Representing diversity in a liberal democracy: a case study of Australia

ABSTRACT

Political representation of ethnic minorities in a liberal democratic system is a crucial step towards having the interests of ethnic minority groups heard. The under-representation of ethnic minorities in a political institution can exacerbate inequality between majority and minority populations and increase feelings of alienation among minority groups. As a country observing increasing ethnic diversity, Australia saw record-level ethnic minority candidates elected in the 2022 federal election. However, the shares of candidates and elected Members of Parliaments with ethnic minority backgrounds are still much lower than their relative shares in the population. In this regard, Australia has lagged behind other major settler countries. In this paper, we examine the political representation of ethnic minorities in Australia’s federal election. Drawing on data from the 2022 federal election and 2021 population census, we find a positive association between ethnic minority concentration and ethnic representation. However, for the two major parties, ethnic minority candidates are less likely to be in safe seats, even when the seat observes high ethnic minority concentration. Findings suggest that ethnic voting is evident but it is perhaps too early to celebrate higher levels of ethnic representation in Australian politics as political parties act as gatekeepers in safe seats.

KEYWORDS:

• Political representation
• ethnic minority
• ethnic vote
• population composition
• party strategies

Introduction

The 2022 Australian federal election saw an increase in ethnic minority candidates elected. However, the presence of candidates and elected Members of Parliaments of ethnic minority backgrounds is still much lower than their relative shares in the population. In 2021, more than 28 per cent of the Australian population reported having non-Australian-and-European (hereafter non-European) ancestries¹ (Australian Bureau of Statistics (ABS) 2022), yet only 9 per cent of the House of Representatives members have non-European backgrounds. We define ethnic minorities as those from non-European backgrounds in this paper. This is significant, because in the Australian context, non-European as a potential category of racial discrimination has rarely been acknowledged since the White Australia policy ended in 1975 (Jupp 1995). In this regard, Australia has lagged behind other major settler countries in not only failing to acknowledge the existence of non-European minority discrimination but also in research on the under-representation of non-European ethnic minorities in political institutions.

The under-representation of non-European ethnic minorities in Australia’s political institutions can exacerbate inequality between majority and minority populations and increase feelings of alienation among minority groups, as political representation is a crucial step towards having the interests of the represented groups heard in the democratic system (Pantoja and Segura 2003; Pietsch 2018). In Australia, the population and ethnic compositions have changed with the country prioritising skilled migration since the mid-1990s and the main source countries of its new immigrants shifting from Europe to Asia (Hugo 2004; Raymer et al. 2018, 2020). New immigrants are becoming more dispersedly distributed across the country (Guan 2019). In the 2022 federal election, even though under-represented as a share of the population, more ethnic minorities were elected to the House of Representatives than ever before. There is no doubt that the increasing size of ethnic minority groups within electorates and the increased political mobilisation of ethnic communities helped to play a role. However, while ethnic voting may have increased in electorates with ethnic candidates, political parties have continued to act as gatekeepers in safe seats, including those with high levels of ethnic concentration due to the persistent influence of factions (Jupp 2003) and perceptions of discrimination (Pietsch 2018).

In this paper, we aim to understand the under-representation of ethnic minorities in Australia’s latest federal election and the Parliament, and whether political parties played a key role in shaping the under-representation of ethnic minorities. The first research question we explore is the extent to which electorates with higher concentrations of ethnic minority populations are more likely to have ethnic minority candidates. Our second research question examines whether ethnic minority candidates are more (or less) likely to be elected than ethnic majority candidates, particularly in areas with high ethnic minority concentrations. Our third and final research question examines whether ethnic minority candidates are more (or less) likely to be selected for unwinnable or precarious seats by the two major parties in Australia.

This research has several innovations and contributions. We begin our analyses with a theoretical and contextual background on ethnic minority under-representation with a focus on the role of party gatekeepers in shaping ethnic minorities’ political under-representation. While there has been research on the role of group representation in terms of gender, economic or professional background (Norris and Lovenduski 1995), less is known about the role of political parties in ethnic minority under-representation, especially in terms of how parties respond to changing attitudes towards race in different societal contexts. Our work fills the research gap and also contributes to understanding the under-representation of ethnic minority groups in Australia’s federal political institution. Methodologically, we augment the traditional sources on candidates’ ethnicity – e.g. publicly available information including newspaper articles, candidates’ websites, interviews, and speeches – by incorporating model-based ethnicity predictions. Also importantly, while most prior studies on ethnic representation or ethnic voting have used survey data and modelled party preference as the outcome variable (Teney et al. 2010; Huber 2012; Carlson 2015; Houle, Park, and Kenny 2019; Pietsch 2017a; 2017b),² we examine results from the actual election.

Theoretical and contextual background

There has been a great deal of research on the significance of political representation within liberal democracies (Hibbing and Theiss-Morse 2002; Andeweg and Thomassen 2005; Rehfeld 2011; Mansbridge 2011). These theoretical debates have become increasingly important, especially in the current context of debates over the existence of a ‘democratic deficit’ and declining public trust in representative institutions (Pharr, Putnam, and Dalton 2000; Norris 2011; Cameron 2020). Pivotal to theoretical studies of representation is the distinction between descriptive and substantive representation in national parliaments (Pitkin 1967; Phillips 1995; Krobe 2018). The distinction is based on the actual numbers counted (descriptive) and the effects of minorities in parliament (substantive). The argument for descriptive representation comes from the idea that political institutions should reflect the social composition of the populations they serve. In other words, legislatures in Western democracies should aim as close as possible to mirror the social characteristics of their electorates. Substantive representation, however, demands that the policy preferences of the electorate be translated into legislative behaviour (Pitkin 1967; Phillips 1995; Pietsch 2018).

The two concepts of descriptive and substantive representation are often intertwined. For example, Phillips (1995) has argued for the importance of a ‘politics of presence’ where the parliamentary presence of ethnic minority groups enhances the quality of democracy. The ‘politics of presence’ argument suggests that there is a link between descriptive and substantive representation. This is because elected minorities are best equipped to deal with the issues affecting minority groups. Other researchers internationally have similarly demonstrated the importance of minority group presence in national parliaments for democracy. For example, research has shown how the election of members from traditionally under-represented groups has a positive impact on policy-making and overall feelings of trust in the political system (Pande 2003; Bird, Saalfeld, and Wüst 2011; Crisp et al. 2018; Pietsch 2018).

In recent decades, the scholarly debate has paid increasing attention to the political representation of ‘visible minority’ and/or ‘non-white’ ethnic minority groups in politics (Lien, Conway, and Wong 2004; Bird, Saalfeld, and Wüst 2011; Pietsch 2018). Much of the literature stems from the observation that ‘non-white’ ethnic minorities are under-represented in formal political institutions despite having a long presence in the country. Pietsch (2018) observes that ethnic minority political under-representation undermines trust in political institutions and an overall sense of belonging among ethnic minority groups. Their lower levels of trust and belonging may also impact their successful social and economic integration. This is because their lack of presence in the political system may result in the policy process not addressing their needs (Pietsch 2022). Therefore, achieving representation of ethnic minorities has important implications for the overall quality of democracy and is an expected outcome of pluralist models of democracy.

There are several barriers to the political representation of ethnic minorities such as generation, where participation increases with each generation as well as the duration of stay in the country of destination and recent immigrants are less likely to participate (Ramakrishnan and Espenshade 2001). Thus, older generations or recent immigrants are less likely to participate in party politics and thus be selected as candidates. Other barriers include the pan-ethnic mobilisation (or lack of mobilisation) of candidates (Lien, Conway, and Wong 2004) and the political origin of immigrants in terms of formative experiences of democratic or authoritarian socialisation. Those from authoritarian systems may know less about the democratic process and feel less engaged (Bilodeau 2008; Bilodeau, McAllister, and Kanji 2010; Pietsch and McAllister 2016; Ruedin 2016). Finally, the legal-institutional context can shape political representation with differences in integration policies, access to citizenship and voting eligibility (Bloemraad 2015; Manatschal and Bernauer 2016), as well as different electoral systems that are more or less favourable for ethnic representation (Bird 2016; Bochsler 2007; Geese and Schacht 2019).

Two possible factors may explain the under-representation of ethnic minorities that remain under-explored: ethnic voting³ and party-political factors such as the selection of ethnic candidates to (un)winnable seats. Public attitudes of the majority group towards minority groups in certain electorates may become an important factor in these decisions. Therefore, those involved in the screening and selection of candidates are likely to be strategic in deciding whether it is electorally beneficial to put forward a candidate with an ethnic minority background (Fieldhouse and Cutts 2008; Bloemraad and Schönwälder 2013). In terms of party strategy, the group size and geographical concentration of ethnic communities also matter a great deal in these decisions. Studies show that ethnic communities with higher geographic concentrations are more likely to realise their descriptive representation (Bochsler 2010). With the continued growth of immigrant and ethnic communities across Australia (Raymer et al. 2018, 2020), the percentage of voters with ethnic minority backgrounds is expected to increase over time as they naturalise and become more integrated (Tillie 2004; Ager and Strang 2008; NASEM 2015). Voting is compulsory for all Australian citizens (Australian Electoral Commission (AEC) 2011). Over time, as ethnic minorities become more integrated (naturalised) and mobilised politically, it is expected that political representation will become more achievable (Bloemraad and Schönwälder 2013). These observations lead us to expect that ethnic minority representation will increase naturally over time as more ethnic candidates are integrated into the political system, achieving parity and successful political incorporation of ethnic minority concerns (Hochschild and Mollenkopf 2009).

While these are our theoretical expectations, in the Australian context there are a few contextual explanations that could have led to a change in the national discourse on ethnic minorities and representation leading up to the 2022 federal election. First, the Australian Human Rights Commission (2018) report highlighted that even though Australia is widely celebrated as a multicultural triumph, concerns about cultural diversity had been assimilated into concerns about gender diversity, while concerns about culture and race had been downplayed or ignored. In the following years, it became clear that these concerns could no longer be ignored. The Black Lives Matter movement in the United States was beginning to gain support in Australia, with non-European Australians expressing shared experiences of discrimination in protests across capital cities in Australia (United Nations Association of Australia (UNAA) 2021). Likewise, Indigenous Australians shared their experiences of Black deaths in custody and police brutality (UNAA 2021). It had become clear that culture and race could no longer be downplayed or ignored in Australia.

Leading up to the 2022 federal election, Australia had more than 28 per cent of its population from non-European backgrounds, political parties could no longer select white candidates in minority-majority seats without justification. This was evident in the lead-up to the 2022 election when the Australian Labor Party (ALP) Senator Kristina Keneally announced her plan to move to the lower house by running in a safe, ethnic-minority majority seat, displacing a local candidate with a Vietnamese refugee background, Tu Lee. It had become clear that in Australia, after nearly 50 years since the White Australia Policy was officially renounced, political parties, as gatekeepers, could no longer exert as much control over political outcomes and were under increasing pressure to diversify. To understand the extent of the influence of these party political factors, we first begin with two hypotheses grounded in theories of political representation:

H1. Electorates with a higher concentration of ethnic minorities are more likely to have ethnic minority candidates running for election (not necessarily from the major parties)

H2. Electorates with a higher concentration of ethnic minorities are more likely to have ethnic minority candidates elected

These recent party political behaviours that marginalise ethnic minority candidates suggest that the increase in ethnic minority representation at the 2022 federal election may have had more to do with the progressive or liberal wave involving more positive attitudes towards minorities in the electorate, rather than any conscious effort on the part of the major parties. For example, in 2022, survey findings showed that the proportion of people who agree or strongly agree with the statement ‘we should do more to learn about the customs and heritage of different ethnic and cultural groups in the country’ increased from 59 per cent in 2018 to 69 per cent in 2022 (Scanlon Institute 2022). Even though broader social attitudes have changed within the electorate, ethnic minority candidates continue to be relegated to marginal seats or unwinnable seats. One reason that explains this is that political parties often rely on perceptions that can be quite different from the reality of discrimination within the electorate (Jupp 2003; Pietsch 2018). If a candidate is pre-selected to contest a safe and contestable seat, they still need to deal with discrimination and racial prejudice in the electorate throughout the election campaign. However, without detailed polling in each electorate, there is a lack of clarity around whether putting forward an ethnic minority candidate would result in a political party being penalised. Consequently, members of ethnic minority groups are often selected for unwinnable seats and, therefore, seldom elected to the Federal House of Representatives. Safe seats, on the other hand, are often occupied by or kept for cabinet or shadow cabinet ministers (Denemark 2000; Pietsch 2016), as in the aforementioned case of Kristina Keneally.

The role of party gatekeepers in Australia resonates with studies of other under-represented groups. For instance, the ‘sliding glass cliff’ effect describes a real-world phenomenon in which women as an under-represented group are more likely to be appointed to precarious leadership positions in poorly performing organisations, while men are more likely to be appointed to stable leadership positions in successful organisations (Ryan and Haslam 2005; Ryan, Haslam, and Kulich 2010). Research on the representation of women in politics similarly shows that women inevitably hit a ‘glass ceiling’ because of invisible boundaries over which their advancement becomes more challenging (Krobe 2018; Martinez I Coma and Lago 2021; Martinez I Coma and McDonnell 2021). So far, research on the political incorporation of ethnic minority candidates in a white-majority parliament has been limited, therefore, we examine whether there is evidence of gatekeeping within the Australian political system – in other words, whether the ‘glass cliff’ phenomenon is similarly experienced by ethnic minority groups within the Australian national parliament. Thus, we establish the third hypothesis for examination:

H3. Low descriptive representation of ethnic minority populations in elections is related to the marginalisation of ethnic minority background candidates.

Data and analytic strategy

Data sources

In the 2022 federal election, 1,203 candidates were running for election in 151 electorates (seats) in the House of Representatives while 421 candidates were running for election for 79 seats in the Senate. In Australia, Senate relies on a Proportional Representation voting system using a Single Transferrable Vote to elect its members. Voting for Senate candidates is state-based: each of the six states in Australia is represented by 12 senators and the two territories are represented by two senators each. House of Representatives relies on a preferential voting system to elect candidates in single-member electorates where voters are asked to ‘mark a preference for every candidate’ running for the seat (AEC 2022). The latest population data are used to distribute and redistribute House of Representative seats across states and territories (AEC 2023). The scope of our analysis is constrained to the House of Representative candidates. Senate candidates are excluded from the analysis because over 98 per cent of Australian voters choose to hand control to parties and vote for parties rather than individual candidates in the Senate (Green 2017).

We consult three data sources to analyse the descriptive representation of ethnic minorities in the 2022 Australian federal election. Election data are sourced from AEC. Both candidate-level statistics and electorate-level statistics are used to inform the analysis. Available candidate-level variables from AEC include surname, given name, party, 2022 election result, historical election result, and electorate. Available electorate-level variables include the number of candidates running for election, the number of enrolled voters, the geographic typology of the electorate, and the seat status. The numbers of enrolled voters for the 151 electorates range from 71,843 for the Solomon electorate to 133,408 for the Macarthur electorate.

The AEC defines seat status based on two-party preferred (TPP) vote counts and the composition of the final two candidates in the last election. TPP vote is for the two major parties: the ALP and the Liberal-National Coalition (LNC).⁴ The 2022 election seat status is derived from the results of the 2019 federal election. Non-classic seats are electorates where the final two candidates in the 2019 election did not comprise both an ALP candidate and a LNC candidate. Of all other seats (i.e. classic seats), TPP vote counts are used to define the seat status. TPP vote counts consider only the preferential votes for the ALP and LNC parties. Safe seats are where the leading party received more than 56 per cent of the TPP vote in the 2019 election. A safe seat for one party is an unsafe seat for the other party. Marginal seats are when the leading party receives less than 56 per cent of the TPP vote. In our analysis, we further group non-classic seats, marginal seats, and unsafe seats for the ALP and the LNC as precarious seats to analyse party strategies in putting forward an ethnic minority background candidate in the election.

Additional electorate-level variables are sourced from the ABS. The 2021 Australian Population Census is used to inform the population compositions of each electorate.⁵ A set of variables are examined to see if these measures represent very different population compositions across electorates. These include the percentage of the resident population born overseas, the percentage using a language other than English at home, the percentage without Australian citizenship, the percentage having at least a parent born overseas, the percentage with ethnic minority ancestries, the percentage with tertiary qualifications, the percentage earn higher than the national median personal income⁶, the percentage not in the labour force, and the old-age dependency ratio. The above percentages are also obtained for the voter population in each electorate, i.e. for those with Australian citizenship and who were at least 17 years old at the time of the Census.⁷ Population sizes for the 151 electorates range from 109,554 for the Solomon electorate to 205,508 for the Macarthur electorate. Voter population sizes for the 151 electorates range from 63,824 for the Solomon electorate to 128,126 for the Macarthur electorate.

The candidates’ ethnic minority backgrounds are identified using data sourced from an ethnic minority candidate list compiled by the Centre of Multicultural Political Engagement, Literacy, and Leadership (COMPELL 2022). The COMPELL data are compiled using public information and are freely available from their website. Noises and errors in the COMPELL data are checked and corrected, including removing three candidates who had withdrawn from the election and correcting one candidate who was categorised as running for the Senate in the COMPELL list but ran for the House of Representatives (and within the scope of our study). After the process, 109 candidates are marked as ethnic minority background candidates. The term COMPELL used to describe ethnic minority background candidates refers to ‘candidates of colour’, which includes candidates of Indigenous backgrounds. The list does not distinguish the detailed ethnic minority backgrounds of candidates. We interpret the list as candidates from a non-white background and not including candidates of European origins. We use a consistent typology to determine our description of an ethnic minority population in the rest of the paper unless otherwise specified. Ethnic minority candidates in this paper refer to candidates who have publicly identified themselves as from non-European ethnic backgrounds.

Predicting ethnicity based on the names of candidates

To further validate the cleaned COMPELL list and increase our confidence in coding ethnic minority backgrounds for the candidates, we augment the list with the predicted ethnicity of candidates using an R package predictrace (version 2.0.0; Kaplan 2021). The package is based on the United States (US) Census data (Tzioumis 2018) to predict the most common ethnicity of a person’s surname, and the US Social Security Administration data to predict the most common ethnicity of a person’s first name. Thus, the typology of predicted ethnicity is based on the US definition and categorisation of race. To our best knowledge, there is no comparable Australian data-based prediction model to perform similar analyses. We consider the predicted white and non-white categorisation to be consistent with the COMPELL typology of ethnicity and thus our definitions of ethnic majority and minority candidates.

In Table 1, the comparison between the candidates’ ethnicity in the cleaned COMPELL list and the predicted ethnicity is presented. The bold cells are where the predicted ethnicity (race) does not align with records in the cleaned COMPELL list. We further verify the ethnicity of the 71 candidates in these cells using publicly available information, including candidates’ websites, online articles interviewing the candidates, and speeches or videos on online platforms. The first two pages of Google search results with each candidate’s first name and surname are used in this verification process. The search is further augmented by using candidates’ party affiliation and the keywords ‘migrant’ or ‘ethnic’. After the search, six candidates are re-coded to ethnic minority backgrounds. For the inconsistent italic cell, we put high confidence in the cleaned COMPELL list⁸ and keep their coding as ethnic minority background candidates. In the end, 115 candidates are coded as ethnic minority candidates.

Table 1. Assessing the ethnicity of candidates in the cleaned COMPELL list and the predicted ethnicity based on candidates’ names.

Predicted ethnicity	Records in the COMPELL list	No record in the COMPELL list	Total
White	53	1,023	1,076
Asian	42	6	48
Black	2	17	19
Hispanic	3	16	19
No predictions	9	32	41
Total	109	1,094	1,203

Note: No predictions are where ethnicity cannot be predicted based on the first name or surname of the candidate.

Analysing ethnic minority representation

Ethnic minority representation in the 2022 Australian federal election is analysed in three steps to answer our research questions. First, Pearson’s correlation coefficients and a scatter plot are used to examine the correlations between the population compositions and the presence of ethnic minority background candidates in an electorate (H1). Second, ordinary least squares regression, logistic regression, and path analysis models are used to test the association between population compositions in an electorate and whether the electorate has an ethnic minority background candidate elected in the 2022 election (H2). These models are built in R using the lm and glm functions in the stats package (version 4.1.0), and the sem function in the lavaan package (version 0.6.12). Third, Fisher’s Exact tests and a boxplot are used to examine the correlation between seat status and the presence of ethnic minority background candidates for the two major parties in Australia (H3).

Results

Ethnic representation in the election

Ethnic minority candidates are not uncommon in the 2022 federal election. Of the 151 electorates, 72 had at least one ethnic minority candidate (Table 2). The number of ethnic minority candidates in the 72 electorates ranges from one to six. Most electorates with ethnic minority candidates had only one ethnic minority candidate (N = 41).

Table 2. The ethnic composition of House of Representative candidates nominated, 2022 Australian federal election.

	Number of ethnic minority candidates
Number of candidates	0	1	2	3	4	6
5	3	3	2*	0	0	0
6	11	9	2	0	0	0
7	23	4	4*	0	1	0
8	19	8**	4	2*	0	1*
9	9	11*	5*	2*	0	0
10	8	4	3*	2*	0	0
11	6	2	2*	0	0	0
12	0	0	1	0	0	0
Number of electorates	79	41	23	6	1	1
Number of electorates with ethnic minority candidates elected	0	4	5	3	0	1

* One of the electorates has an ethnic minority candidate elected

** Three of the electorates have an ethnic minority candidate elected

However, most of the electorates have lower levels of descriptive representation of ethnic minorities. Compared to the percentage of ethnic minority candidates in the 72 electorates, which ranges from 9 per cent to 75 per cent, their percentage of the voter population whose reported ancestries⁹ are not Australia, Europe, New Zealand, United States or Canada ranges from 11 per cent to 81 per cent (Figure 1). Of the 82 electorates with zero ethnic minority candidates, the percentage of ethnic minority voters ranges from 8 per cent to 69 per cent.

Figure 1. The percentage of ethnic minority background candidates in the electorate in the 2022 election (x-axis) by the percentage of ethnic minority ancestries reported in the electorate in the 2021 Census (y-axis): a fitted curve in blue with the confidence interval in grey.

Albeit the overall evidence of under-representation of ethnic minorities in the election, the relationship between the two percentages in Figure 1 is positive. We further examine this positive relationship using Pearson's correlation coefficients and assess if different population composition measures represent similar (or different) patterns across electorates.

Table 3 presented Pearson's correlation coefficients calculated between the percentage of ethnic minority candidates (PEMC) in each electorate and the population compositions of the electorate in different forms. The ten ethnic minority and immigrant population composition variables are all strongly correlated with each other and have moderate to strong¹⁰ positive correlations with PEMC. The highest correlation coefficient is 0.54 between PEMC and the percentage of voters’ reported ancestries not being Australia, Europe, New Zealand, United States or Canada (V8). The correlation tests support the first hypothesis that electorates with a higher concentration of ethnic minority populations are more likely to have ethnic minority candidates running for election. Given the high levels of statistical correlations across the ten ethnic and migrant population composition variables (V1 to V10), we present only results from the percentage of the voter population using a language other than English at home (V4) and the percentage of voters’ reported ancestries not being Australia, Europe, New Zealand, United States or Canada (V8) in the rest of the paper to have a closer look at the descriptive representation of the language-diverse population and the non-European ethnic minorities. Results of other ethnic minority and immigrant population composition variables are presented in online Appendix A as robustness checks.

Table 3. Summary statistics and Pearson’s correlation coefficients between variously defined population composition in each electorate in the 2021 Census and the percentage of ethnic minority candidates (PEMC) in the electorate in the 2022 election.

Population composition	PEMC	V1	V2	V3	V4	V5	V6	V7	V8	V9	V10	V11	V12	V13
PEMC		0.49	0.53	0.52	0.53	0.53	0.53	0.54	0.54	0.48	0.41	0.16	0.06	−0.34
V1: % Born overseas			0.97	0.90	0.86	0.89	0.89	0.88	0.88	0.98	0.82	0.59	0.30	−0.65
V2: % Born overseas (voters)				0.91	0.90	0.90	0.90	0.90	0.90	0.98	0.70	0.46	0.19	−0.60
V3: % Using a language* other than English at home					0.99	0.98	0.98	0.98	0.98	0.89	0.76	0.45	0.10	−0.63
V4: % Using a language* other than English at home (voters)						0.98	0.98	0.98	0.99	0.87	0.68	0.35	0.03	−0.59
V5: % Non-Australia^/British ancestries							1.00	0.99	0.99	0.89	0.73	0.48	0.18	−0.65
V6: % Non-Australia^/British ancestries (voters)								0.99	0.99	0.89	0.71	0.48	0.18	−0.64
V7: % Non-Australia^/Europe/New Zealand/United States/Canada ancestries									1.00	0.88	0.71	0.41	0.13	−0.64
V8: % Non-Australia^/Europe/ New Zealand/United States/Canada ancestries (voters)										0.89	0.70	0.42	0.13	−0.63
V9: % Parent born overseas											0.74	0.57	0.31	−0.65
V10: % Non-Australian citizens												0.50	0.26	−0.66
V11: % Tertiary qualification (age 25+)													0.78	−0.43
V12: % Income above national median personal income														−0.51
V13: Old-age dependency ratio
Mean	9.71	28.36	25.16	22.68	18.34	43.49	43.60	32.02	30.89	49.97	15.80	32.63	50.14	27.63
Standard Deviation	12.90	13.48	13.68	17.73	16.69	13.64	13.83	17.21	17.68	18.95	5.68	14.57	7.00	8.87

* Including Aboriginal and Torres Strait Islander languages used by the Indigenous people

^ Those who reported an Indigenous origin are counted in the non-Australia statistics.

Three other key socio-demographic variables that traditionally have explained electoral behaviour (Butler and Stokes 1969; Campbell et al. 1960) and voting behaviours in Australia (Pietsch, Graetz, and McAllister 2010; Biddle and McAllister 2022) are also examined in Table 3. Old-age dependency ratio of the electorate has a weak to moderate level of negative correlation with PEMC (r = −0.34). This needs to be interpreted in relation to the younger age profile of migrants: the correlation coefficient between the old-age dependency ratio and the ten ethnic minority population composition variables is strong (r = −0.59 ∼ −0.66 in Table 3). The two socioeconomic variables measuring education and income compositions have only a weak or negligible relationship with PEMC which is a surprising finding given the prior significance of socioeconomic factors in explaining electoral behaviours in Australia (see e.g. Pietsch, Graetz, and McAllister 2010). Perhaps, the weak correlation needs to be attributed to factors other than ethnic population composition, as the percentage of tertiary educated residents (age 25+) shows a moderate to strong positive correlation with the ten ethnic minority population composition variables (r = 0.35 ∼ 0.59).

Ethnic votes

Our second hypothesis posits that ethnic votes exist such that electorates with a higher concentration of ethnic minority populations are more likely to have ethnic minority candidates elected. The hypothesis is tested using regression models to predict the ethnicity of the elected candidate in each electorate by the population composition of each electorate.¹¹ Results of model estimates are presented in Table 4, Model set A, for two population composition variables and two socio-demographic variables.

Table 4. Coefficients from separate regression models: selected independent variables (rows), dependent variables (columns), and sample sizes.

Model index and independent variable	Dependent variable: Ethnicity of the elected candidate 0 = Ethnic majority (ref.), 1 = Ethnic minority					Presence of at least an ethnic minority candidate 0 = No (ref.), 1 = Yes			Dependent variable: % Ethnic minority candidates
	Model set A: N = 151			Model set B: N = 72		Model set C: N = 151			Model set D: N = 72
	odds ratio (95% CI)		AIC	odds ratio (95% CI)	AIC	odds ratios (95% CI)		AIC	estimates (95% CI)		Adj. R2	AIC
V4: % Using a language* other than English at home (voters)	1.0388	**	86	1.0166	71	1.0659	***	184	0.2538	***	0.17	547
	(1.009, 1.07)			(0.985, 1.049)		(1.039, 1.098)			(0.123, 0.384)
V8: % Non-Australia^/Europe/ New Zealand/United States/Canada ancestries (voters)	1.0405	**	86	1.0173	71	1.0607	***	183	0.2577	***	0.18	545
	(1.011, 1.072)			(0.986, 1.05)		(1.037, 1.089)			(0.131, 0.384)
V11: % Tertiary qualification (age 25+)	1.0161		92	1.0101	72	1.0153		211	0.1527		0.02	558
	(0.977, 1.054)			(0.967, 1.054)		(0.993, 1.039)			(−0.039, 0.345)
V13: Old-age dependency ratio	0.9400		90	0.9661	71	0.9526	*	207	−0.5378	***	0.16	547
	(0.867, 1.008)			(0.887, 1.037)		(0.915, 0.989)			(−0.819, −0.256)

Significance levels: *** 0.001; ** 0.01; * 0.05. Note: N = 72 are electorates with at least one ethnic minority candidate. CI = confidence interval

Note: We also test alternative models with both an ethnic minority population composition variable and other electorate-level control variables including the percentage of the electorate’s 25 years old and above population with tertiary qualifications, the percentage of the electorate’s population whose income are above the national median personal income, old-age dependency ratio, the number of candidates running for election, the number of enrolled voters, and geographic typology of the electorate. Including or excluding these control variables only slightly change the values of odd ratios for population composition variables but not the direction of relationship between variables or the statistical significance of the relationships. Of all the control variables tested, the number of candidates is sometimes statistically significant and positively associated with having ethnic minority candidates presented.

This set of models uses all 151 electorates to estimate the relationship between population compositions and the election results. Higher proportions of ethnic minority populations in Model set A predict higher odds of having an ethnic minority candidate than an ethnic majority candidate elected in the 2022 federal election. The socio-demographic variables show no statistically significant associations with the election outcome variable.

Note that the first set of models did not account for the fact that there were 79 electorates with zero ethnic minority candidates (Table 2) and thus were guaranteed to have an ethnic majority candidate elected regardless of the electorate’s population compositions. To take this into account, we test a second set of models using only the 72 electorates with at least one ethnic minority candidate running for the election (Table 4, Model set B). For electorates with ethnic candidates running for election, none of the tested models has statistically significant associations with the election outcomes. Comparing Model set A and Model set B, it seems to suggest that ethnic minority population compositions of the electorates are predictors of having ethnic minority candidates running for election rather than predictors of ethnic minority candidates attracting ethnic votes and being elected.

To test such possibilities, we construct a third set of models predicting the presence of ethnic minority candidates in the 151 electorates (Table 4, Model set C) and a fourth set of models predicting the percentage of ethnic minority candidates in the 72 electorates with at least one ethnic minority candidate (Table 4, Model set D). Both sets of models suggest statistically significant positive associations between higher language diversity or higher non-European ethnic minorities in the voter population and higher ethnic representation in the election.

Negative associations between old-age dependency ratios and the two ethnic representation variables are observed with strong statistical significance. Note that we only included one independent variable in each model to predict a dependent variable. The statistical significance associated with the old-age dependency ratio could possibly be attributed to the young age profiles of migrants (Table 3; also see Raymer et al. 2018). In online Appendix A Table A1, we present the full list of models with all 13 population composition variables and the above findings hold.

We further examine the contested relationship between ethnic minority population composition, ethnic representative in the election, and election results using a set of path analysis, including controlling for socio-demographic variables. Results are presented and discussed in online Appendix C. It suggests that the positive associations between a higher concentration of ethnic minority population and a higher probability of having an ethnic minority candidate elected (H2) are established through the presence of higher percentages of ethnic minority candidates running for election in places with higher ethnic concentrations. In other words, higher ethnic concentration predicts higher ethnic engagement in the election but not necessarily higher chances for ethnic minority candidates to be elected. If we further take into account the historical election results, online Appendix B Table B1b shows that the 2022 election results and historical election results are statistically correlated for both ethnic majority and ethnic minority candidates. It suggests an incumbency factor for candidates regardless of their ethnic backgrounds. These all point to an important underlying factor determining ethnic representation in the election and in the parliament such that parties strategically screen and select ethnic minority candidates based on the benefits of putting them forward in the election (Fieldhouse and Cutts 2008; Bloemraad and Schönwälder 2013). We analyse the party strategies in the next section to answer the third research question proposed at the beginning of the paper: are ethnic minority candidates more likely to be selected for precarious seats?

Party strategies

Of the 72 electorates with at least one ethnic minority candidate running for the election, there are 13 electorates whose elected candidates are of ethnic minority background. Table 2 shows a general pattern that the higher the number of ethnic minority candidates running for election, the higher the chances of having an ethnic minority candidate elected in the electorate. For instance, of the 41 electorates with one ethnic minority candidate, four electorates (or 10 per cent) observed the ethnic minority candidate being elected in 2022. Of the 23 electorates with two ethnic minority candidates, five electorates (or 22 per cent) observed an ethnic minority candidate being elected. The percentages are 50 per cent for electorates with three ethnic minority candidates and 100 per cent for the electorate with six ethnic minority candidates. The positive association is further substantiated in the path analysis models (online Appendix C Table C1).

There is one exception, Chifley (in New South Wales), where four out of seven nominated candidates were ethnic minority candidates and over 60 per cent of its voter population are of ethnic minority (non-European) ancestries. However, none of the four ethnic minority candidates is from the ALP whose candidate was re-elected for the seat in 2022. The other major parties, including the LNC and The Greens, both have an ethnic minority running for the seat. The four ethnic minority candidates combined received 39 per cent of the first preference vote and the ALP candidate received 53 per cent. It points to the possibility of ethnic voting for parties, as the ALP has historically attracted more ethnic minority votes, rather than ethnic voting for ethnic minority candidates.¹² This further raises our question of the possibility that the major parties put ethnic minority candidates in precarious seats, making them less likely to win the election. We investigate this by examining the correlation between seat status and the presence of ethnic minority background candidates for the ALP and the LNC.

The distribution of candidates’ ethnicity and seat status are presented in Table 5 for the two major parties in Australia. ALP had two ethnic minority candidates running for their safe seats and 16 for precarious seats. LNC had one ethnic minority candidate running for their safe seats and 24 for precarious seats. Compared to the distribution of ethnic majority candidates, 5 per cent of ALP’s safe seats had an ethnic minority candidate representing the party and 2 per cent of LNC’s safe seats. Overall, 15 per cent of ALP’s candidates and 25 per cent of LNC’s candidates who ran for precarious seats were of ethnic minority backgrounds. Fisher’s Exact test statistics show that there is a statistically significant association between the seat status (precarious or safe) and the ethnicity of candidates for those from LNC (p < .001). The association is not statistically significant for candidates from the ALP. Comparing the two major parties, Pearson’s Chi-square test statistics show that there is a statistically significant association between party affiliation and the ethnicity of candidates for those who ran for precarious seats (p < .1).

Table 5. Ethnic minority candidates and seat status by party.

Seat status in the 2019 election	Australian Labor Party (ALP)			Liberal-National Coalition (LNC)
	Ethnic majority candidates	Ethnic minority candidates	% Minority	Ethnic majority candidates	Ethnic minority candidates	% Minority
LNC safe	45	7	13.46	54	1	1.82
ALP safe	38	2	5.00	26	14	35.00
Marginal	35	8	18.60	34	7	17.07
Non-classic	14	1	6.67	13	3	18.75
Total precarious	94	16	14.55	73	24	24.74
Fisher’s Exact test on total precarious/safe seat status and candidate's ethnicity	p = .1566			p = 1.49E-04 (p < .001)
Chi-square test on party affiliation and candidate's ethnicity for precarious seats	X^2(df = 1,N = 207)= 2.8149, p = .0934 (p < .1)

To understand the strategies of the two major parties in putting forward an ethnic minority candidate, we examine the seat status and candidate’s ethnicity by the electorate’s population composition. The boxplot in Figure 2 shows that the ALP has the majority of its safe seats in electorates with higher concentrations of ethnic minority populations compared to the LNC. This aligns with the ALP’s attraction to ethnic votes in history (Pietsch, Graetz, and McAllister 2010). The ALP has its ethnic minority candidates running for election in electorates with relatively higher percentages of ethnic minority voters regardless of the seat status (safe or precarious). However, for both the ALP and the LNC, and particularly the ALP, there are many safe seats in electorates with very high percentages of ethnic minority populations but having ethnic majority candidates running for the election. For the LNC, only one ethnic minority candidate ran for election in a safe seat with less than 30 per cent ethnic minority population. This LNC candidate, Ian Goodenough, was a Member of the House of Representatives in the previous government. For precarious seats, the LNC put ethnic minority candidates in seats with higher concentrations of ethnic minority voters. This, without further information on the candidates’ history of residency in these electorates and other auxiliary including qualitative information of the candidates and the electorate, seems to suggest the party’s strategy of putting forward ethnic minority candidates to win ethnic votes in precarious seats where there is a high concentration of ethnic minority voters. The analysis seems to suggest that the low descriptive representation of ethnic minority populations in the parliament is relevant to the ‘glass cliff’ effect and the marginalisation of ethnic minority background candidates in precarious seats by both major parties in Australia.

Figure 2. Ethnic population composition of electorates (y-axis) by seat status and candidate’s ethnicity (x-axis), and party affiliation.

Discussion and conclusion

In 2022 Australia celebrated a ‘record level’ of ethnic minority candidates elected to Parliament, which we argue was too early to celebrate, given the gatekeeper role of political parties leading up to the election. Instead, we suggest this had more to do with a liberal wave among voters and the increasing influence of ethnic minorities, both as voters and in shifting attitudes towards diversity and inclusion more generally. Nevertheless, this election showed a positive change in the right direction in the sense that before 2022 studies had shown that Australia lagged behind other settler countries in electing a parliament that reflected the diversity of the ethnic composition of their population (Pietsch 2018). In fact, up until 2022, the House of Representatives at the national level was effectively an all-white legislature with one or two exceptions at different elections. Even with these rare exceptions, an elected ethnic minority candidate in the House of Representatives usually struggled with incumbency as they were usually elected in precarious unsafe seats. This prompts the question of whether it is too early for Australia to celebrate.

Political parties might still be concerned about the ethnic status of candidates having an impact on the incumbent’s share of the vote within safe seats, thus political parties tend to relegate ethnic minority candidates to unwinnable or precarious seats where they are less likely to be successful over the longer term. Our study suggests that political parties might be out of touch with Australia’s increasingly diverse electorate and fail to see the added symbolic and substantive value of descriptive representation in terms of legislative and policy outcomes. To explore this in greater detail, our research explored three over-arching questions: 1) are electorates with a higher concentration of ethnic minority populations more likely to have ethnic minority candidates running for the election? 2) compared to ethnic majority candidates, are ethnic minority candidates more likely to be elected in areas with high ethnic minority concentrations, in other words, is there evidence of an ethnic vote? And finally, 3) are ethnic minority candidates more likely to be selected for unwinnable or precarious seats?

To answer the first question, we hypothesised that in electorates with high concentrations of ethnic minorities, there would be more ethnic minority candidates running for the election, though not necessarily representing the major parties. Our analyses supported this hypothesis showing a moderate to strong correlation between a higher concentration of ethnic minority populations and a higher share of ethnic minority candidates. However, this did not necessarily result in the election of ethnic minority candidates. To explore this further, we tested whether ethnic minority candidates are more likely to be elected, particularly in electorates with a high proportion of ethnic minorities, thus demonstrating evidence of ethnic voting within the electorate. Our findings showed that though there is no statistically significant association between ethnicity and election results at the candidate level, there is a positive association between higher ethnic minority concentration and higher likelihood of an ethnic minority candidate being elected at the electorate level. This positive association is channelled through the presence of higher shares of ethnic minority candidates running for the election in places with higher ethnic minority concentrations. In other words, higher ethnic concentration predicts higher ethnic engagement in the election but not necessarily higher chances for ethnic minority candidates to be elected.

Finally, we examined whether ethnic minority candidates are more likely to be selected for unwinnable or precarious seats, thus making them less likely to be elected. Our hypothesis was supported with evidence. However, our study found significant differences between the major parties. For the LNC, there is a significant association between the LNC candidate’s ethnicity and their seat status. While the LNC put forward ethnic minority candidates, they were more likely than ethnic minority candidates in the ALP to be selected for unwinnable or precarious seats. The differences, however, might be explained by the higher shares of ethnic minority voters in the ALP-held safe seats. Importantly, our study found that for both of the major parties, most safe seats with higher proportions of ethnic minority concentrations were still represented by ethnic majority candidates. Where ethnic minority candidates are successful in elections there is likely to be a ‘glass cliff’ effect, which leads to the long-term marginalisation and exclusion of ethnic minorities in the political system.

In conclusion, our findings show that ethnic voting is evident but it is perhaps too early to celebrate higher levels of ethnic representation in Australian politics. The descriptive representation of ethnic minority groups is far from being met. An incumbency factor is also evident for candidates from both ethnic minority and ethnic majority backgrounds. Even though there is evidence of ethnic voting, ethnic minority candidates are routinely being disproportionately relegated to unsafe precarious seats. If ethnic voting is likely to lead to a successful outcome for an ethnic minority candidate in a safe seat, then these observations suggest a hidden barrier that prevents political parties from selecting ethnic minority candidates for safe seats. Political parties continue to act as gatekeepers in safe seats, often holding onto safe seats for ethnic majority candidates. It seems the Australian case provides an ideal case study to observe that it is often parties that discriminate against ethnic minority candidates. We have therefore argued that it is a little premature for Australia to be celebrating the election of ethnic minority candidates if their election is rarely to a safe seat that is likely to lead to a successful and impactful career in politics.

Acknowledgements

The authors would like to thank the two reviewers and participants at the 20th Australian Population Association Conference (the Social Demography session) for their comments and suggestions on the paper.

Data availability statement

The data that support the findings of this study are openly available from Australian Electoral Commission at https://results.aec.gov.au/ and Australian Bureau of Statistics at https://www.abs.gov.au/census.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 This includes Australian residents who reported Australia, New Zealand, United States, Canada, and European ancestries. Note that a person can report up to two ancestries in the 2021 Australian Census where the statistics is sourced from. The 28 per cent is calculated from dividing the number of reported non-European ancestries by the total population size. This is slightly different from the other ancestries variables used in the rest of this paper. See footnote 9.

2 We also noted that there are studies using stated choice experiments (e.g. Dunning and Harrison 2010; Carlson 2015).

3 In Australia, ethnic voting refers to votes/preferences for ethnic candidates rather than voting for ethnic parties which are not a main feature of Australia’s two-party system.

4 ALP refers to the Australian Labor Party and the LNC refers to the Liberal-National coalition which is sometimes informally referred to as the LNP. However, this acronym is often confused with the Liberal National Party (LNP) of Queensland, who is also counted as part of LNC.

5 Note that the 2021 Census was conducted at a time when the Australian international border was still closed to non-citizens and non-permanent residents. Thus, we found very high correlation between the composition of the resident population and the composition of resident citizen population in the 2021 Census compared to previous censuses. The correlation coefficient between the percentage of overseas-born residents and the percentage of overseas-born residents who have Australian citizenship and are aged 18 years old and above is 0.97.

6 The national median personal weekly income is $805, calculated from the 2021 Census (ABS 2022). The calculation is based on the census population aged 15 years old and above. Those who are unemployed or retired are all included.

7 The 2021 Census was conducted on 10 August 2021 and the 2022 Australian federal election was held on 21 May 2022. We approximate a one-year increment in census-recorded age to reflect the voter population (age 18+) in the election. We acknowledge that this method is not perfect as changes in residence, education, labour force status, and income level between the Census data and the election data are not accounted for. Self-reported age can be inaccurate sometimes. The gaps between the number of enrolled voters in AEC statistics and the voter population in the Census may also be contributed to Australian citizens who are 16 and 17 years old and can enrol in the electorate roll list but cannot vote until turning 18 years old. Pearson’s correlation coefficient between the number of enrolled voters in AEC statistics and the voter population in the Census is 0.98.

8 COMPELL relied on public website and public information to compile the list and our best check would only replicate their data compiling process, if not missing anything. Each of the names on the COMPELL list had a link to the candidate’s website. These were active links during the 2022 election, but some were deactivated after the election (for those who did not win the seat). The list also provides a photo of the candidate and thus we are further reassured if the candidate is an ethnic minority from their physical characteristics.

9 Note that up to two ancestries can be reported for each person in the Australian census. Unless otherwise noted, ancestry percentages used in this paper are calculated from dividing the specified ancestry group by the total number of ancestries reported in each electorate. In another word, the denominators of these ancestry percentages are not the population size of an electorate but the total number of ancestries reported in an electorate.

10 Thresholds described in Akoglu (2018) are used.

11 In online Appendix B, we also examine the correlation between candidates’ ethnic background and their election outcomes at the candidate level.

12 It holds for all but two electorates (Wentworth and Durack) that when an ethnic majority candidate was elected, the first preference votes for ethnic minority candidates combined were lower than that for the elected candidate. Thus we did not further question whether ethnic minority voters split their votes but the party strategies.