How people elect parliaments
There are many possible voting systems, many ways of controlling who can nominate and how people can cast votes, and many ways of administering elections. Features of electoral systems are often designed to suit various interests, or solve various problems as they appear from time to time. But not all voting systems and administrative options are equally legal within Australian constitutional law, because the Constitution creates the Parliament with the intention that it function within a holistic constitutional order, one in which representative and responsible government comes into existence and endures, ensuring that that system is democratically accountable to the electors.
What, then, causes a Parliament to have a representative character? Clearly, that character will arise primarily from the design of the electoral system by which its membership is chosen. How can we determine whether a particular electoral system has brought into being a representative assembly? What features of an electoral system result in successful representation?
I propose that there are five relevant features:
The first and perhaps most basic feature is the level of participation in the election. This in turn is built up in several layers, including who is entitled to vote, who is registered to vote, who actually votes, and whose votes are counted as formal. An outline of the numbers of Australian people showing participation in the 2013 election is as follows:
The above diagram uses voting figures for the election of members of the House of Representatives in 2013. It shows that of an estimated population of around 16.178 million adult citizens eligible to vote, only 12.915 million actually cast valid votes. The real participation rate of Australians electing members to the House was therefore around 79.8%. The submission to JSCEM by the Democratic Audit of Australia (Professor Brian Costar) reaches the same conclusion (see JSCEM submission 116, para 1.2, page 2)
Two other features of representation are the directness of election and the exercise of choice. Both these concepts are explicitly mentioned in sections 7 and 24 of the Constitution.
Direct election can be defined as the absence of any intervening third party between the elector and the elected. For a relationship to be direct there must be no third party (or institution) which intervenes in the process by which the voter chooses the representative; no-one who can exercise control over which candidates are successful in competition with the choices of the electors. This straightforward definition is supported by the text of the Constitution, which refers to only two actors – “the people” and “members”/“Senators” – describes their relationship as “direct”, and mentions no-one else.
Appointed upper houses are clearly not directly elected by the people. Similarly, the appointment of Senators to the United States Senate by State congresses or governors (as was the practice until the early 20th century, and is still the case for casual vacancies) and appointment to the Australian Senate in the case of casual vacancies, are not direct elections by the people.
Similarly the party list electoral systems (common in Europe, Asia and Latin America) are also not systems of direct election, because the individual parliamentarians are chosen by political parties, not by voters. In these systems, voters influence how many appointees each party has, but cannot directly control the identity of the individuals who become members of parliament. In his ruling in the McKenzie case, which upheld the 1984 amendments to the ballot structure for Senate elections, Chief Justice Gibbs stated:
“…”it is right to say that the electors voting at a Senate election must vote for individual candidates whom they wish to choose as Senators, but it is not right to say that the Constitution forbids the use of a system which enables the elector to vote for the individual candidates by reference to a group or ticket.” 
On this understanding, ballot paper features such as displaying party names, grouping candidates under party headings, and even the group voting ticket (GVT) device are permissible, so long as the fundamental vote counting method remains individual-candidate based, and does not become one of voting for parties in themselves.
(Debatably, some ‘open list’ party list systems might be considered partially direct and partially non-direct, and their acceptability will depend on whether a given constitutional system requires ‘direct election’ to be unambiguous – as the Australian Constitution would appear to do.)
The question of whether choice is available to the electors is also fundamental. Without choice, there is no credible election, and the resulting Parliament cannot be said to provide meaningful representation. But there are many shades of choice that might be available, both in terms of the degree of choice and also of nature of it.
Not everyone’s hopes or ideas of appropriate representation can always be accommodated. The world’s electoral systems are all premised on processes of nomination. If none of the nominees are to an electors liking, but they are still provided with the opportunity to cast a ballot differentiating between them, we would still say that a degree of choice has been offered. There is a difference between facing unsatisfying choices and having no choice.
The more common problem is that choices are limited to a specific range of candidates, usually by the device of placing voters in geographical divisions. The idea here is that limiting voter choice only to nominees for that division means that the resulting members represent the electors who are resident in that geographical area. But such limits deny choice to voters whose preferred representative – perhaps an individual of national prominence within a party, or merely a prominent person known for their performance on a key public issue – happens to live elsewhere from the voter.
The above three principles of representation – participation, direct election and free choice – are often guaranteed in national constitutions to some extent or other. Less often guaranteed are my final two important criteria, which are the extent to which each of the electors has an equal influence in the election of the representatives, and what proportion of the electorate achieves actual representation in Parliament by representatives whom they support..
Equality of voting influence is a complex subject. It is possible to measure influence, and thus compare whether it is equal for different voters, by using at least these three distinct analyses:
Political research and system reform in relation to voter equality in the last half-century has been mainly concerned with seeing equality only in terms of the first of the above issues, equality in the numbers of people enrolled in electoral divisions. Reforms have been achieved in various democratic countries to make such enrolments similar in size, a cause that has gone under the banner ‘one vote one value’.
Australia has among the most thorough approaches to equalising divisional enrolments. For the House of Representatives the boundaries within each state are reviewed at least every 7 years, and there are specific triggers for earlier reviews. The result is that most boundaries are reviewed at least every second election. Only a review every single electoral cycle would provide greater frequency.
Furthermore, the equalization standards used for the Australian House of Representatives are fairly tight, requiring individual division enrolment variations of no more than 3.5% from the mean enrolment of the state as at the mid-point of the coming 7-year period.
One of the common ways to measure the degree of variation in a set of values is the co-efficient of variation; the ratio between the mean of those values and the average variation from that mean. The results of this measurement for the enrolments in several major democracies in recent elections are set out in a linked table showing Inequality of representation results by nation (third column).
Clearly the Australian results are the best in the sample of single-member division systems shown, no doubt because of our fairly robust and frequent systems of redistribution of divisional boundaries. This is commendable, and yet the deeper problem is that this work only addresses one aspect of the problem of voter equality, and addressing this factor alone cannot result in equal influence for all electors.
As we will see, the main cause of poor results in the equality of influence of electors is the use of single-member divisions (SMD). 
This point is easily highlighted by the continuing problem of dramatically different influence on total election outcomes of electors enrolled in marginal and safe seats seen under the single-member division system. In such electoral systems individual voters have influence over the election of members in proportion to the marginality of each division’s election result.
In the 2013 election, seat marginality ranged from just 53 votes (Fairfax) to 40,066 votes (Maranoa), with a mean winning margin of 14,955 votes. The summary table World assemblies – key statistical results clearly shows the high level of variation in the effective influence of voters, and note that there is no noticeable association between low variations in enrolment numbers and the variations seen in effective influence.
These results demonstrate the widely known fact that marginal seat voters have an excessive influence on election results. 32% of the electorate has more than double the average influence on election outcomes. These are the marginal seat voters who determine who controls the House and therefore who forms governments in Australia. In fact just over 1 million electors in the 11 most marginal seats have greater than 10 times the effective influence of the average Australian elector.
By contrast around 45% of our electors had low levels of influence.
We can also see that single-member division systems make voters unequal in influence on the basis of which political party they prefer. On average, in 2013 every 70,300 Australian voters who supported a major party elected one representative to the House of Representatives. Most advantaged of all, it took less than 52,000 supporters of the Queensland Liberal National Party to elect each of its successful members. In sharp contrast, it took 1.1 million Green voters to elect just one representative, and 700,000 PUP voters to do the same. Around 900,000 voters who supported other micro-parties elected were left unrepresented. Under single-member division systems it is very clear that the influence of different voters varies greatly depending on who they support.
If the above results look poor, what would success look like in providing equality? True success would be that every voter had identical influence in electing representatives. Below we will discuss the quota-based voting systems that were designed to achieve just that, by dictating that the number of electors necessary to support any elected member be identical. The original use of the term ‘proportional representation’ in the 19th century was related to this idea – that the voter base of every elected member was the same, thus giving all voters a universally proportional influence on the results. (The modern use of this term to compare outcomes between political parties has a significantly different meaning.)
All systems which allow candidates to be elected by different numbers of supporting voters fall short of that standard of equality. Inequality is, in particular, the primary failure of systems based on single-member electoral divisions. The members of such parliaments have often been elected by sharply different numbers of voters. The inequality of influence of such voters on two of the measures mentioned above – different ability to elect representatives for supporters of different parties, and different degrees of marginality of the electoral divisions – are always substantial. Differences in enrolment equality can often be substantial, and indeed have been deliberately manipulated in many countries over the years.
Let’s focus more closely on the relationship between enrolment equality and influence equality. Some data on inequality of division margins in a selection of recent elections in single-member division countries is set out in Table 1 (4th column). We can see clearly that, contrary to common understanding, striving towards equality in the first of the three issues – equality of enrolments in the electoral divisions – does NOT solve the whole problem of inequality of influence. Of the countries shown in the Table above, we see that Australia has the lowest levels of enrolment inequality of the SMD systems. This is probably because of two factors: firstly the compulsory enrolment laws and the resulting very high election participation rates in our country, and secondly because Australia probably has the best division redistribution rules to strive for division population equality.
But these arrangements are not enough to overcome the powerful inequality effects generated by the other two aspects of inequality. Across the nations shown in Table 1, while enrolment variation results range from the very low CofV value of 2.3 (in NSW) to the mid-20s, peaking at over 40 (in Malaysia), all the results for influence variation are much higher, with most CofV values ranging from over 50 to nearly 80.
The clear conclusion is that in the presence of single-member divisions inequality of voter influence is always substantial, and cannot be controlled through enrolment equality measures.
The fifth principle of representation is the notion of actual representation – the outcome for any given elector of achieving a representative in Parliament whom they support, and whom they can safely regard as representing their views and as inclined to vote on legislation and other matters in a manner that the voter would wish. We can measure the level of actual representation as the proportion of the enrolled electorate that achieves it in any given election.
We can regard votes cast for a single candidate (in plurality elections) as well as first preference votes in preferential voting systems, as indicating a candidate who, if elected, gives actual representation to their voters. We could also stretch the definition so that in multi-member divisions, if a voter supports a political party that has at least one candidate elected, then that voter is represented even if their first preference vote was for a different candidate of the party.
In addition, the idea of actual representation can be stretched to accommodate preference voting, so that if the candidate of a voter’s first preference fails to be elected, at least some next preferred candidate might do so, and if that happens then this voter can also be counted among those who have won actual representation. However this assumes that all subsequent preferences are indications of positive support, which cannot be taken for granted in compulsory preferencing systems (which oblige voters to mark a sequence of preferences for all candidates). In such systems, many of the preferences given for lower-ranked candidates may only reflect a relative ranking of disapproved candidates. For this reason, measurements of actual representation using preferences can only be soundly based if preferencing is optional.
Note that it is impossible to give 100% of voters a successful outcome of actual representation, simply because some voters fail to vote, have their votes invalidated, or choose to support small parties and independents who fail to get elected. This is true of all electoral systems.
As with problems of inequality, low rates of actual representation are endemic in single-member division systems, especially non-preferential ones. Results for the proportion of voters who had actual representation is also shown in Table 1 (in the 5th and 6th columns). Not very good, are they? And note that even Australia’s very high participation rates do not generate results of even 50% of the electorate. Not that 50% is any sort of target here – we should be aiming for actual representation rates much higher than half the voters; we should be looking for actual representation for the highest proportion of the electorate that is possible.
Note also that Australia’s preferential voting system, while it gives higher results, does not increase the results by a very great margin.
The direct voting system that will give the highest results in terms of actual representation is the single transferable vote system (STV); we will discuss this in more detail below. Table 1 shows rates of actual representation achieved in some jurisdictions electing assemblies by STV: the Australian Senate, the Tasmanian and ACT Legislative Assemblies, the NSW Legislative Council, the Irish Dáil Éireann, and the Maltese Kamra tad-Deputati (House of Representatives). It is apparent that results near or over 80% actual representation are quite achievable, and the Maltese result demonstrates that representation of over 90% is possible.
It is immediately clear that STV gives voters far higher levels of actual representation than SMD systems.
The five key voting system goals outlined above are significantly interlinked. For example, actual representation outcomes will be lowered whenever participation is lowered. Actual representation also loses its genuine character if voter choice is limited. Electors who have their participation limited, or their choices constrained, can consequentially have a less equal influence, or possibly no influence at all, on election outcomes. Overall, success in achieving each of these five goals appears to support success with others.
Note that these voting system principles are all addressed to the interests of the electors, not the interests of political parties. Political parties were not mentioned in the Constitution as drafted, and indeed in the late 19th century the idea that parties were entitled to have interests in their own right would have been controversial, and most likely a minority view at best, for certainly the ‘influence of faction’ was strongly opposed by many constitutional founders and other political thinkers. Political parties are not an integral element of the constitutionally mandated system of representative and responsible government.
April 2014, revised February 2016
 Derived from the ABS population value for Australia as at 30 June 2013 – 23,130,900 – increased by 7 weeks’ growth (to mid-August 2013) at the ABS annual growth rate of 1.8%
 This is an estimate derived using a 91% rate of enrolment, which is within the range estimated by the AEC in recent years.
 McKenzie v Commonwealth  HCA 75
 The latter principle has at various times been termed real representation and also effective representation.
 The present procedures for redistributions were adopted in 1984; variations in House division enrolments were greater up to the election of 1983.
 Single-member division electoral systems are very common. Most nations which have their origins as British colonies and/or empire use such systems. As at 2012, around 42% of the world’s population live in democracies which elect the lower house of their national parliament by a single-member division system of direct election. Around 20% live in systems which elect assemblies by methods of indirect, party list voting (although some of these will directly elect a president to lead their executive government.), and a further 13% are in countries which use composites of party list and single-member division systems. Fewer than 2% of people live in countries which use non-single-member division systems of direct election. The remaining 22% of the world’s population live in countries which are not democracies.
 At least, it is impossible in ordinary elections to elect Parliaments, with the complex variety of patterns they throw up. 100% actual representation might be possible in abstract examples with small voter pools and highly fortuitous ratios between voters, candidates and seats.
 Only from 1977 were political parties referred to in the Constitution, and only on the specific issue of integrity in the filling of casual vacancies.