Social and Political Context

PR Related Issues

• Thresholds
• Apparentemente
• Open, Closed and Free Lists
• District Magnitude

Cost Considerations

• The Drawing of Electoral Boundaries
• The Design of Ballot Papers
• Voter Education
• The Number and Timing of Elections

The Count

Electoral system design consultants rightly shy away from the "one-size-fits-all" approach of recommending one system for all contexts. Indeed, when asked to identify their 'favourite' or 'best' system, constitutional experts will say "it depends" and the dependants are more often than not variables such as:

• What does the society look like?
• How is it divided?
• Do ethnic or communal divides dovetail with voting behavior?
• Do different groups live geographically inter-mixed or segregated?
• What is the country's political history?
• Are they an established democracy, a transitional democracy, or a re-democratising state?
• What are the broader constitutional arrangements that the legislature is working within?

When assessing the appropriateness of any given electoral system for a divided society, three variables become particularly salient:

Knowledge of the nature of societal division is paramount-the nature of group identity, the intensity of conflict, the nature of the dispute, and the spatial distribution of conflictual groups. The nature of the political system, i.e., the nature of the state, the party system, and the overall constitutional framework. The process which led to the adoption of the electoral system, i.e. was the system inherited from a colonial power, was it consciously designed, was it externally imposed, or did it emerge through a process of evolution and unintended consequences?

The Nature of Group Identity

Appropriate constitutional design is ultimately contextual and rests on a nation's unique social nuances. Division within a society is revealed in part by the extent to which ethnicity correlates with party support and voting behavior. That factor will often determine whether institutional engineering can dissipate ethnic conflicts or merely contain them. There are two dimensions to the nature of group identity:

one deals with foundations-is the society divided along racial, ethnic, ethno-nationalistic, religious, regional, linguistic, lines?

while the second deals with how rigid and entrenched such divisions are.

Scholarship on the later subject has developed a continuum with the rigidity of received identity (primordialism)on one side and the malleability of constructed social identities (constructivist) on the other.

Intensity of Conflict

A second variable, in terms of the nature of any given conflict and its susceptibility to electoral engineering, is simply the intensity and depth of hostility between the competing groups. It is worth remembering that, although academic and international attention is naturally drawn to extreme cases, most ethnic conflicts do not degenerate into all-out civil war. While few societies are entirely free from multiethnic antagonism, most are able to manage to maintain a sufficient degree of mutual accommodation to avoid state collapse. There are numerous examples of quite deeply divided states in which the various groups maintain frosty but essentially civil relations between each other despite a considerable degree of mutual antipathy-such as the relations between Malays, Chinese and Indians in Malaysia. There are other cases (e.g., Sri Lanka)-where what appeared to be a benign inter-ethnic environment and less pronounced racial disputes nonetheless broke down into violent armed conflict- but where democratic government has nonetheless been the rule more than the exception. There are also cases of utter breakdown in relations and the 'ethnic cleansing' of one group by another typified, most recently and horribly, by Bosnia.

The Nature of the Dispute

Electoral system design is not merely contingent on social issues but also, to some extent, on cultural differences as well. The classic dispute is that of group rights and status in a multiethnic democracy-a system characterised both by democratic decision-making institutions and by the presence of two or more ethnic groups. This is defined as a group of people who see themselves as a distinct cultural community; who often share a common language, religion, kinship, and/or physical characteristics (such as skin colour); and who tend to harbour negative and hostile feelings towards members of other groups. The majority of this paper deals with this fundamental division of ethnicity.

Other types of disputes often dovetail with ethnic ones, however. If the issue that divides groups is resource-based, for example, then the way in which the national parliament is elected has particular importance since disputes are managed through the central government allocation of resources to various regions and peoples. In this case, an electoral system, which facilitated a broadly inclusive parliament, might be more successful than one, which exaggerated majoritarian tendencies, or ethnic, regional, or other divisions. This requirement would still hold true if the dispute was primarily cultural, such as protecting minority languages and culturally specific schools. Other institutional mechanisms, such as cultural autonomy and minority vetoes, would be at least as influential in alleviating conflict.

Disputes over territory often require innovative institutional arrangements that go well beyond the positive spins that electoral systems can create. In Spain and Canada, asymmetrical arrangements for respectively, the Basque and Quebec regions, have been used to ease calls for secession, while federalism has been promoted as an institution of conflict management in countries as diverse as Germany, Nigeria, South Africa, and Switzerland.

Spatial Distribution of Conflictual Groups

When looking at different electoral options, a final consideration concerns the spatial distribution of ethnic groups, particularly their relative size, number, and degree of geographic concentration or dispersion. The geographic location of conflicting groups is often related to the intensity of conflict between them. Frequent inter-group contact from geographical intermixture may increase mutual hostility, but it can also act as a moderating force against the most extreme manifestations of ethnic conflict. Familiarity may breed contempt, but it also breeds a certain degree of acceptance as well. Intermixed groups are therefore less likely to be in a state of civil war than those that are territorially separated from each other. Conversely, territorial separation is sometimes the only way to manage the most extreme types of ethnic conflict-that which requires some type of formal territorial devolution of power or autonomy. In the extreme case of 'ethnic cleansing' in Bosnia, areas which previously featured highly intermixed populations of Serbs, Croats, and Muslims are now predominantly monoethnic.

Understanding the demographics of any ethnic conflict is particularly important for attempts at institutional remedies. The number and distribution of ethnic groups is a key variable for electoral engineering for divided societies. According to Lijphart, the optimal number of 'segments' is three or four, and conditions become progressively less favourable as more segments (or groups) are added. Another approach, by contrast, requires a degree of proliferation of ethnic groups (or, at least, ethnic parties) to present the essential preconditions for vote-pooling to take place. Chances for success will typically improve as the number of segments increase. Another factor is the relative size of ethnic groups: consociationalism favours groups of roughly equal size, although 'bicommunal systems', in which two groups of approximately equal sizes coexist, can present one of most onfrontationalist formulas of all. For centripetalism the crucial variable is not size so much as the geographic concentration or dispersion of ethnic groups. When ethnic groups are geographically concentrated in one or two areas, any electoral strategy for conflict management should be tailored to the realities of political geography. Territorial prescriptions for federalism or other types of devolution of power will usually be a prominent concern, as will issues of group autonomy. Indigenous and/or tribal groups tend to display a particularly strong tendency towards geographical concentration. African minorities, for example, have been found to be more highly concentrated in single contiguous geographical areas than minorities in other regions. This means that a single ethnopolitical group will control many electoral constituencies and informal local power bases. This has considerable implications for electoral engineers: any system of election that relies on single-member electoral districts will likely produce 'ethnic fiefdoms' at the local level. Minority representation and/or power sharing under these conditions would probably require some form of multi-member district system-particularly Proportional Representation (PR).

Contrast this with colonial settlements or labour importation such as the vast Chinese and Indian diasporas found in some Asia-Pacific-Singapore, Fiji, Malaysia; and Caribbean-Guyana, Trinidad, and Tobago-countries, in which ethnic groups are more widely inter-mixed and, consequently, have more day-to-day contact. Here, ethnic identities are often mitigated by other disputes, and electoral districts are likely to be ethnically heterogeneous.

Therefore, centripetal electoral systems, which encourage parties to seek the support of various ethnic groups, may well break down inter-ethnic antagonisms and promote the development of broad, multi-ethnic parties. After a year-long review of their Constitution, Fiji has just adopted the Alternative Vote AV) as part of a new, non-racial constitution for this very reason.

Another scenario is where there are so many ethnic groups that some types of electoral systems are naturally precluded. Such a social structure typically revolves around small, geographically-defined tribal groups-a relatively unusual composition in Western states, but common in some areas of central Africa and the South Pacific. This typically requires Single-Member Representation to function effectively. In the extreme case of Papua, New Guinea, there are several thousand competing clan groups speaking over 800 distinct languages. Any attempt at proportional representation in such a case would be almost impossible, as it would require a parliament of several thousand members (and, because parties are either weak or non-existent in almost all such cases, the list-PR system favoured by consociationalists would be particularly inappropriate). This dramatically curtails the range of options available to electoral engineers.

Nature of the State

Institutional prescriptions for electoral engineering need to be alert to the different political dynamics that distinguish transitional democracies from established ones. Transitional democracies, particularly those moving from a deep-rooted conflict situation, typically have a greater need for inclusiveness and a lower threshold for the robust rhetoric of adversarial politics, than their established counterparts. Similarly, the stable political environments of most Western countries-where two or three main parties can often reasonably expect regular periods in office via alternation of power or shifting governing coalitions-are very different from the type of zero-sum politics which often characterise divided societies. This is one of the reasons that 'winner take all' electoral systems such as First-Past-The-Post (FPTP) have so often been identified as a contributor to the breakdown of democracy in the developing world: such systems tend to lock out minorities from parliamentary representation and, in situations of ethnically-based parties, can easily lead to the total dominance of one ethnic group over all others. Democracy, under these circumstances, can quickly become a situation of permanent inclusion and exclusion, a zero-sum game, with frightening results.

For this reason, many scholars see a need for some type of power-sharing government featuring all significant groups as an essential part of the transition from authoritarian rule to democracy. The power-sharing model is usually associated with PR, as this is the surest way of guaranteeing proportional results and minority representation. It is instructive to note that almost all of the major transitional elections in recent years have been conducted under some form of PR. In fact, recent transitional elections in Chile (1989), Namibia (1989), Nicaragua (1990), Cambodia (1993), South Africa (1994), and Mozambique (1994) all used a form of regional or national list PR for their founding elections. Some scholars have identified the choice of a proportional rather than a majoritarian system as being a key component of their successful transitions to democracy. By bringing minorities into the process and fairly representing all significant political parties in the new legislature, regardless of the extent or distribution of their support base, PR has been seen as being an integral element of creating an inclusive and legitimate post-authoritarian regime.

There is also mounting evidence that while large-scale list PR is an effective instrument for smoothing the path of democratic transition, it is less effective at promoting democratic consolidation. Developing countries, in particular those which have made the transition to democracy under list PR rules, have increasingly found that the large, multi-member districts required to achieve proportional results also create difficulties with political accountability and responsiveness between elected politicians and voters. Democratic consolidation requires the establishment of a meaningful relationship between the citizen and the state, and many new democracies-particularly those in agrarian societies-have much higher demands for constituency service at the local level than they do for representation of all ideological opinions in the legislature. It is therefore increasingly being argued in South Africa, Cambodia, and elsewhere that the choice of a permanent electoral system should encourage a high degree of geographic accountability, by having members of parliament who represent small, territorially-defined districts who service the needs of their constituency, to establish a meaningful relationship between the rulers and the ruled. While this does not preclude all PR systems-there are many ways to combine single-member districts with proportional outcomes-it does rule out the national list PR systems.

Nature of Party System

The conventional wisdom amongst electoral scholars is that majoritarian electoral rules encourage the formation of a two-party system (and, by extension, one-party government), while Proportional Representation leads to a multi-party system (and coalition government). While there remains agreement that majority systems restrict the range of legislative representation and PR systems encourage it, the conventional wisdom of a causal relationship between an electoral system and a party system is becoming dated. In recent years, FPTP has facilitated the fragmentation of the party system in established democracies such as Canada and India, while PR has seen the election of what look likely to be dominant single-party regimes in Namibia, South Africa, and elsewhere.

One of the basic precepts of political science is that politicians and parties will make choices about institutions such as electoral systems that they believe will benefit themselves. Different types of party systems will thus tend to produce different electoral system choices. The best-known example of this is the adoption of PR in continental Europe in the early years of this century. The expansion of the franchise and the rise of powerful new social forces, such as the labour movement, prompted the adoption of systems of PR that would both reflect and restrain these changes in society. More recent transitions have underlined this 'rational actor' model of electoral system choice. Thus, threatened incumbent regimes in Ukraine and Chile adopted systems which they thought would maximise their electoral prospects: a two-round runoff system which over-represents the former Communists in the Ukraine, and a unusual form of PR in two-member districts which was calculated to overrepresent the second-placed party in Chile. An interesting exception that proves the validity of this rule was the ANC's support for a PR system for South Africa's first post-apartheid elections. Retention of the existing FPTP system would almost undoubtedly have seen the over-representation of the ANC as the most popular party, but it would also have led to problems of minority exclusion and uncertainty. The ANC made a rational decision that their long-term interest would be better served by a system which enabled them to control their nominated candidates and bring possibly destabilising electoral elements 'into the tent' rather than giving them a reason to attack the system itself.

Overall Constitutional Framework

The efficacy of electoral system design should be judged in the broader constitutional framework of the state. This paper concentrates on elections that constitute legislatures. The impact of the electoral system on the membership and dynamics of legislatures will always be significant, but the electoral system's impact upon political accommodation and democratization more generally is tied to the amount of power beholden in the legislature and that body's relationship to other political institutions. The importance of electoral system engineering is heightened in centralised, unicameral parliamentary systems, and is maximised when the legislature is constitutionally obliged to produce an executive cabinet of national unity drawn from all significant parties that gain parliamentary representation.

Similarly, the efficacy of electoral system design is incrementally diminished as power is eroded away from the parliament. Thus, a number of constitutional structures will proportionately distract attention away from elections to the legislature and will require the constitutional designer to focus on the inter-relationships between executives and legislatures; between upper and lower houses of parliament; and between national and regional and local government. This is not to diminish the importance of electoral systems for these other institutions (how to elect presidents and federal legislatures); rather, it highlights how constitutional engineering becomes increasingly complex as power is devolved away from the centre. Each of the following institutional components of the state may fragment the focal points of political power and thus diminish the significance of electoral system design on the overall political climate:

• a directly elected president;
• a bi-cameral parliament with a balance of power between the two houses;
• a degree of federalism and/or regional asymmetrical arrangements.

Political institutions shape the rules of the game under which democracy is practised, and it is often argued that the easiest political institution to be manipulated, for good or for bad, is the electoral system. This is true because in translating the votes cast in a general election into seats in the legislature, the choice of electoral system can effectively determine who is elected and which party gains power. Even with exactly the same number of votes for parties, one electoral system might lead to a coalition government while another might allow a single party to assume majority control. The two examples below illustrate how different electoral systems can translate the votes cast into dramatically different results.

But a number of other consequences of electoral systems go beyond this primary effect. The type of party system which develops, in particular the number and the relative sizes of political parties in parliament, is heavily influenced by it. So is the internal cohesion and discipline of parties: some systems may encourage factionalism, where different wings of one party are constantly at odds with each other, while another system might encourage parties to speak with one voice and suppress dissent. Electoral systems can also influence the way parties campaign and the way political elites behave, thus helping to determine the broader political climate; they may encourage, or retard, the forging of alliances between parties; and they can provide incentives for parties and groups to be broad-based and accommodating, or to base themselves on narrow appeals to ethnicity or kinship ties. In addition, if an electoral system is not considered Fair and does not allow the opposition to feel that they have the chance to win next time around, an electoral system may encourage losers to work outside the system, using non-democratic, confrontationalist and even violent tactics. And finally the choice of electoral system will determine the ease or complexity of the act of voting. This is always important, but becomes particularly so in societies where there are a substantial number of inexperienced or illiterate voters.

However, it is important to note that a given electoral system will not necessarily work the same way in different countries. Although there are some common experiences in different regions of the world, the effects of a certain electoral system type depends to a large extent upon the socio-political context in which it is used. Electoral system consequences depend upon factors such as how a society is structured in terms of ideological, religious, ethnic, racial, regional, linguistic, or class divisions; whether the country is an established democracy, a transitional democracy, or a new democracy; whether there is an established party system, whether parties are embryonic and unformed, and how many "serious" parties there are; and whether a particular party's supporters are geographically concentrated together, or dispersed over a wide area.

Let us take a hypothetical election (of 25,000 votes contested by two political parties) run under two different sets of electoral rules: a plurality-majority First Past The Post system with five single member districts, and a List PR election with one large district.

Key: P-M= Plurality-Majority system (FPTP), PR = Proportional Representation system.

In our example, Party A with 43% of the votes wins far fewer votes than Party B (with 57%) but under a Plurality-Majority system they win four out of the five seats available. Conversely, under a proportional system Party B wins more seats (three) against two seats for Party A. This example may appear extreme but similar constituency results occur quite regularly in plurality-majority elections.

In our second example the distribution of the votes is changed and there are now five parties contesting the election, but the two hypothetical electoral systems remain the same.

Key: P-M= Plurality-Majority system (FPTP), PR = Proportional Representation system (using the Largest remainder method of seat allocation with a Hare quota).

In the second example five parties are competing. Under the PR system, every party wins a single seat despite the fact that Party A wins nearly three times as many votes as Party E. Under a FPTP system the largest Party (A) would have picked up a majority of the five seats with the next two highest polling parties (B and C) winning a single seat each. The choice of electoral system thus has a dramatic effect on the composition of the parliament and, by extension, the government.

The consequences of different proportional representation systems for parliamentary and government representation are not merely influenced by the "type" of PR system used but also a number of other technical issues concerning the design of the PR electoral system, see PR Systems. File Thresholds details the important impact of thresholds for the parliamentary representation of political parties. File Apparentement discusses the opportunities for parties to come together for the sake of pooling their votes for seat allocation (apparentement). File Open, Closed and Free Lists looks at the voter's ability to choose between candidates and parties on a PR ballot paper-whether the lists are "open, closed, or free." While District Magnitude analyzes the crucial variable of "district magnitude," or how many legislators are elected from each district.

Thresholds

All electoral systems have thresholds of representation: that is, the minimum level of support which a party needs to gain representation, either legally imposed (formal), or merely mathematically de-facto (effective). In some cases, these thresholds are a by-product of other features of the electoral system, such as the number of seats to be filled and the number of parties or candidates contesting the election, and are thus categorized as "effective" thresholds. In other cases, however, these thresholds are written into the electoral law, which defines the PR system, and are therefore "formal".

In Germany, New Zealand, and Russia, for example, there is a five percent threshold: parties which fail to secure five percent of the vote are ineligible to be awarded seats from the PR lists, see Germany: The Original Mixed Member Proportional System, see New Zealand: A Westminster Democracy Switches to PR. This provision had its origins in the German desire to limit the election of extremist groups, and is designed to stop very small parties from gaining representation. However, in both Germany and New Zealand there exist "back-door" routes for a party to be entitled to seats from the lists; in the case of New Zealand a party must win at least one constituency seat, and in the case of Germany three seats, to by-pass the threshold requirements. In Russia in 1995 there were no "back-door" routes, and almost half of the party-list votes were wasted.

Elsewhere, legal thresholds range from 0.67 percent in the Netherlands to 10 percent in the Seychelles. Parties, which gain less than this percentage of the vote, are excluded from the count. In all of these cases the existence of a formal threshold tends to increase the overall level of disproportionality, because votes for those parties who would otherwise have gained representation are wasted. In Poland in 1993, even with a comparatively low threshold of five percent, over 34 percent of the votes were cast for parties, which did not surmount it. But in most other cases thresholds have a limited effect on overall outcomes, and some electoral experts therefore see them as unnecessary and often arbitrary complications to electoral rules, which in most cases are best avoided.

Apparentement

High effective thresholds can serve to discriminate against small parties- indeed, in some cases this is their express purpose. But in many cases an in-built discrimination against smaller parties is seen as undesirable, particularly in those cases where several small parties with similar support bases "split" their combined votes and consequently fall beneath the threshold, when one aligned grouping would have gained enough combined votes to have won some seats in the legislature. To get around this problem, many countries that use list PR systems also allow small parties to group together for electoral purposes, thus forming a "cartel" or apparentement to contest the election. This means that the parties themselves remain as separate entities, and are listed separately on the ballot paper, but that votes gained by each party are counted as if they belonged to the entire cartel, thus increasing the chances that their combined vote total will be above the threshold and hence that they may be able to gain additional representation. This device is a feature of a number of List PR systems in continental Europe, Chile before 1973, Brazil after 1979, and in Uruguay, Argentina, and Israel.

Open, Closed and Free Lists

There are a number of important variations in ways of voting between the various List Proportional Representation (PR) systems. One of the most important is whether lists are open, closed, or free in terms of the ability of electors to vote for a preferred candidate as well as for a party.

The majority of List PR systems in the world are closed, meaning that the order of candidates elected by that list is fixed by the party itself, and voters are not able to express a preference for a particular candidate. The List PR system instituted for the first democratic South African elections in 1994 was a good example of a closed list The ballot paper contained the party names and symbols, and a photograph of the party leader, but no names of individual candidates. Voters simply chose the party they preferred; the individual candidate elected as a result was pre-determined by the parties themselves. This meant that parties could include some candidates (perhaps members of minority ethnic and linguistic groups, or women) who might have had difficulty getting elected otherwise.

One negative aspect of closed lists is that voters have no say in determining who the representative of their party will be. Closed lists are also extremely unresponsive to changes in events. In East Germany's pre-unification elections of 1990, the top-ranked candidate of one party was exposed as a secret-police informer only four days before the election, and immediately expelled from the party; but because lists were closed, electors had no choice but to vote for him if they wanted to support his former party.

Many of the List PR systems used in continental Europe therefore use open lists, in which voters can indicate not just their favoured party, but their favoured candidate within that party. In most of these systems the vote for a candidate as well as a party is optional and, because most voters plump for parties rather than candidates, the candidate-choice option of the ballot paper often has little effect. But in some cases (Finland is one-see Finland: Candidate Choice and Party Proportionality) this choice becomes highly important, because people must vote for candidates, and the order in which candidates are elected is determined by the number of individual votes they receive. While this gives voters much greater freedom over their choice of candidate, it also has some less desirable side effects. Because candidates from within the same party are effectively competing with each other for votes, this form of open list can lead to intra-party conflict and fragmentation. It also means that the potential benefits to the party of having lists, which feature a diverse slate of candidates, can be overturned. In open-list PR elections in Sri Lanka for example, the attempts of major Sinhalese parties to include minority Tamil candidates in winnable positions on their party lists have been quashed because many voters deliberately voted for lower-placed Sinhalese candidates instead.

Some other devices are used in a small number of jurisdictions to add additional flexibility to open-list systems. In Luxembourg and Switzerland electors have as many votes as there are seats to be filled, and can distribute them to candidates either within a single party list or across several party lists as they see fit, see Switzerland. The capacity to vote for more than one candidate across different party lists (known as panachage), or to cast more than one vote for a single highly-favoured candidate (known as cumulation), both provide an additional measure of control to the voter and are categorized here as free list systems.

District Magnitude

There is near-universal agreement among electoral specialists that the crucial determinant of an electoral system's ability to translate votes cast into seats won proportionally is the district magnitude; i.e., the number of members to be elected in each electoral district. Under a single-member system such as First Past The Post (FPTP), or the Two-Round System (TRS), there is a district magnitude of one; voters are electing a single representative. Under a multi-member system, by contrast, there will by definition be more than one member elected in each district. Under any proportional system, the number of members to be chosen in each district determines, to a significant extent, how proportional the election results will be.

The systems, which achieve the greatest degree of proportionality, will utilise very large districts, because such districts are able to ensure that even very small parties are represented in the legislature. For example, a district in which there are only three members to be elected means that a party must gain at least 25 percent +1 of the vote to be assured of winning a seat. A party, which has the support of only ten percent of the electorate, would not win a seat, and the votes of this party's supporters could therefore be said to have been wasted. In a nine-seat district, by contrast, ten percent +1 of the vote would guarantee that a party wins at least one seat. This means not only that the results are more proportional, but that there is also more chance that small parties will be able to be elected. The problem is that as districts grow larger -both in terms of the number of seats and often, as a consequence, in their geographic size as well so the linkage between an elected member and their constituency grows weaker. This can have serious consequences in societies where local factors play a strong role in politics, or where voters expect their member to maintain strong links with the electorate and act as their "delegate" in the legislature.

Because of this, there has been a lively debate about the best level of district size. Most scholars agree, as a general principle, that district magnitudes of somewhere between three and seven seats per district tend to work quite well, and there is also general agreement that odd numbers like three, five, and seven work better in practice than even numbers, particularly in a two-party system. But this is only a rough guide, and there are many situations where a higher number may be both desirable and necessary to ensure satisfactory representation and proportionality. In many countries, the electoral districts follow pre-existing administrative divisions, perhaps state or provincial boundaries, which means that there may be a wide variation in their size. Numbers at the high and low ends of the spectrum tend to deliver more extreme results. At one end of the spectrum, a whole country can form one electoral district, which normally means that the quota for election is extremely low and even very small parties can gain election. In the Netherlands, for example, the whole country forms one district of 150 members,which means that election results are extremely proportional, but also means that parties with extremely small vote shares, even less than one percent, can gain representation, and that the link between an elected member and a geographic area is extremely weak.

At the other end of the spectrum, PR systems can be applied to situations in which there is a district magnitude of only two. A system of List PR is applied to two-member districts in Chile, for example, and as the Chilean case study indicates, this delivers results which are quite disproportional, even though a proportional formula is used, because only two parties can gain representation in each district. This has tended to undermine the benefits of PR in terms of representation and legitimacy.

Both of these polarized examples serve to underline the crucial importance of district magnitude in any system of proportional representation. It is arguably the single most important institutional choice when designing a PR electoral system, and is also of crucial importance for a number of non-PR systems as well. In majoritarian systems, as district magnitude increases, proportionality is likely to decrease. To sum up, when designing an electoral system, the district magnitude is in many ways the key factor in determining how the system will operate in practice, the extent of the link between voters and elected members, and the overall proportionality of election results.

The choice of electoral system has a wide range of administrative consequences, and is ultimately dependent not only on a nation's logistical capacity to hold elections, but also on the amount of money that the country can spend. Simply choosing the most straightforward and least expensive system may well be a false economy in the long run, since a dysfunctional electoral system can have a negative impact on a nation's entire political system and its democratic stability. The choice of electoral system will affect a wide range of administrative issues set out in the following paragraphs.

The Drawing of Electoral Boundaries

Any single-member district system requires the time-consuming and expensive process of drawing boundaries for small constituencies defined by population size, cohesiveness, "community of interest," and contiguity. Furthermore, this is rarely a one-time task since boundaries are regularly adjusted to reflect population changes. First Past The Post (FPTP) and Two-Round System (TRS) systems provide the most administrative headaches on this score. The Mixed-Member Proportional (MMP), and Single-Transferrable Vote (STV) systems also require electorates to be demarcated, but are easier to manage because they use fewer and larger multi-member districts.

At the other end of the scale, List PR systems are often the cheapest and easiest to administer. This is because they either use one single national constituency requiring no boundaries to be drawn, or they use very large multi-member districts that dovetail with pre-existing state or provincial boundaries.

The Design of Ballot Papers

Ballot papers should be as friendly as possible to all voters, to maximize participation and reduce spoilt or "invalid" votes. This often entails the use of symbols for parties and candidates, photographs, and colours. TRS ballots are similarly easy, but in many cases new ballots have to be printed for a second round of voting, thus effectively doubling the production cost. MMP systems usually require the printing of at least two ballots, even though they are both for a single election. STV ballots are slightly more complex than FPTP ballots because they will have more candidates, and therefore more symbols and photographs (if these are used). List PR ballot papers can span the continuum of complexity. They can be very simple, as in a closed list system, or quite complex as in a free list system such as Switzerland's, see Switzerland.

Voter Education

Clearly, the nature of and need for voter education will vary dramatically from society to society, but when it comes to educating voters on how to fill out their ballots, there are identifiable differences between each system. The principles behind voting under preferential systems such as STV are quite complex if they are being used for the first time, and voter education must address this issue, particularly if there are compulsory numbering requirements. The same is true of MMP systems: after over 50 years of using MMP, many Germans are still under the misapprehension that both their votes are equal, when the reality is that the second "national PR" vote is the overriding determinant of party strength in parliament, see Germany: The Original Mixed Member Proportional System. By contrast, the principles behind categorical, single-vote systems such as FPTP are very easy to understand.

The Number and Timing of Elections

FPTP, List PR, and STV electoral systems all generally require just one election on one day, see Parliamentary Size. However, MMP systems essentially mix two (or more) very different electoral systems together, and therefore have logistical implications for the training of election officials and the way in which people vote. Two-Round Systems are perhaps the most costly and difficult to administer, because they often require the whole electoral process to be repeated a week or a fortnight after the first try.

FPTP and simple closed-list PR systems are easiest to count as only one vote total figure for each party or candidate is required to work out the results. The MMP systems nearly always require the counting of two ballot papers. AV and STV, as preferential systems requiring numbers to be marked on the ballot, are more complex to count, particularly in the case of STV, which requires continual re-calculations of surplus transfer values and the like.

Primarily history, context, experience, and resources will determine the stresses, which any electoral system places on a country's administrative capacity. In the abstract, the table below offers some clues to the potential costs of various systems. If equal weight is given to each of the six factors examined in the table (which, it must be said, is unlikely to be the case), a cursory glance at the totals for each system shows that List PR systems, especially national closed-list systems, are the cheapest to run and require fewest administrative resources. Next come FPTP systems followed by the STV and MMP. According to our calculations, the system, which is most likely to put pressure on any county's administrative capacity, is the Two-Round System.

How votes become seats

In proportional representation systems there are five key mathematical formulae through which votes are translated into seats. In list PR systems, see List PR, there are "highest average" methods (D'Hondt and Sainte-Lagu ) and "largest remainder" methods (Hare and Droop). The single transferable vote form of PR, see Single Transferable Vote, almost always uses the Droop quota.