YOUTH EDERAL ELECTION VOTING INTENTIONS: A Statistical and Graphical Analysis of Newspoll Quarterly Data 1996-2010 Dr Ron Brooker June 2011
Acknowledgement The Whitlam Institute, with various partners, including the oundation for young Australians, has been examining questions of young people and the future of Australian democracy and the particular question of political participation by young people - for several years. This analysis of Youth ederal Election Voting Intentions was conducted for the Whitlam Institute by Dr Ron Brooker at the University of Western Sydney. It was only possible due to the generous support of Newspoll by the way of the provision of the data from relevant quarterly surveys for the period 1996-2010. We are particularly indebted to artin O Shannessy, CEO at Newspoll, for his enthusiastic support and advice. Initial work for the Whitlam Institute by Dr Ann Deslandes on the specific question of the influence of young people beyond pioneering new forms of political participation was supported by the NSW Commission for Children and Young People. Thanks are also due to Graham West who, as NSW inister for Youth, took a direct interest in this project and continues to encourage the pursuit of these critical questions. Prepared by Dr. Ron Brooker Editing: The Whitlam Institute ISBN: 978-1-74108-224-1 Copyright: The Whitlam Institute within the University of Western Sydney, 2011 ii
oreword In late 2007, the Whitlam Institute embarked on a modest project to explore the question of community engagement and political participation by younger Australians. This work was driven by a desire to understand how young Australians participate in the democratic life of the country, how they imagined democracy might work better for them, and indeed all citizens, and the potential implications for policy makers. One of the questions emerging from this work was the extent, or otherwise, to which young people are exerting an influence on Australian politics and policy making. It has proved to be an obvious question to ask and a frustratingly difficult one to answer. One relatively fruitful line of inquiry appeared to be the electoral impact of younger voters. Dr Brooker s analysis of the voting intentions of young people at the ederal level is an indepth probe of the Newspoll Quarterly Data over the fourteen-year period from 1996 to 2010. It is worth stressing that this data is voter intention rather than actual votes. It does not allow drilling down to the electorate-by-electorate numbers that occupy so much attention by political strategists and commentators. While these limitations need to be kept well in mind - as does the fact that the analysis of this particular data does not address the impact of the level of voter registrations, informal votes or preferences, our interest is in trends and their implications. In this respect the data does offer valuable insights. The broader implications of the analysis will be the subject of a forthcoming paper from the Whitlam Institute. In the meantime, there are several observations worth making by way of commentary. The youth vote refers to those voters from age 18-34 but it is important to recognise differences evident in the data between those 18-24 and those 25-34. Throughout this 14 year, five-election period younger voters were consistently at the extremes in terms of their frequency and scale of varying voter intentions: as a group they are not so much swinging but rather in a state of constant electoral motion. This finding may be indicative of a more fundamental change in the body politic which could be described as the emergence of a fluid electorate. 1 This fluidity of the electorate has a number of implications not least of which is that it accentuates the potential electoral impact of political moments: the capacity for a particular event to ignite a quick and dramatic shift in political sentiment. These observations are consistent with other research pointing to the fact that young people are attached to issues and are not joiners of the established political organisations such as political parties or trades unions. 2 1 A fluid electorate is characterised by a willingness by a growing number of voters in an increasing number of seats to be either politically unaligned or only marginally attached to a particular party such that their vote cannot be assured. Indicators of a fluid electorate would be: more dramatic swings; less uniformity in the nature and extent of changes between electorates; more frequent shifts in voter support between elections. 2 See, for example, Collin, P. (2008) Young People Imagining a New Democracy: Literature Review, Whitlam Institute, Sydney. iii
Gender matters: while young people broadly tend to support of the progressive parties, young women tend to do so more consistently while at times the split is more marked as was the case in the case in the dramatic shift of support towards the Coalition in 2001 among young men. The relative voter stability of older age cohorts (50+s) makes the fluid nature of the youth vote more electorally significant. Young voters represent some 30 percent of the electorate. This proportion of voters means that a major shift in the youth vote (which this analysis suggests is not uncommon) will be sufficient to change an election outcome notwithstanding the lower level of voter registration of the 18-24 s. Bearing these factors in mind, a conservative reading of this Newspoll data indicates that the youth vote had a substantial, possibly determinative, impact on the outcomes of the last four (2001, 2004, 2007 and 2010) ederal elections. Earlier work at the Whitlam Institute indicates that young people are strongly values driven and their attachment is to issues rather than traditional political organisations. Young people, by extension, tend to shop around : they will make decisions based on whose proposal or offer best fits their values on their issue of priority at a given time. This contention finds support in the current analysis; its footprints can be seen, for example, in such shifts as the voting intentions of young men towards the Coalition in 2001, the dramatic move to Labor in 2007 followed by the equally dramatic shift from Labor to the Greens very soon after. In fact the first major shift in this direction appears within the first quarter following the Ruddslide and not after the abandonment of the Carbon Pollution Reduction Scheme in 2010 as is widely believed. The shift in 2010 was a second decisive move by 25-34 year olds, belatedly bringing them back in line with their younger electoral cousins. This analysis suggests that Australian ederal politics has entered uncharted electoral waters in which no political party will be able to assume a substantial core of primary voter support. While much is being made of the emergence of the Greens as a potential third force, this analysis, if correct, suggests that such a conclusion is premature. While the voter support for the Greens is trending up, it is substantially reliant on the youth vote the most fluid segment of the electorate. A look at the youth vote for the Greens over the whole period reflects shifts at different points both for and against- that are just as dramatic as those affecting the Labor Party, for example. The earlier work has suggested that young people may well be lead indicators, rather that rather than being entirely distinct from other age cohorts - with respect to their issues-orientation, diminished organisation allegiance and forms of political participation- they are canaries in the coalmine. There is some evidence in this analysis to suggest that such an observation could be extended to voting behaviour. iv
The implications of such an analysis are profound, not just for political parties but more broadly for policy development, decision-making processes and Australia s democratic institutions. These are matters being explored further in a companion paper being developed by the Whitlam Institute. Eric Sidoti Director Whitlam Institute within the University of Western Sydney v
Youth ederal Election Voting Intentions: A Statistical and Graphical Analysis of Newspoll Quarterly Data 1996 2010 Table of Contents Executive Summary... 2 Introduction... 3 Statistical Analyses of Voting Intentions of Youths and Other Ages in Calendar Years prior to ederal Elections 1998 2010.... 5 Graphical Analysis of Trends in Youth Voting Intentions 1996 2010... 23 Newspolls January/ebruary1996 October/December 1998... 24 Newspolls October/December 1998 October/November 2001... 30 Newspolls October/November 2001 September/October 2004... 35 Newspolls September/October 2004 October/November 2007... Newspolls October/November 2007 July/August 2010... 45 Chasing the cohorts... 50 Gender Analyses of Youth Voting Intentions (Newspoll Data 1996 2010)... 52 Youth Voting Intentions for the ajor Parties and the Greens... 52 Gender Differences in Voting Intentions for ajor Parties and the Greens Prior to the ederal Elections in 1998, 2001, 2004, 2007, and 2010... 56 A Coefficient of Determination Analysis of the Youth Gender Differences.... 81 Appendix A: AEC Reports of ederal House of Representatives Election Outcomes 1996 2010. (Source: Australian Electoral Commission publications)... 87 References:... 90 1
Executive Summary Youth voting intentions over the most recent 15 years of Newspoll quarterly data were investigated using both statistical and graphical means. An initial approach (after Watson, 2008) separated the various age groups, 18 24s, 25 34s, and all others (35+), on their voting intentions for Labor, the Coalition and the Greens in calendar years leading to a ederal Election in order to compute confidence intervals and t tests on any age group differences. Graphs of the youth voting intentions for the parties were contrasted with those of the others and consistent statistically significant differences were found for the vote for the Coalition and the Greens between these groups. While Labor was distinguished on the voting intentions support of the 18 24s from the others in 2010, 2007, and 2004, it was the 25 34s who had the greater voting intentions percentages in 1998 and 1996. A second approach aggregated the age groups, 18 24, 25 34, 35 49, and 50+, and constructed line graphs for voting intentions in the electoral cycles up to 2010. ajor parties trendline graphs were supplemented by specific graphs for Labor, the Coalition and the Greens which were further analysed by correlational analyses to establish the youth groups commonality with the other groups. Youth voting intentions were here confirmed from those of other age groups. Although the 25 34s have appeared to take the lead on separate occasions, it was the 18 24s who were consistently at the extremes in their voting intentions for all parties. The latter s support though was seen to be extremely volatile but generally in the direction of the progressive left oriented parties. A final section used both graphs and a correlational analysis to identify gender differences amongst the youth groups and the others. All age/gender groups were distinguished from the others to a greater or lesser degree, but the 18 24 females were seen to be the most unique. Their consistent variation from the voting intentions of the other age/groups highlights how they must have different priorities in the values they hold concerning their voting preference. Overall this study has confirmed the youth groups, especially the neophyte electors (18 24s), have supported the progressive small l parties over the conservative ones. They have been the most volatile of the groups but nonetheless the most consistently anti conservative. A gender effect has been apparent amongst the 18 24s in which the females support of parties with green or socially supportive policies has outstripped all others. 2
Introduction Stimulated by the findings of the Young People Imagining a New Democracy project 1 conducted by the Whitlam Institute, the oundation for Young Australians and the University of Western Sydney Office of University Engagement, the present research has sought to compare the size and direction of ederal Election voting intentions reported by younger Australians with those of the older age groups. With the generous support of Newspoll s artin O Shannessy, S Excel spreadsheets of the quarterly cumulative polling records of voting intentions at a ederal Election have been provided electronically. These data detail the weighted percentage voting intentions of various age 2 and gender groupings for all the political parties, both major and minor, for each quarter from January 1996 through to the polling just prior to the August 2010 Election. Total percentages for the whole sample, for males, and for females, and the actual sample sizes were also provided. To balance the effects of the various age, gender and education groupings that exist in the electorate population, Newspoll develops figures labelled WTD. RESP. (000s) for each aggregated polling period. These Weighted 3 Respondents numbers that are reported in thousands reflect the numbers in each of the nominated age and gender groups of the ederal Electorate. or example, the Total WTD. RESP. (000s) figure for 30th Jul 1st Aug & 6th 8th Aug is 14906 which estimates that the total electorate comprises 14,906,000 people. The Australian Electoral Commission (AEC) statistics that are found on the Enrolled Electors by Division, Age Groups and Gender for Election 2010 page 4 state that the Grand Total for 2010 enrolment after high court decision was 14,088,260. On this occasion, Newspoll s estimate of the numbers in the ederal Electorate was 5% greater due to their policy of asking respondents are they eligible to vote rather than are you enrolled or do you vote? 5. Notwithstanding this small sampling variability, Newspoll s estimates of the Electorate populations in the various age groupings are generally in close proximity to both AEC and Australian Bureau of Statistics (ABS) data and they provide a clear indication of the influence of each age x gender group. With respect to the age x gender differences and their impact on the 2010 election, the Weighted Respondents in thousands were converted to averaged percentages across the periods of aggregated polling. These are reported in the following Table. 1 Arvanitakis, J. and arren, S. (2009). Putting the politics back into Politics: Young people and democracy in Australia. Discussion Paper. Available at http://www.whitlam.org/news_and_events/news/?a=82776. Accessed 2 nd ebruary, 2011. 2 The gender divisions, male and female, were sub divided by the age groupings of 18 24, 25 34, 18 34, 35 49, and 50+. 3 Newspoll weights its results by age/sex/area/age left school to best match the known characteristics of each category. 4 http://www.aec.gov.au/enrolling_to_vote/enrolment_stats/elector_count/index.htm (Accessed December 6, 2010) 5 Newspoll report their polling accuracy falls well below an expected sampling error of +/ 2 percentage points. http://www.newspoll.com.au/system/files/f7/o137/2010%2009%20september%20 %20Newspoll%20topic%20paper%20election%202010.pdf (Accessed December 6, 2010) 3
Table 1: Age x Gender Aggregated Election Percentages prior to 2010 ederal Election 18 24 ales 25 34 ales 35 49 ales 50+ ales TOTAL ales 18 24 emales 25 34 emales 35 49 emales 50+ emales TOTAL emales 6.15% 8.80% 14.09% 20.29% 18 34 ales 35 50+ ales 49.34 % 6.02% 8.74% 14.03% 21.88% 18 34 emales 35 50+ emales 50.66 % 14.95% 34.38% 14.76% 35.91% 18 34s Total 35 50+s Total 29.71% 70.29% As the bottom row in Table 1 shows the Youth vote comprises approximately 30% of the electorate, that is, it contributes three tenths to the overall outcome of an election. A not insignificant proportion. While the percentages of Youths 18 34 (males and females) are nearly equal (approximately 15%), the differences between the 25 34 males and females and the 18 24 males and females groups percentages [(8.8+8.74) (6.15+6.02)=5.37%] are apparently due to the different age ranges of 9 and 7 years, and a more diligent enrolment by the older group. Leading up to the 2010 Election, enrolment of the 18 25s was 2.5% below the target of 80%. As outlined by the AEC 6 in the Annual Report 2009 2010, 77.5% of 18 25s were enrolled to vote at 30 June 2010. This target of 80% is well below the overall expected national enrolment in Australia s compulsory voting 7 environment that the AEC has set at 95% 8. Similarly, the AEC s 2004 Report which uses adjusted ABS data to provide overall estimates of participation, indicates that a rate of 81.6% for 18 25s is the norm. luctuations in participation rates from a low 58.29% for 18 year olds to a high of 89.08% for 25 year olds have resulted in this substantially lower (by 15 percentage points) participation rate for 18 25s. The recommendations of the Young People and Democracy in Australia Discussion Paper need to be implemented to arrest this lack of engagement. With such a rich resource as the cumulative polling records of Newspoll, it decided that a variety of approaches would be required to best reflect the characteristics and the trends of the youth voting intentions. There are three separate sections that follow in which different perspectives of the youth voting intentions over the fifteen years of Newspoll polling are provided. 6 http://www.aec.gov.au/about_aec/publications/annual_reports/2010/files/aec ar 10 Outcome 1.pdf (Accessed December 8,2010) 7 Some analysts say compulsory voting and its corollary of full enrolment are supported by the Progressive parties, Labor and the Greens, but opposed by the Conservatives. (Bennett, 2008) 8 Hallett (2007, p. 21) 4
Statistical Analyses of Voting Intentions of Youths and Other Ages in Calendar Years prior to ederal Elections 1998 2010. The first approach involved the computation of means and confidence limits (confidence intervals) of the 1996 2010 percentage data provided. Using the icrosoft Excel software, these data were graphed to ascertain if particular age groups, namely, the Youth groups of 18 24 years, 25 34 years, and 18 34 were significantly different in their voting intentions to the total sample minus these age groups. This approach is modelled on that used by Watson (2008) to examine the conservative voting intentions of all age groups from 1987 to 2007. As such it also comes with the caveat that it s worth reflecting on the difficulty of separating age effects, cohort effects and period effects 9. In statistical terms, the outcome of interest voter support is the product of the additive effects of age, cohort and period. As such, these three effects are always confounded, and without additional outside information, attempts to precisely disentangle these three effects are formidable. (Watson, 2008, p.5). The sampling variability resulting from the use of survey data is managed by the construction of 95 percent confidence intervals about the figures (the point estimates ) provided by Newspoll. These confidence intervals based on the individual sizes of the sampled groups were integrated with the point estimates to construct lines in graphs that showed age group differences in their voting intentions. Preliminary analyses of the most recent data established that a systematic approach which included data from a full calendar year before a ederal Election would be representative of the differences between various age groups voting intentions and their respective influence the eventual outcome. The first of the analyses included the October December polls in 2009 and the four aggregated summaries 10 of polls prior to the August 21 st Election in 2010. Graphs that incorporated the mean percentage support of each major party by their Newspoll grouping, namely the Coalition, Labor, the Greens, and Others, along with their corresponding 95% confidence intervals 11 were generated for inspection. This visual examination of the point estimates (means) and their associated confidence intervals follows a procedure that Watson (2008) describes as common practice. If the two confidence intervals do not overlap, then the difference between the two point estimates can be regarded as statistically significant (p.11). Whenever confidence intervals did not overlap, t tests were conducted to confirm whether these differences between the means had reached significance at p<0.05 or greater 12. Situations where the overlap was visually only marginal were also included in this analysis to provide a mathematical check of whether means with overlapping confidence intervals were not statistically different from each other (Wolfe & Hanley, 2002, p.65). 1998 The ederal Election on 3 rd October, 1998 was won by the Coalition. There were five quarters of polling in the calendar year leading to that Election and a summary of the voting intentions over those quarters follows. 9 Italics added. 10 Namely, six fortnightly polls in the January to arch quarter, five polls in the April to June quarter, and four polls in July and August. 11 Confidence intervals were computed using the sample sizes for each of the age groupings. 12 Degrees of freedom (df=) will only be reported when they fall below 120. 5
45 V I % 35 30 ean Upper CL Lower CL 25 Other Ages Total for Coalition 18 24s for Coalition 25 34s for Coalition 18 34s for Coalition Coalition Voting Intentions 1998 igure 1: Youth Groups and Other Ages Coalition Voting Intentions prior to 1998 ederal Election Both the confidence intervals shown in igure 17 and t tests conducted on the differences with the Other Ages total 13 confirm that all the Youth groupings were statistically significant. While the differences between the 18 24s and the Other Ages total only reached the p<0.01 level, the 25 34s and the combined 18 34s were also both statistically different from the Other Ages total but at the higher p<0.001 level. The 25 34s had pulled the combined 18 34s mean into this similar variation from the Total of the Other Ages. The 18 24s differed from the Total (t=3.24, p<0.01), while the 25 34s difference was a most substantial (t=5.36, p<0.001) and the combined 18 34s difference (t=5.77, p<0.05) was equally significant. The Youth voting intentions here have indicated that they have a lesser amount of support for the Coalition than Other Ages in 1998. 13 Total of all age groups excluding the 18 34s (Other Ages) 6
45 V I % 35 30 ean Upper CL Lower CL 25 Other Ages Total for Labor 18 24s for Labor 25 34s for Labor 18 34s for Labor Labor Voting Intentions 1998 igure 2: Youth Groups and Other Ages Labor Voting Intentions prior to 1998 ederal Election The corresponding graph (igure 2) for Labor indicates that the 25 34s were leading the support for that party. The 18 24s were below the Other Ages total s average and, as their confidence interval overlaps the Other Ages, a t test was used to confirm that this difference was not significant. The combined 18 34s just reach significance (t=2.23, p<0.05) while the 25 34s have the strongest difference (t=3.27, p<0.01). 8 6 ean Upper CL Lower CL V I % 4 2 0 Other Ages Total for Greens 18 24s for Greens 25 34s for Greens 18 34s for Greens GreensVoting Intentions 1998 igure 3: Youth Groups and Other Ages Greens Voting Intentions prior to 1998 ederal Election The missing support for Labor amongst the 18 24s during this Election cycle appears to have been directed to the Greens. The 18 24s mean voting intentions percentage of 4.5 is substantially higher than either the 25 34s or the Other Ages total. These statistically significant differences were found 7
between the two Youth groupings and with the differences with the Other Ages total, the Other Ages total Vs 18 34s (t=10.98, p<0.001), the Other Ages total Vs 18 24s (t=7.89, p<0.001), and the Other Ages total Vs 25 34s (t=7.74, p<0.001). The Greens had made their first impact with the Electorate by winning support from the 18 24s, their 4.5% was statistically different from even the 25 34s at 3.1% (t=3.03, p<0.01). eanwhile, the older age groups (Other Ages total) at 1.2% had failed to see the Greens as a voting option. or a further insight of the age differences, the trends between the parties during the aggregation of the five polls recorded during the calendar year were included in order to compare the 18 34s with the Other Ages total with horizontal line graphs. Line graphs of the five polling periods from November 1997 to the poll just prior to the ederal Election are presented below in pairs in order that a visual inspection of voting intentions between these age groups can be made. The comparisons of the Youth voting intentions and older age groups (Other Ages total) in igures 4a and 4b follow a similar pattern to that displayed in igures 1 3 above. 50 30 Labor Coalition Greens Others 20 10 0 (97#4) (98#1) (98#2) (98#3) (98#4) 18 34s igure 4a: Youth Group s Voting Intentions prior to 1998 ederal Election 8
50 30 Labor Coalition Greens Others 20 10 0 (97#4) (98#1) (98#2) (98#3) (98#4) The Other Ages igure 4b: The Other Age s Voting Intentions prior to 1998 ederal Election While the Labor support around % appears the same between the groups, the older ages have a consistently higher support for the Coalition. Support for the Others (Australian Democrats, Pauline Hanson, and the Independents) is somewhat lower amongst the 18 34s, but it is their Greens support which, although fluctuating, lifts just prior to the Election. Similar systematic investigations of voting intentions for the other political parties using graphs follow. 2001 The ederal poll in 2001 took place on 10 th November of that year. The five quarters prior to the election included the poll in October/December 2000. The Coalition was seeking re election for the third time. 45 V I % 35 30 ean Upper CL Lower CL 25 Other Ages Total for Coalition 18 24s for Coalition 25 34s for Coalition 18 34s for Coalition Coalition Voting Intentions 2001 igure 5: Youth Groups and Other Ages Coalition Voting Intentions prior to 2001 ederal Election 9
On this occasion (see igure 5), the 18 24s Youth voting intentions for the Coalition were surprisingly comparable with the Other Age electors. All of the confidence intervals, with the exception of the 25 34s, are overlapping each other and a t test (t=4.51 p<0.001) found that the 25 34s were, in fact, statistically different from the total of older voters intentions. Their lack of support for the Coalition was also successful in dropping the combined 18 34s mean and confidence interval and it too was found to be statistically significant (t=3.18) at the p<0.001 level. 45 V I % 35 30 ean Upper CL Lower CL 25 Other Ages Total for Labor 18 24s for Labor 25 34s for Labor 18 34s for Labor Labor Voting Intentions 2001 igure 6: Youth Groups and Other Ages Labor Voting Intentions prior to 2001 ederal Election While igure 6 shows the combined Youth voting intentions for Labor were extremely close to that of the Other Ages total, it is clear that the youngest, 18 24s, ranged more widely, and that they were on average lower than all of the other groups. None of the differences were statistically significant. 10
8 6 V I % 4 2 ean Upper CL Lower CL 0 Other Ages Total for Greens 18 24s for Greens 25 34s for Greens GreensVoting Intentions 2001 18 34s for Greens igure 7: Youth Groups and Other Ages Greens Voting Intentions prior to 2001 ederal Election As displayed with the non overlapping confidence intervals in igure 7, all three groupings of the Youth vote are significantly different (t=3.72, 4.69, and 3.31, respectively) in their intentions to vote for the Greens from the Other Ages total. All three reach the p<0.001 level, but it should be noted that the scale on the above graph is a fraction of the scales for the graphs for the major parties and that the sizes of the percentage differences between the Greens and the major parties are similar. It should also be noted that the confidence intervals have been stretched by the vertical axis being reduced to 8 percentage points. igures 8a and 8b provide an interesting comparison between the various parties during the calendar year preceding the 2001 Tampa ederal Election, which shows the Youth voting intentions for Labor dropping in the last quarter while the Coalition s support was increasing. The Greens and the Others remained constant throughout all of the polls and they were not at this time boosted by the loss of support for Labor. 11
50 30 Labor Coalition Greens Others 20 10 0 (00#4) (01#1) (01#2) (01#3) (01#4) 18 34s igure 8a: Youth Group s Voting Intentions prior to 2001 ederal Election 50 30 Labor Coalition Greens Others 20 10 0 (00#4) (01#1) (01#2) (01#3) (01#4) The Other Ages igure 8b: Other Age Groups Voting Intentions prior to 2001 ederal Election Voting intentions of the older respondents (igure 8b) have ultimately favoured the Coalition but surprisingly supported Labor for three quarters after the 1998 election. In the final poll of the Other Ages respondents, the Others, and quite noticeably the Greens, remained at the bottom. 2004 The ederal Election prior to 2007 took place on 9 th October, 2004. Data for the full calendar year begins with 4 th poll in 2003 and ends with the 4 th poll in 2004. The Coalition going for its 4 th term had variable support from the different age groups. The graphs follow. 12
50 45 ean Upper CL Lower CL V I % 35 30 Other Ages Total for Coalition 18 24s for Coalition 25 34s for Coalition 18 34s for Coalition Coalition Voting Intentions 2004 igure 9: Youth Groups and Other Ages Coalition Voting Intentions prior to 2004 ederal Election Clearly the Youth vote was again failing to support the conservative parties. igure 9 shows all three divisions, 18 24s, 25 34s, and the combined 18 34s, were significantly different to the Other Ages total (t=7.1, t=7.4, and t=10.03, respectively). 50 45 V I % 35 ean Upper CL Lower CL 30 Other Ages Total for Labor 18 24s for Labor 25 34s for Labor 18 34s for Labor Labor Voting Intentions 2004 igure 10: Youth Groups and Other Ages Labor Voting Intentions prior to 2004 ederal Election While igure 10 indicates that most of the Labor voting intentions are not differentiated between the age groups, the 18 24s stand out with their confidence interval not overlapping with the Other Ages confidence interval (t=2.92, p<0.01). They again appear to be Labor s most ardent supporters by exceeding all the other groups. 13
The Greens with a much lower percentage of the first preference voting intentions do not appear to differ greatly between the age groups but they are nonetheless statistically significant. The graphs and the t test results follow. igure 11 shows the propensity for the 25 34s to be better supporters of the Greens than any other group but these differences come from a very low base. 20 ean V I % 15 10 Upper CL Lower CL 5 0 Other Ages Total for Greens 18 24s for Greens 25 34s for Greens GreensVoting Intentions 2004 18 34s for Greens igure 11: Youth Groups and Other Ages Greens Voting Intentions prior to 2004 ederal Election The Other Ages total Vs 18 24s (t=5.29, p<.001); the Other Ages total Vs 25 34s (t=7.96, p<.001); and the Other Ages total Vs 18 34s (t=9.2, p<.001). The trend graphs (igures 12a and 12b) for 2004 indicate a similar narrowing of the differences between the major parties as the election looms with the Youth voters being the stronger Labor supporters. The Greens again trail the other lesser parties (the Others) aggregated totals but there is a noticeable difference in the size of the Youth support at approximately 10% and the Other Ages around 5%. 14
50 30 Labor Coalition Greens Others 20 10 0 (03#4) (04#1) (04#2) (04#3) (04#4) 18 34s igure 12a: Youth Groups Voting Intentions prior to 2004 ederal Election 50 30 Labor Coalition Greens Others 20 10 0 (03#4) (04#1) (04#2) (04#3) (04#4) The Other Ages igure 12b: Other Age Groups Voting Intentions prior to 2004 ederal Election 2007 The ederal Election prior to 2010 took place on 24 th November, 2007. Data for the full calendar year begins with 4 th poll in 2006 and ends with the 4 th poll in 2007. Beginning with the Coalition there is once again a significant lack of support from the Youth compared to the older voters. The following graph (igure 13) shows that the Youth confidence intervals do not overlap with the Other Ages total s confidence intervals. A series of t tests confirmed that the Youth voting intentions were significantly different to the older age groups, the Other Ages total percentage Vs 18 24s (t=7.59, p<0.001), the Other Ages total percentage Vs 25 34s (t=11.08, p<0.001), and the Other Ages total percentage Vs 18 34s (t=12.70, p<0.001). 15
55 50 VI for Coalition % 45 35 ean Upper CL Lower CL 30 Other Ages Total for the Coalition 18 24s for Coalition 25 34s for Coalition 18 34s for Coalition Coalition Voting Intentions 2007 igure 13: Youth Groups and Other Ages Coalition Voting Intentions prior to 2007 ederal Election By contrast, the voting intentions for Labor across the age groups were opposite. igure 14 shows the 18 24s leading the charge with an average of 50.44% intending to give Labor their first preference and, with the 25 34s 48.4%, they had a combined 48.62%. The Other Ages total of 45.69% was significantly different to all three combinations of the Youth vote. The 18 24s and the combined 18 35s were statistically significant at the p<0.001 level (t=4.21 and t=4.1, respectively), while at t=3.17, the 25 34s and the Other Ages total were still statistically different at the p<0.01 level. The statistical differences of the non overlapping confidence intervals (see igure 14 below) have been confirmed by these statistical analyses. 16
55 50 VI for Labor % 45 35 ean Upper CL Lower CL 30 Other Ages Total for the Coalition 18 24s for Labor 25 34s for Labor 18 34s for Labor Labor Voting Intentions 2007 igure 14: Youth Groups and Other Ages Labor Voting Intentions prior to 2007 ederal Election The Greens analysis shows a similar picture. While none of the Youth voting intentions confidence intervals overlap the Other Ages, the 25 34s are this time leading the charge. Their difference reached significance on a t test (t=8.2, p<0.001). The 18 24s are similarly different to the Other Ages (t=4.34, p<0.001) but the graph indicates that the 25 34s are ahead in their support for the Greens. 17
20 Greens Voting Intentions 2007 ean VI for Greens % 15 10 Upper CL Lower CL 5 0 Other Ages Total for Greens 18 24s for Greens 25 34s for Greens 18 34s for Greens igure 15: Youth Groups and Other Ages Greens Voting Intentions prior to 2007 ederal Election The trend analyses for the five polls prior to the 2007 ederal Election are presented in the following graphs. 60 50 30 Labor Coalition Greens Others 20 10 0 (06#4) (07#1) (07#2) (07#3) (07#4) 18 34s igure 16a: Youth Groups Voting Intentions prior to 2007 ederal Election 18
60 50 30 Labor Coalition Greens Others 20 10 0 (06#4) (07#1) (07#2) (07#3) (07#4) The Other Ages igure 16b: Other Age Groups Voting Intentions prior to 2007 ederal Election The above graphs indicate that a higher proportion of the 18 34s was intending to vote Labor than there was intending to vote for the Coalition. The separation between the two major parties was accentuated in the sample of the 18 34s where there was over 20 percentage points between them in the April/June polls (07#2) of 2007. During this time series, there was little difference between the Greens and the other smaller parties in both age groups. Notably, the Greens fell below the aggregated voting intentions for the Others. 2010 igure 17 compares the means and confidence intervals of the Coalition voting intentions of the Youth groupings of 18 24 years, 25 34 years, and 18 34 years with the Other Age groupings total. As the graph in igure 17 shows none of the Youth groupings confidence intervals overlap the Other Ages total s confidence interval. The t test comparison of the mean for 18 24 s voting intentions with the Other Ages total mean resulted in t=10.05, p<0.001. During 2010, there was plainly a significant difference between the Coalition voting intentions of 18 24s and all Other Age voters, one where the youngest ones had a much lower support for the Liberals, Liberal nationals, and the Nationals. As observed in igure 17, a less pronounced difference but nonetheless a quite significant one was also found in the difference between 25 34s and the Other Ages total (t=6.89, p<0.001). A similar relationship exists between the combined Youth vote (18 34s) and all Other Ages. The t test conducted showed there was still a significant difference (t=11.79, p<0.001) even though the 25 34 component of this group had been weighed down by the substantial lack of support for the Coalition by the 18 24s. 19
45 ean Upper CL Lower CL VI for Coalition % 35 30 25 Other Ages Total for Coalition 18 24s for Coalition 25 34s for Coalition 18 34s for Coalition Coalition Voting Intentions 2010 igure 17: Youth Groups and Other Ages Coalition Voting Intentions prior to 2010 ederal Election 45 VI for Labor % 35 30 ean Upper CL Lower CL 25 Other Ages Total for Labor 18 24s for Labor 25 34s for Labor 18 34s for Labor Labor Voting Intention 2010 igure18: Youth Groups and Other Ages Labor Voting Intentions prior to 2010 ederal Election As the above graph (igure 18) shows all of the age groupings overlap. In fact, their means are relatively similar and this outcome was confirmed by the relevant t tests. The Labor voting intentions of all age groups were quite close except for the two Youth groupings which appear to have counter balanced each other and brought their combined mean and confidence interval in line with the Other Ages total mean. 20
25 20 VI for Greens % 15 10 ean Upper CL Lower CL 5 Other Ages Total for Greens 18 24s for Greens 25 34s for Greens 18 34s for Greens Greens Voting Intention 2010 igure 19: Youth Groups and Other Ages Greens Voting Intentions prior to 2010 ederal Election The Greens voting intentions shows a considerable disparity between the Youth vote lead by the 18 24s and those voters over 35 years. These two groups mean difference is nearly ten percentage points and a t test confirmed that the Other Ages total was quite significantly different to that for the 18 24s (t=10.74, p<0.001). Similarly, t tests confirmed that mean differences between the Other Ages total and the 25 34s, and the Other Ages total and the combined 18 34s were statistically significant (t=6.97, t=15.23, respectively). igure 20a shows that during the year, the Youth vote for the Greens has climbed to a peak of 21.8% at the Election Poll. The Youth voting intentions for Labor were in a steady decline throughout the year and they were overtaken by an increase in the voting intentions for the Coalition in the July/August polls. 21
50 V I % 30 20 Labor Coalition Greens Others 10 0 ('09#4) ('10#1) ('10#2) ('10#3) ('10#3a) (Election Poll) 18 34s igure 20a: Youth Groups (18 34s) Voting Intentions prior to 2010 ederal Election V I % 50 30 20 Labor Coalition Greens Others 10 0 ('09#4) ('10#1) ('10#2) ('10#3) ('10#3a) (Election Poll) The Other Ages igure 20b: Other Age Groups Voting Intentions prior to 2010 ederal Election The older voters, 35 years and above, show a growing support for the Coalition along with a possibly event related 14 decline in voting intentions for the Labor party. The Greens increased support by this group was marginal (2.3%) when compared with the increase of the Youth voting intentions for the Greens (5.25%). This first approach which chose the calendar year time period for analysis has shown the 18 24s and the 25 34s each showing the way in different elections. The last three comparisons of voting intentions for the major parties, namely, 2004, 2007, and 2010, were highlighted by the 18 24s 14 Events such as a change of leadership or a stance on a particular issue affect the polling at the times when they are most current. 22
leadership in rejecting the Coalition in favour of Labor and the Greens. The earlier elections in 1998 and 2001 saw the emergence of the 25 34s as leaders in their support of Labor 15. The comparisons of the voting intentions trends consistently show both the variability (from the other age groups) and the volatility of the youth vote. Whereas the Other Age groups were relatively static (flat lining), the 18 34s were highly fluid within this set of relatively short time frames. Here is evidence that the youth groups, albeit the smallest proportion of the ederal electorate, were the ones that changed the most. With the other ages being static, the youth effect on the outcomes is accentuated. Graphical Analysis of Trends in Youth Voting Intentions 1996 2010 To differentiate in greater detail between the various age groupings, a second approach was undertaken in which trends of voting intentions were compared across the entire period between elections. Using the Newspoll data of ederal Election voting intentions accumulated quarterly, a series of graphs and statistical analyses for the periods between elections were generated for the two major parties and the Greens. The dataset provided by Newspoll consisted of 61 polling periods with 13 aggregated EDERAL VOTING CUULATIVE STUDY spreadsheets for the quarters between the six Elections during the time from arch, 1996 to August, 2010. To determine the trends in youth voting intentions, this naturally occurring three year time frame was chosen because the Newspoll reported age groups were variable, one limited to seven years (18 24s) with approximately two Elections, and another limited to ten years (25 34s) with approximately three Elections. Newspoll also reported the consolidated age group of 18 34 that covered the first 17 years of a person s voting life which accounts for the five ederal Elections covered by the dataset. These Youth age groups were also being contrasted against the middle 16 years (35 49s) and the seniors 30 and beyond years (50+s). The Australian Electoral Commission (AEC) website reports that in 2010 the 18 34s comprise 27% of the national electorate, of which 11% are 18 24 year olds and 16% are 25 34. The other age groups, 35 49 and 50+ comprise 27% and 46%, respectively. In each of the chosen time frames, an initial graphical comparison of the age groups voting intentions for both Labor and the Coalition is followed by the line graphs for each major party and the Greens. Correlational 16 analyses are then used to identify those age groups that were contributing most to the Total Voting Intention percentages for each party. Gender differences are reported in the next section. 15 In these elections, the Greens had only begun to contest seats in the House of Representatives. 16 Correlational analyses do not determine cause and effect. In this context, the Pearson s r correlation (r) measures how two or more factors (variables) are related to one another and whether they vary in similar ways. When the correlation coefficient is squared, it indicates how much of the variance or variability in each factor is shared. Plainly, the Total Percentage of voting intentions for a particular party will be highly related to all of its contributors, which, in this case, are the age groups voting intentions for that party. While the Total is not being predicted by the age groups scores (percentages), it does share more or less similarity with each one and r 2 represents that percentage of sharing that is accounted for by a particular contributor. The higher the percentage the more they move (vary) together and the lower the more they vary differently. 23
Newspolls January/ebruary1996 October/December 1998 The first figure represents the differences between the Newspoll s age groups of 18 24, 25 34, 35 49, and 50+ in each group s voting intentions for either Labor or the Coalition during the period between the 1996 and the 1998 Elections. Although this first graph, igure 21, does not include value labels because of its size limitations, polynomial trendlines of best fit 17 for the data were included for the extreme age groups (18 24 and 50+) voting intentions for Labor and the Coalition along with the trendlines corresponding R squared values. R 2 values above 0.7 indicate the trendlines are good representations of the data movements between the quarterly polling periods. Separate figures, igures 22, 23 and 24, were created for the individual parties voting intentions of all the Newspoll age groups. igure 21 indicates that the majority of the 50+ group consistently favoured the Coalition ahead of Labor from 1996 through to 1998. The 18 24s meanwhile had a variable amount of support for both major parties, their trendlines intersect on three occasions over the 31 month period. Notably, the youngest group favoured Labor at the polls that coincided with the Elections. Another feature of this period of polling was that, at the 1997 April/June quarter consolidation, there was a coming together of the 18 24s, 25 34s, and 35 49s age groups voting intentions for both Labor and the Coalition (the 50+ groups maintained their support for the Coalition at Labor s expense). This ambivalence was apparently caused by the major events of this time. Events like the founding of Pauline Hansen s One Nation party (April, 1997), the Sorry business (ay, 1997), and the first restructuring of industrial relations in January, 1997 which may have affected all of the younger (younger than 50) voters. igure 22 has clearly demonstrated that there was a slide in the voting intentions for the Coalition. The 18 24s were the most irregular around a gradual decline while the 50+s dropped the most before a last minute recovery prior to the 1998 Election. At r=.98 (r 2 =.96) the older age group almost mirrors the variability in the Total Voting Intention percentage, while the 18 24s r=.76 (r 2 =.58) showed the greatest difference. Interestingly, when the 18 24s were combined with the 25 34s that divergence disappeared and the 18 34s correlation was r=.95. igure 23 confirms the volatility of the Labor vote from arch 1996 to October 1998. Interestingly, the 18 24s were generally less supportive of Labor than both the 25 34s and the 35 49s until the poll just before the Election. With the exception of the 18 24s, the Youth voting intentions represented by the 25 34s followed the volatility in the other age groups. The 35 49s voting intentions percentage for Labor correlated the highest with the Total Labor percentage (r=.97) while the 18 24s correlated the lowest (r=.61). Their r 2 of 94% and 37%, respectively, indicates the large differences in the variability of these age groups voting intentions. The Greens vote during this period was quite small but the 18 24s were consistently the strongest supporters. Their support at 5% contrasted with the less than 1% by the 50+s. In this situation, the 17 A polynomial trendline is a curved line that is used when data fluctuates. It is useful, for example, for analysing gains and losses over a large data set. (http://office.microsoft.com/en us/help/choosing the besttrendline for your data HP005262321.aspx). The order of the polynomial with the highest R squared represents the line of best fit for that data. 24
18 24s and the 25 34s combined to share the most substantial variability of the Total Voting Intention percentage for the Greens (All 18 34s r=.92, r 2 =.85). In summary, the lead up to the 1998 ederal Election won by the Coalition was characterised by the diversity between the extreme age groups voting intentions. The older group were consistently more conservative, while the youngest showed lesser support for the Coalition and some movement towards the Greens 25
60 Labor and Coalition Voting Intentions of Age Groups between 1996 and 1998 Elections 50 R² = 0.9738 R² = 0.8106 R² = 0.7115 R² = 0.8802 30 18 24s Labor 25 34s Labor 35 49s Labor 50+ Labor 18 24s C'tion 25 34s C'tion 35 49s C'tion 50+ C'tion Poly. (18 24s Labor) Poly. (50+ Labor) Poly. (18 24s C'tion) Poly. (50+ C'tion) 96#1 96#2 96#3 96#4 97#1 97#2 97#3 97#4 98#1 98#2 98#3 98#4 igure 21: Voting Intentions for ajor Parties by Age Groups 1996 1998 26
60 Coalition Voting Intentions of Age Groups between 1996 and 1998 Elections 58.05 58.1 56.85 57.75 55 50 49.55 49.15 R² = 0.9738 50.45 48.55 45.25 46.4 46.5 45.05 45.85 45.6 47.3 46.2 47.45 45.95 42.95 42 43.9 42.9 43.05 42.25 42.5 42.6 43.3 43.9 42.7 42.6.2 39.9 37.7 39.15 38.8 38.1 35.8 35.3 38.9 38.45 38.55 36.35 35.25 R² = 0.7115 36.3 35.8 35.25 37.35 18 24s C'tion 25 34s C'tion 35 49s C'tion 50+ C'tion Poly. (18 24s C'tion) Poly. (50+ C'tion) 30 96#1 96#2 96#3 96#4 97#1 97#2 97#3 97#4 98#1 98#2 98#3 98#4 igure 22: Voting Intentions for the Coalition Parties by Age Groups 1996 1998 27
50 Labor Voting Intentions of Age Groups between 1996 and 1998 Elections 47.95 45.35 44.25 44.35 45.6 44.65 45.05 44.5 43.6 42.15 41.6 41.6 41.8 41.65 R² = 0.8106 43 42.65 42.55 39.8 39.3.05.45.4.45 39.35.15 39.5 38.35 37.9 38.3 39.05 37.8 39.15 38.95 38.25 36.45 35.95 36.2 R² = 0.8802 36.8 35.35 33.4 33 33.5 33.25 33.15 33.9 31.6 18 24s Labor 25 34s Labor 35 49s Labor 50+ Labor Poly. (18 24s Labor) Poly. (50+ Labor) 30 96#1 96#2 96#3 96#4 97#1 97#2 97#3 97#4 98#1 98#2 98#3 98#4 igure 23: Voting Intentions for the Labor Party by Age Groups 1996 1998 28
8 Greens Voting Intentions of Age Groups between 1996 and 1998 Elections 18 24s Grns 25 34s Grns 35 49s Grns 50+ Grns Poly. (18 24s Grns) 6 4 2 0 Poly. (50+ Grns) 5.65 5.85 5.35 5.3 5.35 5.3 4.85 4.8 5 R² = 0.5153 3.95 4.15 3.45 3.55 3.2 3.2 3.3 3.05 2.85 2.85 2.7 1.95 1.9 2 2.3 2.35 2.2 2.25 2.2 2.15 1.9 1.65 1.75 1.3 R² = 0.5142 0.9 0.7 0.65 0.65 0.6 0.45 0.55 0.5 0.55 0.65 0.65 96#2 96#3 96#4 97#1 97#2 97#3 97#4 98#1 98#2 98#3 98#4 igure 24: Voting Intentions for the Greens by Age Groups 1996 1998 29