INEQUALITY IN HOUSING AFFORDABILITY: MEASUREMENT AND ESTIMATION

Transcription

1 INEQUALITY IN HOUSING AFFORDABILITY: MEASUREMENT AND ESTIMATION Danny Ben-Shahar and Jacob Warszawski* ABSTRACT This research proposes and examines new measures for assessing the state of housing affordability inequality. We employ a large micro-level dataset by which we estimate and evaluate the time-varying housing affordability inequality in Israel over the period Results show that our developed housing affordability inequality Gini coefficient has considerably increased in the past decade. Moreover, controlling for changes in net income inequality and macroeconomic conditions, housing affordability inequality is found to positively correlate with average housing prices (computed in net-income terms). Furthermore, our method allows for an examination of segmentation in housing affordability. We find that segmentation particularly prevails across the household head s gender, family status, working status, and the number of income providers in the household. Research outcomes may direct decision-makers in designing policies aiming to reduce inequality and segmentation in housing affordability. Current Version: September 18, 2014 Key Words: Housing affordability; Inequality; Gini; Atkinson; Decomposition; Segmentation. JEL Codes: I32, R31, Z13 * Danny Ben-Shahar, Alrov Institute for Real Estate Research, Faculty of Management, Tel Aviv University, Tel Aviv, , Israel, bdanny@post.tau.ac.il; and Jacob Warszawski, Faculty of Architecture and Town Planning, Technion Israel Institute of Technology, Technion City, Haifa 32000, Israel; kobiw@technion.ac.il. The authors thank Roni Golan, Ofer Huberman, and Doron Sayag for their invaluable assistance in generating the dataset and Shlomo Yitzhaki and seminar participants at the Technion and the 2014 International AREUEA meetings and the 2014 Israel Regional Science meetings for helpful comments. 1

2 1 INTRODUCTION Housing is commonly the single largest expenditure item for most households, while poor and near-poor families often devote half their income to housing (Quigley and Raphael, 2004). It is not surprising, then, that the recent social protests occurring in many Western cities around the world were largely incited by requests for a supply of housing at affordable prices. This further explains the major interest that the general public, policymakers, and regulators have in the discussion of housing affordability. Various ratios are found in the literature for measuring housing affordability. Among these are housing-loan-repayment-to-income, ongoing-housing-cost-toincome, debt-to-housing-price, and housing-price-to-income (see, for example, Myer and Engelhardt (1996), Thalmann (1999), Quigley and Raphael (2004), Brounen et al. (2006), Stone (2006), Norris and Shiels (2007), Kim and Cho (2010), and Haffner and Heylen (2011). 1 Also, while the state of housing affordability is commonly explored by focusing on an average and/or median figure, some studies further explore affordability among populations stratified by socio-economic and demographic characteristics such as income, poverty status, race, and ethnicity (see, for example, Quigley and Raphael, 2004; Meen, 2011). Surprisingly, however, to the best of our knowledge previous research has never attempted to develop a measure that summarizes and examines the state of housing affordability inequality. 2 In this study, we assume this task by adapting a widely accepted measure for estimating income equality to the context of housing affordability thereby developing a novel approach for assessing the state of housing affordability inequality. Further, we empirically examine the factors that associate 1 Studies of housing affordability alternatively adopt the residual income approach, by which they examine how costs of basic goods net of housing costs associate with income (e.g., Whitehead, 1991; Stone, 2006; Kutty, 2005; & Chen et al., 2010). One of the controversies surrounding both the ratio and the residual income approaches concerns the threshold above which affordability may become increasingly subjective, as well as the definition and measurement of affordability below that threshold (Stone, 2006). 2 Matlack and Vigdor (2008) associate the rise in household income inequality with the deterioration of housing affordability for poor households by focusing on residual income and the rent-per-income ratio. Our study thus differs from that of Matlack and Vigdor (2008) on three central issues. First, we do not include subjective estimates of household basic needs for goods. Moreover, we consider the entire distribution of household housing affordability as opposed to focusing on the lower tail of the affordability distribution (the latter also associates with Whitehead, 1991; Stone, 2006; & Kutty, 2005). Finally, we develop a Gini measure that summarizes the level of housing affordability inequality. 2

3 with the time-varying dynamics of the derived housing affordability inequality measure. The Gini coefficient is commonly used in socio-economic literature to estimate the state of income inequality (see, for example, Alderson and Nielsen, 2002; Jäntti and Jenkins, 2010; Leigh, 2007; Frank, 2009). 3 The methodology for estimating the income Gini coefficient has been extended and implemented, however, to measure the state of inequality in other areas such as education and human capital (Földvári and Leeuwen, 2011), fossil resource consumption (Papathanasopoulou and Jackson, 2009), ecological entitlements (Ruitebeek, 1996), innovative activity states and R&D spillovers (Audretsch and Feldman, 1996), firm size across industries and locations (Jovanovic, 1982), and child achievements (Sastry and Pebley, 2010). In the housing literature, the Gini coefficient approach has been applied by Buckley and Gurenko (1997) to measure the effect of housing subsidies on living space inequality; by Landis et al. (2002) to measure inequality in housing values, housing costs, and monthly rent; and by Henley (2003) to study changes in the distribution of housing wealth. Robinson et al. (1985) applied the Gini coefficient to measure inequality in housing consumption. Also, studies by Tilly (2006) and Matlack and Vigdor (2008) discuss the association between income inequality and housing affordability challenges at the bottom of the income distribution. More recently, Dewilde (2011), Dewilde and Lancee (2012), and Norris and Winston (2012a,b) relate income inequality to homeownership and homeownership inequality. In this study, we propose and compute a Gini coefficient of housing affordability inequality based on the net income-to-housing price ratio. Intuitively, a household net income-to-housing price ratio measures the share (portion) of the housing unit that a household s periodic net income could purchase, had it designated its entire net income to the purchase of a housing unit identical to the one in which it resides. 4 Similar to the income Gini coefficient, our derived housing affordability 3 Recall that the Gini coefficient measures the area between the Lorenz curve and a hypothetical line of absolute equality, and is expressed as a percentage of the total area under this line. The Gini coefficient thus commonly ranges from zero (perfect equality) to one (complete inequality). For alternative measures of inequality as well as criticism and limitations of the Gini measure, see, for example, Cowell (2000). 4 Intuitively, a household housing price-to-net income ratio can be interpreted as the number of time periods it would take one to complete the purchase of a dwelling unit, had the household earmarked its entire net income to purchase a unit identical to the one in which it resides. Formulation of the Gini 3

4 inequality measure thus indicates on the micro-level degree of inequality in household ability to purchase a housing unit measured by the portion of the housing unit that a household s periodic net income could purchase. 5 We further estimate the factors that associate with our proposed housing affordability inequality measure and, following Yitzhaki and Lerman (1991), Lambert and Aronson (1993), and Cowell (2000), we examine segmentation in housing affordability by stratifying the household sample according to various socio-demographic and locational characteristics. The analysis is based on a large micro-level sample representative of all households in Israel over the period The data include 156,583 observations of household socio-economic, demographic, locational, and dwelling unit characteristics (see further details in the Data section below). 6 Results show that housing affordability inequality, as assessed by our proposed housing affordability Gini coefficient, exhibits a substantial increase over the past decade. This increase is accompanied by a more moderate increase in equivalent net income Gini coefficient. Empirical examination of the time-varying housing affordability inequality measure reveals a highly significant positive correlation with average housing prices (in net income terms), controlling for changes in net income inequality and macroeconomic conditions. In addition, we find that while housing affordability has considerably dropped for all socio-demographic segments over the period, meaningful segmentation in housing affordability particularly prevails across household head gender, household family status, household head working status, household number of income providers, and household geographical location. Finally, we show that our coefficient method prevents us, however, from applying it directly to the housing price-to-net income ratio. Instead, the net income-to-housing price ratio may both conform to the Gini methodology and be consistent with the literature that measures housing affordability. 5 Ultimately, in computing the housing affordability inequality, one might wish to focus on household required amount of housing services (given the household demographic characteristics) and thereby avoid potential noise in focusing on the actual housing unit that one occupies (i.e., omitting possible under- and over-consumption of housing). This exercise is, however, beyond the scope of this study. 6 The interest in studying housing affordability of homeowners is threefold. First, as homeowners comprise a dominant share of the housing market, examining their incomes and the values of dwelling units (thus studying their housing affordability status) indicates the likelihood of non-owners attaining ownership. Also, it particularly indicates the potential of homeowners to filter-up in the housing market. Finally, examining a time-series of housing affordability figures of homeowners allows policymakers to draw conclusions with respect to the time-varying level of housing prices in income terms (see, among others, Fingleton, 2008; Fisher et al., 2009). 4

5 evidence on housing affordability inequality is robust to replacing the Gini coefficient method with the Atkinson index approach. The main contribution of the research is twofold. First, we propose and implement a novel approach for assessing the state of housing affordability inequality based on the Gini coefficient methodology. Moreover, by exploring the factors that correlate with the level of inequality and specifying the population segments that particularly contribute to the level of inequality, we provide decision makers regulating housing prices and housing welfare with a toolbox and practical evidence that can motivate the design of effective policies aimed at promoting housing affordability where it is especially needed. The organization of the paper is as follows. The next section describes the methodology, while Section 3 presents the data, including variable definitions and related summary statistics. Section 4 presents related statistical results on the correlation between housing affordability inequality and average housing prices, and Section 5 assesses the robustness of the outcomes to the Gini coefficient specification. Section 6 examines socio-demographic segmentation in housing affordability, and, finally, Section 7 provides a summary and concluding remarks. 2 METHODOLOGY Given the individual household net income variable, we first compute the quarterly net income Gini coefficient (denoted by G W ) for the period We then compute the quarterly housing affordability Gini coefficient (G H ) based on household i net income-to-housing price ratio (W i /P i ). In deriving the housing affordability Gini coefficient, we adapt the method of computing the income Gini coefficient by substituting the individual net income variable with the individual net income-tohousing price ratio. Inequality in the net income-to-housing price variable essentially indicates inequality in the share of a household s own housing unit that the individual household can afford, given its individual periodic net income. 7 7 Formally, the housing affordability Gini coefficient is then computed as follows:, where and The weighted average of all surveyed household housing affordability is estimated with:, where is the observation weight of household (percentage weight), is the number of members in household, is household i s net income-to-housing price ratio, and is the number of households surveyed (in a given quarter). For more on the Gini coefficient computation, see Araar and Duclos (2013). 5

6 (1) Following the derived time-series of G W and G H coefficients, we can write G Ht = α! + α! G Wt + α! PW t + α! CONTROLS t + u t, where t is a time period index (in quarters), G W and G H are net income and housing affordability Gini coefficients, respectively, P/W is the average housing price-to-net income ratio (across all households at time t), CONTROLS is a matrix of macroeconomic variables, including gross domestic product (GDP), housing construction ends (HE), housing construction starts (HS), unemployment rate (UR), long-term (20 25-year) price-level-adjusted mortgage rate (MR), short-term (1-year) nominal bond rate (BR), construction price index (CI), and new Israeli sheqel-todollar exchange rate (EX). Also, α! α! are parameters, α! is a vector of parameters, and u is a random disturbance term. According to equation (1), time-varying inequality of housing affordability not only associates with net-income inequality, but also may correlate with the average housing price (in net income terms, i.e., average housing price-to-net income ratio) as well as changes in macroeconomic indicators. (For more on the correlation between macroeconomic indicators and housing prices see, for example, Adams and Füss, 2010; Kim and Cho, 2010; Schnure, 2005; Ortalo-Magné and Rady, 2006; Case and Quigley, 2008; Sutton, 2002; and Poterba, On the correlation between macroeconomic indicators and income inequality, see, for example, Achdut, 1996; Blejer and Guerrero, 1990; Milanovic, 2002; Heshmati, 2004; and Jäntti and Jenkins, On the correlation between macroeconomic indicators and housing affordability see, for example, Malpezzi, 1999; Mostafa et al., 2006; and Ben-Shahar and Warszawski, 2011.) In the last part of our study, we stratify the household sample by sociodemographic and locational characteristics. We then decompose the derived housing affordability Gini coefficient (G H ) into within-segment, between-segment, and overlap components of inequality. In other words, for a given stratification, we compute (a) the weighted-average of the housing affordability Gini coefficient by segments (the within-segment component); (b) the weighted-average Gini coefficient if each household in a segment had maintained the segment average housing affordability level (the between-segment component); and (c) the sample Gini coefficient net of the within- and between-segments, that is, the residual part (the overlap component) 6

7 following Yitzhaki (1994), the overlap component is a measure for segmentation in the population. (For a complete discussion on the properties of the overlap component, see Frick et al., 2006.) 3 DATA Data for this study come from three sources. First, 156,583 micro-level observations on individual household socio-economic, demographic, and dwelling unit characteristics are provided by the Household Income and Expenditure Surveys conducted by the Israel Central Bureau of Statistics for the years Each quarterly independent cross-section sample has between 1,041 and 2,921 observations and is representative of all households in Israel (see Central Bureau of Statistics, ). Table 1 shows the number of cross-sectional observations per quarter for the period. Table 2 presents the household socio-demographic and dwelling unit characteristics in the dataset and the 1992 and 2011 shares of each characteristic within the sample. Socio-demographic characteristics include family status, the household head s gender, the number of members in the household, the household head s years of education, last formal education, and age, the number of household income providers, and the household head s working status, occupation, and industry. Dwelling unit characteristics include the number of rooms and location. Another source of data is all housing transactions in Israel for the period , recorded by the Israel Tax Authority a total of 729,505 observations of transacted dwelling unit prices and attributes. Based on this data, we generate a quality-adjusted price for the dwelling unit of each household in the Household Income and Expenditure Surveys. (A detailed description of the procedure by which we produce the quality-adjusted price that is matched to each household is provided in the appendix.) A final source of data includes macroeconomic indicators for the period obtained from the Bank of Israel and the Israel Central Bureau of Statistics. These include the Gross Domestic Product in dollars (GDP), number of housing construction ends (HE), change in the unemployment rate (ΔUR), change in the mortgage rate (ΔMR), change in the rate of change in the construction index (ΔCI), 7

8 change in the one-year bond rate (ΔBR), change in the new Israeli sheqel-to-dollar exchange rate (ΔEX), and change in the number of housing construction starts (ΔHS). 8 Table 3 presents summary statistics for household net income, net income-tohousing price ratio, housing price-to-net income ratio, and the macroeconomic variables. It follows that the mean monthly net income was 2,053 dollars, with a standard deviation equal to 554 dollars (1USD 4 New Israeli Sheqels). The mean housing price-to-net income and net income-to-housing price were about 112 and , respectively, with standard deviations equal to and , respectively. (See Table 3 for summary statistics of the macroeconomic variables.) Exhibit 1 displays the sample quarterly average and deciles of the housing price-to-net-income ratio (P/W) over the period Interestingly, the average ratio approximately overlaps the seventh decile (note that a higher decile associates with a greater housing price-to-net income ratio, i.e, a lower level of affordability). Moreover, while the top decile of the housing price-to-net income ratios (10% of the population with greatest affordability) ranges from 31 to 46 over the examined period, the lowest decile (lowest 10%) experiences equivalent figures in the range. Namely, had households earmarked their entire net income for purchase of a housing unit, the top (bottom) housing affordability decile would have needed ( ) months to complete the purchase. At the same time, the housing price-to-net income ratio of the third, fifth, and seventh deciles ranges in the 48 71, , and figures, respectively. Exhibit 2 shows the time-varying series of the sample quarterly average housing price-to-net income ratio over the period Interestingly, the average price-to-net income ratio not only exhibits a positive slope over the examined period (as seen by the linearized trend); in addition, it experiences a sizable shift from the levels in the years to the levels in the post-2007 period. Note in particular the sharp increase from a level of 98 in 2008Q1 to 137 in 8 The unit root hypothesis could not statistically be rejected for the time-series of the unemployment rate (UR), mortgage rate (MR), rate of change in the construction index (CI), the one-year bond rate (BR), the new Israeli sheqel-to-dollar exchange rate (EX), and the number of housing construction starts (HS). We thus specified these non-stationary control variables in first difference terms (difference between their time t and t-1 values) for which the unit root hypothesis was statistically rejected. 9 In the derivation of the average housing price-to-net income ratio, the lowest and highest 0.5 percent of the observations were omitted to mitigate the effect of outliers. Results, however, are robust to the inclusion of these outliers (available by request). 8

9 2011Q4. In the next section, we examine the relationship between this phenomenon and the attained level of inequality in housing affordability. 4 RESULTS Based on the series of individual household net income W and net income-to-housing price ratio W/P, we first compute the annual net income Gini and housing affordability Gini coefficients, G W and G H respectively, for the period (see Table 4 for summary statistics of G W and G H ). Exhibit 3 shows the computed housing affordability and net income Gini coefficients for the years Note that the level of housing affordability inequality has not only experienced an upward trend since the year 2000, peaking in 2008 with a Gini coefficient of 0.352, but also that its increase has exceeded that of the net income Gini coefficient in the post-2000 decade. Columns (1) and (2) in Table 5 present the outcomes from the estimation of equation (1) (full model and stepwise regression, respectively). Results show that the housing affordability Gini coefficient is, as expected, positively correlated to a high degree with the net income Gini. 10 Interestingly, however, the housing affordability Gini is also positively correlated with average housing price-to-net income ratio at the 5% significance level (t-value equals 2.36). As the standard deviation of the housing affordability Gini coefficient equals (see Table 4), the estimated coefficient on the housing price-to-net income variable indicates that a ten-unit increase in the housing price-to-net-income ratio associates with a rise in the housing affordability Gini coefficient equal to just over 10% of its standard deviation. Put differently, the sharp rise in the average price-to-net income ratio over the period (from 98 in 2008Q1 to 137 in 2011Q4 see, once again, Exhibit 3) associates with an increase in housing affordability Gini coefficient equal to 40% of its standard deviation. 11 Finally, concerning the control variables, it follows that the housing affordability Gini negatively correlates with the number of housing construction ends 10 Also note that the computed Pearson correlation between W and W/P is equal to 0.65 and between W and P is equal to Note that, while the derived Gini coefficient is based on the individual household net-income and housing price, it does not immediately follow that the level of the inequality measure should correlate (either positively and or negatively) with the average housing price-to-net income ratio. Interestingly, however, as described above, our findings indicate that, when average housing price-to-net income rises, households at the lower tail of the housing affordability distribution are adversely affected at a greater degree than those at the higher tail of the distribution. 9

10 and the change in the new Israeli sheqel-to-dollar exchange rate at the 1% and 10% significance levels, respectively, and positively correlates with the level of gross domestic product with a five-period lag at the 1% significance level ROBUSTNESS TEST: HOUSING AFFORDABILITY ATKINSON INDEX In this section we assess the robustness of our findings by generating and estimating an alternative to the housing affordability Gini coefficient. Specifically, we derive a housing affordability Atkinson index and re-estimate the model in (1). Recall that the Atkinson index had originally been developed to measure income inequality (see Atkinson, 1970), allowing for varying levels of inequality intolerance. It was later adapted to measure inequality of, for example, ecological entitlements (Ruitebeek, 1996) and the geographical distribution of general practitioners (Gravelle and Sutton, 2001). Also, Robinson et al. (1985) applied the Atkinson index method to measure inequality in housing consumption. 13 Similar to our derived housing affordability Gini coefficient, we derive an Atkinson index of housing affordability inequality A H (ε), where ε is an inequality aversion parameter based on the net income-to-housing price ratio. 14 In the context of housing affordability, the Atkinson index can be interpreted as the share of individual housing affordability (in W/P terms) that may be disposed so as to generate the same level of social welfare that could be achieved if the mean level of W/P were to be equally distributed among all households While other lags in the level of Gross Domestic Product were found to correlate with the housing affordability Gini coefficient, the highest statistical significance was obtained for the five-period lag reported above. Results obtained for other lags are available from the authors upon request. Also, based on macro-level data, Ben-Shahar and Warszawski (2011) find a negative correlation between the average housing price-to-net income ratio and lagged unemployment rate (also see Jäntti and Jenkins, 2010). 13 As noted by Robinson et al. (1985), the Atkinson index is particularly important in the case of a commodity such as housing which is widely regarded as a necessity and, as such, the plight of those at the bottom end of the distribution is likely to be of special concern (p. 251). 14 Formally, the housing affordability Atkinson index is computed as follows:, where and and where ε is an inequality-aversion parameter. All other variables are as described above. For more on the Atkinson index computation, see Araar and Duclos (2013). 15 Recall that according to Yitzhaki (1983), we have, where is the average income, is the equallydistributed-equivalent income, and is the Atkinson index as a function of the level of inequalityaversion parameter, ε. 10

11 (2) In accordance with equation (1), we then estimate A H t = β! + β! A W t + β! PW t + β! CONTROLS t + v t, where t once again denotes a time-period index (in quarters), A H and A W are the derived housing affordability and net income Atkinson indices, respectively, for ε=0.5, 16 β! β! are estimated parameters, β! is a vector of parameters, v is a random disturbance term, and the remaining variables are as described above. Similar to equation (1), in equation (2) we focus on estimating the correlation between housing affordability inequality (this time estimated by the housing affordability Atkinson index) and the average housing price (in net income terms). Again, we control for changes in net income inequality and time-varying macroeconomic conditions. Outcomes are robust to the Atkinson specification. In particular, Exhibit 4 shows the computed housing affordability and net income Atkinson indices over the period A couple of points are worth noting. First, while in 2000Q4 the housing affordability Atkinson index equaled 0.074, its level rose by about 27% to equal in 2011Q4. Following Atkinson (1970), the latter implies that all households could have disposed 7.4% and 9.4% of the total net income-to-housing price W/P in 2000Q4 and 2011Q4, respectively, so as to generate the same level of social welfare that could have been achieved if the mean level of W/P had been equally distributed among all households (see, for example, Yitzhaki, 1983). Moreover, the pattern previously observed with the Gini coefficient now repeats with the Atkinson index. That is, the sharp increase in the housing affordability Atkinson index in the last decade led it to generally exceed the equivalent increase in the level of the net income Atkinson index. Further, columns (3) and (4) in Table 5 present the outcomes from the estimation of equation (2) (full model and stepwise regression, respectively). Results indicate that the housing affordability Atkinson index is positively correlated not only with the net income Atkinson index but also with average housing price (in net income terms) both at the 1% significance level. Specifically, a ten-unit change in 16 Empirical studies commonly focus on the inequality-aversion parameter ε ranging between 0.5 and 2.5 (e.g., Biewen and Jenkins, 2006). Consistent with the literature, we compute A H for ε=0.5. Unreported results are qualitatively robust to increasing the level of ε. 11

12 the average housing price-to-net income ratio corresponds to a change in the housing affordability Atkinson index equal to about 21% of its standard deviation. Results further show that the housing affordability Atkinson index negatively correlates with the number of housing construction ends and the change in the new Israeli sheqel-todollar exchange rate at the 1% and 10% significance levels, respectively, and positively correlates with gross domestic product (with a five-period lag) at the 1% significance level. 6 SEGMENTATION IN HOUSING AFFORDABILITY We stratify the household sample by socio-demographic characteristics and decompose the derived housing affordability Gini coefficient into within-segment, between-segment, and overlap components (see Yitzhaki and Lerman, 1991; Lambert and Aronson, 1993; Cowell, 2000). A relatively low overlap component in the decomposition of the Gini coefficient implies a relatively heterogeneous population (i.e., high level of segmentation) across the stratified characteristic. Table 6 shows the decomposed components of the housing affordability Gini coefficient by sociodemographic segments and the computed housing affordability Gini coefficient and the average housing price-to-net income ratio by segments for the years 1992 and It follows that in 1992 and 2011, a relatively high level of segmentation (i.e., low overlap component) persists for the household head gender stratification, 18 household family status (divides into married, divorced, widowed, single, and living separately), the household head working status (divides into wage-employee, selfemployed, and non-worker), the number of household providers (divides into zero-, one-, two-, three-, and four-provider segments), and the geographical region where the household resides (divides into nine regions as determined by the Israel Central Bureau of Statistics see regions in Table 2). It also follows that male-headed households experience a sharp increase in the average housing price-to-net income ratio during the examined period (from 79 in 17 On social inequality in the context of homeownership in Israel, see, for example, Lewin-Epstein et al. (2004). 18 Household head gender is generally determined by the gender of the person who is the main income provider in the household. See Israel Central Bureau of Statistics (2013a) for further details. 12

13 1992 to 116 in 2011) that is accompanied by an increasing housing affordability Gini coefficient (from 0.30 to 0.34, respectively). Also, the traditional discrepancy in the price-to-net income ratio between male- and female-headed households maintains over the examined period, where female-headed household figures rise from 125 in 1992 to 139 in While experiencing an increased average housing price-to-net income ratio from 77 to 116 over the period, married households capture the top spot in housing affordability among the family status segments. The divorced segment experiences increasing levels of inequality and average housing price-to-net income ratio (from 0.32 and 113, respectively, in 1992, to 0.34 and 137, respectively, in 2011). 20 The share of single household heads in the population almost doubles from 7% to over 13% over the period (see Table 2), while both its housing affordability Gini and average housing price-to-net income ratio sharply increase (from to 0.37 and from 100 to 206, respectively). With respect to household head working status, average housing price-to-net income rises over the period from 65 to 106 for wage employees, from 77 to 119 for the self-employed, and from 148 to 221 for non-workers. These are coupled with an increased housing affordability Gini coefficient, most notably for the nonworker segment rising from in 1992 to 0.37 in 2011, while the wage-employee and self-employed segments increase from 0.26 and 0.29 to 0.32 and 0.34, respectively, during that period. Note the considerable difference among the zero-, one-, and two or more provider segments. The two or more provider subgroup maintains not only the lowest level of average housing price-to-net income ratio (at and in 1992 and 2011, respectively), but also a relatively low level of housing affordability inequality 19 Compare to the findings of Laux and Cook (1994), and Saegert and Clark (2006), who show that female-headed households in the U.S. exhibit both lower levels of housing affordability and greater income inequality (also see Edwards, 2001). 20 Weiss (1984) and others find that income drops considerably after marital dissolution and maintains its low level some five years following the dissolution, while expenditures at the same time maintain their pre-dissolution levels. For more on single-parent household housing consumption see, for example, Saegert and Clark (2006). 13

14 (Gini coefficient of in 2011). 21 In contrast, as expected, the no-provider household segment experiences the greatest levels of both average housing price-tonet income ratio and housing affordability inequality, reaching figures of 221 and 0.37, respectively, in Correspondingly, in 2011, the one-provider segment attains an average price-to-net income ratio of 150 and a housing affordability Gini of Interestingly, inequality in housing affordability among geographical regions has dramatically increased over the period (the overlap component in the decomposition of the Gini coefficient has decreased from 53% to 38% over this twenty-year period). While in 2011 the North and Krayot regions maintain the greatest level of affordability with an average housing price-to-net income ratio equal to 73 and 69, respectively (up from 58 and 67, respectively, in 1992), Jerusalem and Tel Aviv sit at the bottom of the list with an average ratio equal to 206 and 182 (up from 108 and 111, respectively, in 1992). The sharp increase in the average housing price-to-net income ratio in Jerusalem and Tel Aviv (as compared to North and Krayot) can be attributed, among other possible reasons, to the sharp rise in real estate prices in those specific regions since Jerusalem and Tel Aviv further lead all regions in the level of inequality with a Gini coefficient of 0.32 in Several final points are worth noting. First, a relatively low level of segmentation is found for the number of members in household and the household head s age, years of education, last formal education, occupation, and industry. Further, for some of these household stratifications, we find that segmentation has somewhat diminished (the overlap component has increased) during the period (the exceptions are the household head years of education, last formal education, and the number of members in household, whose overlap component experienced a marginal decline). These trends are nonetheless generally accompanied by a sharp increase in segments average housing price-to-net income ratio. Finally, it should be stressed that the segmentation in affordability analysis proposed here allows 21 This outcome is consistent with the assertion of Cancian and Reed (1998), Heathcote et al. (2010), and others that the net effect of the increased participation of females in the labor force reduces overall inequality. 22 Specifically, while average housing prices increased in the Krayot and North regions by 40% and 57%, respectively, over the period , in Tel Aviv and Jerusalem they increased by 80% and 61%, respectively (Israel Central Bureau of Statistics, ). 14

15 for the identification of population subgroups (according to socio-demographic and locational characteristics) that particularly contribute to the level of housing affordability inequality and may thus provide policymakers with high-resolution information on where policy measures that promote equal housing affordability opportunities are to be implemented. 7 SUMMARY AND CONCLUSIONS Housing is commonly the single largest expenditure item for most households and generally demands more than half the income of poor and near-poor families. In this study, we develop measures (for the first time, to the best of our knowledge) by which the state of housing affordability inequality could be summarized and examined. Particularly, we adapt the Gini coefficient and Atkinson index methods for estimating income inequality to the context of housing affordability and estimate the factors that associate with the time-varying housing affordability inequality measures in Israel over the period Results show that the time-varying household housing affordability Gini coefficient and Atkinson index are positively correlated with average housing price (in net income terms), controlling for changes in net income inequality and macroeconomic conditions. Further, we detect considerable segmentation in housing affordability for household head gender, household family status, household head working status, number of household providers, and household geographical residence. As the state of housing affordability is a constant source of political interest, our developed method for estimating housing affordability inequality and the outcomes on its positive correlation with average housing prices and, further, its capacity to identify socio-demographic segmentation in housing affordability may assist policymakers in designing policies that would increase housing affordability where it is particularly needed, thereby effectively reducing inequality and segmentation in housing affordability. Several major tasks face future researchers in this area. Methodologically, the estimation of housing affordability potentially suffers from a bias caused by household over- and under- consumption of housing. Refined measures of housing affordability inequality may attempt to address this possible bias. Further, while there exist various approaches for computing housing affordability, none of the prevailing 15

16 methods points to a formally derived threshold (benchmark) below which a household is categorized as sub-affordable. Finally, empirically, a panel analysis of housing affordability Gini coefficient across different economies may shed light on the regimes and specific policies that are most effective in promoting housing affordability equality. 16

17 References Achdut, L. (1996). Income inequality, income composition and macroeconomic trends: Israel, Economica, S1 S27. Adams, Z., & Füss, R. (2010). Macroeconomic determinants of international housing markets. Journal of Housing Economics, 19(1), Alderson, A. S., and Nielsen, F. (2002). Globalization and the great U-turn: Income inequality trends in 16 OECD countries. American Journal of Sociology, 107(5), Araar, A., and Duclos, J.Y. (2013). DASP: Distributive Analysis Stata Package User Manual. Université Laval PEP, CIRPÉE and World Bank. Audretsch, D. B., and Feldman, M. P. (1996). R&D spillovers and the geography of innovation and production. American Economic Review, 86(3), Ben-Shahar, D., and Warszawski, J. (2011). Partly cloudy to clear: The Israeli economy and the local housing market under the storm of the world financial crisis. In Bardhan, A., Edelstein, R., and Kroll, C. (Eds.), Global housing markets: Crises, policies, and institutions. Hoboken, NJ: Wiley. Biewen, M., and Jenkins, S. P. (2006). Variance estimation for generalized entropy and Atkinson inequality indices: The complex survey data case. Oxford Bulletin of Economics and Statistics, 68(3), Blejer, M. I., and Guerrero, I. (1990). The impact of macroeconomic policies on income distribution: an empirical study of the Philippines. Review of Economics and Statistics, Bogdon, A. S., and Can, A. (1997). Indicators of local housing affordability: Comparative and spatial approaches. Real Estate Economics, 25(1), Buckley, R. M., and Gurenko, E. N. (1997). Housing and income distribution in Russia: Zhivago s legacy. World Bank Research Observer, 12(1), Cancian, M., and Reed, D. (1998). Assessing the effects of wives earnings on family income inequality. Review of Economics and Statistics, 80(1), Case, K. E., and Quigley, J. M. (2008). How housing booms unwind: Income effects, wealth effects, and feedbacks through financial markets. International Journal of Housing Policy, 8(2), Central Bureau of Statistics ( ). Statistical Abstract of Israel, publication number 42 63, Jerusalem, Israel. Central Bureau of Statistics (2008). Peripherality index of local authorities 2004 New development. Press Release, Jerusalem, Israel. 17

18 Central Bureau of Statistics (2013a). Income survey 2011, publication number 1524, Jerusalem, Israel. Central Bureau of Statistics (2013b). Characterization and classification of geographical units by the socio-economic level of the population Publication number 1530, Jerusalem, Israel. Chen, J., Hao, Q., & Stephens, M. (2010). Assessing housing affordability in postreform China: A case study of Shanghai. Housing Studies, 25(6), Cowell, F. A. (2000). Measurement of inequality. In Atkinson, A. B., and Bourguignon, F. (Eds.), Handbook of income distribution. Amsterdam & New York: Elsevier. Cowell, F. A. (2009). Measuring inequality. Part of the series LSE Perspectives in Economic Analysis, published by Oxford University Press. Dewilde, C. (2011). The interplay between economic inequality trends and housing regime changes in advanced welfare democracies: A new research agenda. Amsterdam, AIAS, GINI Discussion paper 18. Dewilde, C., and Lancee, B. (2012). Income inequality and access to housing in Europe: A new research agenda. Amsterdam, AIAS, GINI Discussion paper 32. Edwards, M. E. (2001). Home ownership, affordability, and mothers changing work and family roles. Social Science Quarterly, 82(2), Fingleton, B. (2008). Housing supply, housing demand, and affordability. Urban Studies, 45(8), Fisher, L. M., Pollakowski, H. O., and Zabel, J. (2009). Amenity- based housing affordability indexes. Real Estate Economics, 37(4), Földvári, P., and van Leeuwen, B. (2011). Should less inequality in education lead to a more equal income distribution? Education Economics, 19(5), Frank, M. W. (2009). Inequality and growth in the United States: Evidence from a new state- level panel of income inequality measures. Economic Inquiry, 47(1), Frick, J. R., Goebel, J., Schechtman, E., Wagner, G. G., and Yitzhaki, S. (2006). Using analysis of Gini (ANOGI) for detecting whether two subsamples represent the same universe the German Socio-Economic Panel Study (SOEP) experience. Sociological Methods & Research, 34(4), Gravelle, H., and Sutton, M. (2001). Inequality in the geographical distribution of general practitioners in England and Wales Journal of Health Services Research & Policy, 6(1), Heathcote, J., Perri, F., and Violante, G. L. (2010). Unequal we stand: An empirical analysis of economic inequality in the United States, Review of Economic 18

19 Dynamics, 13(1), Henley, A. (2003). Changes in the distribution of housing wealth in Great Britain, Economica, 65(259), Heshmati, A. (2004). Regional income inequality in selected large countries (No. 1307). IZA Discussion paper series. Jäntti, M., and Jenkins, S. P. (2010). The impact of macroeconomic conditions on income inequality. Journal of Economic Inequality, 8(2), Jovanovic, B. (1982). Selection and the evolution of industry. Econometrica, Kim, K., and Cho, M. (2010). Structural changes, housing price dynamics and housing affordability in Korea. Housing Studies, 25(6), Kutty, N. K. (2005). A new measure of housing affordability: Estimates and analytical results. Housing Policy Debate, 16(1), Lambert, P. J., and Aronson, J. R. (1993). Inequality decomposition analysis and the Gini coefficient revisited. Economic Journal, 103(420), Landis, J. D., Elmer, V., and Zook, M. (2002). New economy housing markets: Fast and furious but different? Housing Policy Debate, 13(2), Laux, S. C., and Cook, C. C. (1994). Female-headed households in nonmetropolitan areas: Housing and demographic characteristics. Journal of Family and Economic Issues, 15(4), Leigh, A. (2007). How closely do top income shares track other measures of inequality? Economic Journal, 117(524), F619 F633. Lerman, R. I., and Yitzhaki, S. (1989). Improving the accuracy of estimates of Gini coefficients. Journal of Econometrics, 42(1), Lewin-Epstein, N., Adler, I., and Semyonov, M. (2004). Home ownership and social inequality in Israel. In Kurz and Blossfeld (Eds.), Home ownership and social inequality in comparative perspective, Stanford, CA: Stanford University Press. Malpezzi, S. (1999). A simple error correction model of house prices. Journal of Housing Economics, 8, Matlack, J. L., and Vigdor, J. L. (2008). Do rising tides lift all prices? Income inequality and housing affordability. Journal of Housing Economics, 17(3), Meen, G. (2011). A long-run model of housing affordability. Housing Studies, 26(7 8),

20 Milanovic, B. (2002). True world income distribution, 1988 and 1993: First calculation based on household surveys alone. Economic Journal, 112(476), Mostafa, A., Wong, F. K., & Hui, C. M. (2006). Relationship between housing affordability and economic development in mainland China Case of Shanghai. Journal of Urban Planning and Development, 132(1), Norris, M., and Winston, N. (2012a). Home ownership and income inequalities in Western Europe: Access, affordability and quality. Amsterdam, AIAS, GINI Discussion Paper 41. Norris, M., and Winston, N. (2012b). Home-ownership, housing regimes and income inequalities in Western Europe. Amsterdam, AIAS, GINI Discussion Paper 42. Ortalo-Magné, F., and and Rady, S. (2006). Housing market dynamics: On the contribution of income shocks and credit constraints. Review of Economic Studies, 73, Papathanasopoulou, E., and Jackson, T. (2009). Measuring fossil resource inequality A case study for the UK between 1968 and Ecological Economics, 68(4), Poterba, J. M. (1984). Tax subsidies to owner-occupied housing: An asset market approach. Quarterly Journal of Economics, 99(4), Quigley, J. M., and Raphael, S. (2004). Is housing unaffordable? Why isn t it more affordable? The Journal of Economic Perspectives, 18(1), Robinson, R., O Sullivan, T., and Le Grand, J. (1985). Inequality and housing. Urban Studies, 22(3), Ruitenbeek, H. J. (1996). Distribution of ecological entitlements: Implications for economic security and population movement. Ecological Economics, 17(1), Saegert, S., and Clark, H. (2006). Opening doors: What a right for housing means for women. In Bratt, Stone, and Hartman (Eds.), A right to housing: Foundation for a new social agenda. Philadelphia, PA: Temple University Press. Sastry, N., and Pebley, R. (2010) Family and neighborhood sources of socioeconomic inequality in children s achievement. Demography, 47(3), Sayag, D. (2012). Measuring the local house price movements and estimating the price elasticity. Israel Economic Review, 10(1), Schnure, C. (2005). Boom-bust cycles in housing: The changing role of financial structure. International Monetary Fund. Stone, M. E. (2006). Housing affordability: One-third of a nation shelter-poor. In Bratt, Stone, and Hartman (Eds.), A right to housing: Foundation for a new social agenda. Philadelphia, PA: Temple University Press. 20

21 Sutton, G. D. (2002, September). Explaining changes in house prices. BIS Quarterly Review, Thalmann, P. (1999). Identifying households which need housing assistance. Urban Studies, 36(11), Tilly, C. (2006). The economic environment of housing: Income inequality and insecurity. In Bratt, Stone, and Hartman (Eds.), A right to housing: Foundation for a new social agenda. Philadelphia, PA: Temple University Press. Weiss, R. S. (1984). The impact of marital dissolution on income and consumption in single-parent households. Journal of Marriage and the Family, Whitehead, C. M. (1991). From need to affordability: An analysis of UK housing objectives. Urban Studies, 28(6), Yitzhaki, S. (1983). On an extension of the Gini inequality index. International Economic Review, 24(3), Yitzhaki, S. (1994). Economic distance and overlapping of distributions. Journal of Econometrics, 61(1), Yitzhaki, S., and Lerman, R. I. (1991). Income stratification and income inequality. Review of Income and Wealth, 37(3),