The Determinants of Homeownership Affordability in Greater Sydney: Evidence from a Submarket Analysis

Mustapha Bangura & Chyi Lin Lee (2021) The determinants of homeownership affordability in Greater Sydney: evidence from a submarket analysis, Housing Studies, DOI: 10.1080/02673037.2021.1879995 (accepted version).

ABSTRACT

Recognising the existence of socio-economic and demographic disparities across metropolitan cities such as Greater Sydney, this study gauges the determinants of homeownership affordability in the different regions of Greater Sydney using local government area (LGA) data over 1991–2016 with a system generalised method of moments (GMM) and a panel error correction model (ECM). The results of the study showed that the determinants of homeownership affordability vary across the regions of Greater Sydney. Although house price and median personal income are the key drivers of homeownership affordability across all regions, the difference in the magnitude of these determinants between regions have also been documented. Specifically, Western Sydney is more sensitive to income and house price change than the other regions. In addition, Western Sydney is also sensitive to other determinants (i.e. housing supply, residential population, median rent, and housing investors), while no comparable evidence is found for the other regions. This clearly highlights the differences across regions and the importance of submarket considerations in the analysis of homeownership affordability. The implications of the study have also been discussed.

Keywords: Greater Sydney; homeownership affordability; determinants of homeownership affordability; regional policy

1 | Page 1.0 INTRODUCTION

Homeownership affordability has attracted extensive research interest in recent years. A number of factors are involved here. First, a deterioration of homeownership affordability has been observed in many metropolitan cities. For instance, using Demographia (2019) median multiple index, the housing affordability of Hong Kong, has further declined to 20.9 in 2018 from 19.4 in 2017. A similar trend is observed in Vancouver (12.6), San Jose (9.4), Los Angeles (9.2) and some Australian cities, particularly Sydney, where house prices have increased at a faster pace than income growth (Healey 2016; Bangura and Lee 2019). Second, the deterioration of homeownership affordability has a significant ripple effect on households from various aspects, ranging from economic to social (Schwartz 2016). Further, the decline in affordability also has direct and indirect repercussions on the broader economy (Lee and Reed 2014).

Housing affordability index is therefore an important benchmark for evaluating households’ ability to meet their housing expenses. Although previous studies such as Muellbauer & Murphy (2008), Yates (2008), Chakraborty et al. (2010), Kim & Cho (2010), and Duffy-Jones (2018) have enhanced our understanding of the issues surrounding homeownership affordability, they generally employed a narrative approach and their findings are mixed. As reported by De Bruyne and Van Hove (2013), housing affordability varies geographically, even between neighbouring local councils. They attributed this variation to the differences in local socio-economic variables. Therefore, an examination of the relationship between affordability index and local factors is critical to developing an effective housing policy in addressing the deterioration of homeownership affordability (Gabriel et al. 2005; Yates 2008).

However, the literature that links affordability index and local factors is limited. This is in spite of the socio-economic variation within metropolitan cities (Baker et al. 2016). In Greater Sydney, for example, Bangura and Lee (2019) found that the deterioration of housing affordability is more obvious in the low-income regions of Greater Sydney. The disparities between low-income and high-income regions could be supported by Stone’s (1990) shelter poverty theory in which low-income households have lower disposable income and are more likely to be “shelter poor” (i.e. with housing but without adequate non-shelter resources) compared with high-income households. This highlights the geography differential of housing affordability. Nevertheless, current government policies on housing in Australia, particularly housing affordability, are not well tailored to adequately address affordability for low-income

2 | Page earners (Beer et al. 2007; Costello 2009). Further, the empirical evidence on the determinants of homeownership affordability is mixed and it depends on the sample being examined. This can be attributed to the ignorance of housing submarkets.

Acknowledging the importance of disaggregated housing analysis, this study is therefore a clear departure from previous studies on the determinants of homeownership affordability. Unlike previous studies, a disaggregated approach was utilised to identify the key drivers of homeownership affordability in Greater Sydney. It allowed us to gauge the sensitivity of the determinants of homeownership affordability in the different regions of Greater Sydney - western, inner-west, southern, eastern and northern regions1. To the best of our knowledge, this study is the first dedicated sub-city housing analysis to examine the drivers of homeownership affordability. As highlighted by Randolph and Tice (2014) and Bangura and Lee (2019, 2020), Sydney is characterised by diverse socio-economic and demographic mix. These features make Sydney an ideal case study for a sub-city modelling of homeownership affordability. An empirical analysis of local demand and supply-sides drivers of affordability could offer more information to policymakers for informed decision-making on housing affordability.

The study contributes to housing literature in the following ways. First, this is the first submarket analysis to identify the main drivers of homeownership affordability in each region of Greater Sydney. We provided an enhanced understanding of the sensitivity of these factors to the residents in different regions. Specifically, we found that Western Sydney, the most socio-economically disadvantaged region, is mostly affected by changes in the drivers of affordability. Moreover, we examined the role of housing investors in determining housing affordability and found a direct and significant relationship between housing investment and affordability in Western Sydney, while no similar evidence is available in other regions. This intuitively explains the role of housing investors in worsening homeownership affordability in Western Sydney. These findings address a significant knowledge gap in housing literature in general and submarket or regional literature in particular. Unlike previous studies (e.g. Lee and Reed 2014), our findings have provided empirical evidence on how exactly these determinants impact homeownership affordability in the different regions of Greater Sydney. Our study has contributed to the debate on the effectiveness of a uniform housing policy, through which enhanced housing policy could be formulated. These findings could be of interest to

1 See Appendix 1 for the LGAs that make up each region

3 | Page policymakers, investors and other housing stakeholders for better analysis and policy formulation.

Second, this study did not only examine whether there are variations in affordability, but also investigate the factors that cause the variation in affordability and how changes in these factors will affect various households in the different regions of Greater Sydney within Stone’s (1990) shelter poverty framework. The shelter poverty theory asserts that housing affordability should consider both households’ housing decision and their non-housing consumption. Importantly, housing affordability problem would be more severe among lower income households, and this could be partly attributed to a problem of widening income inequality (see the hypothesis section for the details). Recognising the income-inequality and other socio-economic discrepancies across regions, the unique research design of this study, a disaggregated approach, allows us to compare the sensitivity of changes in key drivers of homeownership affordability such as income, price, rent, housing supply, population and the role of housing investors on homeownership affordability within a metropolitan city, for the first time. Specifically, our results found that homeownership affordability in Western Sydney region is very sensitive to changes in these determinants compared with other regions especially the high-income eastern and northern regions. The results reflect the assertion of Stone (1990) which highlighted that households in low-income Western Sydney region would have lower residual incomes (remaining income after housing expenses) than their high-income counterparts. Thus, we support the validity of Stone’s Shelter poverty model and illuminate a way to apply the theory in a real practical setting of housing affordability analysis. This information could be used by policymakers to address the skewed effect of deteriorating homeownership affordability in a metropolitan city.

Lastly, to the best of our knowledge, the study is the first to employ the relatively recent Westerlund (2007) error correction panel cointegration test together with a pooled mean group (PMG) estimator to establish the long run relationship between homeownership affordability and its determinants in a panel form. This cointegration model has better size accuracy and superior power than the residual-based tests. As reiterated by Hsaio (2014), the use of panel data increases degrees of freedom and reduces collinearity among explanatory variables. Further, the use of panel data is generally more efficient in conducting unit root and cointegration tests compared with time series data (Mikhed & Zemčík 2009). This shows the efficiency of our econometric estimations. Therefore, our analyses have provided a more solid empirical ground for analysing the key drivers of housing affordability within a metropolitan

4 | Page city. These results have offered a more optimal empirical evidence and tools for evaluating the impact of key determinants on households’ housing expenses in a metropolitan city that is characterised by socio-economic divergences. These results could be used by policy makers and homebuyers for well-versed decision-making on homeownership affordability.

2.0 LITERATURE REVIEW

Three strands of the literature on homeownership affordability were examined in this study – the conceptual framework of affordability, the determinants of ownership affordability, and the definition of submarket.

2.1 Conceptual Framework of Housing Affordability

The first strand of the literature examined some of the widely used definitions of housing affordability. The Urban Development Institute of Australia (UDIA), the peak body representing the interests of the development industry around Australia defined housing affordability at its most basic level to mean the level of income required to attain a reasonably adequate standard of housing. According to UDIA, housing may be unaffordable if it requires a high proportion of household income (above 30 percent is a common guideline) or if the level of housing expenditure impacts on the ability of households to meet other basic needs (UDIA 2014). According to the Revised European Social Charter (RESC), a dwelling is considered affordable if it costs less than 30 percent of the household’s pre-tax income (Maschaykh 2016).

From the US Department of Housing and Urban Development (HUD), total housing costs at or below 30 percent of gross annual income are considered affordable (Hamidi et al. 2016). The 30/40 rule, a preferred measure of housing affordability in Australia, defined housing as being unaffordable if a household in the bottom forty percent of income distribution spends more than 30 percent of their income on housing costs (Beer et al. 2007; Yates 2007; Costello 2009). This implies that low-income households have lower residual income and are more likely to be shelter poor as posited by Stone (1990). This generally shows the lack of a standard definition of affordability. However, our study adopted an index that integrates standard annuity that relates cost of financing of homeownership to income.

2.2 Determinants of Ownership Affordability

The second strand of the literature reviews the determinants of affordability, drawing from both the supply and demand sides of housing market. From the supply side, McLaughlin (2011)

5 | Page found that an increase in housing supply does lead to an improvement of housing affordability. Lee and Reed (2014) found that housing supply does improve housing affordability of Australian first-time buyers, but it did not do so to a statistically significant extent. These results did not only suggest that housing supply does not have a discernible impact on ownership affordability in Australia, but also highlighted the complexity of the nexus between housing supply and ownership affordability. Yates (2008) argued that the sluggish growth of housing supply relative to its demand would have an adverse effect on affordability. These studies generally highlighted a positive relationship between housing supply and housing affordability.

From the demand standpoint, the Productivity Commission Report ((PCR), 2004) noted that an increase in income has a positive impact on affordability. Ying et al. (2013) found that an improvement in permanent and transitory incomes of the lower middle class of Guangzhou in China would also have a positive effect on homeownership affordability. Comparable evidence is also found by Muellbauer and Murphy (2008) in the UK. However, Lee and Reed (2014) did not find evidence to support the notion that this variable is statistically significant in explaining ownership affordability of first home buyers in Australia. Overall, these findings generally suggest that income growth would have a positive effect on affordability.

In addition, Worthington and Higgs (2013) found that one percent increase in population is expected to improve affordability by 8.7 percent in the short run. However, Gitelman and Otto (2012) found population variable to be statistically insignificant, highlighting the complex nature of the relationship between population and affordability particularly in cities and regions (PCR, 2004). It found that if the increase in population growth is not sourced from immigration, it will offset the upward pressure on house prices coming from, amongst other things, economic growth and cheaper finance, which will improve affordability. This view is supported by Haylen (2014), who found that the impact of housing demand emanating from overseas and interstate migration is more immediate as they require accommodation upon arriving, whether owner occupied or rental.

Importantly, Stone (2004) identified rent as another key determinant of affordability especially for potential first homebuyers. He found that rent growth would adversely affect the chances of low-income renters entering into the housing market. Further, PCR (2004) and Yates (2008) also reported that rent has an inverse relationship with entry to the housing market for first homebuyers through its effect on savings to make a deposit for a home.

6 | Page Yates (2007) and Lee and Reed (2014) also demonstrated that higher housing prices have a strong inverse relationship with housing affordability, particularly the level of housing affordability of first-time buyers. The results are intuitively appealing and indicate that higher house prices eventually would make housing less affordable for first-time buyers. Holmes et al. (2008) and Ruming et al. (2011) have found that any increase in house price would have a negative effect on affordability. Chakraborty et al. (2010) also found similar results in six metropolitan areas in the US. Nevertheless, the results of these studies are generally aggregated and highly narrative despite the fact that local demand and supply factors are fundamental drivers of housing affordability as argued by Kyung-Hwan and Cho (2010), Kutty (2005) and De Bruyne and Van Hove (2013).

While there is no generally accepted definition of affordability, some of the inputs in the widely used measures of ownership affordability include house price, income and mortgage lending rate. In addition, several key determinants of housing affordability have been identified by previous studies. These determinants include house price, income, housing supply, resident population and rent. However, most of these studies on affordability are highly narrative and the empirical evidence is somewhat mixed. This could be attributed to the use of aggregated datasets that failed to capture local demand and supply factors. As such, a disaggregated study of ownership affordability using a sub-city approach is required.

2.3 Definition of Submarkets

Due to the aggregated nature of previous studies on the determinants of housing affordability, this third strand of the literature review focuses on submarkets. The importance of housing submarket analysis in metropolitan areas has long been raised by Grigsby (1963). Grigsby et al. (1987) also applied submarket hedonic models and they found significant variation in house prices across submarkets. Bourassa et al. (1997) deployed several statistical methods and established the existence of housing submarkets within the metropolitan cities of Melbourne and Sydney in Australia. Worthington and Higgs (2013) argued that housing analysis on a broader geographical scale such as national, state and city levels often provide suboptimal results in light of local dynamics that are often largely ignored. Kutty (2005) asserted that regional and locational variables play a critical role in determining affordability. Specifically, Randolph and Holloway (2005) found that the dynamics in the industrial and employment markets that are often triggered by global economic activities, often result in varying and diverse effects on the metropolitan cities of Australian. They, therefore, argued for region-

7 | Page specific housing policies. Collectively, the concept of housing submarket, whether spatial or structural, has now been empirically tested to a point that it is no longer contestable (Costello et al. 2019).

Consequently, housing submarkets have been examined in several dimensions. Palmer (1978) examined housing submarkets using real estate board jurisdictions, racial-ethnic composition of neighbourhoods, and the average house price of neighbourhoods. Fik et al. (2013) and Costello et al. (2019) examined the drivers of spatial change in metropolitan housing submarkets. Goodman and Thibodeau (2007) examined housing submarkets in the construction industry. Biswas (2012) examined housing submarket in the context of foreclosures and found that a distant foreclosure within the same submarket will lower market value of a single-family home. Chiang (2016) explored the interaction among residential, office and retail markets in China and found the three submarkets to be distinctive. Bangura and Lee (2019) used convergence tests to examine house price diffusion between low-priced and high-priced submarkets in Greater Sydney. They found a large degree of diffusion flowing from the low- end to the high-end submarket. Recently, in another disaggregated study, Bangura and Lee (2020a) found evidence of housing bubbles in the western Sydney region of Greater Sydney, while no comparable evidence is found in other regions. Wen and Tao (2015) examined six administrative districts in Hangzhou using hedonic price and spatial models. They found that the city is moving towards a three-centre urban spatial structure. Fernandez and Bucaram (2019) explored how the capitalisation of amenities affect house prices of the various submarkets of Auckland. These studies have shown how housing submarkets have been examined from diverse perspectives. However, no study has examined the determinants of homeownership affordability in a submarket version.

Despite these studies, defining housing submarkets has never been straightforward. In other words, there is no generally accepted definition of housing submarkets (Bourassa et al., 1999; Leishman, 2009). This indicates that the delineation of submarkets is still characterised by several theoretical and operational challenges (Bourassa et al., 1999). Watkins (2001), for example, outlines spatial variation in the urban area under study, differences in timeframe, impact of changes in economic fundamentals, and the divergence in methodologies for testing the existence of submarkets as key challenges in developing housing submarket models. DTZ Pieda Consulting (2003) argued that cultural clustering could result in submarket formation. Coombes (2009) asserted that the allocation of new housing should be based on housing market assessments that consider the variation between and within these housing markets using

8 | Page housing supply-demand balance. Hincks and Wong (2010) and Jones et al. (2012) found a fine nexus between housing and labour market and asserted that the interaction between these two markets often affect household housing decisions. Bibby et al. (2020) found that the growth of uneven soft residential densification has varying effect on neighbourhoods that were already characterised by lower-income residents. They reported that this could widen the level of inequality in the distribution of residential space and result in various housing arrangements.

These findings have highlighted a range of factors that could lead to housing submarket formation. This supports the basis for considering localism in housing analysis and delivery (Maclennan and O’Sullivan 2013). Due to the complexities in metropolitan housing markets, housing submarket analysis has more relevance in these markets than broader regional markets (Jones and Leishman 2006). Greater Sydney is ideal for housing submarket analysis as the city is largely socio-economically diverse (Randolph and Tice 2014; Bangura and Lee 2019). This diversity calls for the need to understand how prosperity in the city could affect its different regions (Mee 2002). Bunker et al. (2005) reiterated the polarised spectrum of housing opportunities in Sydney in the sense that higher income households mainly live in waterfront and inner-city areas, while their lower income counterparts live in the middle and outer suburbs. This polarisation has resulted in diverse household living arrangements leading to various housing submarkets across Greater Sydney.

Again, previous studies have shown how housing submarkets could be examined from different perspectives such as spatial, structural, nested spatial/structure, and housing price. These studies offer a solid basis for our definition of housing submarkets. We consider a combination of both pricing and non-pricing approaches in delineating submarkets within Greater Sydney. The pricing factors include the degree of house price substitutability (Gilber and Tyvimma, 2014) and entry-level affordability (Bangura and Lee, 2019), while non-pricing clusters include spatial delimitation (Jones and Leishman, 2006) and socioeconomic characteristics (Chen et al., 2009; Ling and Hui, 2013). Following Bangura and Lee (2020a), this approach in defining submarkets allows us to present the vignette of ‘socioeconomic localisation’ of Greater Sydney. To ensure the expediency of the identified submarkets, we also conducted several statistical (parametric and non-parametric tests) and cointegration tests of key housing variables (e.g., housing affordability levels, house price, income) to examine relationship between LGAs in region and across regions. These tests show strong association between LGAs in a region but no comparable evidence is found across regions. In other words, the housing affordability levels within a region are homogenous and relatively similar. However, strong variations are

9 | Page found across regions2. These discrepancies across these regions have also been noted in the empirical studies of Bangura and Lee (2019), Randolph and Tice (2014), and Bunker et al. (2005). This suggests the appropriateness of defining these 5 regions. Importantly, this also indicates that housing affordability issues in Greater Sydney are complex and non- homogenous across the city. It reveals that housing affordability issues could be more acute in some regions than in other regions, reflecting the view that these should be analysed in a submarket form. This might also explain the mixed results of previous aggregated studies on the determinants of housing affordability.

3.0 THEORETICAL FRAMEWORK AND HYPOTHESIS DEVELOPMENT

The previous sections highlighted the importance of submarket housing analysis. This section examines the theoretical framework that underpins such micro-analysis, which leads to the formulation of the hypothesis of this study. Specifically, our study seeks to assess how the determinants of homeownership affordability vary across different regions to provide an implicit test of the shelter poverty model that is postulated by Stone (1990).

Stone (1990) shelter poverty model is, in part, a problem of worsening income inequality among those at the bottom of the distribution. As discussed by Stone (2004), shelter poverty refers to a household who is paying "more than they can afford" for housing and leaving them with insufficient money to meet their non-shelter needs adequately. As such, one of the central premises of shelter poverty model is that it is a sliding scale, with the maximum affordable percentage varying with income. As such, the housing affordability problem could be more severe among lower income households and large households (Stone, 1990; page 10). Higher income families can afford to pay a higher percentage of their income for housing compared with lower income families. Further, smaller households can afford to pay a higher percentage than larger households with the same income level.

Households who spend more than they can afford on this sliding scale are shelter poor - the squeeze between their limited incomes and excessive housing costs leave them with not enough money to address their non-housing needs at a minimum adequate level (Stone, 2004; pp. 109). Specifically, the model demonstrates that low-income households with three or more persons can spend less than 25% of their combined income on housing-related expenses but are

2 The results are not reported for brevity.

10 | Page nonetheless “shelter poor” insofar as their residual income (remaining household income after housing expenses) may not be adequate to meet their non-shelter necessities. That is, they have shelter but remain poor or in poverty – they are “shelter poor”. On the other hand, the model shows that high-income households can spend more than 25% of their income on housing related expenses yet are still able to consume non-shelter necessities adequately and are therefore not shelter poor.

As exhibited in Appendix 2, some submarkets (i.e. Eastern and Northern regions) have higher income compared with Western Sydney and Southern Sydney. Further, larger household size is also evident in Western Sydney. This raises the question of whether housing affordability would vary across different regions. By using the cost-to-income ratio approach to measure affordability, we examine the determinants of housing affordability across different regions (e.g. high-income Eastern & Northern regions versus low-income Western Sydney region). Specifically, if high-income families in the Eastern and Northern regions have higher residual income, they are less likely to be shelter poor compared with low-income households in the Western region. This is largely due to the differences in the income levels across these regions of Greater Sydney. Therefore, the determinants of housing affordability are expected to be heterogeneous and vary across different regions. In other words, the entry-level housing affordability levels in different regions should be different, resulting in different residual incomes due to varying household income levels3. As such, it is reasonable to expect that housing affordability levels in low-income regions are more sensitive to key determinants than high-income regions with respect to households in low-income regions having lower residual income to cover their non-shelter expenditures. This study is unique as we consider the income- inequality across the different regions in understanding housing affordability. This provides some empirical evidence on Stone’s (1990) shelter poverty model. Following the shelter poverty model, we formulate the following hypothesis:

Hypothesis: The determinants of homeownership affordability are more significant and sensitive in Western Sydney than the other regions.

This hypothesis suggests that households from relatively low income would be more affected by changes in these determinants than those of higher income status. This can be attributed to the fact that low-income households will have lower residual income than high-income

3 Yates (2008) and Kutty (2005) also reported that household consumption of non-housing goods and services potentially depends on the proportion of income spent on housing-related expenses. As such, low-income families often have limited household non-shelter expenditure due to their low residual income.

11 | Page families. This disparity in income among households in Greater Sydney is also documented in Randolph and Tice (2014), with the Eastern and Northern regions on the higher end of the income scale. Therefore, this model is well situated for examining local drivers and the magnitude of their effect on the affordability of the different regions of Greater Sydney4.

4.0 DATA AND METHODOLOGY

4.1 Data

The study used annual data at the local government area (LGA) level over 1991–2016 to examine the determinants of homeownership affordability for the five regions of Greater Sydney: western, inner-west, southern, eastern and northern. Data on median house price for all housing types (all dwellings, strata and non-strata) and median rent were collected from Housing NSW (2016)5. The median individual income per LGA was obtained from the Australian Bureau of Statistics (ABS 2020). Housing stock was derived from the various ABS census reports of the annual number of private residential properties for each LGA within the study period and we interpolate between two census periods using building approvals.6 ABS defines building approvals as the number of residential (dwelling) building permits issued by local government authorities and other principal certifying authorities in a given period. The number of second home buyers is used to proxy the number of investors and it is obtained from the ABS. Data on estimated resident population was also obtained from ABS regional publications. This is defined by the ABS as the official estimate of all people, regardless of nationality, citizenship or legal status in the LGA, which links people to a place of usual residence for six months or more in each reference year. Mortgage lending rate was collected from the Reserve Bank of Australia.

Due to the importance of local effects, this study disaggregated Greater Sydney into five major regions to examine the key drivers of homeownership affordability in each region. The LGAs that make up these regions are reported in Appendix 1. This disaggregation stems from the notion that regional and locational variables can result in differences in housing market conditions (Kutty 2005). Therefore, these regional variations may have diverse implications

4 If the results show that the determinants of homeownership affordability are homogenous and do not vary across different regions, it refutes the Shelter Poverty Theory. It also highlights the complexity in explaining housing affordability using the Shelter Poverty Theory. 5 We use the median house price because, as reported by Housing NSW (2016), this measure of central tendency is not significantly affected by unusually high or low values. 6 For intercensal periods, the aggregate of building approvals was added progressively to the reported census housing stock. A similar interpolation technique was used by Liu and Otto (2017).

12 | Page for homeownership affordability in different geographical areas (Dufty-Jones 2018). This is particularly the case with Greater Sydney, where homeownership affordability varies across the regions of the city (Bangura and Lee 2019). As such, local demand and supply factors are crucial in determining homeownership affordability (Kyung-Hwan and Cho 2010). Hence, panel data of the five regions of Greater Sydney was utilised.

4.2 Methodology

The study employed a two-staged methodology. First, the study estimated the homeownership affordability index for the different LGAs that make up the different regions of Greater Sydney.

The study utilised a cost-to-income affordability index (HAICtI) to compute entry-level homeownership affordability in the different regions. The second stage of the study involves a panel analysis. Specifically, the study examines the determinants of homeownership affordability in different regions of Greater Sydney using the system generalised method of moments (SGMM). The SGMM allows us to assess Stone’s (1990) shelter poverty theory in which low-income households are more likely to be shelter poor due to lower residual income. Hence it is expected that the magnitudes of determinants in low-income regions are stronger compared with high-income regions. These panel estimators are preceded with a panel unit root test to establish the stationarity of the data. To shed more light on it, we extend the analysis by examining the long run relationship between entry-level affordability and its determinants using the Westerlund (2007) Panel Cointegration and Panel Error Correction Model. The tests allow us to further confirm that the affordability in low-income regions (e.g. Western Sydney) is more susceptible to changes in key housing market variables even in the long run.

Cost-to-income Affordability Index

A cost-to-income affordability index (HAICtI) index incorporates the market value of the property, down payment, and mortgage lending rate, and is expressed as a percentage of median income in an LGA. This cost-to-income formula is given as:

HAICtI = × (MV × LTV × 𝑟𝑟 ) × 12 (1) 100 12 1 I 𝑟𝑟 1−�� � � 1+12 12𝑛𝑛 From (1), HAICtI denotes homeownership affordability index, I denotes median personal income, MV represents the market value of the property, LTV denotes loan-to-value ratio, r is the mortgage lending rate, and n is the term of the loan. The formula is the annualised effective

13 | Page cost of servicing the loan (in parenthesis) and it is computed by means of a standard 25-year annuity formula (i.e. n = 25) with monthly compounding, where the loan is fully amortised over the term and then compared to the annualised median income (Bentzien et al. 2012). In Australia, the homeownership affordability threshold is 30% (Yates 2008), LTV is often 80% implying 20% down payment (Housing Industry Association), the mortgage lending rate is variable, and the loan term is 25 years (BIS Shrapnel Home Loan Affordability). A higher index in a region suggests housing is less affordable. The graphs of the cost-to-income affordability index of the LGAs of the different regions are reported in Appendix 3(a-e). The visual presentation suggests that homeownership housing affordability varies across regions, reflecting the existence of housing submarkets in terms of housing affordability. The preliminary results suggest the importance of a submarket analysis for housing affordability.

System Generalised Method of Moments (SGMM) Model Specification

Once the stationarity of the variables has been established, we assess the determinants of homeownership affordability with a system generalised method of moments (SGMM) model. Given that the generalised method of moments (GMM) addresses potential model misspecifications and gives more consistent estimates in the presence of endogenous regressors (Hansen and Tarp, 2001) than the simpler instrumental variables (IV), we utilise this SGMM in our analysis. The null hypothesis of no autocorrelation was tested using the Arellano–Bond and natural logarithms were introduced in the model to address any scaling effects in the data.

The general model of the SGMM is given as:

lnHAIit = β0 + β1lnHAit-1 + β2lnRPit + β3lnHSit + β4dlnHPit + β5lnMPIit + β6lnMRit + β7 lnINVt

+ εit (2) where εit = αi + µit and E(αi) = E(µit) = E(αiµit) = 0

From Equation (2), the disturbance term has two orthogonal components: the fixed or LGA

effect (αi) which is a time-invariant error term that could represent variables such as the

geophysical characteristics and the social ledger of the LGA; and the idiosyncratic shocks (µit) which captures all other factors that influence homeownership affordability other than the

specified regressors. HAIit denotes homeownership affordability index for LGAi at time t;

HAIit-1 denotes lagged homeownership affordability for LGAi at time t; RPit denotes resident

population for LGAi at time t; HSit denotes housing supply for LGAi at time t; HPit denotes house price for LGAi at time t; MPIit denotes median personal income for LGAi at time t; MRit

14 | Page denotes median rent for LGAi at time t; INVt denotes the number of housing investors at time t.

The coefficients of the lagged dependent variables cannot be determined a priori. β2 is expected to be positive since rising population will escalate housing demand (Yates 2008, Haylen 2014);

β3 is expected to be negative as an increase in housing supply is expected to improve affordability (McLaughlin 2011; Al-Masum and Lee, 2019); β4 is expected to be positive as an

increase in house price is expected to worsen affordability (Yates 2008, Berry 2003); β5 is hypothesised to be negative, as an increase in income is expected to improve affordability

(Gitelman and Otto 2012; Healey et al. 2008); β6 is hypothesised to be positive as a higher median rent leads to a deterioration of homeownership affordability of potential first home buyers (Yates 2008); and β7 is hypothesised to be positive, as the increase in investor activities will increase housing demand (Bangura and Lee 2020b; Pawson and Martin 2020).

The documented results from the system GMM allow us to examine our hypothesis. As hypothesised, it is expected that the magnitudes of determinants in Western Sydney would be stronger compared with other regions particularly Eastern and Northern Sydney. Further, Western Sydney is affected by more determinants of homeownership affordability than the other regions. This reflects the shelter poverty approach as households from Western Sydney would be more sensitive to changes in these housing variables, increasing housing expenses and reducing non-housing expenses.

Westerlund (2007) Panel Cointegration and Panel ECM Model Specification

We extend the second stage of the methodology by testing the presence of a long run relationship between homeownership affordability and its regressors. Following the IPS panel unit root test, we employed the Westerlund (2007) panel cointegration test to formally check the existence of cointegration in Equation (2). This is an error correction-based (ECM) panel cointegration test. Westerlund (2007) proposed four panel cointegration tests that are intended to test if the error correction term in a conditional error correction model is equal to zero.

Once cointegration is established, we estimate the panel ECM. We re-parameterised Equation (2) into an ECM to examine a stable long run relationship among the variables in Equation (2) that change over time. The ECM contains both the long run equilibrium relationship and a short run equation that describes how the long run solution is derived through an error correction (Hoesli et al. 2007). Therefore, the long run relationship becomes:

15 | Page HAI*it = β0 + K (3) 𝑛𝑛 ∑𝑖𝑖=1 β𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 where HAIit is homeownership affordability and K is a vector of the determinants of

homeownership affordability such as RPit, HSit, HPit, MPIit, MRit and INVit.

It is hypothesised that Western Sydney would have higher magnitudes of housing variables in the long run compared with other regions particularly eastern and northern Sydney.

5.0 RESULTS AND DISCUSSION

5.1 Key Determinants of Homeownership Affordability in Different Regions

System Generalised Method of Moments

This section examines the effect of the determinants of homeownership affordability in each region of Sydney. Using robust standard errors, the SGMM results of the five regions of Greater Sydney are presented in Table 17.

(Table 1)

As hypothesised, house price is directly and adversely related to homeownership affordability in all regions of Greater Sydney, suggesting that an increase in house price will cause a deterioration of homeownership affordability. The results show that house price is a key determinant of homeownership affordability across all regions of Greater Sydney at the 1% significance level. The negative impact of house price on homeownership affordability is also reported by Berry (2003), Yates (2008) and Healey (2016).

The coefficient of median personal income is negative and statistically significant at the 1% level, reflecting that income is another statistically significant determinant of homeownership affordability across all regions of Greater Sydney (a higher median income indicates an improvement of housing affordability). These results are consistent with the findings of the Productivity Commission Report (2004) and Yates (2008), who reported that income is a key driver of homeownership affordability respectively. The results suggest that an increase in income will improve affordability, thereby enhancing the residual income.

Housing supply has the expected sign in all regions, supporting the literature that an increase in housing supply will improve affordability. Despite the fact that housing supply slowly

7 The panel unit root results show that the series are first-differenced stationary or I(1)

16 | Page responds to its demand in all regions, the results show that housing supply is a statistically significant driver of homeownership affordability in the western region at 5% significance level. However, the coefficient of housing supply is negative and statistically insignificant in the other regions. This suggests that housing supply plays a marginal role in explaining ownership affordability in the inner-west, southern, eastern, and northern regions. This also confirms the differential effect of housing supply across regions. One caveat of this result is that increasing housing supply could improve affordability in the most socio-economically disadvantaged region of Greater Sydney. However, no comparable effect is observed in the other regions of the city. This supports Stone’s (1990) shelter poverty assertion in that higher- income households are less susceptible to changes in this determinant of homeownership affordability.

Resident population has a positive sign in all regions, implying that affordability is worsened with rising population growth. However, resident population is only statistically significant in the western region at the 10% significance level. This highlights that population growth does have a discernible effect on ownership affordability in western Sydney region, while no comparable evidence is available in the other regions. Again, this reveals the differential geography of the determinants of housing affordability across regions and supports the notion that households in Western Sydney are more sensitive to changes in the determinants of ownership affordability. As discussed by PCR (2004), the increase in immigration is more likely to have an immediate impact on housing demand than natural increase of population. Haylen (2014) also found that the impact of housing demand arising from overseas and interstate migration is instantaneous as they require accommodation upon arriving, whether owner occupied or rental. More specifically, Burnley (2005) asserted that western Sydney has become an attractive settlement for lower-income immigrants. Due to the slow response of housing supply to its demand (Yates 2008), this rise in demand will eventually cause a deterioration of ownership affordability.

In 2019, for example, 11,190 new arrivals were resettled in NSW, accounting for 43% of Australia’s total humanitarian intake (NSW State Government 2020). Moreover, Greater Sydney is home to more than 65% of people residing in the state of NSW (ABS 2020). More importantly, the population of Greater Sydney is projected to grow by 1.6 million in the next 20 years and about 900,000 of this growth is projected to occur in Western Sydney (Griffith 2015). This certainly explains the strong coefficient of 1.77% of population on affordability in the long run as revealed by the ECM results reported in Table 3. This concentration of NSW’s

17 | Page population in Greater Sydney has strong housing implications which explained the positive relationship with homeownership affordability and why population is statistically significant in Western Sydney.

Median rent is included in Equation (2) to assess its effect on the affordability of potential first home buyers. As reported by Stone (2004), renters are more likely to be shelter poor than homeowners. An increase in rent has an inverse effect on entry to housing market for first home buyers through its effect on savings to make a deposit for a home (Productivity Commission Report 2004; Yates 2008). Further, higher rental income will stimulate housing investment, which will create competition among the limited housing supply thus driving house prices. Those who own an investment property benefit from price increases while the savings of low- income renters will be affected adversely (Healey 2016). As expected, median rent has a positive sign across all regions, meaning as rent increases, it minimises the likelihood of potential first home buyers entering the housing market. The net losers of this result are renters or prospective homebuyers. Median rent is statistically significant at 1% in western and inner- west regions, 5% in the southern region but insignificant in the eastern and northern regions. The divergence of the results across regions further supports the notion that low-income households are more likely to be in shelter poverty.

The number of investors was included in Equation (2). The results show a positive relationship between housing investment and homeownership affordability in each region, suggesting that an increase in housing investment would result in a deterioration of affordability in the region. However, the number of investors is only statistically significant in Western Sydney at the 1% but insignificant in the other regions. Again, as housing supply is not catching up with its demand (Yates 2008), the influx of housing investment activities in the western region will heighten competition between homebuyers and investors, causing affordability to decrease. The increase in housing investment in the western region has been recorded in the literature. Martin (2018) found that the desire of landlords to expand their housing portfolio is increasing investment in relatively low value areas. Pawson and Martin (2020) also reported the disproportionate increase in property investment in socio-economically disadvantaged areas such as western Sydney. Bangura and Lee (2020b) found that the rise in property investment in Western Sydney is a causative agent of housing bubbles in the region. All these findings espouse the positive and significant impact of housing investment on affordability in the western region. The lack of comparable evidence in other regions further confirms the differential geography of housing affordability.

18 | Page A dummy variable is introduced in the model to assess the effect of key policy changes between 1999 and 2000 (i.e. introduction of the FHOG and GST; Lee and Reed 2014). The dummy variable evaluates the effect of the introduction of the GST in July 1999 and the FHOG in July 2000 on homeownership affordability. The results show that homeownership affordability deteriorates after 2000 in all regions, indicating that the effect of FHOG in improving affordability could not effectively cushion the impact of the GST on affordability. The net effect of the FHOG and GST on affordability is adverse. As shown by the house price elasticity, it means western, inner-west and southern regions are more sensitive to these policy changes than the eastern and northern regions. This again highlights the spatial polarisation of Greater Sydney.

A closer look at the magnitude of the regression coefficients reveals an asymmetry effect. House price, for example, shows significant variation across the regions of Greater Sydney. House price has a stronger effect on affordability in the western region than the other regions. There is also variation in the magnitude of the coefficients across the inner-west, southern, eastern and northern regions. This means the impact of house price on affordability in each region is distinct. A percentage change in house price has the greatest impact on homeownership affordability in the western region with 1.33%, followed by the inner-west region with 0.72% and the southern region with 0.66%. The eastern region has a coefficient of 0.28%, while it is 0.35% in the northern region. Therefore, a significant difference in the effect of a change in house price on homeownership affordability is observed across all regions of Greater Sydney.

As the Western Sydney region is characterised by several measures of deprivation such as low- incomes, high unemployment rate and heavy reliance on public housing (Baum 2004; Bangura and Lee 2019), a coefficient of 1.33% highlighted the significant effect of rising house prices on entry-level affordability for residents in the region. This indicates that residents in Western Sydney would spend more on housing expenses than those in other regions when house prices increase. Haylen (2014) found that housing prices have greater adverse effect on homeownership affordability in low socio-economic areas than high-income regions. This also reinforces the findings of Yates (2008), who reported that low-income groups are mostly hit in the face of an increase in house price than high-income regions due to differences in residual income (income after housing expenses). These results clearly highlight the existing differences in the effect of a change in house price on homeownership affordability across the

19 | Page regions of Greater Sydney. This is a further demonstration of the shelter poverty model as it highlights how changes in house price affects the affordability of low-income households more.

The sensitivity of affordability to changes in median personal income also varies across the regions of Greater Sydney. Income has the strongest impact on homeownership affordability in the western region with an elasticity of 1.28%, followed by the southern region with 0.82% and the inner-west region with 0.66%. The northern region has an elasticity of 0.44%, while the eastern region has 0.32%8. This suggests that residents in the western region are more sensitive to income change than those in other regions. The low rate of income growth of residents in Western Sydney means they will continue to experience severe affordability in the face of existing uniform housing policies that pay little attention to the socio-economic imbalances across the regions of the city. Berry (2003) asserted that low-income households do not have sufficient income for securing and maintaining a mortgage for a standard, average price three-bedroom house in almost any area of Sydney. The results again highlight the notion of the shelter poverty model, as residents from low-income regions are generally more susceptible to affordability pressures when their income changes.

In addition, an elasticity far below one shows that the rate of increase in housing stock is slow across all regions of Greater Sydney. This supports the findings of Yates (2008), asserting that the response of housing supply is not proportionate to housing demand. This can be attributed to the nature and unique characteristic of housing supply. Notably, housing supply is not instantaneous due to factors such as decision lags, longer construction periods, and meeting consumers’ tastes and preferences. This slow response of housing supply can also be linked to the cost of land development processes and policies, levies, construction cost and property- related taxes (Yates and Milligan 2007). One caveat from this result is that even though increasing housing supply is expected to improve affordability, the inelasticity of housing supply creates undue bottlenecks towards the improvement of affordability.

Resident population also recorded very low elasticities in all regions with 0.34% in western region, 0.12% in the inner-west region, 0.19% in the southern region, 0.06% in the eastern region, and 0.27% in the northern region. As discussed earlier, the positive effect of population on affordability can be explained via its effect on housing demand. This reveals that a percentage increase in population growth in Western Sydney would worsen affordability in the

8 Elasticity measures the level of change in affordability when there is a change in any of the regressors (e.g. income), holding everything else constant. For example, a 1% increase in income would improve affordability by 1.21% in the western region.

20 | Page region by 0.34% ceteris paribus. Median rent is also inelastic across all regions with the highest in the inner-west region with 0.37%. This is trailed by the western and southern regions with 0.20% each. This further highlights the uneven effect of median rent across households in Greater Sydney. It indicates that potential first home buyers, particularly those in western, inner-west and southern regions, are more likely to be affected by higher rents than those in the eastern and northern regions. Finally, the number of investors is also inelastic across all regions. However, western Sydney has the highest coefficient with 0.46%, suggesting that an increase in housing investment activities in Western Sydney would contribute to the deterioration of housing affordability in the region. This also explains the skewed effect of housing investment activities across the regions of Greater Sydney with western being the most affected region9.

Overall, our results show that households in the western, inner-west and southern regions, particularly Western Sydney are more sensitive to changes in key determinants than their counterparts in the eastern and northern regions. Specifically, housing ownership affordability in the eastern and northern regions is only influenced by house price and income, whilst housing ownership affordability in the other regions, especially the western region, is driven by these factors and other factors such as housing supply, rent, population and housing investment activities. These findings are generally in conformity with the hypothesis of the study and reflect the central idea of the shelter poverty model, as the western region is more susceptible to changes in key housing market variables than all the other regions. The differences in income level, unemployment rate and general socio-economic characteristics across the regions of Sydney largely accounted for the disparity in the effect of these determinants. These findings have shown that changes in these determinants could translate to rising household expenses which could majorly affect non-housing consumption of households from low-income Western Sydney, as posited by the shelter poverty model.

Westerlund (2007) Cointegration and Panel ECM

The previous sections highlighted the directions and magnitudes of the effect of the key drivers of affordability in each region of Greater Sydney. To further authenticate our SGMM results, we extend the analysis by examining the long run relationship in Equation (2) using Westerlund

9 We re-run the tests with a disaggregated dataset (SA3). Again, this robustness check shows that the determinants of homeownership affordability vary across the regions of Greater Sydney. This reflects the importance of submarket considerations in homeownership affordability analysis. The results are not reported for brevity. Thanks to the referee for highlighting this point.

21 | Page (2007) and employing another estimator, panel ECM. The results of Westerlund (2007) are reported in Table 2.

(Table 2)

The panel cointegration test results show the existence of an error correction for both the individual group and for the panel as a whole for all regions at the 1% significance level. The cointegration test confirms the existence of a long run relationship between affordability and the explanatory variables modelled in Equation (2). These results further validate the baseline and robustness check results. With the existence of cointegration, we estimated the ECM to examine the long run drivers of affordability in each region using mean group (MG) and pooled mean group (PMG) estimators. The ECM results are reported in Table 3.

(Table 3)

The Hausman test shows that the PMG is a more consistent estimator for all the regions. The results of the ECM reveal that house price and income are statistically significant at the relevant level in all regions. They remain the key drivers of affordability across the regions of Greater Sydney. The results are in line with our baseline results in which house price and income are more sensitive in the western, followed by inner-west and southern regions and then the eastern and northern regions. Further, the ECM also reveals that Western Sydney is the only region whose affordability is affected by all the regressors. This indicates that affordability in Western Sydney, which is characterised as a low-income region, is more prone to changes in key housing market variables even in the long run. The ECM results largely support our hypothesis and the shelter poverty model as there is enormous evidence of variation in the effect of key housing market variables on households across the regions of Greater Sydney.

To sum up, our long run results are in line with our baseline findings in which a geographical differential is evident among different regions of Greater Sydney. Importantly, affordability in the western region is more predisposed to changes in housing variables. The results have provided some empirical evidence of shelter poverty model as households, especially those in Western Sydney, would generally have lower disposable and residual incomes and would be more likely to be affected by changes in housing variables compared with those in the eastern and northern regions.

22 | Page 5.2 Robustness Checks

Robustness checks were conducted to assess the strength of our baseline findings. This investigates whether our regression coefficients will be consistent if we modify our model. One could argue that, based on similarities in income, house price and socio-economic characteristics, the inner-west and southern regions could be considered as a medium-income region, while the eastern and northern regions a high-income region. The western region remains a low-income region. We follow the SGMM discussed in Section 4.2 to estimate the regressions of low-income, medium-income and high-income regions.

Our regression results suggest households in the low- and medium-income regions, particularly the low-income region, are more sensitive to changes in key determinants than their counterparts in the high-income region. Housing affordability level in the high-income region is only caused by house price and median income. Furthermore, the elasticities of house price and median income in both low-income and medium-income regions are greater than the high- income region. These results confirm that our baseline results are robust to the five regions of Greater Sydney. This offers further evidence to support that homeownership affordability in the western, inner-west and southern regions of Greater Sydney is affected by both demand- side and supply-side factors as residents in these regions have lower residual income.

As different housing types usually have different market determinants in metropolitan areas (Chakraborty et al. 2010; Lee 2017), we further validate our baseline results using different housing types. We computed homeownership affordability index for strata and non-strata dwellings and regressed each housing type against the determinants using Equation (2). Strata title includes townhouses, terraces/villas and flats/units, whereas non-strata title properties refer to detached houses (Housing NSW 2016). The results of strata and non-strata analysis are generally consistent with our baseline findings.10 This confirms that our baseline results are robust to different housing types.

6.0 CONCLUSION AND IMPLICATIONS

This paper considered the existence of disparities in the income and socio-economic characteristics of Greater Sydney to gauge the determinants of homeownership affordability in the different regions of Greater Sydney – western, inner-west, southern, eastern and northern

10 The system GMM results of low-income, medium-income, and high-income regions as well as strata and non-strata housing types are not reported for brevity but they are available from the authors.

23 | Page over 1991 to 2016. Specifically, a disaggregated approach was utilised for the first time to identify the key drivers of homeownership affordability in Greater Sydney.

Several findings have been identified. Firstly, house price and median personal income are the major causes of changes in homeownership affordability across all regions, but the glaring difference in the magnitude of these determinants between regions have also been documented. Specifically, western region, followed by inner-west and southern regions, are more sensitive to changes in income and house price than eastern and northern regions. This clearly highlights the differential geography of homeownership affordability. Secondly, affordability in western region is particularly more sensitive to changes in other determinants (such as housing supply, residential population, median rent and housing investors), while no comparable evidence is found in other regions. This supports our hypothesis in the sense that changes in the affordability of Western Sydney are caused by more factors and the region is also more sensitive to changes in these factors than the other regions. These findings clearly reveal that changes in these factors will result in rising household expenses which will impact on household non-housing consumption especially those from low-income Western Sydney as posited by Stone’s (2004) shelter poverty model. This also gives credence to submarkets in the analysis of homeownership affordability.

These findings have important policy implications. Policy makers should take into consideration the varying sensitivity of house price across Greater Sydney in the design of policies purported to improve homeownership affordability. Particularly, policy makers should consider a more targeted housing policy for low-income households as they are more likely to be in shelter poverty due to rising housing expenses as these determinants change. Besides, the differential geography of homeownership affordability has also provided sufficient evidence to policy makers for the formulation of regionally balanced housing policies to improve homeownership affordability within metropolitan cities. We therefore argue for a more targeted housing policies and actions at all levels of government to promote homeownership especially in low-income regions as the findings show that households in different regions are affected in varying ways.

24 | Page REFERENCES

Adair, A, S, Berry, J, N, McGreal, W, S, J, Murtagh, B & Paris, C 2000, The local housing system in Craigavon, N. Ireland: ethno-religious residential segregation, socio-tenurial polarisation and sub-markets, Urban Studies, 37(7), pp.1079–1092

Al-Masum, M.A. and Lee, C.L. 2019, Modelling housing prices and market fundamentals: evidence from the Sydney housing market, International Journal of Housing Markets and Analysis, 12(4), pp.746-762.

Australian Bureau of Statistics 2013, Estimates of personal income for small areas, cat.no.6524.0.55.002

Australian Bureau of Statistics 2013, Perspectives on regional Australia: housing arrangements - home ownership in local government areas, cat.no.1380.0.55.010

Australian Bureau of Statistics 2016a, Regional population growth, Australia, cat.no.3218.0

Australian Bureau of Statistics 2016b, Australian Statistical Geography Standard (ASGS): Volume 1 - Main Structure and Greater Capital City Statistical Areas, cat.no.1270.0.0.55.001 Australian Bureau of Statistics 2017, Residential Property Price Indexes: Eight Capital Cities, cat.no.6416.0 Canberra

Australian Bureau of Statistics 2018, Housing Finance, Australia, cat.no.5609.0

Australian Bureau of Statistics 2020, Data by region, cat.no.1410.0

Baker, E., Bentley, R., Lester, L. & Beer, A. 2016, ‘Housing affordability and residential mobility as drivers of locational inequality’, Applied Geography, 72, pp.69-70

Bangura, M & Lee 2019, The differential geography of housing affordability in Sydney: a disaggregated approach, Australian Geographer, 50(3).pp.295-313

Bangura, M & Lee, C, L 2020a, House price diffusion of housing submarkets in Greater Sydney, Housing Studies, 35(6), pp. 1110-1141

Bangura, M & Lee, C, L 2020b, Housing price bubbles in Greater Sydney: evidence from a submarket analysis, Housing Studies, DOI: 10.1080/02673037.2020.1803802

Baum, S 2004, Measuring socio-economic outcomes in Sydney: An analysis of census data using a general deprivation index, Australasian Journal of Regional Studies, 10(1),pp.115–120

Beer, A, Kearins, B & Pieters H 2007, Ownership affordability and Planning in Australia: The Challenge of Policy Under Neo-liberalism, Housing Studies, 22(1),pp. 14–21

Bentzien, V, Rottke, N & Zietz, J 2012, Affordability and Germany’s low homeownership rate, International Journal of Housing Markets and Analysis, 5(3),pp. 289–312

25 | Page Berry, M 2003, Why it is important to boost the supply of affordable housing in Australia – and how can we do it?, Urban Policy and Research, 21, pp. 413–435

Bibby, P, Henneberry, J & Halleux, J-M 2020, Incremental residential densification and urban spatial justice: The case of England between 2001 and 2011. Urban Studies. Birrell, B & Healy, E 2003, Migration and the ownership affordability crisis, People and Place, 11(3), pp.55

Biswas, A 2012, Housing submarkets and the impacts of foreclosures on property prices, Journal of Housing Economics, 21, pp.235-245

Bourassa, S, C, Hamelink, F & MacGregor, B, D 1999, Defining housing submarkets, Journal of Housing Economics, 8, pp.160-183

Bramley, G, Leishman, C & Watkins, D 2008, ‘Understanding neighbourhood housing markets: regional context, disequilibrium, submarkets and supply’, Housing Studies, 23(2), pp. 179–212.

Brown, R, Brown, R, O’Connor, I, Schwann, G & Callum, S 2011, ‘The other side of ownership affordability: The user cost of housing in Australia’, The Economic Record, 87(279),pp.558

Bunker, R, Holloway, D & Randolph, B 2005, Building the connection between housing needs and metropolitan planning in Sydney, Australia, Housing Studies, 20(5), pp.771–794

Burnley, I. H. 2005. “Immigration and Housing in an Emerging Global City, Sydney, Australia.” Urban Policy and Research, 23(3), pp.329–345.

Chakraborty, A, Knaap, G, Nguyen, D & Shin, J, H 2010, The effects of high-density zoning on multifamily housing construction in the suburbs of six US metropolitan areas, Urban Studies, 47, pp.437-451

Chiang, S 2016, Interaction among real estate properties in China using submarket panels, Habitat International, 53, pp.243-253

Coombes, M 2009, English rural housing market policy: some inconvenient truths? Planning Practice & Research, 24, pp.211-231

Costello, L. 2009, ‘Urban–rural migration: housing availability and affordability, Australian Geographer, 40, pp.224

Costello, G, Leishman, C, Rowley, S & Watkins, C 2019, Drivers of spatial change in urban housing submarkets, The Geographical Journal, 185, pp.432–446.

De Bruyne & Van Hove, J 2013, Explaining the spatial variation in housing prices: an economic geography approach, Applied Economics, 45, pp.1673-1689

26 | Page Dowling, R & Mee, K 2000, ‘Tales of the city: Western Sydney at the end of the millennium’, in J Connell (ed.), Sydney: the emergence of a world city, Oxford University Press, Melbourne

DTZ Pieda Consulting (2003) Merchant City: Towards a Sustainable Future (Edinburgh: DTZ Pieda Consulting).

Dufty-Jones, R 2018, ‘A historical geography of housing crisis in Australia, Australian Geographer, 49, pp.5-8

Fernandez, M, A & Bucaram 2019, The changing face of environmental amenities: Heterogeneity across housing submarkets and time, Land Use Policy, 83, pp.449-460

Fik, T, J, Ling, D, C & Mulligan, G, F 2003, Modelling spatial variation in housing prices: A variable interaction approach, Real Estate Economics, 31, pp.623-646

Jones, C & Leishman, C 2006, Spatial dynamics of the housing market: an interurban perspective, Urban Studies, 43, pp.1041–1056 Gabriel, M, Jacobs, K, Arthurson, K, Burke, T & Yates, J 2005, ‘Conceptualising and measuring the housing affordability problem’, Research Paper no. 1, Australian Housing and Urban Centre Institute.

Galster, G 1996, William Grigsby and the analysis of housing submarkets and filtering, Urban Studies, 33, pp. 1797– 1806

Gibler, K, M & Tyvimaa, T 2014, The potential for consumer segmentation in the Finnish housing market, Journal of Consumer Affairs, Summer, pp.351–352

Gitelman, E & Otto, G 2012, ‘Supply elasticity estimates for the Sydney housing market’, The Australian Economic Review, 45, pp.179–190

Gleeson, B 2006, ‘Desocialising space: the decline of the public realm in Western Sydney’, Social & Cultural Geography, 7(1), pp.23–24

Goodman, A, C & Thibodeau, T, G 2007, The spatial proximity of metropolitan area housing markets, Real Estate Economics, 35(2), pp.209-232

Griffith, G 2015, 'The Greater Sydney Commission', NSW Parliamentary Research Service, no. 20. Grigsby, W, G 1963, Housing markets and public policy. Philadelphia, PA: University of Pennsylvania Press

Grigsby, W, Baratz, M, Galster, G, & Maclennan, D 1987, The Dynamics of neighbourhood change and decline, Progress in Planning, 28(1), pp.1–76

27 | Page Hansen, H & Tarp, F 2001, ‘Aid and growth regressions’, Journal of Development Economics, vol. 64, pp. 554–558

Haylen, A 2014, ‘House prices, ownership and affordability: trends in New South Wales’, Briefing Paper, NSW Parliamentary Research Service, pp. 1–31

Healey, J 2016, Ownership affordability, Spinney Press, vol. 400, pp. 2–11

Hincks, S and Wong C 2010, The spatial interaction of housing and labour markets: commuting flow analysis of North West England, Urban Studies, 47(3), pp.620-649

Holmes, S, London, K & Sheehan, L 2008, Ownership affordability in Australia: A supply- side analysis, Discussion paper, The Australian Competition Policy Research Alliance, pp. 1– 45

Housing NSW 2016, Rent and Sales Report, viewed 22 July 2016, http://www.housing.nsw.gov.au/about-us/reports-plans-and-papers/rent-and-sales- reports/issue-116

Jones, C & Leishman, C 2006, Spatial dynamics of the housing market: an interurban perspective, Urban Studies, 43(7), pp.1041–1056 Jones C, Coombes M, Dunse N, Watkins D, Wymer C 2012, Tiered housing markets and their relationship to labour market areas, Urban Studies, 49(12), pp.2633-2650.

Kutty, N, K 2005, ‘A new measure of ownership affordability: Estimates and analytical results’, Housing Policy Debate, 16(1), pp.113–142

Kyung-Hwan, K & Cho, M 2010, ‘Structural changes, housing price dynamics and ownership affordability in Korea’, Housing Studies, 25(6), pp.839–856

Latham, M 1992, ‘Urban policy and the environment in Western Sydney’, Australian Institute of Policy and Science, 64(1), pp.72

Lee, C, L 2017, An examination of the risk-return relation in the Australian housing market, International Journal of Housing Markets and Analysis, 10(3), pp. 431-449

Lee, C, L & Reed, RG 2014, ‘The relationship between housing market intervention for first- time buyers and house price volatility’, Housing Studies, 29(8), pp. 1082–1092

Liu, X & Otto, G 2017, ‘Housing supply elasticity in local government areas of Sydney’, Applied Economics, 49(53), pp.5441–5461

Maclenna, D & O’Sullivan, A 2013, Localism, devolution and housing policies, Housing Studies, 28(4), pp.599-615

McLaughlin, R, B 2011, ‘Metropolitan growth policies and new housing supply: evidence from Australia’s capital cities’, Australasian Journal of Regional Studies, 17(1), pp. 70–72

28 | Page Mee, K 2002, ‘Prosperity and the suburban dream: Quality of life and affordability in Western Sydney’, Australian Geographer, 33(3), pp. 339–349

Meen, G 1996, Spatial aggregation, spatial dependence and predictability in the UK housing market, Housing Studies, 11(3), pp. 345–372

Muellbauer, J & Murphy, A 2008, ‘Housing markets and the economy: the assessment’, Oxford Review of Economic Policy, pp. 1–33

NSW State Government 2020, Refugee settlement, available at: https://www.nsw.gov.au/projects/refugee-settlement Palm, R 1978, Spatial segmentation of the urban housing market, Economic Geography, 54(3), pp. 210–221

Pawson, H & Martin, C 2020, Rental property investment in disadvantaged areas: the means and motivations of Western Sydney’s new landlords, Housing Studies. doi: 10.1080/02673037. 2019.1709806

Productivity Commission Inquiry Report 2004, First Home Ownership, accessed 20 August 2016, http://www.pc.gov.au/inquiries/completed/first-home-ownership/report/housing.pdf

Randolph, B., & Holloway, D. (2005) ‘Social disadvantage, tenure and location: an analysis of Sydney and Melbourne’, Urban Policy and Research, 23, pp.174-197

Randolph, B, Pinnegar, S & Tice, A 2013, ‘The first home owner boost in Australia: A case study of outcomes in the Sydney housing market’, Urban Policy and Research, 31, pp. 57–60

Randolph, B & Tice, A 2014, ‘Suburbanizing disadvantage in Australian cities: Sociospatial change in an era of neoliberalism’, Journal of Urban Affairs, 36, pp. 392–395

Ruming, K, Gurran, N & Randolph, B 2011, ‘Ownership affordability and development contributions: new perspectives from industry and local government in New South Wales, Victoria and Queensland’, Urban Policy and Research, 29, pp. 258–262

Schwartz, D 2016, ‘The importance of affordable housing to economic competitiveness’, Economic Development Journal, 15(1), pp. 40

Stone, M, E 1990, One third of a nation: a new look at housing affordability in America, Economic Policy Institute, Washington, DC 20036

Stone, M, E 2004, ‘Shelter poverty: the chronic crisis of housing affordability’, New England Journal of Public Policy, 20, pp. 107–117

Stone, M, E 2006, A housing affordability standard for the UK, Housing Studies, 21, pp.453- 476,

Taylor, J & Wren, C 1997, ‘UK regional policy: an evaluation, Regional Studies, 31, pp.835- 848

29 | Page Watkins, C, A 2001, The definition and identification of housing submarkets, Environment and Planning A, 33, pp. 2235 Westerlund, J 2007, ‘Testing for error correction in panel data’, Oxford Bulletin of Economics and Statistics, 69, pp. 709–748

Worthington, A, C & Higgs, H 2013, ‘Macro drivers of Australian ownership affordability, 1985–2010, an autoregressive distributed lag approach’, Studies in Economics and Finance, 30, 347–365

Yates, J 2007, ‘The polarisation of ownership affordability’, Research Paper no. 8, AHURI, Melbourne, pp.12–20

Yates, J 2008, ‘Australia’s ownership affordability crisis’, The Australian Economic Review, 4, pp. 200–214

Yates, J & Milligan, V with Mike Berry, Terry Burke, Michelle Gabriel, Peter Phibbs, Simon Pinnegar and Bill Randolph 2007, ‘Ownership affordability: a 21st century problem National Research Venture 3: Ownership affordability for lower income Australians, AHURI Final Report no. 105, pp.13

Ying, Q, Luo, D & Chen, J 2013, ‘The determinants of homeownership affordability among the ‘sandwich class’: empirical findings from Guangzhou, China’, Urban Studies, 50,1870– 1888

30 | Page