in search of a compensatory mixed electoral system … › ... · 2016-10-12 · to make the final...

SUMMARY

Louis MassicotteAssociate professor

Political Science DepartmentUniversité de Montréal

IN SEARCH OF A COMPENSATORY MIXED

ELECTORAL SYSTEM FOR QUÉBEC

Graphic designRM communication designSecrétariat à la communication gouvernementale

Legal deposit — 2004Bibliothèque nationale du QuébecISBN 2-550-43380-7© Gouvernement du Québec, 2004

Electoral System Review

TABLE OF CONTENTS

Preliminary note 5

FIRST PART EXPERIENCES OF OTHER COUNTRIES 6

SEAT DISTRIBUTION 6

Choosing a ratio 6

The level of compensatory seat distribution 6

Methods for distributing compensatory seats 8

The problem of excess seats 9

Thresholds 9

VOTER CHOICE 10

One vote or two? 10

Should voters be allowed to indicate preferences among list candidates? 12

DESIGNATING AND REPLACING ELECTED MEMBERS 12

Two classes of members? 12

How will compensatory seats be filled? 13

Double candidacy, the key to compensatory systems 14

Women's representation in parliaments 15

How should vacancies be filled between general elections? 15

In Search of Compensatory Mixed Electoral System for Québec SUMMARY

SECOND PART SIMULATIONS BASED ON THE 1998 AND 2003 QUÉBEC ELECTIONS 17

METHODOLOGY 17

Constituency/list ratio 17

Territorial subdivision 17

Seat distribution method 18

Computation techniques 18

Other variables 18

OUTCOMES 18

KEY FINDINGS 19

Overall level of distortion and majority bonus 19

Reversals of party standings and asymmetry of the different approaches 20

Thresholds and small parties 22

Identity of list members 22

THIRD PART CONCLUSION AND RECOMMENDATIONS 23

ADVANTAGES 23

ELECTORAL SYSTEM CHARACTERISTICS 23

4


Preliminary note

In July 2003, the Minister for the Reform ofDemocratic Institutions, Jacques P. Dupuis,mandated Louis Massicotte, associate pro-fessor in political science at the Universitéde Montréal, to investigate the possibleapproaches for a compensatory mixedelectoral system that reflects Québec’sspecific conditions. It was agreed that the mandate would focus on the workingmodel put forward by Prof. Massicotte andhis colleague, Prof. André Blais.1 In thisreport In Search of a Compensatory MixedElectoral System for Québec, Prof. Massicottehas brought his thoughts on this matter to fruition. A summary is provided in thefollowing pages. Any reader interested inexamining all of the supporting data willhave to refer to the report itself.

In this summary, the expression “compen-satory mixed electoral system” refers to anelectoral system where the candidate whoreceives the most votes is elected to repre-sent a single-member constituency andwhere the parties are then assigned compen-satory seats, to ensure that each party’stotal number of seats is close to its share of the popular vote.

In the report, the first part covers whatmay be learned from the experience ofsocieties where a compensatory mixed elec-toral system already exists. The focus is onthe experiences of Germany (at the federaland Land levels), New Zealand, Scotland, andWales. The different legislative approachescurrently in force are an interesting source

of inspiration and show the variety ofpossible arrangements for such a system in all its facets: constituency/list seat ratio;geographic framework for compensation(territory-wide or regional); seat distributionmethods; different computation techniquesfor seat distribution; and excess seats.

The first part also covers the choices thatvoters may be asked to make, i.e., the possi-bility not only to vote for a constituencycandidate, but also to vote separately for apolitical party, as well as to express prefer-ences for certain candidates on a party list.This part also addresses the issue of coexis-tence between constituency members andlist members in a single Parliament, as wellas the issues of women’s representationand replacement of elected members.

The second part is devoted to a series ofsimulations that project the outcomes ofthese different possible approaches whenapplied to the 1998 and 2003 Québecelections.

The third part encompasses the conclusionand the author’s recommendations on themodalities of electoral reform.

This summary was made with the assistanceof André Lemelin, of the firm La productionde l’espace-temps Inc. The Secrétariat à laréforme des institutions démocratiquesassumes full responsibility for the finalversion.

5


1. Louis MASSICOTTE and André BLAIS. "La réforme électorale : profil d’un mode de scrutin mixte approprié auQuébec", in Jean Crête (ed.), Hommage à Vincent Lemieux. La science politique au Québec. Le dernier des maîtresfondateurs, Québec, Les Presses de l’Université Laval, 2003, pp. 247-269.

FIRST PART Experiences of other countries

SEAT DISTRIBUTIONIn a compensatory mixed system, seat distribution is a complex equation withmany variables that come into play. We willexamine the variables one by one.

Choosing a ratio

The ratio is the number of elected consti-tuency members and the number of compen-satory seats. A compensatory mixed systemdoes not absolutely require an equal numberof direct seats and compensatory seats—also called list seats. A ratio equal or veryclose to 50/50 is certainly used in Germanyat the federal level and in eight Länder.But in the other Länder and countriesunder study the list seats make up 29% to43% of the total. Such compensation willvery effectively reduce the distortions ofthe first-past-the-post system, as long asthe distortions are not too large. The initialratio has seldom been substantially alteredover the years. No government has takenthe risk of reducing the total number ofmembers by cutting back exclusively on thenumber of list seats, which are largely heldby opposition members.

If the ratio is biased more to constituencyseats, the distortions may be less completelycorrected. The further one goes below thebar of 40%, the likelier there will not beenough compensatory seats to correct thedistortions. The ratio one chooses willdepend to some degree on whether thereare regional subdivisions. Territory-widecompensation gives lawmakers much leewayin setting the ratio because the compensa-tory seats are not tied to any specific region.For a 125-seat Parliament, the ratio mayeasily range from 70/25 to 80/45 or 63/62.Regional compensation narrows the rangeof possible ratios, all the more so if the

regions are small in size and large in number.If each region has 5 seats, the only possibleratios are 3+2 or 4+1.

The smaller the proportion of list seats, thelikelier it is that the party that wins themost constituencies will have excess seats.The total number of seats will thus exceedthe statutory total. This clearly emergesfrom a review of the 131 elections heldunder personalized proportional represen-tation in Germany from 1947 to September2004. Wherever list seats made up 40% orless of the total, and partly for this reason,excess seats were produced on averageover half the time, versus less than a quarterof the time wherever list seats made up ahigher proportion of the total. The morethe ratio was biased to constituency seats,the lower the likelihood that the strongestparty would receive list seats.

The level of compensatory seat distribution

Four options:

1. Territory-wide compensation is the simplestapproach. A comparison is made, foreach party, between the number of seatsit would receive under proportional repre-sentation (PR), based on its nationalshare of the popular vote, and the totalnumber of seats it actually won. The secondnumber is subtracted from the first togive the number of compensatory listseats the party is entitled to. The electedmembers for these seats are not tied toany specific territorial subdivision. Undera system of territory-wide compensation,nothing requires parties to produceregionally balanced lists, but in practicethis is dictated by political commonsense. New Zealand and most Länderhave opted for territory-wide compensa-tion.

6


2. If one wants list members to be anchoredto more concrete and smaller territories,the solution is to group the single-memberconstituencies into a certain number ofregions and to compute compensationseparately in each region. This is regionalcompensation, which is notably used inthe Land of Bavaria, in Scotland, and inWales. It was in force in the first twoBundestag elections (1949 and 1953).

3. German lawmakers chose in 1956 tointroduce a new seat distribution method,to make the final outcome more propor-tional. Territory-wide compensation isfollowed by redistribution of party seatsamong the regions. This method is still in force and uses two successive distribu-tions. First, seats are distributed amongthe parties at the national level, usingthe total number of valid second votesfor each party that has cleared the 5%threshold. Next, for each party, the seatsreceived at the national level are redis-tributed among the Land lists, based onthe Land’s share of all valid second votesfor the party. In the second operation,each Land is competing with the otherLänder for a share of each party’s seats.

4. Regional redistribution may be optional,i.e., left to the discretion of the parties.This variant is notably found in the Landof Berlin. Each party may present eithera list for the whole territory or severalregional lists. Seats are first distributedamong the parties for the whole territory.The parties that so wish may then go onto a second distribution among theirregional lists. If a party wishes to anchorits list members to limited territories, itcan do so without affecting anotherparty’s right to present a national list.

The choices in this area have three majorimplications. First, territory-wide compen-sation reduces to a minimum the spreadbetween each party’s percentage of thevote and its number of seats. This methodmakes it likelier for small parties to berepresented in Parliament because theentire territory forms one constituency forthe purposes of computation—the surestway to minimize distortions. Conversely,the distortions are greater if the compen-sation is done regionally rather than terri-tory-wide, and greater still if the regionsare small in size and large in number.

Second, territory-wide compensation reducesthe possibility of excess seats. Such seatsare likelier to arise if the compensation isdone within regions, and even more so ifthere are many small regions. The moreregional entities there are, the likelier aparty will be overrepresented in someregions. To offset overrepresentation byone party, these regions have to receivemore representation to compensate theother parties and the total number ofelected members will increase.

Third, there is a fear that a national list willgive party leaders too much clout in choosinglist candidates and that they might use thispower to impose their favourites. With anational list, different geographic areasmay still be equitably represented, but thisoutcome is less certain than it is withregional compensation.

7


Methods for distributing compensatoryseats

The German method is by far the mostcommon one. All seats are fictitiouslydistributed according to the vote for eachparty. From these seat standings are sub-tracted the number of constituency seatsactually won by each party. The differenceis the number of compensatory seats andas long as enough are available this ope-ration will produce an overall outcome inproportion to each party’s vote. With thismethod, a party may manage to win moreconstituency seats than the total numberof seats it should get under PR (excessseats). So the total number of electedmembers has to increase.

The method used in Scotland and Walesavoids such an increase. To calculate PRwithin each region, this method first consi-ders the constituency seats already won byeach party. Each party’s number of votes isdivided by its number of constituency seatsplus one, this being the basis for assigningthe regional seats up to the statutory limit.This method prevents the final number ofelected members from being higher thanexpected, but it slightly increases thestrongest party’s majority bonus in theregion, since it is authorized to keep theexcess seat or seats it may have won.

The loser and surplus method, a variant ofwhich is used in Italy, adds up votes for thelosers and winning margins2 of the winnersamong each party’s constituency candidates.The total, which overrepresents the weakestparties, becomes the basis for any of theusual techniques of PR seat distribution. In Germany, this method was discontinuedwherever it had been used because it pro-duced too much distortion. It does not,however, increase the total number ofelected members.

Whichever of the three methods is chosen,the seats will be distributed among theparties by one of the traditional computationtechniques: the Largest Average technique(D’Hondt3), the Sainte-Laguë technique4;and the Largest Remainder technique.5It matters little which technique is chosenif seat distribution is calculated at theprovince-wide level. But when applied tosmaller regional entities, these techniquesmay produce somewhat different outcomes.The D’Hondt technique will favour thestrongest parties, whereas the Sainte-Laguëand Largest Remainder techniques will giverise to a more proportional distribution,despite the existence of regions. The LargestRemainder technique tends to favour theweakest parties, all the more so if thereare many regions. In Germany, at both the federal and Land levels, the LargestRemainder technique has superseded theD’Hondt technique. The Sainte-Laguë tech-nique is used in New Zealand.

8


2. The difference between the votes for the winning candidate and the votes for the runner-up.3. The D’Hondt technique divides the number of votes for each party successively by 1, 2, 3, etc. and gives the seats

to the parties with the highest quotients.4. The Sainte-Laguë technique divides the number of votes for each party successively by odd numbers

(1, 3, 5, 7, etc.) and gives the seats to the parties with the highest quotients.5. The Largest Remainder technique first requires calculating a quotient: the total number of votes for the eligible

parties divided by the number of seats. The votes for each party are then divided by this quotient. The wholenumber produced by the division will be that party's number of seats. If this operation fails to distribute all of theseats, the unassigned seats will go to the parties that, after division by the quotient, have the largest remainders,until all of the seats have been filled.

The problem of excess seats

With the German method, as mentionedabove, the number of seats may end uphigher than expected and may vary fromone election to the next.

Excess seats are not unusual. In Germany,they have arisen in 11 of the 15 federalelections since 1949, as they have in 41 ofthe 116 Land elections held under person-alized PR from 1947 to September 2004.Excess seats, especially if their appearanceleads to a concomitant creation of additionalseats to keep the outcome proportional,sometimes create the possibility of a verysubstantial increase in the total number ofseats. In such a case, if a party wins excessseats, additional compensatory seats arecreated for the other parties to make theoutcome fully proportional. This approachcreates the possibility of an even largerincrease in the total number of seats.

Previously, there were very few excess seatsat the federal level. Their numbers have risenwith the incorporation of the eastern Länder.Nonetheless, the increase in the statutorysize of the Bundestag, due to excess seats,has never exceeded 2.4% and has been onaverage less than 1%. On the other hand,in the Länder, the increase has sometimesbeen markedly higher (up to 37% moreelected members than expected), partlybecause of the compensatory additionalseats that result from excess seats.

Based on our analysis of elections in othercountries, excess seats most often ariseunder three conditions: the spread is largebetween the strongest party’s percentageof the vote and its percentage of the consti-tuency seats; the constituency/list ratio isbiased to constituency seats; and compen-sation is done within regions rather thanterritory-wide. The last condition is all themore important if the regions have only afew seats.

Thresholds

With the experience of the Weimer Republicin mind, the Germans appear to have beenamong the first to impose an explicit statu-tory threshold to limit the fragmenting ofparliamentary representation. The situationof the postwar era and the pressures of theoccupying powers led Länder in the Britishzone to require a minimum number ofvotes before a party could receive list seats.This requirement was imposed as early asthe first Bundestag election, in 1949. It hassince spread to all Länder. A figure of 5%of all valid ballots has over the yearsbecome the standard everywhere. NewZealand has opted for a 5% threshold atthe national level. No statutory thresholdexists in Scotland or Wales, where the frag-mentation of the territory into regions limitsthe chances of small parties anyway. Insome societies, the threshold may be waivedfor parties that fail to reach it but stillmanage to win one or more constituencyseats.

9


Thresholds may stray from the proportion-alist ideal but they are fairly common. Arecent survey shows that, out of 29 demo-cratic countries using PR for election by list,no fewer than 19 impose some kind ofthreshold (national or regional) on seat distribution. Turkey has the highest one(10%), followed by Poland (7%). In contrast,Israel sets its threshold at 1.5% and theNetherlands at only 0.67%.

The real consequences of a threshold areunpredictable because it is impossible toknow how the vote will break down infuture elections. In some societies, theconsequences can be severe, with manyparties failing to clear the thresholddespite receiving a high percentage of thevote. If we look at parties excluded fromParliament by a 5% threshold, their votescame to about 20% of the total in the1990 Czechoslovak election and reached a staggering 49.5% in the 1995 Russianelection. In 2002, Turkey’s 10% thresholdexcluded all but two parties from Parliament,the leading party winning 65.6% of the seatswith only 43.2% of the vote. In Germany andNew Zealand, the consequences have notbeen considerable, the excluded partiesoften receiving a tiny number of votes.

VOTER CHOICE

One vote or two?

A compensatory mixed system may workquite well when a single vote is cast for aconstituency candidate and is also considereda vote for that candidate’s party. Eachparty receives compensatory seats simplyon the basis of the total votes for all of itscandidates. This was the form in which thecompensatory mixed system was introducedinto Germany.

It is also possible to give each voter twovotes: the first vote will be cast for one ofthe constituency candidates and the secondfor a party list. The first vote will be usedto choose the constituency candidate andthe second to determine the overall distri-bution of seats among the parties, thisbeing the most important choice. Votersmay, for example, support the party oftheir choosing while not backing a candidatewho seems unsatisfactory.

When a voter has two votes, the constituencyvote matters only in the constituency whereit is cast. It may be interesting, however, toadd up nation-wide the constituency votesfor a party’s candidates. Analysts will some-times look at a constituency to comparethe vote for a party’s constituency candi-date and the vote for the party to see howpopular or well known a politician is (orhow unpopular and unknown). The officialterminology for the two votes varies fromone place to another. Sometimes, the twovotes are cast on a single ballot, sometimeson two different ballots.

10


Most compensatory mixed systems hadonly one vote when introduced, but theoverwhelming majority now have two votes.

What do the experiences of foreign countriestell us on this point?

• The existence of a second vote simplifiesthe task for small parties. If voters can casta second vote for a list, a small party maypresent a list with only several namesand field only very few or no constituencycandidates, since such candidates clearlyhave no chance of getting elected. Thisfactor seems to have led to successiveintroduction of the second vote in theLänder from the 1970s on.

• It is difficult to estimate the number ofvoters who will split their allegiances andto foresee the consequences of this splitvoting on each party’s relative strength.In German federal elections from 1953 on,the differences were initially small betweenthe distributions of first and second votes.Beginning in the 1970s, more voters beganto split their allegiances. “Split ballots”reached about 20% of the total in 1998and 22% in 2002. Wherever personalizedPR has been more recently introduced,split voting has soon caught on with voters,reaching respectively 21% and 19% inScotland and Wales in 1999. In NewZealand, it has taken on exceptional proportions: 37% in 1996; 35% in 1999;and 39% in 2002.

• When voters have two votes, the big partiesgenerally do better on the first vote thanon the second, the reverse being true forthe small parties. This phenomenon arisespartly because some of the small partiesfield few or no constituency candidateswhile presenting a list.

• The existence of a second vote is associatedwith higher rates of rejected ballots(roughly twice as many). This phenome-non has been rigorously demonstrated.

• The two votes open the door to voteswapping between parties. Some voterswho support a big party will choose tocast their list vote for an ideologicallyclose small party, to put it over the 5%threshold. They reason that a parliamentwith three parties, where their party isthe dominant partner in a winning coali-tion, is better than one with two parties,where their party sits in the opposition.

• The existence of a second vote may pro-foundly upset the normal operation of a compensatory system. This was Italy’sexperience in the 2001 election. The listseats were assigned on the basis of thesecond vote, minus the winning marginsof members elected in local constituen-cies, so as to create a compensatory effect.Silvio Berlusconi’s party resorted to theshrewd ploy of giving most of its consti-tuency candidates a party label differentfrom the one for its list candidates. Thecompensatory effect was almost cancelledout and the distortions were higher.

11


Should voters be allowed to indicate preferences among list candidates?

This question implies that the voters havetwo votes and that the compensatory seatsare distributed on the basis of lists presentedby political parties. If so, there are twooptions. Under the “closed list” system, list candidates may be elected in the orderthey appear on the list. Whoever establishesthe list order therefore largely determineswho will be elected. This factor may becountered by preferential voting, whichallows voters to alter the candidate listorder by expressing one or more preferencesfor candidates of the party they support.

Among the compensatory mixed systemswe examined, Bavaria is the only onewhere a voter can vote for a constituencycandidate and also express a preference forone of the list candidates, through a fairlycomplex procedure. In practice, voters takevery seriously the possibility available tothem of demoting less popular candidatesand promoting others to the top of the listwho were initially further down. Each party’sstar candidates tend to receive an exorbitantnumber of individual preferences. One resultof preferential voting is that regional can-didates belonging to the same party willcompete against each other at election time.

In a compensatory system where doublecandidacy is widely practised, the effectof preferences is relative. Indeed, many listcandidates are struck from the party list on election night because they have beenelected in a constituency. Any value, otherthan a moral one, is removed from the listpreferences they may have received.

DESIGNATING AND REPLACING ELECTEDMEMBERSThere are two factors. First, for anyonewho is accustomed to voting for a personand to meeting him or her during a termof office, it does matter how this person ispicked to fill a compensatory seat assignedto a party. Second, elected members areworried about seeing other members along-side them who have been elected in a dif-ferent way. But are there really groundsfor fearing the creation of two classes ofmembers who will be differentiated notonly by their mode of election but also bythe nature of their work, their prestige,their career opportunities, and even theirpay? This fear seems groundless in countriesthat have adopted compensatory mixedsystems.

Two classes of members?

The literature on mixed systems seldomrefers to the existence of “castes” of parlia-mentarians, either list members who aremocked by “real” members or, conversely,list members who snub their lowly counter-parts from the “stagnant ponds.” Instead,it explicitly emphasizes there have been noproblems due to the coexistence of memberswho enter parliament through separatemodes of election and at different territoriallevels.

In Germany, voters tend to communicatejust as well with constituency members aswith list members. For these voters, themember’s party allegiance is the main factorthat determines the priority of contact. Themost that may be said is that list membersare less inclined than their colleagues tothink their reelection chances will dependon securing construction projects for theirconstituencies. They are also a little less

12


inclined to sit on parliamentary committeesthat deal with local issues. It might be pos-sible that this second group of members,who are elected from lists in larger territories,will generate serious conflicts that end upcalling the system into question. But thispossibility has never materialized.

Scotland’s experience appears to have beenmore laborious. Some regional memberstend to compete unfairly with duly electedconstituency members, and for some consti-tuency members it is difficult to accept thatanother member has set up shop in “their”constituency. Still, many constituency membersadmit that the voters win from this rivalry.

The type of parliamentary mandate doesnot affect the actual seating arrangementof members in the house. Everywhere, therule is to group together members fromthe same party. Nowhere are list membersseated together independently of partyaffiliation. Nowhere does a member’s paydepend on his or her type of parliamentarymandate, although allowances may vary insome cases. As for career prospects, bothconstituency members and list membersare found in each of the prestigious officesof parliament: government leader; ministers;speakers; and caucus chairs. In general,constituency members predominate simplybecause the strongest party usually sits onthe government side. Most of its memberscome from constituencies, the reversebeing true for the other parties.

How will compensatory seats be filled?

Two solutions are possible. Compensatoryseats may go to the highest-scoring defeatedconstituency candidates of each party,according to either the number of votesthey received or their percentage of thevote. Compensatory seats may also go tocandidates on lists submitted by each partybefore the election.

The approach whereby defeated consti-tuency candidates are recycled has beenused in Baden-Württemberg since 1956.The winning constituency candidatereceives a “direct mandate.” The popularvote is then used to determine the numberof compensatory seats each party is entitledto. These seats are then assigned to defeatedcandidates who received the highest numberof votes in the region (and not percentageof the vote). A single constituency mayhave two, three, or even four membersand this number may vary from one electionto the next.

For all intents, such a system makes doublecandidacy necessary. A person can hope toreceive a compensatory seat only by runningin a constituency. Compensatory seats willthus be held by elected candidates whohave developed local roots and received ahigh number of votes. As well, party leaderscannot control assignment of these seatssince no list is used.

The approach whereby compensatory seatsgo to candidates on a list is used in almostall other compensatory systems. The list ofcandidates and their order on the list aredetermined by the party. Germany tends to regulate the way lists are prepared toensure they are prepared democratically.

13


The other countries leave it up to the partiesto determine list composition, while leavingit up to the party rank and file to preventtheir leaders from manipulating the listsbehind closed doors and leaving it up to thevoters to punish parties that have resortedto excessively autocratic procedures.

Double candidacy, the key to compensatory systems

The problem of two classes of electedmembers does not materialize in Germany.This is largely due to two reasons. Theoverwhelming majority of elected membersrun both in constituencies and on lists(double candidacy). As well, the vagaries ofshifting voting patterns increase the numberof members who have belonged to bothcategories during their careers.

Electoral laws do not forbid double candi-dacy. In the event of double success, however,they state that the fortunate candidate musttake the constituency mandate. His or hername is automatically struck from the listand the list seat goes to the next non-electedcandidate on the list. Although doublecandidacy is allowed, it is not required.

In theory, each party may field two comple-tely separate teams of candidates. In practice,the top positions on the list go to peoplewho are also candidates in specific consti-tuencies. Double candidacy is thereforewidespread.

This practice does not result from a legisla-tive requirement but from a simple andrational strategic calculation by the partiesand the candidates. In a compensatory system,the more constituency seats a party wins,the fewer list seats it will receive. It istherefore unwise, and even dangerous, forcandidates to rely entirely on being electedfrom a list. If their party sweeps the consti-tuencies, they may very well not be entitled

to any list seats. It is also unwise for a cons-tituency candidate to forego the safety netof being on a list. If the party is swept outof power, most of its elected members willcome from the list. So a candidate shouldavoid putting all of his or her eggs in onebasket and instead run on both levels,since it is impossible to foresee the elec-torate’s verdict several weeks in advance.Therefore, for someone who has tried tobelong to the other category of electedmember, it becomes difficult to denigrate it.

In the last Bundestag election (2002), 35%of the candidates resorted to double candi-dacy. Among the big parties, the only onesin practice with a serious chance of winninga constituency, the proportion was almosthalf (47.6%). The overwhelming majorityof constituency candidates fielded by themajor parties had the safety net of beingon a list: 83.5% of Christian DemocraticUnion (CDU) candidates; and 97% of SocialDemocratic Party (SPD) candidates. A verylarge majority of elected members ran inconstituencies and only a very small minorityof elected list members did not (3.6% inthe last German federal election and about4.5% in the Länder as a whole).

This German practice has spread to countriesthat have adopted such a system. The per-centage of candidates who did not run inconstituencies was only 8.5% and 3.3%respectively in the 2003 Scottish and Welshelections, and only 5.8% in the 2002 NewZealand election.

The historical data available indicate thatdouble candidacy was originally not aswidespread as it is today. The practiceseems to have resulted from lessons learnedover many elections. With experience, peoplerealized it was clearly preferable.

14


There is another factor that reduces thelikelihood of conflict between constituencymembers and list members: the possibilitythat in the course of a political career acandidate will switch from one categoryto the other. Hopefully, members will thinktwice before putting down the other cate-gory if it is highly likely that they oncebelonged to it or will later belong. Thislikelihood is far from small. In the Bundestag,19% of the members who sat between1949 and 2002 alternately held constituencyseats and list seats. In the six Länder withcomparable data, the proportions rangedfrom 8% to 24%. If one looks only at the53 Bundestag members whose careersexceeded 10,000 days, over half of themwere alternately constituency membersand list members. This possibility enabledChancellor Kohl to sit in the Bundestagdespite occasional defeats in his consti-tuency, without his authority being affected.

Women’s representation in parliaments

The percentage of women is high in parlia-ments that use the compensatory mixedsystem. The Bundestag that was elected in 2002 is 33% female. In Länder using the same system, the proportion rangesbetween 27% and 41%. The correspondingfigures for New Zealand, Scotland, andWales are respectively 28%, 39%, and 50%.Most of these percentages are slightlyhigher than the one for the 2003 Québecelection (30.4%).

The opinion that territory-wide compensa-tion is an absolute prerequisite for substantialwomen’s representation is not confirmed byexperience. On average, women make up35.7% of the six regionally subdivided parliaments. On average, they make up32.3% of the ten parliaments where com-pensatory seats are distributed for theentire territory. In the one state that recyclesthe best defeated candidates, womenaccount for only 22% of the total.

How should vacancies be filled betweengeneral elections?

With the first-past-the-post system (singleor double ballot), by-elections are the mostcommon way to replace members whoseseats have become vacant. With PR systems,the most common technique is to choosethe “next person on the list.”

The situation is more complex in a mixedsystem. In the Bundestag and in almost allLänder using personalized PR, a list memberis replaced by the next non-elected candi-date on that party’s list, with the revealingproviso that he or she must still be a memberof the party, in keeping with the spirit ofPR. A vacant constituency seat is also filledby the highest non-elected candidate onthe party list, so as to eliminate the needfor a by-election. This solution has theadvantage of preserving election-nightparty standings for the life of the Parliamentwhile sparing the cost of holding by-elections.It also helps preserve government stabilitywhen many vacancies occur within the caucusof a party or coalition with a very slim par-liamentary majority. Finally, replacementsmay be made without delay. On the otherhand, this solution makes it impossible fora newly chosen minister or party leader toenter Parliament before the next generalelection.

15


The Länder of Hesse and Rhineland-Palatinateuse a different method. If a seat (whethera constituency seat or a list seat) becomesvacant, it is filled by a substitute candidatefrom the same party who is chosen at thesame time as the primary candidate.

In New Zealand, Scotland, and Wales, thenext person on the list is used to fill anyvacancies that occur among list seats. Forconstituency seats, by-elections are preferred.

16


SECOND PART Simulations based on the 1998 and 2003 Québec elections

METHODOLOGYThe above analysis of elections in othercountries provided us with some interest-ing insights. We felt it necessary to verifythese insights by running simulations withQuébec election data.

In keeping with the aims of this study, wesought:

• to estimate the distribution of seatsamong the parties running in the 1998and 2003 elections, based on a compensa-tory mixed system with a constituency/listseat ratio of 60/40;

• to investigate the impact of differentvariables on party representation, on themagnitude of vote/seat distortions and,potentially, on the creation of excessseats;

• to see whether the different approacheswould perpetuate, reduce, or eliminatethe asymmetry that characterizes theoperation of the current electoral system;and

• to see which scenarios are mostfavourable to small parties.

To this end, we ran simulations. Differentmethods and techniques were applied tothe actual results of the last two Québecelections.

Constituency/list ratio

In the case of compensatory mixed systems,simulations pose a special challenge. Itseemed unrealistic to keep the current125 constituencies. With a 60/40 ratio, thiswould mean adding about 80 compensa-tory seats for a total of over 200 seats inthe National Assembly. No one seems tohave countenanced such figures if onegoes by proposals made in recent years.A total of 125 MNAs seemed to reflect the

most popular view and, for our purposes,this would mean subdividing the provinceinto much fewer constituencies. Now it sohappens that Québec has had 75 consti-tuencies on the federal electoral map sincethe last federal election in June 2004. Thismap is based on the 2001 census and wasprepared by an independent electoralboundaries commission, which held regionalhearings on an initial draft version. Its cons-tituencies differ very little in population,making it quite faithful to rep. by pop.

When the 1998 election results were trans-posed into these 75 constituencies, the outcome was 47 seats for the PQ, 28 forthe QLP, and none for the ADQ. When the2003 results were transposed, the outcomewas 47 seats for the QLP, 25 for the PQ,and 3 for the ADQ. This transposition wasused invariably for all of our simulations.The constituency seats were supplementedwith compensatory seats according to different scenarios.

In distributing the compensatory seats, welooked at many different combinations ofthree types of variable: territorial subdivision;seat distribution method; and computationtechnique.

Territorial subdivision

Compensation could be province-wide withall of Québec forming just one constituencyfor this purpose. If compensation is doneregionally, there must be basic single-memberconstituencies distributed among the regions.Five different regional subdivisions wereprepared by consolidating the 75 federalconstituencies in different ways. The provincewas thus subdivided into 26, 16, 13, and 4 regions, whereas with province-widecompensation all of Québec formed butone “region.” These subdivisions weremade strictly for purposes of computation.The 60/40 ratio was respected within eachregion, e.g., a region with a total of 5 MNAs

17


would have 3 constituencies and 2 compen-satory seats. All five territorial subdivisionsproduced a total of 124 seats: 75 consti-tuency seats and 49 compensatory seats.

Seat distribution method

Three methods for seat distribution weretested: German regional compensation;province-wide compensation followed byredistribution of party seats among theregions; and Scottish regional compensa-tion. With the first and third methods,compensatory seats are assigned separatelywithin each region. With the secondmethod, compensatory seats are assignedfirst at the provincial level. The seatsassigned to each party are then redistributedamong the regions according to the party’snumber of votes in each region.

Computation techniques

For computation purposes, the three maintechniques were used: the Sainte-Laguëtechnique; the D’Hondt technique; and the Largest Remainder technique. Each of them is compatible with the Germanmethod, but the Scottish method has todate been combined only with the D’Hondttechnique. For this study, computation procedures were designed (see Appendix VIof this report) to adapt the Scottish methodto the Sainte-Laguë and Largest Remaindertechniques. These procedures were submittedfor discussion to Prof. John Curtice, of theUniversity of Strathclyde in Scotland, whohas judged them to be correct.

Other variables

• Computations were based on the votesfor party candidates in the two Québecelections under study. We assumed thatvoting patterns would have remainedthe same.

• A 5% threshold was applied for Québecas a whole.

• No projection was made to determinethe personal identities of elected members.

OUTCOMESBy combining the five territorial subdivisionassumptions, the three seat-distributionmethods, and the three computation tech-niques, we produced 3 series of 14 outcomesfor a total palette of 42 scenarios.

For each simulation and for each election,a detailed table gives the distribution ofconstituency seats and compensatory seatsby party in each region, as well as thetotals for all of Québec. These tablesappear in the appendices of the report.

Using these tables, we prepared summarytables that bring together the total resultsfor each simulation. They give the totalnumber of seats assigned to each party (in absolute numbers and in percentages of the total) and the number of resultingexcess seats, if any. These tables are printedin the body of the text.

With these data, we designed tables thatgive the overall level of distortion producedby each simulation. We used the indexdeveloped by Michael Gallagher. Tableswere also prepared, showing the majoritybonus produced for each simulation, i.e.,the difference between the strongestparty’s percentage of the valid votes (inballots cast) and its percentage of theseats. Outcomes were also analyzed interms of the presence or absence of

18


regional monopolies, i.e., cases where aparty manages to win all of the availableseats in a region. The number of excessseats produced by each method, if likely to be produced, was also compiled. Wealso sought to see whether a methodwould or would not correct reversals ofparty standings, such as what happened in 1998. Finally, we verified the degree towhich the different approaches work symmetrically or asymmetrically.

KEY FINDINGS

Overall level of distortion and majoritybonus

The simulations produced outcomes thatlargely agree with the lessons from experi-ences in other countries. There were fourmain findings:

• Only electoral system reform will correctthe anomalies that have been pointedout in the current first-past-the-post sys-tem. Even with a complete redrawing ofthe electoral map, such as subdivision ofthe province into 75 constituencies, thedistortions remain considerable and theasymmetry is almost just as pronounced.

• All of the scenarios reduce the distor-tions, i.e., they produce outcomes thatare clearly more proportional to eachparty’s share of the popular vote thanthe current first-past-the-post system.

• The compensatory mixed system wouldend the regional monopolies of the twomain parties and encourage more partypluralism within each region.

• Finally, excess seats may be a serious disadvantage.

If we look at the degree of regional sub-division, the simulations indicate thatprovince-wide compensation would pro-duce a minimal level of distortion and vir-tually eliminate the majority bonus of thebig parties. As soon as compensation isdone within four large regions, majoritybonuses and higher distortions becomeapparent. At this point too, excess seatsbegin to appear. All three trends grow inmagnitude as the number of regions isincreased to 13, 16, and 26.

The effects of the three seat-distributionmethods may be summarized as follows:

German regional compensation proved tobe highly versatile. Even with 26 regions,the distortions and majority bonus may beminimized by using the Largest Remaindertechnique. Using the D’Hondt techniqueincreases the majority bonus. The Germanmethod’s main weakness is the high numberof excess seats it produces—as many as 11 for a total number of 135 seats in thelegislature. Analysis of the outcomes indicatesthat excess seats are a serious problem. First,the legislature may temporarily expand toa size that remains unforeseeable until theelection results are known. Second, theexcess seats also upset interregional balance,since some regions receive them and othersdo not. These seats arise almost alwaysoutside the Island of Montréal and thusincrease the relative weight of the otherregions. The greater the distortion betweenthe distribution of valid votes and the dis-tribution of constituency seats within aregion, the likelier the region will haveexcess seats, this being a sort of distortionbonus. If we wish to know which partiesbenefit, excess seats go more to the PQ(which is more popular outside Montréal)and never to the ADQ (which does notclearly dominate any region).

19


Regional compensation with redistributionof party seats among the regions producesvery similar outcomes for all scenarios,whatever the territorial subdivision or com-putation technique. This method tends toproduce a lower level of distortion andmajority bonus than do the other twomethods. To a lesser degree than Germanregional compensation, it also leads toexcess seats. The simulations reveal twoproblematic consequences of this method.First, each region’s final representationremains uncertain until election night. Anabove-average turnout may give a regionmore seats than expected whereas a rela-tively low turnout may deprive a region of one of the compensatory seats it wasentitled to by virtue of its population. Thechances are high (two out of three) thatthe total number of compensatory seatsassigned to a region will differ from theexpected number. Second, a region whoseturnout was well below average may forthis reason not receive any compensatoryseats.

Scottish regional compensation producesoutcomes similar to those of Germanregional compensation. In regions whereno excess seats arise, the two methods pro-duce identical results. Where they do arise,the Scottish method solves the problem to the detriment of the weakest parties.Distortions and majority bonus are a littlehigher than with German regional com-pensation, but this is the price to be paidto avoid increasing the total number ofelected members and upsetting the regionalbalance of representation.

As for computation techniques, distortionsand majority bonus reach a maximum levelwith the D’Hondt technique and are broughtback to a clearly lower level with the othertechniques. The impact of computationtechniques varies greatly with the degreeof territorial subdivision. The kind of tech-nique matters more if compensation isdone regionally and this factor will increasein importance as the number of regionsincreases. Conversely, the differences vanish if compensation is province-wide,regardless of whether this distribution isfollowed by redistribution among theregions. The simulations also reveal thatthe D’Hondt technique tends to producefewer excess seats than the other two.

Reversals of party standings and asymmetryof the different approaches

The electoral systems differ intrinsically intheir ability to represent parties in Parliamentin the order that voters choose. Reversalsof party standings may be perceived to bean undesirable anomaly. They are notunusual with the first-past-the-post systemand they might arise with a PR system. Forany compensatory system, the Sainte-Laguëand Largest Remainder techniques will produce reversals more often than will theD’Hondt technique. No reversals appearwith province-wide compensation or a 4-region subdivision, but they occur moreoften with other territorial subdivisions.

20


One of the reproaches made against thefirst-past-the-post system, in Québec, isthat it treats the two main parties differ-ently, even when they are equal in thepopular vote. In the 1998 and 2003 elections,the winning party won 76 of the 125 seats.To achieve this result, the QLP needed a 13-point lead in the popular vote in 2003.The PQ achieved the same standing in1998 with a half-point less than its mainrival. In 1994, with a 13,000-vote lead, thePQ won 30 seats more than the QLP did. In1998, with a 27,000-vote lead, the QLP won28 seats less than the PQ did. This type ofspread has been a constant of the Québecpolitical landscape since 1944, and it large-ly explains why a party won three elections(1944, 1966, and 1998) with fewer votesthan its main rival. The same anomaly wouldhave occurred in the 1995 referendumunder a first-past-the-post system. Of course,there were no serious consequencesbecause only the popular vote counted.

In neutral language, the system worksasymmetrically. In more polemical lan-guage, it is biased systematically towardone of the two big parties and against theother. This effect can be measured by projec-tions based on election results. A commonlyused method in the academic literature startswith the results of a given election andgives a party 1% less (and its rival 1%more) in each constituency, with the votefor the other parties remaining identical.The number of total seats each partywould win is then computed.

When projections are made from the 2003results, very clear trends emerge:

• When applied to the current 125 consti-tuencies, the first-past-the-post systemworks very asymmetrically, as could beseen in 1998, and this effect persisted in2003. For each percentage of the vote,the QLP would always win fewer seatsthan would the PQ with the same shareof the popular vote. With 47% of thevote, the QLP would win 60.8% of theseats in 2003 and the PQ 75%.

• When applied to the current 75 federalconstituencies, the first-past-the-post system scarcely works less asymmetrically.The asymmetry would survive even acomplete overhaul of the electoral mapas long as the current electoral system ismaintained.

• With Scottish regional compensation, theasymmetry is very greatly reduced. In oursimulations, this approach always gavethe most seats to the party with themost votes.

• The asymmetry is even weaker with Germanregional compensation and with moder-ate regional PR. It is reduced almost tozero with province-wide compensation.

• The direction of the bias almost alwaysfavours the PQ. With equal shares of thepopular vote, the PQ tends to win moreseats than does the QLP. This is true forall of the approaches, albeit with signifi-cant differences.

21


Thresholds and small parties

The absence of any threshold for represen-tation is a prerequisite for representationof the small parties, since these partiesreceived less than 5% of the valid votes inthe last two elections. The absence of anythreshold does not necessarily mean thatthe small parties would be automaticallyrepresented in the National Assembly.Province-wide compensation, followed ornot by regional redistribution would bethe ideal scenario for these parties, giventhe number of votes they actually receivedin the last two elections.

Identity of list members

Understandably, it was impossible to deter-mine who would be elected from the partylists. If list seats went to the best defeatedcandidates, we could at least predict whichconstituencies would receive such seats.With each party’s defeated candidatesplaced in decreasing order, by either numberof votes or percentage of the vote, thehighest candidates on the list would bedeclared “elected” until each party hadreceived all of the compensatory seats itwas entitled to.

When applied to province-wide compensa-tion, the simulation reveals that compensa-tory seats would go preferentially to someconstituencies and not to others. For allintents, Île de Montréal and Outaouaiswould receive none, because the winningmargins of elected constituency candidatesare higher there. The defeated candidatesthus have too few votes to qualify for com-pensatory seats at the provincial level. Onthe other hand, constituencies in the otherregions, where the fights are closer, wouldtake almost all of the compensatory seats,since the defeated candidates have relativelymore votes.

When applied to the 26-region territorialsubdivision, for the 1998 and 2003 elections,this method would not lead to any seriousregional distortion. Each region wouldreceive an appropriate number of compen-satory seats. Another kind of problem,however, would arise in regions where oneof the parties very clearly outperforms itsrivals. This party may not only win all threeof the constituencies up for grabs but alsobe entitled to a fourth seat, given its highproportion of the vote in the region (e.g.,70%). The winning party has no “bestloser” to recycle, simply because it has nolosers at all. This problem would likely beendemic in Québec. It appears in 7 regionsand in 1998 would have involved 8 seats in5 regions (6 seats in 2003).

22


THIRD PART Conclusion and recommendations

ADVANTAGESWhen applied to Québec, the compensa-tory mixed electoral system would haveseveral advantages. It would reduce therecurring problem of distortions. It wouldbring a party’s percentage of the seats closerto its percentage of the vote. All partieswould be more equitably represented andthe doors of Parliament may be opened tonew parties with a minimum level of popularsupport. There would be more political andparty pluralism not only in the NationalAssembly but also in the regions, and itwould become exceptional for a party tomonopolize representation of a specificregion. A compensatory system wouldwork much more symmetrically and equi-tably than does the current electoral system.A reversal of standings for the main partieswould be much less possible and it wouldbecome extremely unlikely for a party, asin 1998, to win an absolute majority of theseats without even receiving a plurality ofthe popular vote. In addition, these goalswould be reached without losing whatmany consider to be one of the key advan-tages of the current electoral system: theexistence of single-member constituenciesthat help preserve a closer bond with theelectorate.

ELECTORAL SYSTEM CHARACTERISTICSNo one is pushing for an increase in thenumber of MNAs. At 125, the NationalAssembly has more members than anyother Canadian legislature. A 60/40 ratioseems to provide the most balance. Anyratio that is more biased to compensatoryseats would imply a concomitant decreasein the number of constituencies and anincrease in their size. Any ratio that is morebiased to constituencies would increase the

number of excess seats. By choosing a 60/40ratio, we would create about 75 cons-tituency seats and 50 compensatory seats.This would enable us to use the federalelectoral map for election of constituencyMNAs and thereby hasten introduction ofthe new system, in particular for the nextelection.

Under the new system, constituency memberswill be the most numerous MNAs. Howshould they be elected? Under all of thecompensatory mixed systems we examined,constituency members are elected by a plu-rality of the vote. Conceivably, this electioncould take place through a double ballot,as in France, or through alternative voting,as in Australia.

If a double-ballot system is used, we must decide which figures will be used tocalculate the overall seat distribution and,thus, each party’s final number of seats.Depending on whether one uses the resultsof the first ballot or the “decisive ballot”(the ballot where one of the parties winsthe constituency), small parties will eitherbenefit or be excluded when the final PRcomputation is made, since political realismor legal requirements will force them towithdraw their candidates before the secondballot. The same dilemma arises with thealternative voting system. Should one usethe distribution of first preferences, i.e.,the ones counted on election night, or thefinal distribution after transfer of prefer-ences? This analysis leads to the followingconclusion: if the double-ballot system orthe alternative voting system is used toelect constituency MNAs, compensationwill have to be based on a separate supple-mentary vote that the voter casts for apolitical party.

23


By replacing the current electoral systemwith the double-ballot or alternative votingsystem, we would clearly face additionalcomplications in electing constituencyMNAs. It is less clear what advantages wewould gain with either approach. Therewould be no expected gain in terms ofincreased PR. Moreover, the double ballotand alternative voting do not reduce thedistortions of the first-past-the-post system.They may even aggravate them. Theseways of voting are unknown in Québec,and the second type alone would require a campaign to educate the public becauseof the new kind of ballot it entails. Bothsystems would have the effect of guaran-teeing that more constituency MNAsreceive an absolute majority of the vote.Nothing, however, suggests that voters feelan MNA is less legitimate just because heor she was elected with less than half thevotes. How many voters know this detailabout their own MNA?

Some of the 42 scenarios can be ruled outon self-evident grounds. Thus, regionalcompensation, with redistribution of partyseats to the regions, probably suits a highlyproportionalist society where concern for a mathematically exact result trumps allother considerations. But it would leaveQuébec’s political players and citizens inthe dark until election night as to the numberof compensatory seats that will go to thedifferent regions. The chances are twotimes out of three that this number willdiffer from the expected total and someregions may not receive any compensatoryseats—a sure way to get their residentsdeeply upset.

German regional compensation tries to correct this fault by assigning each region a fixed number of compensatory seats. Butlike the previous method it can lead toexcess seats. Such seats are, the author

feels, a major problem. They upset theregional balance of representation andcause the total number of elected membersto vary from one election to the next. Inseveral 1998 election simulations, excessseats gave a plurality of the seats to theparty that came second in the popularvote. This is in fact one of the anomalies to be corrected.

Scottish regional compensation seemspreferable. It does not prevent a partyfrom winning more constituencies in aregion than it is entitled it, but it ensuresthat the region’s total number of seats willremain unchanged, at the risk of penalizingother parties by reducing their representation.The simulations showed that this breach ofproportionality would have had no majorimpact on the 2003 election. With a 26-regionsubdivision of the province, the Scottishmethod would have given the strongestparty 53% of the seats and produced noexcess seats, versus 52% for the Germanmethod—at a cost of nine excess seats.

When we choose the geographic frameworkfor compensation, there is a tension betweentwo opposing principles of representation.Some people would give the priority to territory: for them, it is essential to electthe list MNAs in relatively homogeneousand recognizable territorial entities. Thisview naturally leads to subdividing theprovince into 26 regions, as regional PRproponents had concluded in their time.This subdivision means, however, that someclassic distortions will persist in parliamentaryrepresentation of the parties, to the benefitof the strongest parties.

Consequently, some people would insteadgive priority to mathematical exactness ofrepresentation, i.e., more or less total pro-portionality. The most appropriate solutionin this case would be province-wide com-pensation, with all of Québec forming just

24


one constituency for computation purposes.If we pursue this same logic, no minimumthreshold should keep small parties out ofthe National Assembly.

Which of these two priorities should wechoose? This debate is a completely politicalone in which an expert opinion can onlyhelp clarify the options. These optionsshould be submitted to the governmentand the public for purposes of discussionand decision making. The author feels itpreferable to choose an electoral subdivisionwith small regions, at the risk of bendingthe principle of proportionality, and a 26-region subdivision appears to him to bethe most satisfactory working model. Ofcourse, province-wide compensation wouldreduce the distortions almost to zero. Butsuch a solution may seem undesirable tomost Québeckers, who are accustomed tolinking an MNA to a territory and who,though interested in more electoral fairness,would not go so far as to allow an infiniteproliferation of political parties. A publicdebate would reveal the opinions of MNAsand the public on this subject.

Electoral reform also includes the issue ofhaving two votes. Although a second votehas some advantages, three considerationsargue for a single vote. It is the simplestoption from the voter’s standpoint. It pro-duces fewer rejected ballots. Finally, as hasbeen seen in Italy, a second vote may pavethe way for electoral practices that destroythe very essence of the compensatory system.Proponents of a second vote should suggestmechanisms that will exclude this type ofabuse.

To be on the safe side, lawmakers couldrequire that, for the first election underthe new system, proportionality will bebased on votes for constituency candidates.This is the form in which personalized PRwas introduced into Germany. The harm it

does to small parties is minor. They onlyhave to field candidates in all constituencies(which will be reduced to 75 in number) tomake the most of their potential votes.

Compensatory seats can be filled by recyclingthe best defeated constituency candidatesor by using a closed list. These seem to bethe only options possible.

The second option prevails in almost allsocieties with compensatory mixed systems.From the candidates’ standpoint, it has thedisadvantage of setting a predeterminedorder, which may be flattering for thosenear the top, but not so for those furtherdown. It has the advantage of ensuringmore security for party leaders and giving,for example, more priority to women can-didates. Moreover, it enables an electedmember to preserve a close tie to a regionas a whole, while providing an officialrationale for working in a specific consti-tuency if the member so chooses.

It does not seem advisable for now toallow voters to express preferences forsome list candidates. Introduction of thisoption could be considered in the future.

If the compensatory system is to workproperly, constituency MNAs and list MNAsmust not form two classes in permanentconflict with each other. Some precautionsare therefore in order. Notably, each partyshould fill the top places on its lists with its constituency candidates for that region.This would preserve the advantage of thesystem of recycling the best defeated can-didates, i.e., it would guarantee that theimmense majority of MNAs, whatever theirstatus, will have passed the test of electoralcampaigning in a constituency. Be it forpay, allowances, or responsibilities, theprinciple of equality between constituencyMNAs and list MNAs should prevail.

25


With regard to list preparation, the govern-ment should trust the parties and the votersto avoid any abuses. It would be difficult tofind a uniform approach that would satisfyall political parties. By not meddling in theinternal workings of political parties, Québecwould only be imitating the overwhelmingmajority of democratic countries.

To replace a list MNA who has died orresigned, we should use the highest non-elected candidate on the list. Parties shouldbe asked to submit lists with enough candi-dates to deal with all possible outcomes,on the understanding that the seat willremain vacant if no one eligible is on thelist. If a constituency seat becomes vacant,it seems preferable to call a by-election.

26


For information, please contact your Member of the National Assembly:

• by ordinary mail:

Hôtel du Parlement Québec (Québec) G1A 1A4

• by e-mail, using the list of the Members’addresses on the National Assembly's Web site at: http://www.assnat.qc.ca

You may also get in touch with the Government House Leader and Minister for the Reform of Democratic Institutions:

• by ordinary mail:

Office of the Government House Leader Minister for the Reform of Democratic InstitutionsÉdifice Pamphile-Le May1045, rue des Parlementaires, bureau 1.39Québec (Québec) G1A 1A4

• by e-mail:[email protected]

Finally, you may consult the Web site of the Secrétariat à la réforme des institutions démocratiques at: www.institutions-democratiques.gouv.qc.ca

28


in search of a compensatory mixed electoral system … › ... · 2016-10-12 · to make the final...

Documents