Economic and social change tends to create significant spatial disparities. Over the last decade, researchers in Australia have analysed the spatial impacts of globalisation,¹ increasing income disparities,² increasing incidence of dependency,³ and internal migration.⁴ Regional disparities in the patterns of population growth⁵ and decline⁶ have been identified, as have regional ‘hot spots’ and ‘cold spots’ with increasing or declining shares of employment in economic activity sectors.⁷ It is now apparent that there is a considerable degree of spatial mismatch in the patterns of population growth, employment growth, and the distribution of investment in new construction activity.⁸

Regionally, industrially, occupationally and socially there are ‘winners’ and ‘losers’ in the contemporary era of globalisation. That there is a strong spatial dimension to this increasing social and economic polarisation in Australia is beyond doubt. As a result, disillusionment, despair and alienation are not only widespread throughout Australian society, but also are locationally specific. The electorate — or at least significant sub-sections of the Australian population — are expressing their anger and disillusionment at the ballot box.

The One Nation phenomenon had captured attention across Australia and overseas following the election of Pauline Hanson, an independent candidate, to the traditional Labor seat of Oxley (centred on the old mining and industrial city of Ipswich in what is increasingly a commuter belt in the western growth corridor of the Brisbane-SEQ metro region) at the March 1996 Commonwealth election. This election saw the electorate toss out the Keating Labor government and elect the Howard-Fischer Liberal-National Coalition government with a huge majority in the House of Representatives. Shortly after her election, Hanson formed the One Nation Party (ONP). Seen widely by politicians, political commentators, the media, and many of the professional segments of society as a distasteful apparition of latent racism, the ONP was largely dismissed as an embarrassing temporary nuisance. While national support for the ONP waxed and waned between about four and ten per cent in the opinion polls, in the run-up to the 1998 Queensland State election, it became apparent that its support across Queensland was as high as 16 or 18 per cent. It would be fair to say that the political pundits fell well short of predicting the devastating inroads the ONP made into the primary vote of the traditional political parties at the 13 June 1998 Queensland State election. The ONP momentum can be seen in Table 1 which describes the pre-election polling and the summary results from the Queensland State Election.

The 13 June 1998 Queensland State election provided a magnitude of voter reaction against the long standing hegemony of the Liberal-National Coalition parties and the Australian Labor Party (ALP). This has reverberated across the nation as the ONP captured approximately 23 per cent of the primary votes cast and won 11 of the 89 seats in the Legislative Assembly, Queensland’s only state parliamentary chamber.

Figure 1 shows the geographic distribution of seats amongst the four political parties and the two independents. The traditional division between the rural seats, dominated by the National Party and the ALP strongholds of the Brisbane metropolitan region and the regional centres (excluding Gladstone), is apparent in the map. The liberals maintain a few blue ribbon seats in Brisbane and share some of the Gold Coast and Sunshine Coast seats with the Nationals.

While Labor formed a minority government with the support of one of the two independents, One Nation is the third largest party in the Legislative Assembly, ahead of the decimated Liberals with just nine seats, and not much behind the Nationals with 23 seats. Table 2 presents the seats which changed hands at the 13 June poll.

The success of the ONP at the Queensland State poll has led to a plethora of media commentary and has crystallised the attention of politicians in the major parties across the nation. With such widespread support across Queensland, and with such a strong representation in the State Parliament at its first electoral foray, the ONP suddenly has had to be taken seriously.

What was behind its burgeoning support? Surely it is more than the racist connotations which are sometimes perceived to be behind ONP’s strong stand against immigration, foreign investment and the Aboriginal programs of governments. Perhaps it is a reflection of the wider disenchantment of politicians, a disillusionment with contemporary political processes, and the seemingly unfair benefits conveyed to particular interest groups. And surely it is more than a reaction from a minority of people who have ranted against the new tough gun control measures introduced nationally following the Port Arthur massacre in April 1996. It may also be a reaction to the fear of unemployment and underemployment which has resulted from economic restructuring, to the uncertainties of the impacts of globalisation and increasing competition, to community unease with crime, and to the increasingly apparent socio-economic disparities between the ‘haves’ and the ‘have-nots’.

Skeffington provides a view of the ONP as a new entrant to a market duopoly, carving out a niche market against dominant incumbents.⁹ A Morgan Gallop Poll conducted in the week before the Queensland State election provides some indication of the concerns of Hanson supporters. Table 3 shows the primary reason people indicated for supporting the ONP.

While it is clear from this poll that ONP supporters are dissatisfied with the current political situation, what is less clear is who the people are and what binds them together with an interest in this new political party. Does a One Nation supporter’s income, occupation, industry, education or family situation partially explain their attraction to the views of the ONP? The quantification of this information is of obvious interest to the major political parties that are now anxious to reinvent themselves and rewrite policy in order to attract disenfranchised voters.

From the results of the Queensland State Election, a focus on the actual votes cast at the voting booth level, enables the spatial distribution of the One Nation vote to be analysed through correlation and regression analysis to model its association with a range of socio-economic, demographic and locality attributes. This has been achieved through developing a geographic information system (GIS) whereby polling booth returns are geo-coded and matched with small area census data and other variables. Using this approach, the paper provides a statistical analysis of the possible explanatory variables for the spatially differentiated incidence of One Nation support at the 13 June 1998 Queensland State Election.

An analysis of voting behaviour is understandably an important part of post-election rituals. In most cases, these analyses are anecdotal, based upon the opinions of seasoned commentators and the politicians themselves, or are found in the results of confidential political party exit polling or focus groups. In academic circles, more formalised approaches and attempts to understand voting behaviour and change are attempted. Some approaches, such as the Australian Election Study, involve a detailed survey of voters on their attitude to major issues, their voting intentions and their corresponding economic and demographic characteristics.¹⁰ Other studies, have attempted to examine social structural voting influences through a multilevel analysis of survey data and demographic characteristics.¹¹

Singleton and others examined the ALP’s vote at the 1996 Federal Election to test the proposition that the party’s traditional blue collar supporters deserted it in that poll. In their approach, the 1996 swing was disaggregated to the booth level and then assigned ad hoc to suburbs. Using a simple one-variable linear regression model, they found considerable evidence to suggest that the Labor vote was highly affected by an erosion of blue collar support.¹²

Goot has examined the voter demography of ONP supporters through evidence gathered on the demographic profiles provided through opinion polls. He also suggests that ONP support is primarily male, low-educated, blue-collar workers.¹³

In general, most studies on voting behaviour, including those mentioned above, have been dogged by data limitations and issues of data aggregation, particularly analyses examining underlying demographics as explanatory variables. Voting behaviour is complex, with numerous variables likely to influence a person’s voting decision. The more disaggregated the data, the greater the ability of a model to examine the characteristics of voting behaviour. However, State and Federal electorates are relatively large in area, particularly in regional areas, and often contain dissimilar population groupings. For example, the Queensland State seat of Redlands on the eastern outskirts of Greater Brisbane has 28,701 enrolled voters and a population at the 1996 census of 42,528. Most federal seats contain over 70,000 voters and in excess of 100,000 persons. The differences between areas within the same electorate are very decisive when larger electorates such as the State electorate of Cook (situated on Cape York Peninsula) are considered. At the 1998 Queensland State Election, the vote of the incumbent Steve Bredhauer (ALP) ranged from 2.56 per cent at the booth of Einasleigh to 93.65 per cent at the booth of Napranum. Analyses which attempt to quantify voting behaviour based on the electorate aggregate for both the electoral results and the demographic, social and economic characteristics are suspect.

In this paper a new approach is developed which utilises GIS techniques to compare a wide range of census (plus some locational) variables and electoral results at their smallest possible spatial grouping; these are the census collector district (CD) and individual polling booth respectively. The aim of the analysis is, therefore, to assess the demographic, social and economic information collected at the 1996 census (ABS 1998) as a determinant of voting behaviour, and in particular the vote for One Nation.

The major issue in this analysis is the process of systematically associating the 1,647 polling booths with the 6,448 CDs across Queensland. To conduct the analysis, a spatial polling booth data-set was constructed. Each of the polling booths was geo-coded with the assistance of raster images scanned from the State Electoral Guides and refined with the assistance of geographical feature maps. The central polling booth data set contains all polling booths at both the 1995 and 1998 Queensland State Elections.

The Australian Electoral Commission observes that approximately 90 per cent of electors cast their vote at local polling booths, with the remainder casting their vote through absentee or pre-poll methods.¹⁴ In this analysis, it is assumed that all the eligible residents of a CD cast their votes at the polling booth closest to the centroid of that CD.

Using this method, each CD is assigned the electoral results (available through the Queensland Electoral Commission’s Web Site¹⁵) for its closest booth. This links the data-sets and makes possible an evaluation of any factor included in the 1996 census with any percentage returns for any candidate (party). This process also permits a spatial examination of booth level data to better understand the full effect of the ONP vote on the main electoral parties.

Figure 2 is a map showing the distribution of the ONP vote allocated by the model to each CD in Queensland through its association with the nearest polling booth. It clearly shows the spatial dimension of the ONP’s vote at the 13 June 1998 State election. Areas of high support for the ONP are identified as those where the ONP candidate received 35 per cent or more of the primary vote. With that level of support, the ONP stood a high chance of receiving allocations of preference from candidates polling lesser levels of primary support. These were sometimes sufficient for the ONP candidate to win the seat.

The data show that the variation in the ONP primary vote across the polling booths in Queensland in electorates where the ONP stood a candidate, ranged from a high of 66.3 per cent in the Burra Burri booth in the electorate of Western Downs (South West Queensland) to a low of 0.65 per cent in the Murray Island booth in the electorate of Cook (Far North Queensland) a booth in an Aboriginal community.

Generally, the lowest ONP vote is recorded in inner Brisbane, Cape York and Torres Strait and Western Queensland. For each regional town or city in which ONP ran a candidate, the lowest component of the ONP vote can be found in the town or city centre, rising dramatically to a peak on the outskirts of the city. For example, in Brisbane the ONP’s vote within a five kilometre radius of the central business district is almost non-existent; however, within 40 kilometres of the city centre the ONP vote reached more than 50 per cent! This phenomenon is consistent for Brisbane, the Gold Coast and Sunshine Coast, Townsville, Cairns, Mount Isa, Rockhampton, Bundaberg and Toowoomba. The pattern is particularly evident in Figure 3 which clearly shows the effect for the centres of Brisbane (3a), Cairns (3b), Townsville (3c) and Toowoomba (3d). Although the ONP constant or base vote is higher the smaller the town or city, this inverse distance decay effect around the metropolis and the regional centres is still evident. It identifies what are considerable belts of ONP support in the ‘peri-urban’ or rural-urban fringe residential areas that have been growing rapidly around Brisbane and other established urban centres in the South East Queensland Figure 2 metropolitan region as well as around many of the large coastal non-metropolitan regional cities. We have termed this phenomena as the ‘urban fringe’ effect.

Figure 3: The urban fringe effect

Perhaps the most dramatic example of the urban fringe effect can be seen in the seat of Yeronga. This seat, held by the Labor Party since 1989, runs along the southern bank of the Brisbane River taking in river-side suburbs such as Fairfield (five kilometres from the city centre), Yeerongpilly and Tennyson, and middle-class suburban areas such as Annerley through to the outer border which takes in the industrial suburb of Rocklea (over 10 kilometres from the city centre). The residential values of areas in Yeronga electorate can be over $500,000 for houses on the Brisbane river but are generally in the $170,000-$240,000 range, the exception being Rocklea where the residential values in this primarily industrial area are around $80,000. The average ONP vote in the electorate of Yeronga, excluding Rocklea, was 12.7 per cent, with a 2.7 per cent swing towards the ALP. However, in Rocklea the ONP vote was 24.3 per cent!

In the Greater Brisbane areas, some of the highest ONP votes were achieved on the fringe of the city, particular in the western parts of Logan City and in the south western corridor running towards Ipswich. For example, in the seat of Ipswich held by Labor, the Eastern Heights and Raceview booths returned 43.5 per cent and 41.1 per cent respectively to the ONP. At the Bundamba TAFE and Redbank Plains booths in the electorate of Bundamba east of Ipswich, the ONP captured 41.3 per cent and 38.6 per cent of the vote respectively. In the electorate of Waterford, taking in the areas of Beenleigh and Eagleby in the City of Logan, the ONP vote at the booth of Logan Reserve recorded an ONP vote of 40 per cent. Generally, speaking the ONP vote in the South-Western corridor is in the 30 to 40 per cent range. The other major component of the ONP vote in the Greater Brisbane was in the North-East sector of Brisbane past Redcliffe and Strathpine, and including the areas of Deception Bay and Caboolture. These electorates are traditional ALP strongholds, similar to the electorates in the South-West of Brisbane City described above. Yet they too returned a ONP vote within the 30 to 35 per cent range.

For the rest of Queensland, many of the closer settled rural hinterlands show up in Figure 2 as locations with high levels of ONP primary support. They include, for example, the Lockyer Valley (west of Ipswich) through to Helidon (west of Toowoomba), much of the Darling Downs, and a broad rural belt extending from Boonah and Esk in the western part of the South East Queensland region north through Kingaroy, Murgon and the coastal region to Gympie, Hervey Bay and Maryborough. In the mining areas of Southern and Central Queensland, the ONP vote is strong in the southern parts of the coal areas of the Bowen Basin near Moura, in the Callide Basin near Biloela and along the Moreton Basin west of Ipswich. On the eastern side of Ipswich, the booths around the Amberley air base return a vote to the ONP of over 50 per cent, ONP support was also high in the Thuringowa electorate which includes the areas of outer Townsville and beyond. There are also significant numbers of people working in the large Army defence base in that region. In addition, there are belts of high ONP primary voting along the sugar coast regions stretching north from Bundaberg to Mossman (north of Cairns).

Having established the locations where the ONP vote is the highest, we are now in a position to examine the previous voting patterns in these areas. By comparing 1995 and 1998 booth swings, we are able to identify the parties for which the constituents voted at the previous election. Some direct connections are not possible as booths are added and removed between elections mainly due to population changes. (There were 1,672 booths in 1995 and 1,647 booths in 1998.) Table 4 classifies the swings against the major parties for all of Queensland for the 1,570 polling booths which were in existence at both elections.

Much of the ONP’s vote eventuated from a swing from the conservative parties. The effect on the National Party was particularly significant, with over 89 per cent of the booths in which they ran candidates registering a swing against them of greater than 10 per cent. The swing against the Liberal Party flowed in part to the Labor party, which registered a positive swing in 26.6 per cent of the booths. However, the Labor party also suffered negative swings in more than 73 per cent of their booths, with 42.5 per cent of the booths showing a swing of five per cent or more. The spatial relationship of the swing against Labor is also very interesting and supports our notion of the ‘urban fringe’ effect, particularly when examining booth swings in the Greater Brisbane area. Generally, within a 10 kilometre radius of Brisbane there is a positive swing to Labor. However, past this point there are only a few pockets where the Labor Party improved its vote. These areas include the ‘Koala Road’ areas of Springwood and Mansfield south-east of Brisbane, parts of the ethnically diverse electorate of Sunnybank, the relatively high income areas in the State seat of Aspley, the North-Western suburbs heading into the more populated components of Pine Rivers shire and along the Brisbane river in the wealthy Liberal-held electorate of Moggill.

The swing against the Labor party is very strong (around 10-15 per cent) in most other parts of the Greater Brisbane area and beyond. These comprise most of the western part of Logan city, most of the Ipswich district (with the exception of the centre of Ipswich which recorded a slight swing towards Labor) and Deception Bay and Bribie Island on Moreton Bay, north of Brisbane. A common feature of these areas is their low socio-economic character. They can be characterised as areas of relatively low-incomes, high proportions of unskilled workers and relatively affordable housing As mentioned above these are also areas with some of the highest incidence of ONP voting and the highest swings against the Labor party anywhere in the State. For example, at the Regent Park booth in the State seat of Logan (the former seat of ex-premier Wayne Goss) the swing against the ALP was 18.9 per cent (ONP vote 31.0); at the seaside booth of Toorbul in the electorate of Caboolture the anti-ALP swing was 19.1 per cent (ONP vote 40 per cent), at the Deception Bay south booth in the electorate of Murrumba it was 15.8 per cent (ONP vote 31 per cent), and at Waterford West near Beenleigh in the electorate of Waterford the swing against Labor was 16.1 per cent (ONP vote of 32 per cent). While the ONP vote consists mainly of former supporters of conservative parties, these results indicate that at the urban fringe the component of the ONP vote drawn from former ALP supporters was as high as 50 per cent. In other words, if ONP is viewed as consisting of a conservative ‘cake’, then the former supporters of the ALP are the ‘icing’.

Some areas, such as the area around Gympie, are known to be the ‘gun belts’ characterised by high levels of disenchantment with the new gun control laws.

A multiple regression approach was utilised to explain the underlying socio-demographic and economic factors that might underlie the ONP vote. Twenty dependent variables, taken from the 1996 census, were used in the model. Two other derived variables were included to test the ‘urban fringe’ effect — a measure of population density and a distance dummy variable. The latter variable returns a unity value if the distance of the centroid of a CD to a major centre (Brisbane, the Gold Coast, Maroochydore on the Sunshine Coast, Toowoomba, Bundaberg, Gladstone, Rockhampton, Mackay, Townsville, Cairns and Mount Isa) is less than 10 kilometres from the city centre.

The variables included in the model are taken from the 1996 census and are described in Table 5.

These variables were chosen as representing a range of socio-demographic and economic factors that might characterise groups of voters likely to be either attracted to or repulsed by One Nation’s rhetoric. For example, areas with a high proportion of Asians or Aboriginals are likely to be unenthusiastic about their local One Nation candidate. Other variables, such as household tenure and mobility were chosen as they are proxies for issues such as security and attitudes to change. Blue collar industries and unskilled workers are those facing the greatest negative changes in employment in the contemporary era of economic restructuring. Variables such as membership of fundamentalist Christian religions and family type might be expected to provide some indication of longings for traditional values.

The variables were evaluated through a linear step-wise multiple regression model. Table 6 displays the results of the model. The 20 variables entered into the model account for 42.4 per cent of the total variance in the spatial distribution of the ONP primary vote across Queensland’s polling booths.

The results indicate that the ONP’s support base in Queensland is approximately 15.8 per cent. That is, the ONP will obtain a vote of 15.8 per cent when all the independent variables considered in the model are zero. The ONP support declines most significantly with the distance dummy variable which places a significant spatial context to the ONP vote. The distance dummy variable shows that within 10 kilometres of major centres, there is evidence that the ONP vote will decline by more than three per cent. Other variables showing a negative relationship to the ONP vote are, not surprisingly, the percentage of Aboriginals and Torres Strait Islanders (ABTS), the percentage of Asians (ASIAN) and the percentage of persons born overseas (OS). High income earners (HIGHY) were also less likely to be supporters of the ONP. Single parent families (LONE) and those renting their own house (RENTH) were also negatively related to the ONP vote. The other variable measuring the fringe effect, population density (DEN) provided a slight negative relationship.

All the industry and employment variables used in this model — that is, percentage employed in blue collar industries (BLUE), percentage employed in agricultural (AG) and unskilled workers - are all positively related to the ONP vote. Those who own their own homes or are in the process of buying them (OWNH and BUYH respectively) are also more inclined to the views of the ONP. This is not surprising as the proportion of home ownership has a strong positive relationship to age. Mobile households, defined as those who lived at a different address five years ago (a proxy for population mobility), are positively related to the ONP vote. Finally, areas with higher percentages of non-mainstream (the fundamentalist) Christian groups (F_CHRST) are also more inclined to support the ONP.

Although Queensland’s politics are notoriously different from the rest of Australia, several issues are likely to be relevant for the 3 October 1998 Federal election, even though the magnitude of the vote will differ from state to state.

The main factor in the ONP vote in Queensland at the 13 June 1998 poll was what we call the ‘urban fringe’ effect. Generally, the ONP primary vote is at its strongest at the edge of metropolises and the regional urban centres and in their hinterlands. Where these areas contain unskilled workers in blue collar industries, few indigenous Australians or people born overseas, and have a high number of people either achieving or attempting to achieve the Australian dream of home ownership, then the ONP is likely to do well. Conversely, inner city areas, particularly those areas marked by high levels of multicultural populations or higher incomes, are the safe havens of the major political parties.

Political parties examining policies aimed at winning back ONP supporters need to be wary of simply labelling all regional and rural areas with the same classification. For example, large areas of rural Queensland have provided no more support for the ONP than is the case in the inner and middle areas of the metropolis and the main regional urban centres. The belief that the ONP is solely a problem for the National Party and its followers is a fallacy. Most of the pastoral and mixed farming heartland of the National Party does not show up as regions of high support for the ONP. While there are certainly issues for rural producers, we suggest that the study of policies aimed at ONP supporters will need to reflect the needs and aspirations of population’s living towards and beyond the fringe of the metropolis and the regional urban centres and extending into the rural lifestyle zones of these hinterlands.

Equally, attempts to win over ONP supporters through decentralising services to regional centres may have little impact on appeasing the disaffection of the ONP supporters since they do not live in large numbers at the core of those urban centres. For example, additional government services placed within the centre of Townsville may only further anger the voters of the ONP electorate of Thuringowa, where the real disenchantment lies.

There does, however, seem to be substantial support for the ONP along the coastal rural areas, characterised by small farming and the sugar industry, where there are concerns about long term survival. This is also the case in parts of the coal mining regions.

What does seem to be a characteristic of the core areas of ONP support is that they are places experiencing substantial levels of population change, particularly as locations that attract in-migrants who seek quasi-rural lifestyles, but who remain economically attached to urban type economic pursuits. This is not to say that in some rural areas with high levels of ONP support, actual rural producers are not being attracted to ONP; some are, but our analyses show the employment in agricultural occupations is not as important as employment in blue collar occupations as an explanatory factor in the ONP vote. Moreover, a high incidence of unskilled workers is a particularly important factor related to high levels of ONP support.

These findings present a dilemma for public policy initiatives by the major political parties, as neither a rural development focus nor an urban development focus in regional development programs is likely to directly impact the core areas of the ONP constituency. It is a large and, in all likelihood, a growing component of the community that is both disenchanted, feels disenfranchised, and is relatively highly mobile.

A version of this paper was delivered to the annual conference of the Australian and New Zealand section of the Regional Science Association held at Tanunda, South Australia, 21-23 September 1998.