While there are many factors that could contribute to informal voting (in particular, unintentional informal voting), previous AEC studies have highlighted the influences of:
However, there are likely to be many other factors (such as public commentary in the lead up to an election) that might also influence levels or patterns of informality. The very nature of the secret ballot (and uniqueness of the election environment for each federal election) means that it is difficult to determine what influences a voter to intentionally or unintentionally cast an informal vote.
Previous AEC research into informal voting has found that divisions where high proportions of the population are not proficient in English tended to have higher levels of informal voting. It is also possible that linguistic and cultural barriers experienced by some electors from non-English speaking backgrounds may amplify problems associated with high candidate numbers and state/federal electoral differences (AEC 2003; AEC 2005a; AEC 2006a; AEC 2009).
Appendix C provides informality rates for all divisions in the 2010 House of Representatives election, along with proportions and rankings calculated from the 2006 Census population within these divisions who reported that they did not speak English well, or did not speak English at all. Analysis indicated a moderate correlation between these variables, with the proportion of the population with lower levels of English proficiency explaining about a third of the total variation in informality rates across divisions 9.
A linear regression showed that the proportion of the population within a division with lower levels of English language proficiency was a statistically significant predictor of informality in the 2010 House of Representatives election 10, indicating that divisions where higher proportions of the population have low levels of English proficiency are likely to have higher levels of informal voting. Table 6 shows, for the 10 divisions with the highest and 10 divisions with the lowest informality rates at the 2010 House of Representatives election, the proportion of the population in each of these divisions who did not speak English well or did not speak English at all, along with rankings based on these proportions. This shows that five out of the 10 divisions with the highest informality rates at the 2010 House of Representatives election also had the five highest proportions of their population with low levels of English language proficiency. The remaining five divisions also had relatively high proportions of their populations with low levels of English proficiency.
Table 14 shows the average number of candidates per division and informality rates by state and territory for the 2001, 2004, 2007 and 2010 House of Representatives elections. The highest numbers of candidates on a ballot paper were recorded for the divisions of Bennelong and Greenway (each in NSW, and each with 11 candidates), while the lowest numbers of candidates were in the divisions of Canberra (ACT), Barton (NSW), Bradfield (NSW), Mackellar (NSW), Werriwa (NSW) and Braddon (Tas.), each with 3 candidates. For the 2010 election, the 10 divisions with the highest informality rates included those with the highest (Greenway) and lowest (Barton) numbers of candidates.
AEC research for previous House of Representatives elections indicated that increasing numbers of candidates are positively related to increases in the proportion of informal votes (AEC 2003; AEC 2005a; AEC 2009). While the 2010 House of Representatives election saw increasing informality rates in each state and territory combined with decreases (or, in the case of the Northern Territory, no change) in the average number of candidates, a linear regression indicates that the change in the number of candidates in each division between the 2007 and 2010 elections was still a significant predictor of the change in informality 11. However, the model is a relatively poor fit, with only about 15 per cent of the total variation within changes in informality rates between the 2007 and 2010 elections explained by the change in the number of candidates.
A multivariate regression model analysing the effects of lower levels of English proficiency and changes in candidate numbers on informality rates within divisions shows that the proportion of the population with lower levels of English proficiency was a stronger predictor of informality rates in 2010 12. However, changes in candidate numbers were a stronger predictor of changes (swings) in informality than lower English proficiency 13.
|House of Representatives elections||NSW||Vic.||Qld||WA||SA||Tas.||ACT||NT||National|
|Average number of candidates per division (no.)|
|Informal votes (%) (a)|
(a) Informal votes as a percentage of all votes cast.
Source: AEC 2002; AEC 2005b; AEC 2008; AEC 2010b.
Formality rules for lower house elections vary between federal and state or territory electoral systems. Key formality requirements for the House of Representatives, and within each state and territory lower house are summarised in Appendix A.
Electoral legislation for some states (New South Wales and Queensland) provides for optional preferential voting, and previous AEC research has suggested that some voters who can cast a formal ballot with a number '1' only or with incomplete numbering at a state election may also be more likely to cast such votes at federal elections, not realising that this is informal under the federal system. This confusion between state and federal voting requirements could also be heightened if the federal event is conducted soon after a state event (AEC 2009).
As shown in Table 15, state elections were held in Queensland, South Australia and Tasmania less than 18 months prior to the 2010 federal election. There does not appear to be a clear pattern between the proximity of the most recent state or territory election and informality rates at the 2010 House of Representatives election. For example, while South Australia and Tasmania both had a state election in March 2010, South Australia recorded the third highest informality rate of any state or territory (5.46 per cent of votes cast) and Tasmania recorded the lowest (4.04 per cent).
|State/territory||Most recent state/territory election date|
|NSW||24 March 2007|
|Vic.||25 November 2006|
|Qld||21 March 2009|
|WA||6 September 2008|
|SA||20 March 2010|
|Tas.||20 March 2010|
|ACT||18 October 2008|
|NT||9 August 2008|
Source: AEC 2010c.
Comparison of informality rates by category for the 2010 House of Representatives election (see Table 8) with the formality requirements at state and territory lower houses provides mixed results. New South Wales and Queensland both have state provisions for optional preferential voting, and showed rates of incompletely numbered ballots at the 2010 House of Representatives election that were higher than the national average (2.40 and 1.89 per cent of all votes cast in New South Wales and Queensland, respectively, compared with 1.69 per cent of votes cast nationally). However, while Tasmania uses partial preferential voting for state elections, and held a state election less than 6 months prior to the 2010 federal election, the rate of incompletely numbered ballots in Tasmania (0.81 per cent of votes cast) was substantially below the national average. The Australian Capital Territory is the only other state or territory using partial preferential voting at lower house elections, and also had a below average informality rate for ballots with incomplete numbering (1.30 per cent).
Analysis for states accepting ticks or crosses as valid first preferences (New South Wales, Victoria, Queensland, and South Australia) also shows mixed results. While the 2010 rates of informal ballots with ticks and crosses for New South Wales and South Australian voters (0.94 per cent and 0.70 per cent, respectively) are above the national average (0.65 per cent), the rates of informal ballots with ticks and crosses for Queensland and Victorian voters (0.54 per cent and 0.41 per cent, respectively) are below the national average.
While voter confusion about the differences between state and federal voting systems may still have influenced incompletely numbered ballots or ballots with ticks and crosses in some states, these results suggest that other factors are more significant.
9. As a linear relationship was assumed between these variables, a Pearson product-movement correlation coefficient was calculated. The value of the calculated Pearson's r was 0.58, indicating a moderate positive correlation. Since the calculated p-value was less than 0.001, the result is statistically significant (would be expected to occur by chance less than one time in a thousand). The calculated value for the coefficient of determination (r2) was 0.33, indicating that about a third of the total variation in informality rates was explained by the proportion of the population with lower levels of English language proficiency.
10. ß = 0.402, p = 0.000, R2 = 0.336 (95% level of confidence).
11. ß = 0.198, p = 0.000, R2 = 0.155 (95% level of confidence).
12. Low English proficiency: ß = 0.403, p = 0.000; Change in candidate number: ß = 0.162, p = 0.018; R2 = 0.362 (95% level of confidence).
13. Low English proficiency: ß = 0.132, p = 0.000; Change in candidate number: ß = 0.198, p = 0.000; R2 = 0.302 (95% level of confidence).