)

Ran

ge5

253

848

308

516

869

763

022

490

55O

bser

vati

ons

4040

3540

4035

4040

35R

20

760

750

730

840

830

800

440

430

38

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s












TA

BL

EV

IP

RIM

AR

YS

PE

ND

ING

TR

AN

SF

ER

S

PU

BL

ICG

OO

DS

AN

DE

LE

CT

OR

AL

SY

ST

EM

S(O

EC

DC

OU

NT

RIE

S)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

305

303

20

34(L

ogA

M)

(21

0)

(40

3)

(04

1)L

ogst

and

dist

mag

n2

562

632

028

(Log

SM

)(1

89)

(3

64)

(0

37)

RA

E1

411

622

064

(12

9)(2

56)

(1

14)

Pop

sha

reov

er65

137

115

143

119

095

117

022

025

040

(14

9)(1

17)

(14

0)(2

50)

(1

80)

(1

99)

(0

43)

(04

5)(0

75)

Log

GD

Ppe

rca

pita

171

516

39

157

67

737

056

594

484

574

09(2

36)

(2

23)

(2

03)

(20

5)

(18

0)

(14

8)(1

08)

(11

1)(1

03)

OE

CD

214

210

213

154

212

092

274

42

264

42

254

04

223

17

224

34

224

04

(20

2)

(18

5)

(16

2)(2

04)

(1

70)

(12

6)(0

58)

(06

1)(0

63)

Ran

ge15

28

128

09

1115

18

131

310

45

169

141

413

Obs

erva

tion

s20

2020

2020

2020

2020

R2

034

031

024

061

057

044

20

82

08

20

1

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s




VII4 Robustness




(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD



Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)


Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)





















VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES




































V2 Voting Method










V3 Legal Thresholds





























SV RV

































VII1 The Data












TA

BL

EII

SU

MM

AR

YS

TA

TIS

TIC

S

All

Mea

nA

llS

DA

llM

inA

llM

axO

EC

DM

ean

OE

CD

SD

OE

CD

Min

OE

CD

Max

LA

Mea

nL

AS

DL

AM

inL

AM

ax

Avg

dis

tric

tm

agn

(A

M)

161

307

115

015

323

115

017

329

91

120

Sta

nd

dist

rict

mag

n(

SM

)23

639

71

180

820

133

114

83

272

461

180

8R

AE

34

40

326

12

41

90

36

83

76

50

326

1P

rim

ary

expe

ndG

DP

326

151

103

666

454

94

327

666

198

55

103

298

Tra

nfe

rsG

DP

134

95

14

316

213

63

115

316

55

36

14

146

Pub

lic

good

sG

DP

181

56

71

299

219

42

1629

914

44

17

123

5G

DP

per

capi

ta84

0055

0012

0018

100

1320

028

0068

0018

100

3500

1800

1200

7200

Pop

uls

hare

over

659

75

22

919

514

42

113

195

52

22

911

8E

ffn

umb

ofpa

rtie

s3

51

71

38

43

71

61

98

43

31

71

38

4E

TH

NIC

(fra

cti

ndex

)20

119

81

7522

122

11

7518

117

51

159

9O

PE

N(o

penn

ess)

305

173

74

961

301

141

87

629

309

204

74

961

Cro

ss-s

ecti

onal

aver

ages

sta

nda

rdde

viat

ions

min

imum

and

max

imu

mva

lues

for

all

the

vari

able

sfo

rth

efu

llsa

mpl

eth

eO

EC

Dsa

mpl

ean

dth

eL

atin

Am

eric

ansa

mpl

eS

eeth

eA

ppen

dix

for

vari

able

srsquom

nem

onic

sE

ach

cros

s-se

ctio

nalo

bser

vati

onis

the

1991

ndash199

4(o

rcl

oses

tav

aila

ble

peri

od)a

vera

geof

the

vari

able

for

the

cou

ntry

Num

ber

ofob

serv

atio

ns

40fo

rth

ew

hol

esa

mpl

e(3

5fo

rR

AE

)20

for

the

OE

CD

sam

ple

20fo

rth

eL

atin

Am

eric

ansa

mpl

e(1

5fo

rR

AE

)


















Panel B OECD












VII3 Basic Results






TA

BL

EV

PR

IMA

RY

SP

EN

DIN

G

TR

AN

SF

ER

SP

UB

LIC

GO

OD

SA

ND

EL

EC

TO

RA

LS

YS

TE

MS

(FU

LL

SA

MP

LE)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

105

170

20

60(L

ogA

M)

(11

0)(3

49)

(1

12)

Log

stan

ddi

stm

agn

074

132

20

48(L

ogS

M)

(08

9)(3

06)

(1

03)

RA

E0

320

382

002

(10

5)(2

21)

(0

13)

Pop

sha

reov

er65

113

114

130

125

127

142

20

122

012

20

10(1

86)

(1

87)

(1

95)

(4

04)

(3

96)

(3

81)

(0

34)

(03

5)(0

27)

Log

GD

Ppe

rca

pita

116

083

054

219

169

145

21

642

146

21

72(0

37)

(02

6)(0

16)

(13

7)(1

04)

(07

8)(0

92)

(08

3)(0

93)

OE

CD

dum

my

134

913

82

115

51

201

752

021

108

310

64

106

5(2

10)

(2

14)

(1

57)

(03

7)(0

52)

(00

5)(2

98)

(2

92)

(2

60)

C

onst

ant

275

574

108

62

217

92

174

62

117

529

32

278

328

87

(01

1)(0

23)

(04

2)(1

73)

(1

36)

(08

1)(2

09)

(2

01)

(20

0)

Ran

ge5

253

848

308

516

869

763

022

490

55O

bser

vati

ons

4040

3540

4035

4040

35R

20

760

750

730

840

830

800

440

430

38

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s












TA

BL

EV

IP

RIM

AR

YS

PE

ND

ING

TR

AN

SF

ER

S

PU

BL

ICG

OO

DS

AN

DE

LE

CT

OR

AL

SY

ST

EM

S(O

EC

DC

OU

NT

RIE

S)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

305

303

20

34(L

ogA

M)

(21

0)

(40

3)

(04

1)L

ogst

and

dist

mag

n2

562

632

028

(Log

SM

)(1

89)

(3

64)

(0

37)

RA

E1

411

622

064

(12

9)(2

56)

(1

14)

Pop

sha

reov

er65

137

115

143

119

095

117

022

025

040

(14

9)(1

17)

(14

0)(2

50)

(1

80)

(1

99)

(0

43)

(04

5)(0

75)

Log

GD

Ppe

rca

pita

171

516

39

157

67

737

056

594

484

574

09(2

36)

(2

23)

(2

03)

(20

5)

(18

0)

(14

8)(1

08)

(11

1)(1

03)

OE

CD

214

210

213

154

212

092

274

42

264

42

254

04

223

17

224

34

224

04

(20

2)

(18

5)

(16

2)(2

04)

(1

70)

(12

6)(0

58)

(06

1)(0

63)

Ran

ge15

28

128

09

1115

18

131

310

45

169

141

413

Obs

erva

tion

s20

2020

2020

2020

2020

R2

034

031

024

061

057

044

20

82

08

20

1

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s




VII4 Robustness




(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD



Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)


Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)





















VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES



































V3 Legal Thresholds





























SV RV

































VII1 The Data












TA

BL

EII

SU

MM

AR

YS

TA

TIS

TIC

S

All

Mea

nA

llS

DA

llM

inA

llM

axO

EC

DM

ean

OE

CD

SD

OE

CD

Min

OE

CD

Max

LA

Mea

nL

AS

DL

AM

inL

AM

ax

Avg

dis

tric

tm

agn

(A

M)

161

307

115

015

323

115

017

329

91

120

Sta

nd

dist

rict

mag

n(

SM

)23

639

71

180

820

133

114

83

272

461

180

8R

AE

34

40

326

12

41

90

36

83

76

50

326

1P

rim

ary

expe

ndG

DP

326

151

103

666

454

94

327

666

198

55

103

298

Tra

nfe

rsG

DP

134

95

14

316

213

63

115

316

55

36

14

146

Pub

lic

good

sG

DP

181

56

71

299

219

42

1629

914

44

17

123

5G

DP

per

capi

ta84

0055

0012

0018

100

1320

028

0068

0018

100

3500

1800

1200

7200

Pop

uls

hare

over

659

75

22

919

514

42

113

195

52

22

911

8E

ffn

umb

ofpa

rtie

s3

51

71

38

43

71

61

98

43

31

71

38

4E

TH

NIC

(fra

cti

ndex

)20

119

81

7522

122

11

7518

117

51

159

9O

PE

N(o

penn

ess)

305

173

74

961

301

141

87

629

309

204

74

961

Cro

ss-s

ecti

onal

aver

ages

sta

nda

rdde

viat

ions

min

imum

and

max

imu

mva

lues

for

all

the

vari

able

sfo

rth

efu

llsa

mpl

eth

eO

EC

Dsa

mpl

ean

dth

eL

atin

Am

eric

ansa

mpl

eS

eeth

eA

ppen

dix

for

vari

able

srsquom

nem

onic

sE

ach

cros

s-se

ctio

nalo

bser

vati

onis

the

1991

ndash199

4(o

rcl

oses

tav

aila

ble

peri

od)a

vera

geof

the

vari

able

for

the

cou

ntry

Num

ber

ofob

serv

atio

ns

40fo

rth

ew

hol

esa

mpl

e(3

5fo

rR

AE

)20

for

the

OE

CD

sam

ple

20fo

rth

eL

atin

Am

eric

ansa

mpl

e(1

5fo

rR

AE

)


















Panel B OECD












VII3 Basic Results






TA

BL

EV

PR

IMA

RY

SP

EN

DIN

G

TR

AN

SF

ER

SP

UB

LIC

GO

OD

SA

ND

EL

EC

TO

RA

LS

YS

TE

MS

(FU

LL

SA

MP

LE)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

105

170

20

60(L

ogA

M)

(11

0)(3

49)

(1

12)

Log

stan

ddi

stm

agn

074

132

20

48(L

ogS

M)

(08

9)(3

06)

(1

03)

RA

E0

320

382

002

(10

5)(2

21)

(0

13)

Pop

sha

reov

er65

113

114

130

125

127

142

20

122

012

20

10(1

86)

(1

87)

(1

95)

(4

04)

(3

96)

(3

81)

(0

34)

(03

5)(0

27)

Log

GD

Ppe

rca

pita

116

083

054

219

169

145

21

642

146

21

72(0

37)

(02

6)(0

16)

(13

7)(1

04)

(07

8)(0

92)

(08

3)(0

93)

OE

CD

dum

my

134

913

82

115

51

201

752

021

108

310

64

106

5(2

10)

(2

14)

(1

57)

(03

7)(0

52)

(00

5)(2

98)

(2

92)

(2

60)

C

onst

ant

275

574

108

62

217

92

174

62

117

529

32

278

328

87

(01

1)(0

23)

(04

2)(1

73)

(1

36)

(08

1)(2

09)

(2

01)

(20

0)

Ran

ge5

253

848

308

516

869

763

022

490

55O

bser

vati

ons

4040

3540

4035

4040

35R

20

760

750

730

840

830

800

440

430

38

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s












TA

BL

EV

IP

RIM

AR

YS

PE

ND

ING

TR

AN

SF

ER

S

PU

BL

ICG

OO

DS

AN

DE

LE

CT

OR

AL

SY

ST

EM

S(O

EC

DC

OU

NT

RIE

S)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

305

303

20

34(L

ogA

M)

(21

0)

(40

3)

(04

1)L

ogst

and

dist

mag

n2

562

632

028

(Log

SM

)(1

89)

(3

64)

(0

37)

RA

E1

411

622

064

(12

9)(2

56)

(1

14)

Pop

sha

reov

er65

137

115

143

119

095

117

022

025

040

(14

9)(1

17)

(14

0)(2

50)

(1

80)

(1

99)

(0

43)

(04

5)(0

75)

Log

GD

Ppe

rca

pita

171

516

39

157

67

737

056

594

484

574

09(2

36)

(2

23)

(2

03)

(20

5)

(18

0)

(14

8)(1

08)

(11

1)(1

03)

OE

CD

214

210

213

154

212

092

274

42

264

42

254

04

223

17

224

34

224

04

(20

2)

(18

5)

(16

2)(2

04)

(1

70)

(12

6)(0

58)

(06

1)(0

63)

Ran

ge15

28

128

09

1115

18

131

310

45

169

141

413

Obs

erva

tion

s20

2020

2020

2020

2020

R2

034

031

024

061

057

044

20

82

08

20

1

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s




VII4 Robustness




(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD



Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)


Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)





















VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES




















































SV RV

































VII1 The Data












TA

BL

EII

SU

MM

AR

YS

TA

TIS

TIC

S

All

Mea

nA

llS

DA

llM

inA

llM

axO

EC

DM

ean

OE

CD

SD

OE

CD

Min

OE

CD

Max

LA

Mea

nL

AS

DL

AM

inL

AM

ax

Avg

dis

tric

tm

agn

(A

M)

161

307

115

015

323

115

017

329

91

120

Sta

nd

dist

rict

mag

n(

SM

)23

639

71

180

820

133

114

83

272

461

180

8R

AE

34

40

326

12

41

90

36

83

76

50

326

1P

rim

ary

expe

ndG

DP

326

151

103

666

454

94

327

666

198

55

103

298

Tra

nfe

rsG

DP

134

95

14

316

213

63

115

316

55

36

14

146

Pub

lic

good

sG

DP

181

56

71

299

219

42

1629

914

44

17

123

5G

DP

per

capi

ta84

0055

0012

0018

100

1320

028

0068

0018

100

3500

1800

1200

7200

Pop

uls

hare

over

659

75

22

919

514

42

113

195

52

22

911

8E

ffn

umb

ofpa

rtie

s3

51

71

38

43

71

61

98

43

31

71

38

4E

TH

NIC

(fra

cti

ndex

)20

119

81

7522

122

11

7518

117

51

159

9O

PE

N(o

penn

ess)

305

173

74

961

301

141

87

629

309

204

74

961

Cro

ss-s

ecti

onal

aver

ages

sta

nda

rdde

viat

ions

min

imum

and

max

imu

mva

lues

for

all

the

vari

able

sfo

rth

efu

llsa

mpl

eth

eO

EC

Dsa

mpl

ean

dth

eL

atin

Am

eric

ansa

mpl

eS

eeth

eA

ppen

dix

for

vari

able

srsquom

nem

onic

sE

ach

cros

s-se

ctio

nalo

bser

vati

onis

the

1991

ndash199

4(o

rcl

oses

tav

aila

ble

peri

od)a

vera

geof

the

vari

able

for

the

cou

ntry

Num

ber

ofob

serv

atio

ns

40fo

rth

ew

hol

esa

mpl

e(3

5fo

rR

AE

)20

for

the

OE

CD

sam

ple

20fo

rth

eL

atin

Am

eric

ansa

mpl

e(1

5fo

rR

AE

)


















Panel B OECD












VII3 Basic Results






TA

BL

EV

PR

IMA

RY

SP

EN

DIN

G

TR

AN

SF

ER

SP

UB

LIC

GO

OD

SA

ND

EL

EC

TO

RA

LS

YS

TE

MS

(FU

LL

SA

MP

LE)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

105

170

20

60(L

ogA

M)

(11

0)(3

49)

(1

12)

Log

stan

ddi

stm

agn

074

132

20

48(L

ogS

M)

(08

9)(3

06)

(1

03)

RA

E0

320

382

002

(10

5)(2

21)

(0

13)

Pop

sha

reov

er65

113

114

130

125

127

142

20

122

012

20

10(1

86)

(1

87)

(1

95)

(4

04)

(3

96)

(3

81)

(0

34)

(03

5)(0

27)

Log

GD

Ppe

rca

pita

116

083

054

219

169

145

21

642

146

21

72(0

37)

(02

6)(0

16)

(13

7)(1

04)

(07

8)(0

92)

(08

3)(0

93)

OE

CD

dum

my

134

913

82

115

51

201

752

021

108

310

64

106

5(2

10)

(2

14)

(1

57)

(03

7)(0

52)

(00

5)(2

98)

(2

92)

(2

60)

C

onst

ant

275

574

108

62

217

92

174

62

117

529

32

278

328

87

(01

1)(0

23)

(04

2)(1

73)

(1

36)

(08

1)(2

09)

(2

01)

(20

0)

Ran

ge5

253

848

308

516

869

763

022

490

55O

bser

vati

ons

4040

3540

4035

4040

35R

20

760

750

730

840

830

800

440

430

38

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s












TA

BL

EV

IP

RIM

AR

YS

PE

ND

ING

TR

AN

SF

ER

S

PU

BL

ICG

OO

DS

AN

DE

LE

CT

OR

AL

SY

ST

EM

S(O

EC

DC

OU

NT

RIE

S)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

305

303

20

34(L

ogA

M)

(21

0)

(40

3)

(04

1)L

ogst

and

dist

mag

n2

562

632

028

(Log

SM

)(1

89)

(3

64)

(0

37)

RA

E1

411

622

064

(12

9)(2

56)

(1

14)

Pop

sha

reov

er65

137

115

143

119

095

117

022

025

040

(14

9)(1

17)

(14

0)(2

50)

(1

80)

(1

99)

(0

43)

(04

5)(0

75)

Log

GD

Ppe

rca

pita

171

516

39

157

67

737

056

594

484

574

09(2

36)

(2

23)

(2

03)

(20

5)

(18

0)

(14

8)(1

08)

(11

1)(1

03)

OE

CD

214

210

213

154

212

092

274

42

264

42

254

04

223

17

224

34

224

04

(20

2)

(18

5)

(16

2)(2

04)

(1

70)

(12

6)(0

58)

(06

1)(0

63)

Ran

ge15

28

128

09

1115

18

131

310

45

169

141

413

Obs

erva

tion

s20

2020

2020

2020

2020

R2

034

031

024

061

057

044

20

82

08

20

1

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s




VII4 Robustness




(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD



Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)


Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)





















VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES






























































VII1 The Data












TA

BL

EII

SU

MM

AR

YS

TA

TIS

TIC

S

All

Mea

nA

llS

DA

llM

inA

llM

axO

EC

DM

ean

OE

CD

SD

OE

CD

Min

OE

CD

Max

LA

Mea

nL

AS

DL

AM

inL

AM

ax

Avg

dis

tric

tm

agn

(A

M)

161

307

115

015

323

115

017

329

91

120

Sta

nd

dist

rict

mag

n(

SM

)23

639

71

180

820

133

114

83

272

461

180

8R

AE

34

40

326

12

41

90

36

83

76

50

326

1P

rim

ary

expe

ndG

DP

326

151

103

666

454

94

327

666

198

55

103

298

Tra

nfe

rsG

DP

134

95

14

316

213

63

115

316

55

36

14

146

Pub

lic

good

sG

DP

181

56

71

299

219

42

1629

914

44

17

123

5G

DP

per

capi

ta84

0055

0012

0018

100

1320

028

0068

0018

100

3500

1800

1200

7200

Pop

uls

hare

over

659

75

22

919

514

42

113

195

52

22

911

8E

ffn

umb

ofpa

rtie

s3

51

71

38

43

71

61

98

43

31

71

38

4E

TH

NIC

(fra

cti

ndex

)20

119

81

7522

122

11

7518

117

51

159

9O

PE

N(o

penn

ess)

305

173

74

961

301

141

87

629

309

204

74

961

Cro

ss-s

ecti

onal

aver

ages

sta

nda

rdde

viat

ions

min

imum

and

max

imu

mva

lues

for

all

the

vari

able

sfo

rth

efu

llsa

mpl

eth

eO

EC

Dsa

mpl

ean

dth

eL

atin

Am

eric

ansa

mpl

eS

eeth

eA

ppen

dix

for

vari

able

srsquom

nem

onic

sE

ach

cros

s-se

ctio

nalo

bser

vati

onis

the

1991

ndash199

4(o

rcl

oses

tav

aila

ble

peri

od)a

vera

geof

the

vari

able

for

the

cou

ntry

Num

ber

ofob

serv

atio

ns

40fo

rth

ew

hol

esa

mpl

e(3

5fo

rR

AE

)20

for

the

OE

CD

sam

ple

20fo

rth

eL

atin

Am

eric

ansa

mpl

e(1

5fo

rR

AE

)


















Panel B OECD












VII3 Basic Results






TA

BL

EV

PR

IMA

RY

SP

EN

DIN

G

TR

AN

SF

ER

SP

UB

LIC

GO

OD

SA

ND

EL

EC

TO

RA

LS

YS

TE

MS

(FU

LL

SA

MP

LE)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

105

170

20

60(L

ogA

M)

(11

0)(3

49)

(1

12)

Log

stan

ddi

stm

agn

074

132

20

48(L

ogS

M)

(08

9)(3

06)

(1

03)

RA

E0

320

382

002

(10

5)(2

21)

(0

13)

Pop

sha

reov

er65

113

114

130

125

127

142

20

122

012

20

10(1

86)

(1

87)

(1

95)

(4

04)

(3

96)

(3

81)

(0

34)

(03

5)(0

27)

Log

GD

Ppe

rca

pita

116

083

054

219

169

145

21

642

146

21

72(0

37)

(02

6)(0

16)

(13

7)(1

04)

(07

8)(0

92)

(08

3)(0

93)

OE

CD

dum

my

134

913

82

115

51

201

752

021

108

310

64

106

5(2

10)

(2

14)

(1

57)

(03

7)(0

52)

(00

5)(2

98)

(2

92)

(2

60)

C

onst

ant

275

574

108

62

217

92

174

62

117

529

32

278

328

87

(01

1)(0

23)

(04

2)(1

73)

(1

36)

(08

1)(2

09)

(2

01)

(20

0)

Ran

ge5

253

848

308

516

869

763

022

490

55O

bser

vati

ons

4040

3540

4035

4040

35R

20

760

750

730

840

830

800

440

430

38

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s












TA

BL

EV

IP

RIM

AR

YS

PE

ND

ING

TR

AN

SF

ER

S

PU

BL

ICG

OO

DS

AN

DE

LE

CT

OR

AL

SY

ST

EM

S(O

EC

DC

OU

NT

RIE

S)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

305

303

20

34(L

ogA

M)

(21

0)

(40

3)

(04

1)L

ogst

and

dist

mag

n2

562

632

028

(Log

SM

)(1

89)

(3

64)

(0

37)

RA

E1

411

622

064

(12

9)(2

56)

(1

14)

Pop

sha

reov

er65

137

115

143

119

095

117

022

025

040

(14

9)(1

17)

(14

0)(2

50)

(1

80)

(1

99)

(0

43)

(04

5)(0

75)

Log

GD

Ppe

rca

pita

171

516

39

157

67

737

056

594

484

574

09(2

36)

(2

23)

(2

03)

(20

5)

(18

0)

(14

8)(1

08)

(11

1)(1

03)

OE

CD

214

210

213

154

212

092

274

42

264

42

254

04

223

17

224

34

224

04

(20

2)

(18

5)

(16

2)(2

04)

(1

70)

(12

6)(0

58)

(06

1)(0

63)

Ran

ge15

28

128

09

1115

18

131

310

45

169

141

413

Obs

erva

tion

s20

2020

2020

2020

2020

R2

034

031

024

061

057

044

20

82

08

20

1

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s




VII4 Robustness




(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD



Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)


Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)





















VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES











































TA

BL

EII

SU

MM

AR

YS

TA

TIS

TIC

S

All

Mea

nA

llS

DA

llM

inA

llM

axO

EC

DM

ean

OE

CD

SD

OE

CD

Min

OE

CD

Max

LA

Mea

nL

AS

DL

AM

inL

AM

ax

Avg

dis

tric

tm

agn

(A

M)

161

307

115

015

323

115

017

329

91

120

Sta

nd

dist

rict

mag

n(

SM

)23

639

71

180

820

133

114

83

272

461

180

8R

AE

34

40

326

12

41

90

36

83

76

50

326

1P

rim

ary

expe

ndG

DP

326

151

103

666

454

94

327

666

198

55

103

298

Tra

nfe

rsG

DP

134

95

14

316

213

63

115

316

55

36

14

146

Pub

lic

good

sG

DP

181

56

71

299

219

42

1629

914

44

17

123

5G

DP

per

capi

ta84

0055

0012

0018

100

1320

028

0068

0018

100

3500

1800

1200

7200

Pop

uls

hare

over

659

75

22

919

514

42

113

195

52

22

911

8E

ffn

umb

ofpa

rtie

s3

51

71

38

43

71

61

98

43

31

71

38

4E

TH

NIC

(fra

cti

ndex

)20

119

81

7522

122

11

7518

117

51

159

9O

PE

N(o

penn

ess)

305

173

74

961

301

141

87

629

309

204

74

961

Cro

ss-s

ecti

onal

aver

ages

sta

nda

rdde

viat

ions

min

imum

and

max

imu

mva

lues

for

all

the

vari

able

sfo

rth

efu

llsa

mpl

eth

eO

EC

Dsa

mpl

ean

dth

eL

atin

Am

eric

ansa

mpl

eS

eeth

eA

ppen

dix

for

vari

able

srsquom

nem

onic

sE

ach

cros

s-se

ctio

nalo

bser

vati

onis

the

1991

ndash199

4(o

rcl

oses

tav

aila

ble

peri

od)a

vera

geof

the

vari

able

for

the

cou

ntry

Num

ber

ofob

serv

atio

ns

40fo

rth

ew

hol

esa

mpl

e(3

5fo

rR

AE

)20

for

the

OE

CD

sam

ple

20fo

rth

eL

atin

Am

eric

ansa

mpl

e(1

5fo

rR

AE

)


















Panel B OECD












VII3 Basic Results






TA

BL

EV

PR

IMA

RY

SP

EN

DIN

G

TR

AN

SF

ER

SP

UB

LIC

GO

OD

SA

ND

EL

EC

TO

RA

LS

YS

TE

MS

(FU

LL

SA

MP

LE)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

105

170

20

60(L

ogA

M)

(11

0)(3

49)

(1

12)

Log

stan

ddi

stm

agn

074

132

20

48(L

ogS

M)

(08

9)(3

06)

(1

03)

RA

E0

320

382

002

(10

5)(2

21)

(0

13)

Pop

sha

reov

er65

113

114

130

125

127

142

20

122

012

20

10(1

86)

(1

87)

(1

95)

(4

04)

(3

96)

(3

81)

(0

34)

(03

5)(0

27)

Log

GD

Ppe

rca

pita

116

083

054

219

169

145

21

642

146

21

72(0

37)

(02

6)(0

16)

(13

7)(1

04)

(07

8)(0

92)

(08

3)(0

93)

OE

CD

dum

my

134

913

82

115

51

201

752

021

108

310

64

106

5(2

10)

(2

14)

(1

57)

(03

7)(0

52)

(00

5)(2

98)

(2

92)

(2

60)

C

onst

ant

275

574

108

62

217

92

174

62

117

529

32

278

328

87

(01

1)(0

23)

(04

2)(1

73)

(1

36)

(08

1)(2

09)

(2

01)

(20

0)

Ran

ge5

253

848

308

516

869

763

022

490

55O

bser

vati

ons

4040

3540

4035

4040

35R

20

760

750

730

840

830

800

440

430

38

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s












TA

BL

EV

IP

RIM

AR

YS

PE

ND

ING

TR

AN

SF

ER

S

PU

BL

ICG

OO

DS

AN

DE

LE

CT

OR

AL

SY

ST

EM

S(O

EC

DC

OU

NT

RIE

S)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

305

303

20

34(L

ogA

M)

(21

0)

(40

3)

(04

1)L

ogst

and

dist

mag

n2

562

632

028

(Log

SM

)(1

89)

(3

64)

(0

37)

RA

E1

411

622

064

(12

9)(2

56)

(1

14)

Pop

sha

reov

er65

137

115

143

119

095

117

022

025

040

(14

9)(1

17)

(14

0)(2

50)

(1

80)

(1

99)

(0

43)

(04

5)(0

75)

Log

GD

Ppe

rca

pita

171

516

39

157

67

737

056

594

484

574

09(2

36)

(2

23)

(2

03)

(20

5)

(18

0)

(14

8)(1

08)

(11

1)(1

03)

OE

CD

214

210

213

154

212

092

274

42

264

42

254

04

223

17

224

34

224

04

(20

2)

(18

5)

(16

2)(2

04)

(1

70)

(12

6)(0

58)

(06

1)(0

63)

Ran

ge15

28

128

09

1115

18

131

310

45

169

141

413

Obs

erva

tion

s20

2020

2020

2020

2020

R2

034

031

024

061

057

044

20

82

08

20

1

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s




VII4 Robustness




(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD



Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)


Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)





















VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES



















































Panel B OECD












VII3 Basic Results






TA

BL

EV

PR

IMA

RY

SP

EN

DIN

G

TR

AN

SF

ER

SP

UB

LIC

GO

OD

SA

ND

EL

EC

TO

RA

LS

YS

TE

MS

(FU

LL

SA

MP

LE)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

105

170

20

60(L

ogA

M)

(11

0)(3

49)

(1

12)

Log

stan

ddi

stm

agn

074

132

20

48(L

ogS

M)

(08

9)(3

06)

(1

03)

RA

E0

320

382

002

(10

5)(2

21)

(0

13)

Pop

sha

reov

er65

113

114

130

125

127

142

20

122

012

20

10(1

86)

(1

87)

(1

95)

(4

04)

(3

96)

(3

81)

(0

34)

(03

5)(0

27)

Log

GD

Ppe

rca

pita

116

083

054

219

169

145

21

642

146

21

72(0

37)

(02

6)(0

16)

(13

7)(1

04)

(07

8)(0

92)

(08

3)(0

93)

OE

CD

dum

my

134

913

82

115

51

201

752

021

108

310

64

106

5(2

10)

(2

14)

(1

57)

(03

7)(0

52)

(00

5)(2

98)

(2

92)

(2

60)

C

onst

ant

275

574

108

62

217

92

174

62

117

529

32

278

328

87

(01

1)(0

23)

(04

2)(1

73)

(1

36)

(08

1)(2

09)

(2

01)

(20

0)

Ran

ge5

253

848

308

516

869

763

022

490

55O

bser

vati

ons

4040

3540

4035

4040

35R

20

760

750

730

840

830

800

440

430

38

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s












TA

BL

EV

IP

RIM

AR

YS

PE

ND

ING

TR

AN

SF

ER

S

PU

BL

ICG

OO

DS

AN

DE

LE

CT

OR

AL

SY

ST

EM

S(O

EC

DC

OU

NT

RIE

S)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

305

303

20

34(L

ogA

M)

(21

0)

(40

3)

(04

1)L

ogst

and

dist

mag

n2

562

632

028

(Log

SM

)(1

89)

(3

64)

(0

37)

RA

E1

411

622

064

(12

9)(2

56)

(1

14)

Pop

sha

reov

er65

137

115

143

119

095

117

022

025

040

(14

9)(1

17)

(14

0)(2

50)

(1

80)

(1

99)

(0

43)

(04

5)(0

75)

Log

GD

Ppe

rca

pita

171

516

39

157

67

737

056

594

484

574

09(2

36)

(2

23)

(2

03)

(20

5)

(18

0)

(14

8)(1

08)

(11

1)(1

03)

OE

CD

214

210

213

154

212

092

274

42

264

42

254

04

223

17

224

34

224

04

(20

2)

(18

5)

(16

2)(2

04)

(1

70)

(12

6)(0

58)

(06

1)(0

63)

Ran

ge15

28

128

09

1115

18

131

310

45

169

141

413

Obs

erva

tion

s20

2020

2020

2020

2020

R2

034

031

024

061

057

044

20

82

08

20

1

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s




VII4 Robustness




(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD



Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)


Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)





















VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES









































VII3 Basic Results






TA

BL

EV

PR

IMA

RY

SP

EN

DIN

G

TR

AN

SF

ER

SP

UB

LIC

GO

OD

SA

ND

EL

EC

TO

RA

LS

YS

TE

MS

(FU

LL

SA

MP

LE)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

105

170

20

60(L

ogA

M)

(11

0)(3

49)

(1

12)

Log

stan

ddi

stm

agn

074

132

20

48(L

ogS

M)

(08

9)(3

06)

(1

03)

RA

E0

320

382

002

(10

5)(2

21)

(0

13)

Pop

sha

reov

er65

113

114

130

125

127

142

20

122

012

20

10(1

86)

(1

87)

(1

95)

(4

04)

(3

96)

(3

81)

(0

34)

(03

5)(0

27)

Log

GD

Ppe

rca

pita

116

083

054

219

169

145

21

642

146

21

72(0

37)

(02

6)(0

16)

(13

7)(1

04)

(07

8)(0

92)

(08

3)(0

93)

OE

CD

dum

my

134

913

82

115

51

201

752

021

108

310

64

106

5(2

10)

(2

14)

(1

57)

(03

7)(0

52)

(00

5)(2

98)

(2

92)

(2

60)

C

onst

ant

275

574

108

62

217

92

174

62

117

529

32

278

328

87

(01

1)(0

23)

(04

2)(1

73)

(1

36)

(08

1)(2

09)

(2

01)

(20

0)

Ran

ge5

253

848

308

516

869

763

022

490

55O

bser

vati

ons

4040

3540

4035

4040

35R

20

760

750

730

840

830

800

440

430

38

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s












TA

BL

EV

IP

RIM

AR

YS

PE

ND

ING

TR

AN

SF

ER

S

PU

BL

ICG

OO

DS

AN

DE

LE

CT

OR

AL

SY

ST

EM

S(O

EC

DC

OU

NT

RIE

S)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

305

303

20

34(L

ogA

M)

(21

0)

(40

3)

(04

1)L

ogst

and

dist

mag

n2

562

632

028

(Log

SM

)(1

89)

(3

64)

(0

37)

RA

E1

411

622

064

(12

9)(2

56)

(1

14)

Pop

sha

reov

er65

137

115

143

119

095

117

022

025

040

(14

9)(1

17)

(14

0)(2

50)

(1

80)

(1

99)

(0

43)

(04

5)(0

75)

Log

GD

Ppe

rca

pita

171

516

39

157

67

737

056

594

484

574

09(2

36)

(2

23)

(2

03)

(20

5)

(18

0)

(14

8)(1

08)

(11

1)(1

03)

OE

CD

214

210

213

154

212

092

274

42

264

42

254

04

223

17

224

34

224

04

(20

2)

(18

5)

(16

2)(2

04)

(1

70)

(12

6)(0

58)

(06

1)(0

63)

Ran

ge15

28

128

09

1115

18

131

310

45

169

141

413

Obs

erva

tion

s20

2020

2020

2020

2020

R2

034

031

024

061

057

044

20

82

08

20

1

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s




VII4 Robustness




(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD



Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)


Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)





















VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES



































TA

BL

EV

PR

IMA

RY

SP

EN

DIN

G

TR

AN

SF

ER

SP

UB

LIC

GO

OD

SA

ND

EL

EC

TO

RA

LS

YS

TE

MS

(FU

LL

SA

MP

LE)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

105

170

20

60(L

ogA

M)

(11

0)(3

49)

(1

12)

Log

stan

ddi

stm

agn

074

132

20

48(L

ogS

M)

(08

9)(3

06)

(1

03)

RA

E0

320

382

002

(10

5)(2

21)

(0

13)

Pop

sha

reov

er65

113

114

130

125

127

142

20

122

012

20

10(1

86)

(1

87)

(1

95)

(4

04)

(3

96)

(3

81)

(0

34)

(03

5)(0

27)

Log

GD

Ppe

rca

pita

116

083

054

219

169

145

21

642

146

21

72(0

37)

(02

6)(0

16)

(13

7)(1

04)

(07

8)(0

92)

(08

3)(0

93)

OE

CD

dum

my

134

913

82

115

51

201

752

021

108

310

64

106

5(2

10)

(2

14)

(1

57)

(03

7)(0

52)

(00

5)(2

98)

(2

92)

(2

60)

C

onst

ant

275

574

108

62

217

92

174

62

117

529

32

278

328

87

(01

1)(0

23)

(04

2)(1

73)

(1

36)

(08

1)(2

09)

(2

01)

(20

0)

Ran

ge5

253

848

308

516

869

763

022

490

55O

bser

vati

ons

4040

3540

4035

4040

35R

20

760

750

730

840

830

800

440

430

38

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s












TA

BL

EV

IP

RIM

AR

YS

PE

ND

ING

TR

AN

SF

ER

S

PU

BL

ICG

OO

DS

AN

DE

LE

CT

OR

AL

SY

ST

EM

S(O

EC

DC

OU

NT

RIE

S)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

305

303

20

34(L

ogA

M)

(21

0)

(40

3)

(04

1)L

ogst

and

dist

mag

n2

562

632

028

(Log

SM

)(1

89)

(3

64)

(0

37)

RA

E1

411

622

064

(12

9)(2

56)

(1

14)

Pop

sha

reov

er65

137

115

143

119

095

117

022

025

040

(14

9)(1

17)

(14

0)(2

50)

(1

80)

(1

99)

(0

43)

(04

5)(0

75)

Log

GD

Ppe

rca

pita

171

516

39

157

67

737

056

594

484

574

09(2

36)

(2

23)

(2

03)

(20

5)

(18

0)

(14

8)(1

08)

(11

1)(1

03)

OE

CD

214

210

213

154

212

092

274

42

264

42

254

04

223

17

224

34

224

04

(20

2)

(18

5)

(16

2)(2

04)

(1

70)

(12

6)(0

58)

(06

1)(0

63)

Ran

ge15

28

128

09

1115

18

131

310

45

169

141

413

Obs

erva

tion

s20

2020

2020

2020

2020

R2

034

031

024

061

057

044

20

82

08

20

1

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s




VII4 Robustness




(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD



Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)


Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)





















VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES













































TA

BL

EV

IP

RIM

AR

YS

PE

ND

ING

TR

AN

SF

ER

S

PU

BL

ICG

OO

DS

AN

DE

LE

CT

OR

AL

SY

ST

EM

S(O

EC

DC

OU

NT

RIE

S)

Dep

var

(1

)E

XP

(2)

EX

P(3

)E

XP

(4)

TR

AN

(5)

TR

AN

(6)

TR

AN

(7)

PG

OO

D(8

)P

GO

OD

(9)

PG

OO

D

Log

avg

dist

m

agn

305

303

20

34(L

ogA

M)

(21

0)

(40

3)

(04

1)L

ogst

and

dist

mag

n2

562

632

028

(Log

SM

)(1

89)

(3

64)

(0

37)

RA

E1

411

622

064

(12

9)(2

56)

(1

14)

Pop

sha

reov

er65

137

115

143

119

095

117

022

025

040

(14

9)(1

17)

(14

0)(2

50)

(1

80)

(1

99)

(0

43)

(04

5)(0

75)

Log

GD

Ppe

rca

pita

171

516

39

157

67

737

056

594

484

574

09(2

36)

(2

23)

(2

03)

(20

5)

(18

0)

(14

8)(1

08)

(11

1)(1

03)

OE

CD

214

210

213

154

212

092

274

42

264

42

254

04

223

17

224

34

224

04

(20

2)

(18

5)

(16

2)(2

04)

(1

70)

(12

6)(0

58)

(06

1)(0

63)

Ran

ge15

28

128

09

1115

18

131

310

45

169

141

413

Obs

erva

tion

s20

2020

2020

2020

2020

R2

034

031

024

061

057

044

20

82

08

20

1

Dep

ende

ntva

riab

les

EX

Ps

hare

ofto

tal

prim

ary

spen

ding

inG

DP

(col

umn

s(1

)ndash(3

))T

RA

Ns

hare

oftr

ansf

ers

inG

DP

(col

umn

s(4

)ndash(6

))P

GO

OD

sh

are

ofsp

endi

ng

onpu

blic

good

sin

GD

P(c

olum

ns(7

)ndash(9

))a

vera

ges

1991

ndash199

4or

clos

est

avai

labl

epe

riod

See

App

endi

x3

for

the

de

niti

onof

all

vari

able

sE

stim

atio

nby

ordi

nary

leas

tsq

uar

es(t

-sta

tist

ics

are

inpa

ren

thes

es)

(

)si

gni

cant

atth

e10

perc

ent

(5pe

rcen

t)le

vel

Ran

geis

ran

geof

vari

atio

nof

depe

nden

tva

riab

lea

ssoc

iate

dw

ith

rang

eof

vari

atio

nof

elec

tora

lva

riab

le

hold

ing

con

stan

tal

lot

her

vari

able

s




VII4 Robustness




(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD



Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)


Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)





















VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES





































VII4 Robustness




(1)EXP

(2)EXP

(3)TRAN

(4)TRAN

(5)PGOOD

(6)PGOOD



Pop share over 65 069 049 098 095 2 038 2 052(096) (065) (269) (244) (075) (097)


Constant 2984 2600 2 1212 2 1113 3764 3387(140) (124) (112) (102) (249) (226)





















VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES




















































VIII1 Common Shocks













Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES








































Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



(161) (435)RAE p time dummy 005 006

(184) (356)RAE 2 295 2 293

( 2 143) ( 2 279)Pop share over 65 095 101 095 021 021 025

(222) (243) (228) (108) (106) (123)Log GDP per cap 2 409 2 362 2 384 2 479 2 446 2 443










(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES









































(17)








Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES








































Dep var(1)

EXP(2)

EXP(3)

EXP(4)

TRAN(5)

TRAN(6)

TRAN



RAE 2 014 2 039(2 030) ( 2 153)

RAE p unempl rate 005 006(101) (209)

Unempl rate 042 048 08 027 030 068(204) (238) (444) (238) (274) (680)

Pop share over 65 168 168 177 105 104 115(438) (435) (458) (500) (491) (531)






















APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES




















































APPENDIX 1










k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES







































k1 k2

k3 k2


ik1= ik1




with k1


ik1(2 2 ik1



APPENDIX 2







22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES




































22 Upper Quotas




















APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES












































APPENDIX 3









32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES




































32 Fiscal Variables

OECD Countries


















34 Other Variables









REFERENCES












































34 Other Variables









REFERENCES



































ELECTORAL SYSTEMS AND PUBLIC SPENDINGiversen/PDFfiles/Melesi-FerettiPerottiR… · “greater...

Documents

Transcript of ELECTORAL SYSTEMS AND PUBLIC SPENDINGiversen/PDFfiles/Melesi-FerettiPerottiR… · “greater...