Identity and support for policies towards Indigenous people: evidence from Australia

This paper adds to knowledge on the role of politicians’ and voters’ identities in influencing policy-making in societies marked by ethnic inequality. The outcome we investigate is the initiatives and policies targeting Indigenous populations in the context of Australia. We ask whether and how politicians’ and voters’ identities, defined based on a range of their observable characteristics, shape initiation and support of Indigenous-focused policies. Drawing on data on the voting history of members of the two houses of the Australian parliament, we provide an analysis of law-making behaviour and show that political ideology, and to some extent politicians’ propensity to act rebelliously and diligently, are significant determinants of their agreement on Indigenous policies. Our complementary analysis of voters’ attitudes on issues concerning Indigenous Australians points to the high relevance of political ideology, and in doing so, highlights the alignments in the behaviours of politicians and voters.


Introduction
Ethnic inequalities are endemic and costly (Cederman et al., 2011;Lessmann & Steinkraus, 2019;Fleming et al., 2020). Reducing such inequalities is a global priority and is underscored in the Sustainable Development Goals (SDGs), especially in SDG 10: Reducing Inequality. While there are numerous policy initiatives around the world targeted at reducing ethnic inequalities, knowledge around the way such initiatives come about is largely lacking. This study aims to understand the process of initiation and support of policy initiatives addressing ethnic inequalities. Specifically, drawing on two unique datasets, we provide inferences around the role of politicians' and voters' social and political identities in these processes.
We focus on initiatives targeting the Indigenous populations who have significantly poorer health and wellbeing outcomes relative to non-Indigenous populations across many parts of the world (Stephens et al., 2005;Anderson et al., 2016). One such place is Australia which is home to one of the most disadvantaged First Nations in terms of a wide range of socio-economic measures (Hunter & Daly, 2013;Duncanet al., 2019). Yet, it is only recently that there have been coordinated intergovernmental efforts to address the Indigenous disadvantage in Australia.
Policies related to the Indigenous population took a critical turn in 2008 when Labor Prime Minister Kevin Rudd issued an official apology to Indigenous Australians for forced removals of Indigenous children, known as the "Stolen Generation", and to Indigenous Australians more generally for past policies which "inflicted profound grief, suffering and loss". As part of the apology, the government committed to ensuring that Indigenous Australians have equal life chances to all other Australians and initiated the National Indigenous Reform Agenda, which features a set of targets known as Closing the Gap. However, little is known on the processes underlying the policy initiatives that have emerged since then, including the determinants of support extended by individual politicians and voters. Given the lack of knowledge on these processes, the analysis conducted in this paper is deliberately open-minded and exploratory in focusing on a range of observable characteristics of politicians and voters and how their social and political identities, as captured through such characteristics, shape their support on issues concerning Indigenous Australians.
This study is related to several strands of literature. There is a sizeable body of work on ethnic equality initiatives and policies in various contexts which focuses on analysing their impact (Dunning & Nilekani, 2013;Jensenius, 2015;Bhavnani, 2017;Jensenius, 2017;Gulzar, Haas, and Pasquale, 2020). In contrast, the literature concerned with studying the initiation of such policies is relatively small (Arauco et al., 2014;Vergara-Camus, 2016). Our study adds to the emerging body of work in the latter area. Through focusing on the identity of politicians in support of ethnic equality issues and initiatives, our study also feeds into the growing empirical literature that shows that politicians' identities matter for policy-making (Jones & Olken, 2005;Besley et al., 2011;Bhalotra et al., 2014;Mavisakalyan, 2014;. This paper contributes to expanding this literature by focusing on an under-researched context and outcome.
Through its focus on Indigenous disadvantage in Australia, this study relates to a large body of work that is concerned with documenting Indigenous-non-Indigenous inequalities and their drivers, highlighting the role of colonization, racism, discrimination, and historical trauma (King et al., 2009;Kalb, Le, Hunter, and Leung, 2014;Paradies, 2016;Duncan et al., 2019) and the protective role of cultural attachment (Dockery, 2010(Dockery, , 2020. While there are studies that evaluate the impact of Indigenous-focused policies (Lokuge et al., 2017;Beks et al., 2019;Chando et al., 2021), evidence on the determinants of such policies is largely lacking.
Our study highlights the relevance of political identity for Indigenous-focused policy preferences and initiatives, and in doing so, it also relates to the literature that engages with the links between political ideology and its role in shaping attitudes towards minorities (Rich, 1986;Sidanius et al., 1996;Whitley Jr, 1999;Poteat & Mereish, 2012;Bailey & Williams, 2016). The closest to this study is that by Langford and Ponting (1992), which draws on a nationally representative survey of non-aboriginal Canadians conducted in [1986][1987] to study the willingness to support the government policy in three areas of aboriginal affairs: self-government, special status and level of priority attached to native issues. They show that a moderate to large proportion of the variation in non-aboriginal responses to aboriginal issues can be explained by prejudice, economic conservatism, and perceived conflict of group interests. Our study, however, is distinct from that by Langford and Ponting (1992) in a number of significant ways. First, we combine analysis of behaviours of both politicians and the general population rather than focusing on the attitudes in the general population alone. Second, in studying the drivers of support for Indigenous issues and policies, our focus is on social and political identity of individuals whereas Langford and Ponting (1992) look at inter-related yet distinct processes of group conflict, perceptions of personal threat and "symbolic predispositions" including prejudice, ethnocentrism and economic conservatism. Finally, our focus is on the case of Indigenous Australians, which has received relatively little attention in the context of this literature.
We first discuss the key theoretical propositions of relevance to our work (Sect. 2). Next, Sect. 3 provides an overview of Indigenous disadvantage in Australia and the associated policy responses. In Sect. 4 we present our study of politicians, including a discussion of our empirical approach and results, followed by the analysis of the general population presented in Section 5. We conclude with a discussion of the implications of our findings in Sect. 6.

Conceptual background
Theoretically, our work is inspired by three sets of studies. First, it is related to the concept of identity by Kranton (2000, 2010), defined as a sense of belonging to a social category that has implications for how people behave. Second, it draws on the political economy models that suggest that politicians' decisions reflect the preferences of their constituency (Downs, 1957) or, in the absence of complete political commitment, their own preferences (Osborne & Slivinski, 1996;Besley & Coate, 1997). Finally, our work provides engagement with intergroup contact theory which proposes that the prejudice to minorities is likely to decline with increased opportunities for contact (Allport et al., 1954;Pettigrew & Tropp, 2006). These studies and their implications are discussed in detail next.

Identity and political preferences in the population
This paper posits that identity has implications for political preferences. Kranton (2000, 2010) define identity as a sense of belonging to a social category. A person can be represented in terms of a range of social categories, including their gender, class, religion, political orientation, etc. As Kranton (2000, 2010) argue, identity changes 'payoffs' from different actions since, on one hand, following the behavioural prescriptions associated with a social category affirms one's identity and, on the other hand, violating the prescriptions instigates anxiety and discomfort in oneself and in others. Akerlof and Kranton (2000: 726) note that "politics is often a battle over identity". Certain political or religious affiliations, for example, may at times come with rhetoric that may foster social or ethnic divisions in a society, while others may mitigate these. Moreover, identification with different social groups may mean having different preferences over outcomes. In particular, the literature shows that class, race and gender have high relevance for individuals' political preferences and attitudes towards minorities. Educational attainment -a key dimension of class -is associated with more favourable attitudes to minorities (Nunziata et al., 2016;Cavaille & Marshall, 2019;Margaryan et al., 2021;Scott, 2022;Yang, 2022). One explanation put forward in the literature to explain this finding focuses on the "liberalising effect of education" whereby education changes the values and capacities of individuals, making them less prejudiced (Gaasholt & Togeby, 1995;Pascarella et al., 1996;Lancee & Sarrasin, 2015). Gender is another dimension of identity which appears to have theoretical relevance for individuals' attitudes towards minorities. Based on theories of gender-role socialisation, women are socialized to be more concerned about others' wellbeing (Cross & Madson, 1997) -a proposition that could be extended to the case of minorities (Johnson & Marini, 1998). Based on empirical evidence, women indeed tend to exhibit significantly more favourable outlook towards minorities than men in some contexts (Smith et al., 2013), however in others, there is little or no relationship between gender and attitude to minorities (Hughes & Tuch, 2003).
Not only may different dimensions of social identity shape individuals' beliefs and attitudes towards minorities separately, they may also do so in combination given that individuals hold positions in different systems of stratification simultaneously (Berg, 2010). The literature shows that intersectionality plays out in attitudes towards different social issues (Han, 2006;Foster, 2008). In the context of attitudes to minorities, Berg (2010) shows that important differences between groups and within groups defined by race, class, gender, and social space emerge when particular intersections are considered in the analysis.
The impacts of identity are not restricted to attitudes. They extend to voting behaviour as well-for example, there are systematic differences in voting and political preferences by class, race and gender after measures of economic self-interest are accounted for (Luttmer, 2001;Glaeser & Ward, 2006). Thus, based on their social identities, individuals might choose courses of action that determine the aggregate outcomes in the society (Shayo, 2009).

Identity and political choices of politicians
What about the role of identity in politicians' behaviours and choices? The Downsian model predicts that policy outcomes are determined by the preferences of the median voter, and therefore the politicians' own preferences should not matter (Downs, 1957). On the other hand, "citizen candidates" models by Osborne and Slivinski (1996) and Besley and Coate (1997) propose that in the absence of complete political commitment, politicians implement policies consistent with their preferences.
In line with "citizen candidates" models, a growing body of empirical work has documented the impact of politicians' social identity on policy choices. In particular, existing evidence suggests that women politicians contribute to a range of desirable outcomes including higher spending on health (Mavisakalyan, 2014) and education (Clots-Figueras, 1 3 2012), more stringent environmental policies , higher expenditures relevant to women's needs (Chattopadhyay & Duflo, 2004), better institutions (Swamy et al., 2001), and higher rates of economic growth (Jayasuriya & Burke, 2013). Cast and ethnicity are some of the other dimensions of identity that have received attention in the literature on the substantive impacts of politician identity. Some studies show that minority leaders transfer state resources to members of their own ethnic group (Pande, 2003;Kramon & Posner, 2016), however no such evidence is found in others (Kudamatsu, 2009;Dunning & Nilekani, 2013). Existing evidence also suggests that policy outcomes vary by the educational attainment of leaders with more educated leaders adopting policies that are conducive to economic development (Besley, Montalvo, & Reynal-Querol, 2011;Jain, Kashyap, Lahoti, and Sahoo, 2022).
Political affiliation is another significant determinant of policy choices that politicians make. In the context of the US, Case (1995, 2003) find a significant impact of Democratic party affiliation on taxes, workers' compensation benefits and total spending. Existing studies also show that minorities have better outcomes under Democrats vs. Republicans. Looking at the differences in the distribution of spending between Democrats and Republicans, Hill and Jones (2017) show that school districts with higher shares of minority students receive larger state transfers than other districts under Democratic governors. Beland (2015) finds that Democratic governors lead to an increase in the annual hours worked by blacks relative to whites, which results in a reduction in the racial earnings gap between black and white workers. Similarly, Beland and Unel (2018) shows that immigrants are more likely to be employed, work longer hours and more weeks, and have higher earnings under Democratic governors.

Intergroup contact and attitudes towards minorities
Misperception and prejudice to minorities is likely to be a significant obstacle for willingness to endorse minority-focused policies. What is likely to reduce such attitudes, based on the literature, is the intergroup contact. The intergroup contact theory of majority-minority relations suggests that prejudice to minorities declines with contact with out-group members (Allport et al., 1954;Pettigrew, 1998;Pettigrew & Tropp, 2006). Increased knowledge, anxiety reduction and enhanced empathy are thought of as key mechanisms underlying the phenomenon (Pettigrew & Tropp, 2006;Pettigrew et al., 2011).
The intergroup contact theory has received support in a number of empirical studies. A meta-analysis of empirical studies of contact theory by Pettigrew and Tropp (2006) suggests that contact reduces prejudice unconditionally, but the effect is larger if those in contact have equal status in the particular context, share common goals, are in a cooperative context and if the contact takes place under some form of authority (Pettigrew, 1998). More recent engagements with the theory based on well identified studies using random assignment produce similar results. Based on a field experiment in the Norwegian military, Finseraas and Kotsadam (2017) find that contact affects the views on immigrants' work ethnics, but has small and insignificant effects on support for welfare dualism. Using a similar research design, Finseraas et al. (2019) show that personal contact with minorities increases trust towards a generic immigrant. Based in a riot-prone part of Nigeria, a study by Scacco and Warren (2018) finds that interventions which induce contact between members of religious groups in conflict have little effect on intergroup prejudice but lead to a reduction in discriminatory behavior in heterogeneous group settings. For other recent empirical engagements with the intergroup contact theory see Ghosn et al. (2019), Bursztyn et al. (2021).

Context
This work is based on a case study of Australia whose Indigenous people are among the oldest continuous human cultures in existence (Nagle et al., 2017). In spite of their resilience, the Indigenous Australians have been severely affected by the colonisation that has had long-lasting impacts on their socio-economic outcomes and wellbeing (Griffiths et al., 2016;Paradies, 2016). This section presents an overview of the Indigenous disadvantage in contemporary Australia and the policy context and initiatives to address it.

Indigenous disadvantage in Australian
Australia presents one of the great number of cases where colonization has had dire consequences for the way of life and wellbeing of Indigenous people. Estimates of Australia's Indigenous population prior to European contact vary from 300,000 to over one million, with a figure of 750,000 commonly accepted (Aboriginal and Torres Strait Islander Commission (1998)). Based on the latest Australian Census estimates, the Indigenous population of Australia as at 30 June 2016 was 798,400 people, or 3.3 per cent of the total Australian population.
Indigenous Australians are among the most disadvantaged minority groups in contemporary Australian society, based on figures relating to many important aspects of life. As Fig. 1 demonstrates, in spite of some improvements over time, relative to non-Indigenous Australians, Indigenous Australians have significantly lower levels of life expectancy, education, and employment, and are at disproportionately higher risk of imprisonment. For example, as of 2015-2017, the gap in life expectancy between non-Indigenous and Indigenous populations was 8.6 years for males and 7.8 years for females. Similarly, by 2016, only 63.2 per cent of Indigenous Australians aged 20-24 years had attained Year 12 or a non-school qualification or above, relative to 88.5 per cent among the non-Indigenous population. There are striking differences in employment rates too, with only 51 per cent of Indigenous individuals employed as of 2016, relative to the nearly 78 per cent employment rate among non-Indigenous individuals. Furthermore, the agestandardized rate of Indigenous prisoners was 2222.7 per 100,000 adult population in 2021-a significant over-representation relative to the non-Indigenous population.
Discussions on the wellbeing of Indigenous Australians have highlighted the significance of self-determination and preservation of Indigenous culture alongside achievement of socioeconomic equity. Dockery (2010) argues that Indigenous culture should in fact be seen as part of the solution to Indigenous disadvantage in Australia. He bases his view on an empirical analysis that demonstrates a strong link between attachment to traditional culture and a range of desirable socio-economic outcomes in the Indigenous adult population including better health and a lower likelihood of engaging in risky alcohol consumption. In another study, Dockery (2020) looks at Indigenous children's health and socio-economic adjustment, showing that they are strongly and positively influenced by Indigenous parents' practices with respect to passing on to their children the traditional knowledge, strong kinship traditions and connection to country at early stages of upbringing.

Indigenous policy-making in Australia
In spite of the long-standing nature of Indigenous disadvantage in Australia, it is only in recent years that the situation has been the subject of coordinated intergovernmental attention, with occasional instances of recognition such as the Closing the Gap targets, the Constitutional Reform process including the Uluru Statement from the Heart, and gestures such as the apology to the Stolen Generation in 2008. However, based on an overview of five distinct periods in Australian history, (Davis (2015): 42) highlights "the ad hoc nature of laws, insecure rights and the disruptive manner with which law and policy is applied to indigenous communities".
The policy-making on Indigenous affairs has taken different turns under Australian governments of different party persuasions. The party system in Australia is often described as more ideologically driven relative to the party systems in other countries (Parkin et al., 2002). The two major parties in Australia, the Australian Labor Party and the Liberal Party were created and have historically competed around class interests. In terms of Indigenous policies, Mulgan and Sanders (1996); Sanders et al. (2015) describe the developments under all governments from Whitlam to Keating over the Fig. 1 Indigenous-non-Indigenous gaps in selected outcomes. Source: authors' creation based on Closing the Gap Data Repository accessed at: https:// www. pc. gov. au/ closi ng-the-gap-data period from 1972-1996 as a movement in a "decolonising" direction. This included major steps such as the adoption of "self-determination" as the key term of Indigenous policy, funding of Indigenous organisations for the delivery of services, and development of draft legislation for land rights in the Northern Territory under the Whitlam Labor Government; establishment of a Royal Commission into Aboriginal Deaths in Custody and passing of legislation to establish Aboriginal and Torres Strait Islander Commission under the Hawke Labor Government; and the launch of the International Year of the World's Indigenous Peoples and passing of the Native Title Act under Keating Labor Government. In contrast, Howard's Liberal-National Coalition, which held Government from 1996-2007, rejected the self-determination approach as 'symbolic reconciliation', pursuing instead 'practical reconciliation' approach through economic development. Sanders et al. (2021) notes that under the subsequent labor governments led by Rudd (2007Rudd ( -2010 and Gillard (2010-2013) some elements of Howard's Indigenous policy legacy were overturned, however these did not progress too far before the Coalition government reclaimed the government thereby confirming the retreat from a decolonising approach in Australian Indigenous policy.
Unsurprisingly, the discussions on the limited impact of existing policy initiatives highlight the fact that they were developed and implemented without a focus on Indigenous self-determination, i.e. Indigenous people's genuine participation in decision-making on their own development and outcomes. As (Hunt et al. (2022): 25) note: "A missing piece of infrastructure in Australia is a commitment to Indigenousled participatory engagement and development, including building the institutions and providing resources to support this in whatever form it might take". Such approach to participatory Indigenous development policy would be in line with the Indigenous right to self-determination in the United Nations Declaration of Indigenous Peoples (UNDRIP). While the Australian Government has indicated its support for the UNDRIP, it remains a largely aspirational non-binding declaration (Davis, 2012).
Unlike the Commonwealth, in some jurisdictional responses Indigenous selfdetermination and the right to participate in decision-making has been a longstanding area of focus. For example, through Local Decision Making (LDM) -one of the core initiatives supported by the NSW Government's Opportunity, Choice, Healing, Responsibility, Empowerment (OCHRE) Plan for Aboriginal Affairs in NSW -regional Aboriginal representative bodies negotiate formal and binding agreements with NSW Government agencies. However concerns have been raised over the lack of appropriate resourcing of LDM initiatives which constrain the engagement with regional alliances (O'Bryan et al., 2022). Moreover, Dreise et al. (2021) highlight the potential gains that could result from better alignment of such established regional forms of representation with Commonwealth-level policy priorities and processes.

Study of politicians
We start with an analysis of policy-making on issues concerning Indigenous Australians. To that end, we draw on a unique dataset of voting in the Australian federal parliament and ask the following questions: How have the parliamentarians voted on recent Indigenous-focused policy initiatives? How does the identity of parliamentarians, captured through individual observable characteristics, shape their position on such initiatives?

Empirical approach
Our analysis of Australian parliamentarians' agreement on policies concerned with Indigenous Australians is based on a simple model where the level of agreement on a policy is the dependent variable and individual characteristics of parliamentarians constitute our main explanatory variables of interest. Equation (1) formally presents the model, where AG i,j is the agreement level of parliamentarian i for policy j, X i is the vector of individual characteristics, Geo i and Pol j represent jurisdictional and policy fixed effects respectively, while u ij is the error term: The main source of data is the They Vote For You (TVY) project, which was developed by the OpenAustralia Foundation-an independent, non-partisan not-for-profit organization. The project, founded by a donation from Google, scrapes the data from the Australian Parliament's website and the published transcripts of votes and discussions (called Hansard), and provides a relatively clear voting history for each member of parliament.
TVY groups related motions and divisions into a number of policy groups. Our dependent variable, AG i,j , is specifically captured through the share of supportive votes by member i of the House of Representatives or the Senate for motions and divisions related to a policy j. 1 This implies that the level of agreement varies across individuals when different policies are considered and policies have several agreement levels that could range from 0 to 100.
In addition to the level of agreement, which is our outcome measure, TVY contains several measures on parliamentarians derived from their voting history. These include the number of rebellious votes (also known as crossing the floor), which captures the number of votes against the party whip, 2 and the total number of attended votes. We use these measures to create a ratio of rebelliousness (the percentage of rebellious votes in the total attended votes) and attendance rate (the percentage of attended votes in all possible votes) to proxy for individual characteristics of parliamentarians, such as their predisposition to act rebelliously and their diligence. Another individual characteristic we include in our analysis is the gender of the parliamentarians. We use the data from the US Social Security Administration, which includes the name and gender of all US citizens registered between 1880 and 2017, to assign gender to the politicians based on their first names. 3 Similar to most other countries with the Westminster system, the Australian Parliament has two 'houses': the Senate and the House of Representatives. The Senate is also known as the Upper House, while the House of Representatives is referred to as the 'parliament' or Lower House. TVY provides data on the political party association in both houses. While the parliamentarians are much more heterogeneous (in terms of their political orientation) in the senate, the majority still belongs to one of the major political parties: the Australian Labor Party or the Liberal-National Coalition (LN) with some representatives from the Australian Greens. We consider the political orientation of each parliamentarian as part of the set of their individual characteristics that proxies for political ideology, and as such could have implications for the level of agreement on policies.
While TVY provides voting history data from both the Upper and the Lower House, data on the associated electorates is available for politicians in the Lower House only; for the senators, there is only data on the associated State/Territories. To ensure the representativeness of our analysis over the entire pool of politicians, our baseline analysis includes state fixed effects to capture the jurisdictional differences. It should be noted, however, that there are many localised factors relating to particular circumstances of Indigenous people which may shape or influence the support for specific motions or policies. While we aren't able to capture such specific factors, we are able to explore the impact of a limited set of socio-economic characteristics of electorates on politicians' behaviour limited to the sample of parliamentarians in the Lower House, and we pursue this as a robustness check. Given that our analysis is based on a pooled sample of policies, we additionally control for policy fixed effects in all regressions.

Descriptive analysis
To capture the politicians' views on different topics, the TVY project groups proposals, motions, and questions into groups called 'policies'. Each policy has a common theme and tracks all the votes for each of the motions and proposals within. There are 4,540 divisions and motions dating back to 2006 captured on the TVY website, which are grouped into 269 policies from which just six can be linked to Indigenous Australians. Table 1 lists these six policies, the associated number of divisions and motions and the average level of agreement for each policy. 4 As seen from the Table, there does not appear to be a meaningful relationship between introduced motions for each policy and the associated level of agreement. 5 It is important to note that the numbers presented in Table 1 are a stock-take of all divisions and motions related to the policy over the period from 2006-2021, and do not reflect the specific time dimension. For example, five motions and divisions related to the 'Close the Gap' policy occurred in 2007, 2020, and 2021. As such, our study employs a cross-sectional design. Nonetheless, it should be highlighted that even though what we observe in terms of policy support is one point in time for the data, the motion or policy put forward could be due to a long period of advocacy or call for change and larger external factors which we are not able to fully capture in the current setup.
Policies under consideration are concerned with key issues of relevance to Indigenous disadvantage, including the large gaps in outcomes between Indigenous and Table 1 Policy proposals focused on issues of concern to Indigenous Australians. Source: authors' compilation based on TVY data Change in Australia day The date of Australia Day should be changed from 26 January, which was the day the First Fleet arrived at Port Jackson, Sydney, and is a day of mourning for many Australians.

27
Enrich first nation identity The Constitution should be amended to include a requirement for an Indigenous voice to federal parliament, such as that envisaged by the Uluru Statement from the Heart.
2 48 non-Indigenous populations and the significant number of Indigenous deaths in custody. 6 Other important issues the policy proposals have engaged with include Indigenous land rights, cultural heritage protection, and constitutionally enshrined voice in parliament. To provide further insights on the issues of concern in these policy proposals, Fig. 2 presents the key themes featured in the summaries of policy texts captured through word frequency. Land, health, custody deaths, and rights appear to be dominant themes captured through this exercise. Additionally,Figs. 6,7,8,9,10,11 in Appendix A provide further information on the timeline, initiators, and outcomes of key initiatives raised under each policy proposal. As these figures demonstrate, the policy initiatives included in the sample have been raised by the representatives of the Green Party and, to a lesser degree, the Labor Party, and many of these have failed to pass in parliament. Is our data picking up patterns specific to Indigenous voting, or might it be simply reflective of the political divide in policy preferences more broadly? To engage with this question we check whether the votes on Indigenous-focused policies are correlated with votes on other issues. In particular, how does support for Indigenous-related policies go along with support for multiculturalism and immigration? Is it correlated with support for other groups, such as women? We descriptively engage with these questions in Fig. 3. As can be seen, voting on Indigenous policies appears to follow a distinct pattern and does not correlate with patterns of voting on issues concerning other minority groups. 7 Table 2 presents detailed definitions of the variables included in the analysis along with their summary of statistics. As shown in the table, there is less than 42 per cent agreement on the Indigenous policies across the politicians in the sample, of whom 41.5 per cent are female. Less than 0.03 per cent of the votes observed in the sample have been against the associated party's whip, but the attendance rate among politicians has been high, just under 79 per cent. In our sample, 42.4 per cent of parliamentarians are associated with the LN Coalition, 36.6 per cent are from the Australian Labor Party, and 8.5 per cent are from the Australian Greens, with the rest associated with various other political parties.
As discussed earlier, our sample consists of six policies targeted at issues concerning Indigenous Australians. Within the sample, around 33% of the observations relate to Increase Aboriginal land right policy. Increase heritage protection and Close the Gap policies each present 15% of the sample, whereas each of Enquiry into Aboriginal death and Enrich First Nation identity policies comprise 14% of the sample. The remaining 9% of the observations relate to Change in Australia Day policy. The votes in the Lower House comprise 24% of the sample.

Estimation results
As outlined earlier, the main focus of our analysis is to uncover the determinants of politicians' agreement on policies concerning Indigenous Australians. To that end, we estimate Eq. (1) and present the results in Table 3.
We estimate a positive significant coefficient on the Female dummy in column (1); however, it loses its significance in subsequent models where controls for party association of politicians are included. Rebellious votes appear to be correlated with higher level of agreement with Indigenous-focused policies across most specifications. The results also show that 'diligent' politicians tend to agree more on such policies. More importantly, the impact of political ideology seems to be robust and consistent after controlling for several available individual characteristics, as well as the state jurisdiction of the electorate. Parliamentarians associated with a right-wing ideology (proxied by association to the NL Coalition) are more likely to be in disagreement with Indigenous-focused policies, whereas Labor Party associates, and to a much higher degree the associates of the Greens, are likely to be in agreement.
In columns (6) and (7) we conduct the analysis based on the most extensive model specification (per column (5)) in separate sub-samples of men and women. Some heterogeneities in the results emerge. Among male parliamentarians, attendance and rebellious votes do not have any association with support for policies, while for female parliamentarians there is a highly significant association between rebellious votes and support for policies. On the other hand, the estimated positive coefficients on Labor and Green Party membership are larger in size in the male sub-sample relative to the female sub-sample-that is, the political ideology plays a potentially larger role for male decisionmaking on policies concerning Indigenous Australians. Hence, intersectionality between gender and political ideology plays out in the stances held by politicians on Indigenous policies.
The identity of politicians appears to matter in voting for policies, which is consistent with the 'citizen candidates' models by Osborne and Slivinski (1996) and Besley and Coate (1997). On the other hand, the Downsian model predicts that the policy outcomes are determined by the preferences of the median voter (Downs, 1957). To engage with the possibility that the characteristics of own electorates might contribute to decisions to support Indigenous policies, we restrict the analysis to the Lower House of parliament, where we are able to identify the home electorates of parliamentarians precisely. We then generate a series of measures to capture the demographic and socio-economic characteristics of electorates. Given the cross-sectional design of our study, we draw on the latest 2016 wave of the Australian Census of Population and Housing as a source of the characteristics of the electorate. These include the shares of Indigenous and overseas-born populations of the electorate alongside the shares of different age cohorts in the population. We also control for SEIFA scores-aggregate measures summarizing different aspects of the socio-economic conditions in the area available from the Census; higher scores mean better socio-economic conditions. This is complemented with measures of educational attainment to capture the socio-economic profile of electorates. We control for the shares of builders, We start the analysis by re-estimating the baseline model (as per column (5) of Table 3) in the separate sub-samples of Upper and Lower Houses of parliament, presented in columns (1) and (2) of Table 4. In both sub-samples, the positive significant association between the share of rebellious politicians and support for policies is present. We also observe the strong correlation between the measures of political ideology and support for Indigenous policies. Attendance rate, however, loses its significance in the Upper House sub-sample and is negative and statistically significant in the Lower House sub-sample. The latter is likely to be the outcome of the combination of two factors: the relatively higher representation of right-wing politicians in the Lower House, and the relatively higher probability of attendance of right-wing politicians in the Lower House relative to the Upper Note: standard errors in parentheses. * Significance at 10 per cent; ** 5 per cent; *** 1 per cent levels House. In Table 8 in Appendix A we show descriptive statistics that are consistent with this scenario. In particular, members of the Liberal Party are more likely to be in the Lower House relative to the members of the Australian Greens. On the other hand, members of the Liberal Party are more likely to attend in the Lower House than in the Upper House (79 vs 88 per cent attendance rate). Conversely, the attendance rate of the Australian Greens in the Upper House is 91 per cent, while it is lower in the Lower House at under 87 per cent. The analyses in subsequent columns of Table 4 are limited to the Lower House subsample. In columns (3) and (4) we introduce the measures of demographic characteristics of electorates, the share of Indigenous population first, complemented with measures of overseas-born population and age-based cohort shares next. None of these variables are statistically significantly related to support for policies. In columns (5) and (6) we introduce measures of socio-economic profile of electorates: SEIFA score first, then complemented with measures of educational attainment. Again, none of these are significantly correlated with support for policies. Column (7) presents the results where these electorate characteristics are controlled for jointly, while the analysis presented in column (8) additionally controls for state fixed effects. The key findings of this analysis with electorate characteristics are in line with those without these characteristics, reported in column (2). None of the electorate-level variables appear to be correlated with support for policies, which goes against the prediction of the Downsian model. It should be noted, however, that this analysis is based on a smaller sub-set of politicians, which limits the inferences that can be drawn based on it.

Study of the general population
We find that political ideology is a core determinant of politicians' agreement on policy proposals addressing issues concerning Indigenous Australians. But what are the preferences of the general population on these issues? Is political ideology key to having a supportive attitude on such issues for ordinary Australians? Are there other factors that play a role as well? Next, we turn to addressing these questions in an empirical study of a representative cross-sectional survey of the Australian population.

Empirical approach
Our analysis of individual Australians' views on issues concerning Indigenous Australians is based on a standard model in which the propensity for holding a supportive attitude, Support * i for an individual i is assumed to depend on a series of individual controls K i for party identification and standard demographic and socio-economic characteristics, as well as state fixed effects State i . Unobserved factors i further contribute to the propensity for holding a supportive attitude on issues concerning Indigenous Australians, leading to an equation of the form: Observed support status Support i is assumed to relate to latent propensity through the criterion Support i = 1(Support * i ≥ 0) , so that the probability of holding a supportive attitude under an assumption of normality for i becomes Data for this analysis is derived from the Australian Survey of Social Attitudes (AuSSA)an annual national postal survey of adult Australians randomly drawn from the Australian Electoral Roll, which contains 95 per cent of the citizens aged 18 and above. As such, the respondents are representative of the population of Australian voters. Conducted since 2003, AuSSA is the main source of data for studying the social attitudes, beliefs, and opinions of Australians. Our analysis employs the 2016 wave of the survey conducted between June and October 2016, which contains a set of questions that capture the attitudes on issues concerning Indigenous Australians (AuSSA, 2016). The 2016 survey sampled 5,000 respondents and the response rate was 25 per cent. Given the missing values on variables of interest, the actual sample sizes employed in the baseline analysis is around 1060 observations. The AuSSA 2016 asked individuals' views on the following statements about the situation of Aboriginal people in Australia today, with possible response categories including strongly agree, agree, neither agree nor disagree, disagree, and strongly disagree: 1. Aboriginal people's level of disadvantage justifies extra government assistance. 2. Granting land rights to Aboriginal people is unfair to other Australians. 3. Aboriginal people should not have to change their culture to fit into Australian society. 4. Aboriginal people are now treated equally to other Australians.
We employ the responses to these questions to derive four key outcomes of interest, presented in Table 5. We change the direction of questions (2) and (4) to ensure that the statements and their responses are framed in the same direction, facilitating the interpretation of the results. Accordingly, to ensure that the outcomes are consistently capturing the support on issues concerning Indigenous Australians, we employ either the combination of 'strongly agree' and 'agree' responses (questions (1) and (3)) or the combination of 'strongly disagree' and 'disagree' responses (questions (2) and (4)) in the construction of the binary variables presented in Table 5.
The survey additionally asks individuals whether they usually think of themselves as close to any particular political party and, if yes, which party that is. We use this information to distinguish between individuals who pick one of the three major partiesthe LN Coalition, the Labor Party, and the Greens-vs those who pick other or no party. Additionally, our analysis includes standard demographic and socio-economic information on individuals, including their gender, age, ethnicity, family status, educational attainment, employment status, and religion.

Descriptive analysis
The definitions and sample means of all the variables employed in the analysis are presented in Table 5. Around 50 per cent of individuals in the sample agree that Aboriginal people's disadvantage justifies extra government assistance, while 48 per cent do not believe that granting land rights to Aboriginal people would be unfair to other Australians. Around 57 per cent of individuals agree that Aboriginal people should not have to change their culture to fit and 55 per cent do not believe that Aboriginal people are treated equally to other Australians. (3) In the sample there are individuals who think of themselves as close to the Liberal Party (27.5 per cent), the Labor Party (20 per cent), and the Greens (7.5 per cent), with the rest identifying with other parties. Over 48 per cent of the respondents are female, and the average age in the sample is nearly 54. Around 3 per cent of individuals identify as Indigenous; however, dropping these individuals from the sample makes no substantive difference to the results of the analysis. Nearly 28 per cent in the sample were born in a country other than Australia.
The educational attainment in the sample is rather high, with 36 per cent possessing a tertiary qualification and 35 per cent a post-secondary qualification. Over 61 per cent of respondents are employed, and over 60 per cent have a religion, with the majority of these comprising Christians.
We start by asking whether, similar to politicians, individuals' political ideology contributes to their views on issues concerning Indigenous Australians. In Fig. 4 we look at the question descriptively, presenting the shares of individuals who express agreement on the core set of issues studied: (1) that government assistance to Indigenous people is justified, given the level of disadvantage they face; (2) that granting rights to Indigenous people is not unfair to other Australians; (3) that Indigenous people are not treated equally to other Australians; and (4) that Indigenous people should not have to change their culture to fit into Australian society. We see a clear pattern by political ideology emerge, with individuals who identify with the LN Coalition the least likely to agree with the statements, whereas those who identify with the Greens are the most likely to do so. For example, only 34 per cent of those who identify with the LN Coalition agree that granting rights to Indigenous people is not unfair to other Australians, relative to 58 per cent agreement with the statement among Labor Party supporters and 89 per cent among Greens. Similarly, only 41 per cent of Liberals believe that Indigenous people are not treated equally, compared to 60 per cent among those who identify with Labor and 88 per cent among those who identify with the Greens.
As discussed earlier, the intergroup contact theory of majority-minority relations suggests that prejudice to minorities might decline with contact with out-group members (Allport, Clark, and Pettigrew, 1954;Pettigrew & Tropp, 2006). Are there systematic differences in patterns of interaction with Indigenous Australians by type of party identification? Fig. 5 suggests that individuals who identify with the Greens, and to a lesser extent those who identify with the Labor Party, are more likely to mix with Indigenous people relative to those who identify with the LN Coalition. Looking at interactions across all contexts, 85 per cent of individuals who identify with the Greens report mixing with Indigenous people frequently, sometimes, or occasionally (as opposed to not mixing at all). Only 70 per cent of individuals who identify with the LN Coalition do so. Similarly, the share of individuals who report mixing with Indigenous people at least occasionally in the context of their work is 58 per cent among those who identify with the Greens, 51 per cent for those who identify with Labor, and down at 44 per cent for those who identify with the LN Coalition.

Estimation results
Next, we explore these relationships further in a regression analysis that additionally controls for a wide range of individual characteristics and state fixed effects, as presented in Eq.
(3). The results reported in Table 6 show that the patterns of association between party affiliation and support on issues concerning Indigenous Australians persist. We estimate consistently negative marginal effects on identification with the LN Coalition and positive marginal effects on identification with the Labor and Green Parties. For example, moving from other/no party identification to identifying with the Labor Party is associated with a 17 percentage point increase in the probability of expressing support on the statement that granting land rights to Indigenous people is not unfair to other Australians. However, identifying with the LN Coalition decreases the probability of expressing support on the statement by 6.4 percentage points. Among other characteristics of relevance to individuals' views on issues concerning Indigenous Australians, educational attainment is positively associated with supportive attitudes, with those with tertiary educational attainment significantly more likely to have such attitudes relative to those with secondary or lower educational attainment. On the other hand, religious individuals appear to be less likely to possess supportive attitudes on Indigenous issues relative to their non-religious counterparts.
Next, we turn to the analysis of patterns of interaction with Indigenous Australians in a regression analysis that controls for a range of individuals' observable characteristics and state fixed effects. The results reported in Table 7 suggest that the earlier established positive associations between party identification and interactions with Indigenous Australians are robust to controlling for a range of individual characteristics; however, the estimated marginal effect on identification with the LN Coalition is statistically insignificant. Other individual characteristics that are positively associated with the probability of mixing with Indigenous people include educational attainment, being employed, and coming from an Indigenous background. On the other hand, overseas-born individuals are significantly less likely to have interactions with Indigenous Australians.
Overall, the results of this analysis suggests that individuals identifying with the Greens or the Labor Party are on average more likely to have interactions with Indigenous people and this may, in turn, feed into the more supportive attitudes they exhibit towards issues concerning Indigenous people, as reported in the first part of the analysis. We provide further engagement with this possibility by re-estimating Eq. (3) while controlling for our measure of interaction with Indigenous Australians on the right-hand side, and excluding  Table 9 in the Appendox, and are statistically significant on two instances. Increased interaction with the Indigenous population is associated with increased probability of having supportive attitudes on issues concerning the land rights and cultural identity of Indigenous Australians. We then control for political identity of individuals-the results are reported in the even-numbered columns of Table 9. The significance of the marginal effect on our measure of interactions with the Indigenous population is preserved in the regression, where the dependent variable is individuals' judgements over the question of whether Indigenous people should not have to change their culture to fit into Australian society. In all other instances, the estimated marginal effects on this variable, while positive, are statistically insignificant.

Conclusion
Australia's Indigenous population is significantly disadvantaged in terms of a wide range of observable outcomes, and policy interventions to address the issue are critical. Yet, there is a lack of understanding on how policies dealing with issues concerning the Indigenous population come about-a question that this paper attempts to engage with. In a study of Australia's parliamentarians and voters, we show that identity plays an important role in determining the expressed support for issues concerning the Indigenous population. Political ideology is an important determinant of politicians' and voters' preferences on Indigenous policies, and there is a strong alignment between the two groups in that respect. Politicians' propensity to act rebelliously and diligently also feeds into their stance on Indigenous-focused policies. Among voters, on the other hand, the propensity to have a That politicians' identity determines their voting over policies is consistent with 'citizen candidates' models which predict that politicians implement policies consistent with their preferences in a setup characterized by the absence of complete political commitment  (Osborne & Slivinski, 1996;Besley & Coate, 1997). Moreover, we do not find that characteristics of the electorates have any explanatory power over politicians' voting choices, against the prediction of the Downsian model that policy outcomes should be determined by the preferences of the median voter (Downs, 1957). Our analysis, however, is based on a small sample and is descriptive in nature; therefore our findings need to be taken with a degree of caution. While our analysis shows that various dimensions of identity of politicians and voters might determine their stance on policies concerning the Indigenous population, we are not able to make causal claims in the current setup. The datasets that we have employed, while unique, are lacking important background details on individual politicians and voters, as well as their contexts and circumstances, and omission of such information may confound our estimates. Future research in this area would benefit from the use of richer datasets and quasi-experimental setup to establish causal effects. A dedicated survey of politicians to elicit their attitudes and insights alongside collection of more detailed background information would be particularly valuable in this space.
That certain dimensions of identity might prompt a stance against policies to address the Indigenous disadvantage calls for more research into why that might be the case. Our individual-level analysis highlights the potential role of education in supporting preferences for addressing the issues concerning the Indigenous population, including potentially targeting the racial bias among certain groups in society. Consistent with the intergroup contact theory, the prejudice against the Indigenous population is likely to decline with increased opportunities for contact and Indigenous cultural immersion. Understanding the nature of different contexts and the associated discourse that promote the racial bias among certain groups in society defined by political, religious, or other forms of affiliation or identification is an important area of research.

Appendix A
See Tables 8 and 9. See Figs. 6,7,8,9,10,11, (3) Standard errors in parentheses. * Significance at 10 per cent; ** 5 per cent; *** 1 per cent levels  Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.