«Since the work of Maurice Duverger (1954), one of the most fundamental ideas in political science is that a party system is shaped by the electoral ...»
In Tasmania and the Australian Senate the two largest parties14 typically obtain over 85 percent of the first-preferences, while in Ireland and the Australian House the two largest parties usually receive three-fourths of the total vote.15 In all four systems, the two largest parties possess over eighty-five percent of the seats.16 Consistent with Rae's propositions, the large parties are over-represented and the smaller parties are under-represented. The STV elections present a clear picture with the largest party obtaining a seat bonus from a low of 1.6 percent in Tasmania to a high of 3.5 percent in the Australian Senate. Furthermore, the second-largest party always obtains a seat bonus. Third parties are slightly under-represented (e.g., Irish Labour almost breaks even with an average deficit of -0.6 percent) while smaller parties and independents suffer to a greater degree.
The results in the Australian House (with AV) differ from that of the STV legislatures. Since AV is a more majoritarian system than STV, the third and smaller parties are a bit more under-represented. However, what is surprising is that the largest party is under-represented while the second largest party tends to be over-represented.
The Labour Party (the largest party) has a seat share less than two percent on average than its vote share. In ten of the sixteen elections, the largest party suffered a seat deficit. In accordance with more observations of majoritarian systems, the deviations from the two percent mean are quite extreme. For example, in the 1975 election the Labour Party received 14.5 percent fewer seats than votes, while in the 1983 election they obtained a bonus share of seats of 10.5 percent.
In three of the four legislatures, the seat bonus of the second largest party is greater than that received by the largest party. This aside, in all four legislatures the top two parties, when taken together, always obtain an average seat bonus of near five percent at the expense of the other parties. This even holds true in the Australian House where the seat bonus of the Liberal Party is greater in absolute terms than the seat deficit of the Labour Party. This two-party seat bonus on average does not significantly differ from that reported by Gallagher, Laver and Mair (1995) for the European parliamentary democracies.
TOWARDS A SYSTEM OF PRIf Douglas Rae's propositions are correct, the two largest parties both singularly and collectively would prefer the current system to a system more closely approximating the ideal of Proportional Representation (see Figure 1). This section tests these two propositions by imposing hypothetical PR elections onto the four party systems above. In other words, how would the seat shares of each party change if, holding the voters' decisions constant, the election had been one of a nation-wide party-list election.
I calculate each party's total seat share as being equal to each party's total vote share (i.e., their entire share of the vote across all districts). This approximation of PR assumes only that the actual votes cast display a sincere party preference. Studies on STV have never disputed this claim (Gallagher, 1986; Sinnott, 1992; Farrell, Mackerras and McAllister, 1996). Moreover, my method is consistent with Rae as he implicitly makes a similar assumption in formulating and testing his differential propositions.
One may ask why I have not chosen to keep the districts intact and use a common PR formula (e.g., D'Hondt) to calculate seats. My compelling reason is that STV is the "more proportional" of the two PR formulas. Many scholars have conclusively shown that STV, ceterus peribus including district magnitude, is a more proportional formula than D'Hondt, which is one of the least proportional of all (Blondel, 1969; Loosemore and Hanby, 1971; Lijphart 1986; Cox 1997). Gallagher (1992) makes a strong case that D'Hondt is more favorable to large parties than STV (and most other PR forumale as well). The conclusion of these authors can be confirmed empirically, with D'Hondt producing a greater seat share for the larger parties.17 Thus, running a simulation with another PR formula would bring about either marginal/indeterminate change (in the case of equal or slightly more proportional formulae) or seemingly paradoxical, but empirically accurate, change in a majoritarian direction (in the case of D'Hondt).
Table 1 presents the change in seats that would occur in the four party systems with hypothetical PR elections. To formulate this table, I first calculate the national vote share (first-preferences) for each party. I accomplish this by summing the vote shares of all candidates from a similar party. Given a hypothetical single national district of a size equal to the legislature, I then multiply each party's vote share by the number of seats in the legislature. A seat is awarded for each integer with left-over seats awarded by a "largest-remainder" formula.18 I then sum the results of each election and divide by the number of elections to produce the mean change in seats. The three rows present the best and worst results, as well as the average for all the elections.
Table 1 confirms Rae's propositions, with the exception of the Australian House.
For instance, in Ireland, both large parties (Fianna Fail and Fine Gael) would lose seats in a PR election. Most smaller parties would gain seats (the Workers' Party, the Progressive Democrats, and the Other category) with only the Labour Party losing seats. This may be due to the intermittent electoral alliance of Fine Gael and Labour which increases vote transfers between the parties (Gallagher, 1978; Mair, 1989; Sinnott, 1995). The pattern for the Australian Senate and Tasmania are similar, with the large parties losing seats under PR and the smaller parties and independents gaining seats.
The pattern in the Australian House is quite different. The largest party gains seats and the third party (Country Party) loses seats. Labour would gain seats in ten of the sixteen elections and the Country Party would lose seats in all sixteen. This is clearly divergent from Rae's predictions. Why does such a pattern occur? It is the combination of the electoral system and the alliance system in the Australian House. As in Ireland, the second and third parties often enter into pre-election coalition. They are rewarded for this effort by an electoral system that allows their supporters to transfer their votes from one party to the other. If the voters behave as the party elites want, the sum of the preferences for the second and third parties can overcome the greater first preferences for the largest party.
What is striking is that this happens more often in the single-member transfer system (AV) than in the multi-member transfer system (STV). Perhaps the "winner takeall" nature of a single seat explains the difference between the Australian House's Country Party and Ireland's Labour Party. The former benefits greatly from the current electoral system (AV) while the latter benefits from STV only marginally. Irish Labour is able to win a seat or two in some districts without the help of Fine Gael, but with only a single seat in each district the Liberal and Country parties cannot carve out their own niche, but together they can defeat their rival, the Australian Labour Party.19
TOWARDS A SYSTEM OF PLURALITY
proposes that the two largest parties benefit from overrepresentation and that this overrepresentation is more pronounced in majoritarian than proportional systems. If this is the case, the two largest parties in any party system should benefit from a change of electoral formulae towards a more majoritarian system. In this next section, I compare the effect an alternative electoral system would have on the Tasmanian and Irish elections.20 I compare the actual results of the STV contests with the expected results from using the Single Non-Transferable Vote (SNTV)21 electoral system. This essay shows that the consequences of this change are non-trivial. Only the second largest party in Tasmania (Liberal) and the largest party (Fianna Fail) in Ireland would prefer a change to SNTV. Moreover, in both cases the two largest parties taken together do not have a combined incentive to introduce a more majoritarian system.
The Single Non-Transferable Vote (SNTV) SNTV is an electoral formula similar to STV in district magnitude, but similar to plurality formulae in its seat allocation method. In other words, SNTV also uses multimember districts, but as the name implies, there is no transfer of votes. The seats are allocated in a fashion similar to the "first past the post" plurality system, except in the case of SNTV, the seats are won by the "first 'n' number of candidates past the post" where 'n' equals the district magnitude. For this reason a number of researchers have proposed that SNTV be classified as a plurality system (Rae, 1967) or at best, semiproportional (Lijphart, 1984; Taagepera and Shugart, 1989).
Expanding on Rae's work, I expect that the distribution of legislative seats under an SNTV system would display characteristics closer to simple plurality, single-member district systems than proportional systems. Both large parties, individually as well as collectively, should receive a substantial seat bonus. Consequently, in comparison to STV, large parties should favor a change to SNTV, while smaller parties would favor retaining STV.
I introduce SNTV elections into Ireland and Tasmania in the following manner.
First, I hold static all district magnitudes in each of the elections. Thus, if five seats are allocated in the district of Cork in the Irish 1954 election (with STV), five seats would be allocated by SNTV in the hypothetical election. Second, I hold static the actual candidates (as well as their party affiliations). This has the inherent problem of neglecting the highly important role of party nomination strategies. More precisely, it neglects the effect that over-nomination has on large parties. Studies show that parties in SNTV elections must be concerned with both nominating the proper number of candidates and distributing the votes equally (Cox and Niou, 1994; Cox and Rosenbluth, 1994). Over-nomination can lead to a loss of seats (from a hypothetical "possible" number of seats)22and large parties tend to over-nominate more often than smaller ones.
Other studies have shown neither over-nomination nor vote distribution to be an overwhelming concern for party leadership in STV elections (Cohan, et. al, 1975;
Lijphart and Irwin, 1979; Katz, 1981; Marsh 1996).
Therefore, my proposed method of simulating a move from STV to SNTV elections will exaggerate the potential gains to large parties (and potential losses to small parties) from such an electoral switch. Results confirming Rae's hypotheses are more likely to result from my simulation than is probably empirically valid. Any results where large parties would stand to gain a large number of seats should be taken cautiously.
However, any contrary results (i.e., a large party not gaining seats) would cast a serious doubt on Rae's hypotheses. As such, the above assumption limits the possibility of my hypothesis being correct while strengthening that of the rival hypothesis (Rae).
Third, I assign each first-preference vote (with STV) as the single vote of the SNTV voter. I choose this method for two reasons. First, voters in STV elections should be rewarding thier first-preferences in a sincere manner (as mentioned earlier). Thus, their votes are a good approximation of their actual preferences among the candidates.
Second, from a practical standpoint, it is a more plausible approximation than choosing a different distribution of votes. For example, I could sum up the total first-preference vote for each party and then divide it evenly among the optimal number of candidates (in an approximation of the party perfectly managing the vote). This would not be suitable as it is already clear that parties have a difficult time producing such a distribution and I would throw away needlessly the only actual information that I do have about voter preferences.
While the assumption of the static voter appears to be strong, for the rationalchoice literature points out that as institutional incentives change voters modify their behavior to adapt, nevertheless, it does provide a starting point for my examination. The focus of this essay is to determine if the translation of votes to seats, as performed by different electoral institutions, favors some parties over others. In order to isolate the institutional effect, I choose to hold voting behavior constant.23 In the common trade-off between completeness and simplicity, I have chosen the latter in the hope that the simulation will produce generalizable results without too large a discrepancy from reality.
Fourth, for each district the list of successfully elected candidates is determined using the SNTV formula, and their party affiliations noted. Also, if the results in any district are different from those under STV, the party affiliation of the "new winner" is noted, as well as the party affiliation of the "new loser". Fifth, a hypothetical legislature is formed by aggregating all of the successful candidate lists from every district.
As an example of the above methodology, in the same aforementioned district, 3,052 first preference votes were recorded for Seán Casey (Labour), who was later elected after ten transfer counts with 8,516 final votes (he won the fifth seat). In the hypothetical SNTV election, Mr. Casey would receive only the 3,052 first-preference votes and this figure is compared against the first-preference votes of the other candidates (Mr. Casey would still win the fifth seat).
Table 2 shows the change from the actual elections to those that would result from SNTV elections in Ireland and Tasmania. As in Table 1, the three rows present the best and worst results, as well as the average for all the elections. I also include two separate means for the Irish elections due to a change in the pattern of the seat distribution (explained below).
The results in Tasmania are straight-forward, but in conflict with Rae's hypothesis.