Current Issues Brief Index 2002-03

Current Issues Brief No 15

House of Representatives By-elections 1901-2002

Gerard Newman, Statistics Group
Scott Bennett, Politics and Public Administration Group
3 March 2003

Contents

By-elections are held to fill vacancies in the House of Representatives. This paper gives details of by-elections held from the first for Darling Downs on 14 September 1901 to the most recent held on 19 October 2002 for Cunningham.

Among the major points made in the paper:

• Since September 1901 there have been 140 by-elections, an average of 3.5 per parliament.
• The number of contestants has grown over the years from 2.2 per by-election
• (1901-18) to 8.7 per by-election (1990-2002).
• In only four cases (2.9 per cent) was a by-election contested by just a single candidate.
• An increasing tendency has been for governments to avoid contesting by-elections in their opponents' safe seats. Since 1990 this has occurred in 10 of the 18 by-elections held.
• In only seven cases (5 per cent) has the Opposition party failed to contest a by-election, the most recent case being Bradfield in 1952.

• Sixty-seven (47.9 per cent) of the by-elections occurred because of the death of the member, 67 (47.9 per cent) as the result of the resignation of the member, five (3.6 per cent) because of voided elections, and one (0.7 per cent) because of the expulsion of the Member from the House (Kalgoorlie 1920).
• Since 1949 most by-elections have been caused by the resignation of the sitting member and have occurred in safe seats. Only eight of 52 by-elections in this period (15.4 per cent) have occurred in marginal seats.

• On 34 occasions (24.3 per cent) the party complexion of a seat has altered at a by-election.
• Twenty-four of these (17.1 per cent) have been in seats lost by the government of the day. The most recent case was the loss of Ryan by the Coalition Government to Labor in 2001.
• Four of the losses (2.9 per cent) have been by the Opposition of the day. The most recent case was the loss of Cunningham by Labor to the Greens in 2002.
• The average two-party preferred swing against the government of the day has been 4.0 per cent, while the average swing in government-held seats was 5.0 per cent against the government.
• In the period since 1949, the largest two-party swing against a government occurred in Canberra in 1995 when the ALP Government's vote fell by 16.1 per cent. The largest swing to a government was 16.2 per cent, which occurred in the Coalition Government's seat of McPherson in 1981.

By-elections are held to fill vacancies in the House of Representatives resulting from the death, resignation, absence without leave, expulsion, disqualification or ineligibility of a Member.

• There have been 140 Commonwealth by-elections held since Federation.
• The first was held in the Queensland seat of Darling Downs, on 14 September 1901 barely four months after the opening of the new Commonwealth Parliament.
• The most recent by-election was held in the New South Wales seat of Cunningham on 19 October 2002.

Of the 140 by-elections, 67 (47.9 per cent) have occurred because of the death of the member, 67 (47.9 per cent) as the result of the resignation of the member, five (3.6 per cent) because of voided elections, and one (0.7 per cent) because of the expulsion of the member from the House (Kalgoorlie 1920).

The pattern has altered quite markedly, however:

• From 1901 to 1970, 63.3 per cent of all by-elections were brought about by death.
• By contrast, since 1971 85.7 per cent of all by-elections have been brought about by resignation.
• Goot reports a similar phenomenon in New South Wales by-elections, with 1969 the first year in which more by-elections were brought by resignation than by death-in Commonwealth elections this occurred just five years earlier.⁽¹⁾

Various factors are probably relevant here.

Members today enter the House of Representatives at a significantly younger age than used to be the case. Of the 40 parliaments to date, the second quarter's intake (1929-51) was the oldest, averaging 48.3 years per new Member. By contrast, the fourth quarter's intake (1977-2001) was the youngest at 42.2 years per new Member.⁽²⁾ Another factor is the greater preparedness of Members to leave Parliament before death, quite often to pursue another career. This has been aided by a third consideration, namely the general increase in longevity of male Australians.

The Constitution (s. 33) confers on the Speaker of the House of Representatives the power to issue a writ for the election of a new Member. The Commonwealth Electoral Act 1918 specifies that the election must be held between 33 and 68 days from the date the writ is issued.

Interestingly, there is no prescribed time period within which a by-election writ must be issued, and, in fact, there is no accepted view as to the period that should elapse between vacancy and polling day.

Perhaps surprisingly this has not become a matter of any long-term political debate, allowing governments a great deal of freedom in the setting of by-election dates. The length of time from a vacancy until polling day has therefore varied considerably, with the maximum number of days being the 82 days for Moreton in 1983, and the minimum being the 17 days for East Sydney in 1903.⁽³⁾ The total elapsed time from vacancy to election has increased since 1901. Between 1901 and 1949, the average was 40.3 days, whereas during the period 1949 to 2002 the average has been 51.2 days. The average figure for all by-elections has been 46.7 days.

Ideally, of course, by-elections are held as early as possible after a seat becomes vacant, 'so that the electors are not left without representation any longer than is necessary'.⁽⁴⁾ In fact, such a consideration is occasionally pushed aside by short-term political factors.

On 22 October 1982, the member for Flinders (Phillip Lynch) resigned his seat. The Speaker fixed 4 December as the date for a by-election, but also declared that there would be a delay of three almost three weeks before the writs for the by-election would be issued. According to Summers, this gave the Fraser Government the option of calling a general election for 4 or 11 December. The by-election could therefore have been pushed aside.⁽⁵⁾

In the cases of the eight by-elections held during 1994-5, the delay between the date of the seat becoming vacant and the date of the issuing of the writ varied considerably. The cases of Mackellar (35 days) and Kooyong (27 days) contrasted with the virtually instantaneous issuing of writs for by-elections in Fremantle, Boynthon and Warringah:

The variation that is allowed in regard to by-election dates is thus an interesting anomaly in an electoral system that is generally highly regulated.

Appendix 4 provides details of the timing of all by-elections held between September 1901 and October 2002. It also shows the number of days elapsed between the seat becoming vacant and the by-election date, as well as the number of days elapsed since the previous general election, and the number of days between the by-election and the next general election.

There have been 20 occasions when the Speaker has in fact declined to issue such a writ for a by-election due to a pending general election. The longest period a seat has been without a member prior to a general election was 128 days in the case of Hindmarsh in 1909-10. The shortest period was the 39 days between the plane crash on 13 August 1940 that killed three members and the election of 21 September 1940.

 

Division	Vacant	Next election	Days
Hunter (NSW)	30.09.03	16.12.03	77
Indi (Vic)	12.10.06	12.12.06	61
Northern Melbourne (Vic)	13.10.06	12.12.06	60
Hindmarsh (SA)	06.12.09	13.04.10	128
East Sydney (NSW)	24.12.09	13.04.10	110
West Sydney (NSW)	06.09.28	17.11.28	72
Wimmera (Vic)	14.10.31	19.12.31	66
Martin (NSW)	05.06.34	15.09.34	102
Ballarat (Vic)	31.07.34	15.09.34	46
Werriwa (NSW)	02.08.34	15.09.34	44
Henty (Vic)	13.08.40	21.09.40	39
Flinders (Vic)	13.08.40	21.09.40	39
Corangamite (Vic)	13.08.40	21.09.40	39
West Sydney (NSW)	14.08.46	28.09.46	45
Hindmarsh (SA)	14.08.46	28.09.46	45
McMillan (Vic)	14.10.55	10.12.55	57
Leichhardt (Qld)	11.10.58	22.11.58	42
Warringah (NSW)	03.08.66	26.11.66	110
Bonython (SA)	30.09.77	10.12.77	71
Wills (Vic)	25.11.92	13.03.93	108

The 140 by-elections have been contested by an average of 4.5 candidates. Over the years, however, there has been a steady increase in the number of nominations.

In the 24 first-past-the-post cases between 1901 and 1918 (October), the average number of nominations was 2.2 per contest.

In the 116 preferential voting cases there have been an average of 5.0 nominations per by-election.

• Between the introduction of preferential voting in 1918 (December) to the start of 1970, there was an average of 3.7 nominations per by-election.
• The 1970s and 1980s saw the first of two significant increases in candidate numbers, with 6.2 candidates per by-election over these decades.
• This has been exceeded during 1990-2002, with the average climbing to 8.7 candidates per contest.

The record number of nominations occurred in this latter decade in Wills 1992, when 22 nominations were received to fill the seat vacated by former Prime Minister Hawke.

In only four cases (2.9 per cent) (Kalgoorlie 1913, Dalley 1915, Wide Bay 1928, Cunningham 1956) was a by-election contested by just a single candidate.

In recent years there has been a tendency for governments to avoid contesting by-elections in their opponents' safe seats. Since 1990 the incumbent government has failed to contest 10 of the 18 by-elections held. In all cases bar one this has meant an easy win to the party holding the seat. In the case of Cunningham in 2002, however, the absence of a Liberal candidate helped the Green candidate win the seat from the ALP.⁽⁶⁾

By way of contrast, oppositions have contested every by-election since Bradfield in 1952. In seven cases (5.0 per cent), over the whole period the Opposition has failed to contest a by-election (Darling Downs 1901, Echuca 1907, Dalley 1915, Echuca 1919, Wide Bay 1928, Balaclava 1929, Bradfield 1952).

• During the period of voluntary voting (1901-24) the average turnout was just 56.7 per cent.⁽⁷⁾ The lowest figure was 15.1 per cent (East Sydney 1903).
• Since the introduction of compulsory voting in 1924, the average by-election turnout figure has been 87.8 per cent. This contrasts with a 95.1 per cent turnout in general elections over that period.
• During the compulsory voting years, there has been a slight fall in the by-election turnout figure. Prior to 1970 the turnout in 62 by-elections was 88.5 per cent; the turnout for the 44 by-elections since 1970 has been 86.8 per cent. In the 18 contests between 1991 and 2002 the turnout has been 86.7 per cent.
• The lowest turnout figure for the compulsory voting period has been 69.5 per cent (Wentworth 1981).

In 34 of the 140 by-elections (24.3 per cent) a seat has altered its party complexion:

 

Division	Government	Sitting party	Winning party
Melbourne (Vic) 1904	Prot	Prot	ALP
Riverina (NSW) 1904	ALP	FT	Prot
Adelaide (SA) 1908	Prot	Prot	ALP
Boothby (SA) 1911	ALP	ALP	Lib
Grampians (Vic) 1915	ALP	ALP	Lib
Wide Bay (Qld) 1915	ALP	ALP	Lib
Swan (WA) 1918	Nat	Nat	ALP
Corangamite (Vic) 1918	Nat	Nat	VFU
Echuca (Vic) 1919	Nat	Nat	VFU
Ballarat (Vic) 1920	Nat	Nat	ALP
Kalgoorlie (WA) 1920	Nat	ALP	Nat
Maranoa (Qld) 1921	Nat	ALP	CP
Wide Bay (Qld) 1928	Nat	Nat	CP
Franklin (Tas) 1929	ALP	Ind	ALP
Parkes (NSW) 1931	ALP	ALP	Nat
East Sydney (NSW) 1932	UAP	UAP	LLab
Darling Downs (Qld) 1936	UAP	UAP	CP
Gwydir (NSW) 1937	UAP/CP	CP	ALP
Wakefield (SA) 1938	UAP/CP	UAP	ALP
Wilmot (Tas) 1939	UAP/CP	UAP	ALP
Corio (Vic) 1940	UAP	UAP	ALP
Henty (Vic) 1946	ALP	Ind	LP
Flinders (Vic) 1952	LP/CP	LP	ALP
Calare (NSW) 1960	LP/CP	LP	CP
Dawson (Qld) 1966	LP/CP	CP	ALP
Corio (Vic) 1967	LP/CP	LP	ALP
Bass (Tas) 1975	ALP	ALP	LP
Lowe (NSW) 1982	LP/NP	LP	ALP
Adelaide (SA) 1988	ALP	ALP	LP
Groom (Qld) 1988	ALP	NP	LP
Wills (Vic) 1992	ALP	ALP	Ind
Canberra (ACT) 1995	ALP	ALP	LP
Ryan (Qld) 2001	LP/NP	LP	ALP
Cunningham (NSW) 2002	LP/NP	ALP	Grn

Note: By the time of the Grampians by-election of 1917, the Liberal Party had been subsumed by the Nationalist Party. This by-election is therefore not included as an occasion when the seat changed party hands.

• Twenty-seven (19.3 per cent) by-elections have been lost by a major party to another major party.

Four (2.9 per cent) have been lost by a major party to a minor party (Corangamite 1918, Echuca 1919, East Sydney 1932, Cunningham 2002).

• One (0.7 per cent) has been lost by a major party to an independent (Wills 1992).
• Two (1.4 per cent) previously held by independents were won by a major party (Franklin 1929, Henty 1946).

• Twenty-four by-elections (17.1 per cent) have been lost by the government of the day.
• Four (2.9 per cent) have been lost by the opposition of the day (Riverina 1904, Kalgoorlie 1920, Maranoa 1921, Cunningham 2002).
• Four (2.9 per cent) have been lost by one Coalition partner to another (Wide Bay 1928, Darling Downs 1936, Calare 1960, Groom 1988). Note that the losses of Corangamite in 1918 and Echuca in 1919 by the Nationalist Party were both to the Victorian Farmers Union which shortly thereafter formed the Country Party. However, as there was no formal anti-Labor coalition at this time, these are not included here.
• Two seats (1.4 per cent) that had been held by an independent MP were won by the government of the day (Franklin 1929, Henty 1946).

In terms of winning seats from other parties, over the long haul the major non-Labor parties of the day had not done quite as well as its rivals:

* These figures include Riverina 1904, won by the Protectionists from the Free Traders.

On nine of the 34 occasions where a seat has changed party hands at a by-election (Boothby 1911, Swan 1918, Kalgoorlie 1920, Wakefield 1938, Wilmot 1939, Flinders 1952, Adelaide 1988, Canberra 1995 and Ryan 2001), the party that won the seat at the by-election lost the seat at the next general election.

The fact that relatively few seats have changed party hands in by-elections is probably due more to the propensity for by-elections to occur in safe seats, rather than any other factor. Political parties and individual members are only too aware of the possible political consequences of losing a seat at a by-election, and thus try to ensure that by-elections caused by resignation occur only in relatively safe seats. For example, since 1949, of the 52 by-elections that were caused by the resignation of the sitting member, only eight have been in marginal seats (i.e. seats requiring a swing of less than 6.0 per cent to change hands). By far the largest number of by-elections, 32, have occurred in safe seats (i.e. seats requiring a swing of over 10 per cent to change hands).

Of the 67 seats vacated by death, 15 have been lost by the party holding the seat, identical figures for the 67 seats vacated by resignation.

Apart from a party's success or defeat in a by-election, the most important aspect of a by-election is the swing that takes place. Conventional wisdom holds that there is usually a swing against the government of the day at a by-election. The success or otherwise of a government, opposition, or party leader at the by-election is often measured by the size of the swing in comparison with the average swing recorded in past by-elections. However, by-elections occur in varied and disparate electoral divisions, with different numbers and mixes of candidates, and with a variety of local, state and national issues involved. It could therefore be argued that given the variety of factors involved, there is no 'normal' by-election swing. By-elections are held in such different circumstances that none can be regarded as typical, and it is generally agreed that any swing that occurs, is usually explained by the special factors pertaining to each by-election.⁽⁹⁾

By-election swings can be calculated by comparing the first preference and two-party preferred votes received by the various parties at the by-election, with the votes at the previous general election. The two-party preferred swing is more commonly used, as it overcomes some of the difficulties inherent in using first preference swings. First preference swings can be affected by the number and nature of candidates and parties contesting the by-election when compared with the previous general election.

Average first preference and two-party preferred vote swings for by-elections held between 1949 and October 2002 are given below. Swings cannot be calculated where one of the major parties (i.e. ALP or LP/NP Coalition) did not stand a candidate at the by-election or the preceding general election, or where the division was uncontested at the by-election or the preceding general election.

The average two-party preferred swing against the government of the day in all by-elections held during the period 1949 to 2002 was 4.0 per cent. The swing against ALP governments, 5.4 per cent, was somewhat higher than the swing against LP/NP Coalition governments, 3.4 per cent, while the swing against the government in government-held seats was 5.0 per cent. The average two-party preferred swing in by-elections caused by the death of the sitting member was half the swing in by-elections caused by the resignation of the member.

Two-party preferred swings at by-elections during the period have varied from an anti-government swing of 16.1 per cent (Canberra 1995) to a swing of 16.2 per cent to the government (McPherson 1981). The largest swing against an ALP Government was the Canberra figure, while the largest swing against a LP/(CP)NP Coalition Government was achieved when Gough Whitlam won Werriwa in 1952 with a 12.4 per cent swing. Two-party preferred swings to the government of the day are a rare event in by-elections, with only ten being recorded during the period. The largest swing to an ALP Government was in Wills in 1992 with a 5.9 per cent swing (although the ALP lost the seat to an independent candidate, P. Cleary) while the largest swing to a LP/NP Coalition Government was in McPherson, referred to above. Because the number of candidates contesting a by-election is generally larger than the number contesting the previous general election, first preference swings against the government tend to be higher than two-party preferred swings.

ALP Government			LP/NP Government
Division	Swing		Division	Swing
Parramatta 1973	-6.6		Balaclava 1951	-3.4
Bass 1975	-14.6		Macquarie 1951	+0.1
Wannon 1983	-1.5		Lyne 1952	-8.8
Bruce 1983	-3.8		Flinders 1952	-11.0
Moreton 1983	-1.2		Werriwa 1952	-12.4
Corangamite 1984	-1.3		Bradfield 1952	n.a.
Hughes 1984	-5.0		Dalley 1953	n.a.
Richmond 1984	+0.5		Lang 1953	-2.5
Scullin 1986	-4.4		Corangamite 1953	-6.5
Adelaide 1988	-8.4		Gwydir 1953	-0.4
Port Adelaide 1988	-11.1		Cook 1955	n.a.
Groom 1988	-5.2		Cunningham 1956	n.a.
Oxley 1988	-11.8		Barker 1956	-9.9
Gwydir 1989	n.a.		Wentworth 1956	n.a.
Menzies 1991	n.a.		Richmond 1957	n.a.
Wills 1992	+5.9		Parramatta 1958	-6.4
Werriwa 1994	-6.3		Hunter 1960	n.a.
Fremantle 1994	+1.0		La Trobe 1960	-7.5
Bonython 1994	-7.8		Balaclava 1960	-4.6
Mackellar 1994	n.a.		Bendigo 1960	+0.1
Warringah 1994	n.a.		Calare 1960	+1.0
Kooyong 1994	n.a.		Higinbotham 1960	-9.2
Canberra 1995	-16.1		Batman 1962	n.a.
Wentworth 1995	n.a.		Grey 1963	+4.9
			East Sydney 1963	n.a.
			Denison 1964	-1.9
			Angas 1964	-2.1
			Parramatta 1964	-3.0
			Robertson 1964	-1.4
			Riverina 1965	-0.9
			Dawson 1966	-11.9
			Kooyong 1966	-3.8
			Corio 1967	-11.1
			Capricornia 1967	+1.5
			Higgins 1968	-0.3
			Curtin 1969	-7.1
			Bendigo 1969	+2.1
			Gwydir 1969	-7.7
			Australian Capital Territory 1970	+13.8
			Chisholm 1970	-2.8
			Murray 1971	-0.4
			Cunningham 1977	-0.3
			Werriwa 1978	-11.3
			Grayndler 1979	-6.9
			McPherson 1981	+16.2
			Boothby 1981	-1.2
			Curtin 1981	-7.5
			Wentworth 1981	-6.2
			Lowe 1982	-8.5
			Flinders 1982	-3.3
			Blaxland 1996	n.a.
			Lindsay 1996	+5.0
			Fraser 1997