Heterogeneous Effects of Multidimensional Housing Insecurity on Mental Health by Gender and Region

doi:10.21203/rs.3.rs-3231442/v1

Housing insecurity is a prevalent housing and urban problem worldwide, attracting considerable attention from scholars investigating its impact on mental health. However, little evidence exists that housing insecurity, in its multidimensional nature, affects depressive symptoms heterogeneously by gender and region. To fill research gaps, by analyzing the Korea Welfare Panel Study between 2007 and 2020, we examine three research questions: (1) how do the effects of multiple aspects of housing insecurity, indicated by a series of housing-related indicators, on mental health vary by gender? (2) how do the effects of the different aspects of housing insecurity on mental health vary by region? (3) do significant interaction effects between multiple aspects of housing insecurity on mental health exist? Our research demonstrates that regardless of gender or region, housing affordability stress (HAS) and housing quality negatively affect mental health. HAS has a greater impact on men’s mental health than on women, whereas housing quality has a more substantial impact on women. When men simultaneously experience a positive transition in housing tenure, the effect size of HAS is diminished. For women, the effect of HAS varies based on other facets of housing insecurity, such as overcrowding and housing quality. Additionally, we provide evidence that various aspects of housing insecurity affect mental health differentially based on residence location. These findings emphasize the significance of considering the role of housing insecurity in shaping mental health from multiple perspectives and the need for devising gender- and region-specific strategies to combat housing-related stress.

Housing insecurity

Mental health

Interaction effect

Hybrid model

Panel data

Housing insecurity is a global urban problem. Housing insecurity is often conceptualized as “the state of having difficulty acquiring housing, having minimal control over one’s housing, being at risk of losing housing, being uncertain about tenure, or living in housing that does not meet basic household needs” [1]. Housing insecurity, which encompasses a wide range of housing-related hardships, has been operationalized and measured variedly [2]. However, it typically refers to unaffordability, living in low-quality housing units, doubled-up, overcrowding, and residential instability manifested by forced and/or frequent moves [3]. Across the world, a growing number of households have experienced housing insecurity due to rapidly rising housing prices, stagnated wages, and insecure employment [4,5]. The literature highlights that housing insecurity negatively affects various dimensions of lives, including physical health [6], education outcomes [7], employment [8], and social networks [9].

Among the negative consequences of housing insecurity, one significant relationship is between housing insecurity and mental health [10]. Previous studies have shown that individuals’ mental health tends to deteriorate when they suffer from housing cost burdens [11], when they are monthly-paying renters, not homeowners [12], when they experience residential mobility [13], and when they reside in poor-quality housing [14]. Park & Seo found that these housing-related conditions accumulatively deteriorate mental health [15]. These findings emphasize that housing insecurity is a multifaceted state, and without understanding it, examining the effects of housing insecurity on mental health may provide an incomplete picture.

Previous studies have indicated that mental health manifests differently by gender [16]. Women are more susceptible than men to mental health problems, such as depression. This susceptibility arises due to the combination of biological differences and socio-environmental factors [17]. Typically, women react more strongly to stressors than men, even when they experience the same stressors [18]. Furthermore, women are more exposed to various stressful life events, such as poverty and sexual abuse, which contribute to their poorer mental health than men [19]. Consequently, numerous studies investigate the underlying mechanisms behind the observed differences in mental health between men and women.

Within this context, researchers have directed their attention toward understanding the influence of housing on gender differences in mental health. Firstly, disparities exist in how men and women perceive housing. Men tend to view housing primarily as a socioeconomic indicator, whereas women perceive housing as a sanctuary [20]. Women typically spend more time in domestic labor, including cleaning, cooking, and home maintenance leading to greater sensitivity toward housing conditions [21]. Regoeczi found that women are prone to experiencing higher levels of depression when living in overcrowded housing than men [22]. According to previous studies, housing can significantly contribute to the gender gap in mental health [15].

While housing has been recognized as a significant contributor to mental health disparities between men and women, only a handful of researchers have examined the gender-specific effects of housing insecurity on mental health. Bentley et al. found that the cumulative experience of housing cost burden has differing impacts on men’s and women’s mental health [16]. Lee found that functional problems, such as noise issues, are key risk factors for depression in women, while structural problems, like non-permanent buildings, are key risk factors for men [23]. Park & Seo examined housing insecurity across multiple dimensions and found that individuals experiencing simultaneous challenges in multiple dimensions of housing insecurity face further deterioration in their mental health [15]. Moreover, they emphasized that women, regardless of their initial mental health levels, are particularly vulnerable to housing insecurity-related mental health issues compared to men. However, most studies have yet focused on a single dimension of housing insecurity, such as housing affordability or housing quality, while neglecting the interactions among these dimensions, thus indicating a notable limitation.

Another significant mechanism that has garnered increasing attention from scholars is the influence of the place of residence on mental health [24]. South Korea (Korea) presents an ideal case study for examining distinctive mechanisms through which housing insecurity influences mental health across different regions. In Korea, over half of its total population and the majority of high-quality occupations are concentrated in the Capital region, encompassing Seoul, Incheon, and Gyeonggi province [25]. The concentration of population and employment within the Capital region has exacerbated housing challenges, such as escalating housing costs and substandard living conditions, making the region contrast to the rest of the country [26]. Thus, the socioeconomic imbalance between the Capital and non-Capital regions has become a national concern, making it a compelling study area to examine the mechanisms underlying regional disparities in mental health. Despite the importance, most studies have treated the place of residence as control variables in their analytical models, grouping them by population size [15], or simply comparing mental health in urban and rural areas [27].

To fill these research gaps, we examine the effects of multiple aspects of housing insecurity on mental health through the lens of gender and region based on individuals in Korea. Specifically, this study addresses three research questions: (1) how do the effects of multiple aspects of housing insecurity, indicated by a series of housing-related indicators, on mental health vary by gender? (2) how do the effects of the different aspects of housing insecurity on mental health vary by region? (3) do significant interaction effects between multiple aspects of housing insecurity on mental health exist?

To investigate the associations between housing insecurity and mental health, we analyzed the Korea Welfare Panel Study (KoWePS) data between 2007 and 2020. KoWePS has collected information by tracking nationally representative Korean households annually since 2006. KoWePS collects various information, particularly related to individuals’ residential environments and mental health outcomes. The analytic sample, after we cleaned missing and abnormal values for key variables used for our analyses, includes 153,326 observations from 19,396 individuals for 14 years.

Depressive symptoms are our primary outcome. We relied on the Center for Epidemiologic Studies Depression (CES-D 11) scale available to measure the severity of depressive feelings. The CES-D 11 consists of eleven items, including (1) “I lost my appetite,” (2) “I was relatively OK,” (3) “I was very depressed,” (4) “Everything seemed to be overwhelming,” (5) “I could not sleep well,” (6) “I was lonely and felt I was the only one in the world,” (7) “I did not have any complaint,” (8) “Everyone seemed to be mean to me,” (9) “I was sad,” (10) “Everyone seemed to hate me,” and (11) “I did not have the courage to do anything.” Respondents answered each item on a 4-point scale ranging from 0 (rarely or none of the time, less than a day) to 3 (Mostly, 5 ~ 7 days or more a week). Items 2 and 7 are reverse-coded since they indicate the opposite direction of depression. The severity of the symptom is indicated by a total score ranging from 0 to 33, with higher values reflecting greater severity.

Given our focus on the multidimensionality of housing insecurity, we measured it from six different dimensions: (1) housing affordability, (2) housing tenure, (3) residential mobility, (4) overcrowding, (5) poor housing facility, and (6) poor physical environments. First, we measured housing affordability by calculating the proportion of housing-related costs to household income. The total housing-related costs are derived from mortgage-related costs, other loans to pay security deposits, rent, and utility costs. Following previous studies, we categorized housing affordability problems into two types: when housing-related costs account for 30% or more but less than 50% of household income, it is considered housing affordability stress (HAS); and when the costs account for 50% or more of household income, it is severe HAS [28]. Second, we categorized housing tenure into three groups: monthly-paying renters, deposit-only renters (or Jeonse renters in Korean), and homeowners. Korea has a unique tenure system known as Jeonse, which allows tenants to pay a deposit equal to approximately 50–80% of the housing price upfront without paying monthly rent over a contract period. Since Jeonse is commonly regarded as a stepping stone to advance from monthly-paying renting to homeownership, we set Jeonse renters as the reference group, which helps capture the effects of upward/downward changes along with a housing tenure ladder on mental health outcomes. Third, we measured residential mobility based on whether they moved between KoWePS waves. Fourth, we operationalized overcrowding by measuring whether sample households did not meet the official minimum living space requirements per household member set by the Korean government (see the detailed requirements in Appendix 1). Fifth, we regarded poor housing facilities when the sample households’ housing units do not meet at least one of the following conditions: (1) independent water and sewer systems, (2) cooking facilities, (3) toilet facilities, and (4) bathing facilities with hot water. Sixth, we regarded poor physical environments as the housing units that do not meet any of the following criteria: (1) a permanent building made of durable, heat-resistant, fireproof, and moisture-controlled materials, (2) appropriate soundproofing, ventilation, lighting, and heating equipment, (3) suitable for residential use in terms of noise, vibration, odor, or air pollution, and (4) safe from natural disasters such as storms, floods, landslides, or land collapses.

Following previous studies, we included a set of independent variables [15, 29]. First, we categorized the location of residence into four types: (1) Seoul, (2) the Capital region surrounding Seoul (i.e., Gyeonggi and Incheon), (3) metropolitan cities in non-Capital regions, and (4) other areas. We also include variables related to gender, age, age-squared, education attainments, disability status, existing chronic health problems, marital status, the number of children, log-transformed household income, and employment, all of which may influence mental health.

To exploit the advantages of analyzing panel data, we employed a hybrid regression approach, a specific form of conventional fixed-effects models for panel data. The conventional fixed-effects approach cannot estimate the effects of time-invariant factors such as gender, whereas hybrid models combine the advantages of fixed- and random-effects models and simultaneously estimate the effects of time-varying and invariant factors [30]. In hybrid models, all time-varying variables are decomposed into two variables, between- and within-individual variations, and both variables are included in the models. Typically, the estimated coefficients derived from within-individual variations are similar to the estimates from fixed-effects models, whereas the coefficients for time-invariant variables typically reflect underlying contextual effects (see the example in Kang [31]).

Table 1 presents the descriptive statistics of key variables. We divided the sample into men and women. We conducted a comparative analysis to assess whether the difference in values between the groups was significant by conducting a series of t-tests for continuous variables and chi-square tests for categorical variables. The average CES-D 11 score is 3.129 and 4.353 for men and women, respectively, consistent with previous findings. While the housing cost-to-income ratio does not demonstrate any notable disparity, women tended to face higher rates of HAS and severe HAS compared to men. In comparison to women, men have a higher propensity for homeownership. Women often reside in housing situations characterized by lower-quality facilities, construction, and environments. However, they are less likely to experience overcrowding. In comparison with women, men tend to reside in the Capital region, including Seoul.

Table 1

Descriptive statistics of key variables for the period between 2007 and 2020.
Variable	All sample individuals				Men				Women				Significance of T-tests or chi-square tests
Variable	Mean	Std. dev.	Min.	Max.	Mean	Std. dev.	Min.	Max.	Mean	Std. dev.	Min.	Max.	Significance of T-tests or chi-square tests
Depressive symptoms (CES-D 11)	3.809	4.800	0	33	3.129	4.325	0	33	4.353	5.083	0	33	***
Housing affordability status
Housing cost to income ratio (HIR)	20.260	671.792	0	182973	19.752	711.320	0	182973	20.666	638.429	0.170	182973
No housing affordability stress experience (HIR < 30%)	0.893	0.309			0.910	0.287			0.881	0.324			***
Experiencing housing affordability stress (30% <= HIR < 50%)	0.075	0.264			0.063	0.243			0.085	0.279			***
Experiencing severe housing affordability stress (50% <= HIR)	0.031	0.174			0.027	0.163			0.035	0.183			***
Housing tenure (ref. Jeonse renter)
Homeowner	0.679	0.467			0.690	0.463			0.670	0.470			***
Monthly paying renter	0.181	0.385			0.175	0.380			0.186	0.389			***
Residential mobility
Experiencing residential mobility (yes = 1)	0.111	0.314			0.112	0.315			0.111	0.314
Overcrowding
Living in overcrowded housing (yes = 1)	0.026	0.159			0.027	0.162			0.025	0.157			*
Housing quality
Living in substandard facility housing (yes = 1)	0.048	0.213			0.047	0.211			0.049	0.215			+
Living in substandard construction and environmental housing (yes = 1)	0.160	0.367			0.154	0.361			0.164	0.371			***
Location of residence (ref. Seoul)
Capital region surrounding Seoul (i.e., Gyeonggi and Incheon)	0.219	0.414			0.225	0.418			0.214	0.410			***
Metropolitan cities in non-Capital regions	0.223	0.416			0.219	0.414			0.226	0.418			**
Other areas	0.394	0.489			0.393	0.488			0.395	0.489
Gender (ref. male)
Female	0.556	0.497
Age	53.132	18.132	15	112	52.284	17.630	15	101	53.811	18.496	15	112	***
Education attainments (ref. below elementary school)
Below high school	0.477	0.499			0.532	0.499			0.433	0.495			***
Over university degree	0.248	0.432			0.293	0.455			0.213	0.409			***
Disability status (ref. a respondent without disabilities)
A respondent with disabilities	0.095	0.293			0.117	0.321			0.077	0.266			***
Health status (ref. a respondent who has no chronic diseases)
A respondent who has chronic diseases	0.528	0.499			0.488	0.500			0.560	0.496			***
Marital status (ref. married)
Single	0.163	0.369			0.194	0.395			0.138	0.345			***
Widowed, divorced, or separated	0.190	0.392			0.092	0.290			0.267	0.443			***
Number of children within a household	0.349	0.734	0	5	0.367	0.750	0	5	0.335	0.720	0	5	***
Log(monthly household income)	5.640	0.850	-1.792	10.572	5.730	0.801	-1.792	10.572	5.568	0.881	-1.792	10.572	***
Employment status (ref. regular worker)
Temporary worker	0.200	0.400			0.188	0.391			0.210	0.407			***
Stable self-employed	0.070	0.256			0.122	0.328			0.029	0.167			***
Unstable self-employed	0.075	0.264			0.107	0.309			0.051	0.219			***
Unemployed or economically inactive	0.456	0.498			0.297	0.457			0.583	0.493			***
N (observations)	153,326				68,127				85,199
Note. To test the statistical significance of the difference between men and women, we conducted chi-squared tests for categorical variables and t-tests for continuous variables (p + < 0.1; p* < 0.05; p< 0.01; p* < 0.001).

To examine the association between depressive symptoms and multidimensional housing insecurity, we estimated a series of hybrid linear regression models (Model 1–3). Model 1 presents the results of analyzing the relationship between housing insecurity and depressive symptoms based on all observations. First, we analyzed the between-individual effects, which explain how differences in the variable between individuals, averaged across all periods, were associated with depressive symptoms. Results demonstrate consistent findings with previous research related to mental health; experiencing housing affordability stress (HAS; b = 1.087, severe HAS; b = 1.979), being homeowners (b = -0.269) or monthly paying renters (b = 0.515), living in overcrowded housing (b = 0.603), and living in housing with poor quality (substandard facility housing; b = 0.603, substandard construction and environmental housing; b = 1.324). In terms of location of residence, results indicate that individuals who reside in Seoul had higher CES-D 11 scores than those who reside in the Capital region surrounding Seoul (b = -0.210), metropolitan cities in non-Capital regions (b = -0.416), and other areas (b = -0.489).

We estimated within-individual effects, which refer to changes in depressive symptoms caused by changes in independent variables (from housing security to insecurity or vice versa). Results indicate that CES-D 11 scores significantly increase as individuals experience HAS, severe HAS (b = 0.164, b = 0.192 respectively), overcrowding (b = 0.285), or substandard construction and environmental housing (b = 0.554). CES-D 11 scores significantly decrease when individuals migrate from Seoul to the Capital region surrounding Seoul (b = -0.265) or other areas (b = -0.344).

Table 2

Hybrid regression models that explain CES-D 11 scores.
Variable		Model 1: Total				Model 2: Men				Model 3: Women
		Dependent variable: Depressive symptoms (CES-D 11)
		Between		Within		Between		Within		Between		Within
		Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.
Precarious housing condition variables	Housing affordability status (ref. no housing affordability stress)
	Experiencing housing affordability stress	1.087^***	0.149	0.164^***	0.043	0.820^***	0.214	0.209^**	0.064	1.201^***	0.207	0.139^*	0.059
	Experiencing severe housing affordability stress	1.979^***	0.217	0.192^**	0.065	1.051^**	0.317	-0.027	0.096	2.521^***	0.295	0.334^***	0.089
	Housing tenure (ref. Jeonse renter)
	Homeowner	-0.269^**	0.080	-0.019	0.049	-0.144	0.111	0.080	0.067	-0.398^***	0.112	-0.112	0.070
	Monthly paying renter	0.515^***	0.091	0.074	0.057	0.333^*	0.129	0.122	0.081	0.607^***	0.128	0.020	0.080
	Residential mobility (ref. no experiencing residential mobility)	0.238⁺	0.131	0.004	0.035	0.460^**	0.175	0.058	0.048	0.102	0.191	-0.036	0.049
	Overcrowding
	Living in overcrowded housing (ref. no living in overcrowded housing)	0.209	0.188	0.285^***	0.080	0.376	0.251	0.283^*	0.110	0.026	0.276	0.288^*	0.113
	Housing quality
	Living in substandard facility housing (ref. no living in substandard facility housing)	0.603^***	0.135	0.095	0.067	0.959^***	0.190	-0.205^*	0.094	0.368⁺	0.189	0.311^**	0.093
	Living in substandard construction and environmental housing (ref. no living in substandard construction and environmental housing)	1.324^***	0.103	0.554^***	0.031	1.065^***	0.141	0.459^***	0.044	1.496^***	0.148	0.620^***	0.044
Regional variables	Location of residence (ref. Seoul)
	Capital region surrounding Seoul (i.e., Gyeonggi and Incheon)	-0.210^**	0.070	-0.265^*	0.112	-0.369^***	0.096	-0.282⁺	0.156	-0.065	0.100	-0.237	0.158
	Metropolitan cities in non-Capital region	-0.416^***	0.070	-0.263⁺	0.149	-0.541^***	0.097	-0.524^*	0.211	-0.300^**	0.100	-0.060	0.208
	Other areas	-0.489^***	0.067	-0.344^*	0.133	-0.621^***	0.092	-0.512^**	0.187	-0.364^***	0.095	-0.222	0.187
Sociodemographic variables	Female dummy (ref. male)	0.452^***	0.047
	Age	-0.044^***	0.007			-0.002	0.010			-0.077^***	0.011
	Age square	0.000^***	0.000			0.000	0.000			0.001^***	0.000
	Education attainments (ref. below elementary school)
	High school	-0.358^***	0.065			-0.471^***	0.093			-0.272^**	0.094
	University degree or above	-0.404^***	0.084			-0.448^***	0.113			-0.419^**	0.127
	Disability status (ref. a respondent without disabilities)
	A respondent with disabilities	0.987^***	0.077			0.841^***	0.098			1.135^***	0.119
	Health status (ref. a respondent who has no chronic disease)
	A respondent who has a chronic disease	1.465^***	0.078	0.373^***	0.031	1.113^***	0.102	0.354^***	0.041	1.781^***	0.118	0.386^***	0.045
	Marital status (ref. married)
	Single	0.147	0.095	-0.032	0.100	0.152	0.129	0.188	0.132	0.021	0.144	-0.290⁺	0.150
	Widowed, divorced, or separated	1.110^***	0.067	0.078	0.073	1.458^***	0.114	0.811^***	0.121	0.848^***	0.090	-0.276^**	0.094
	Number of children within a household	0.116^**	0.041	-0.024	0.027	0.264^***	0.057	0.027	0.036	0.056	0.060	-0.052	0.039
Economic variables	Log(averaged monthly household income)	-1.052^***	0.042	-0.380^***	0.029	-0.980^***	0.062	-0.348^***	0.040	-1.066^***	0.058	-0.397^***	0.040
	Employment status (ref. regular worker)
	Temporary worker	0.012	0.094	0.010	0.047	0.170	0.123	0.089	0.062	-0.083	0.151	-0.081	0.070
	Stable self-employed	0.188	0.121	0.078	0.073	0.259^*	0.129	-0.036	0.085	0.239	0.278	0.281^*	0.131
	Unstable self-employed	-0.098	0.129	0.143^*	0.070	0.115	0.158	0.028	0.088	0.016	0.230	0.308^**	0.110
	Unemployed or economically inactive	0.703^***	0.082	0.302^***	0.049	1.371^***	0.118	0.677^***	0.068	0.376^**	0.129	0.062	0.071
Year dummy variables		Included				Included				Included
Intercept		10.208^***	0.376			8.740^***	0.538			11.578^***	0.531
Random coefficient		2.360				2.207				2.468
ICC		0.289				0.296				0.284
N (observations)		153,326				68,127				85,199
Note. p + < 0.1; p* < 0.05; p< 0.01; p* < 0.001

Model 2 and Model 3 present results that estimate the association between housing insecurity and depressive symptoms for men and women, respectively. Regarding between-individual effects, most findings align with those derived from Model 1 for both genders. However, for within-individual coefficients, results reveal that housing insecurity and location of residence have heterogeneous effects by gender. Among men, CES-D 11 scores increase when experiencing HAS (b = 0.209), living in overcrowded housing (b = 0.283), or residing in substandard construction and environmental housing (b = 0.459). Interestingly, their depressive scores significantly decrease when they experience substandard facilities housing (b = -0.205). Furthermore, CES-D 11 scores decrease when they move to metropolitan cities in non-Capital regions or other areas (b = -0.524 and b = -0.512, respectively). For women, CES-D 11 scores increase when they experience HAS or severe HAS (b = 0.139 and b = 0.334 respectively), overcrowding (b = 0.288), substandard facilities housing (b = 0.311), or substandard construction and environmental housing (b = 0.620). However, no statistically significant association exists between the place of residence and CES-D 11 scores.

To examine the cumulative effects between multidimensional housing insecurity, Model 4, Model 5, and Model 6 include several interaction terms between different aspects of housing insecurity. In Model 4, we focused on within-individual effects and found that compared with Jeonse renters, the effect of HAS is weaker for individuals who experience the transition from Jeonse renters to homeowners. Moreover, the results reveal that when individuals experience severe HAS, the effect of living in units with substandard construction and environments is stronger than the effect when they do not experience any HAS. Conversely, the effect of living in overcrowding declines when individuals experience severe HAS. Model 5 and Model 6 present the results of gender-stratified models with interaction terms. The findings indicate that housing insecurity manifests in distinct ways within the mental health mechanisms of men and women. In the case of men, the effect of HAS is comparatively weaker when they experience a transition in housing tenure (i.e., from Jeonse renters to homeowners) simultaneously. Conversely, for women, we observed that the effect of HAS varies depending on other aspects of housing insecurity, such as overcrowding and housing quality. Specifically, the combinations of HAS and overcrowding or severe HAS and substandard construction and environmental housing further dramatically increase CES-D 11 scores. However, the combination of experiencing severe HAS and overcrowding led to a decrease in CES-D 11 scores.

Table 3

Hybrid regression models that explain CES-D 11 scores with interaction terms.
Variable		Model 4: All sample individuals				Model 5: Men				Model 6: Women
		Dependent variable: Depressive symptoms (CES-D 11)
		Between		Within		Between		Within		Between		Within
		Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.
Precarious housing condition variables	Housing affordability status (ref. no housing affordability stress)
	Experiencing housing affordability stress	0.504	0.359	0.428^***	0.118	-0.002	0.523	0.682^***	0.189	0.756	0.493	0.303^*	0.153
	Experiencing severe housing affordability stress	1.683^**	0.505	0.230	0.189	0.150	0.799	0.399	0.310	2.395^***	0.658	0.138	0.241
	Housing tenure (ref. Jeonse renter)
	Homeowner	-0.249^**	0.087	0.008	0.050	-0.154	0.119	0.124⁺	0.069	-0.371^**	0.126	-0.099	0.071
	Monthly paying renter	0.275^**	0.104	0.085	0.059	0.138	0.143	0.134	0.083	0.346^*	0.150	0.035	0.083
	Residential mobility (ref. no experiencing residential mobility)	0.254⁺	0.130	0.003	0.035	0.490^**	0.175	0.056	0.048	0.103	0.191	-0.035	0.049
	Overcrowding
	Living in overcrowded housing (ref. no living in overcrowded housing)	0.223	0.206	0.279^**	0.085	0.417	0.274	0.315^**	0.117	-0.033	0.303	0.251^*	0.121
	Housing quality
	Living in substandard facility housing (ref. no living in substandard facility housing)	0.599^***	0.149	0.125⁺	0.070	0.921^***	0.208	-0.183⁺	0.097	0.388⁺	0.211	0.353^***	0.098
	Living in substandard construction and environmental housing (ref. no living in substandard construction and environmental housing)	1.236^***	0.112	0.520^***	0.033	0.910^***	0.151	0.449^***	0.046	1.468^***	0.163	0.569^***	0.046
Regional variables	Location of residence (ref. Seoul)
	Capital region surrounding Seoul (i.e., Gyeonggi and Incheon)	-0.216^**	0.070	-0.263^*	0.112	-0.374^***	0.096	-0.272⁺	0.156	-0.073	0.100	-0.239	0.158
	Metropolitan cities in non-Capital region	-0.414^***	0.070	-0.259⁺	0.149	-0.533^***	0.097	-0.502^*	0.211	-0.306^**	0.100	-0.065	0.208
	Other areas	-0.493^***	0.067	-0.341^*	0.133	-0.622^***	0.092	-0.495^**	0.187	-0.372^***	0.095	-0.228	0.187
Sociodemographic variables	Female dummy (ref. male)	0.459^***	0.047
	Age	-0.047^***	0.007			-0.003	0.010			-0.080^***	0.011
	Age square	0.000^***	0.000			0.000	0.000			0.001^***	0.000
	Education attainments (ref. below elementary school)
	High school	-0.350^***	0.065			-0.453^***	0.093			-0.265^**	0.094
	University degree or above	-0.400^***	0.083			-0.435^***	0.112			-0.414^**	0.127
	Disability status (ref. a respondent without disabilities)
	A respondent with disabilities	0.986^***	0.077			0.837^***	0.098			1.136^***	0.119
	Health status (ref. a respondent who has no chronic disease)
	A respondent who has a chronic disease	1.457^***	0.078	0.373^***	0.031	1.099^***	0.102	0.353^***	0.041	1.777^***	0.118	0.386^***	0.045
	Marital status (ref. married)
	Single	0.127	0.095	-0.035	0.100	0.131	0.129	0.181	0.132	0.001	0.144	-0.285⁺	0.150
	Widowed, divorced, or separated	1.081^***	0.067	0.077	0.073	1.409^***	0.115	0.815^***	0.121	0.829^***	0.090	-0.282^**	0.094
	Number of children within a household	0.122^**	0.041	-0.023	0.027	0.267^***	0.057	0.028	0.037	0.063	0.060	-0.051	0.039
Economic variables	Log(averaged monthly household income)	-1.066^***	0.042	-0.372^***	0.029	-0.995^***	0.062	-0.336^***	0.041	-1.083^***	0.058	-0.394^***	0.040
	Employment status (ref. regular worker)
	Temporary worker	0.025	0.094	0.009	0.047	0.194	0.122	0.088^**	0.062	-0.078	0.151	-0.081	0.070
	Stable self-employed	0.187	0.120	0.078	0.073	0.255^*	0.129	-0.039	0.085	0.269	0.278	0.286^*	0.131
	Unstable self-employed	-0.079	0.129	0.144^*	0.070	0.138	0.158	0.027	0.088	0.035	0.230	0.313^**	0.110
	Unemployed or economically inactive	0.689^***	0.082	0.300^***	0.049	1.356^***	0.118	0.675	0.068	0.365^**	0.128	0.061	0.071
Interaction terms	*Housing affordability stress Housing tenure**
	Housing affordability stress * Homeowner	-0.195	0.413	-0.359^**	0.128	0.200	0.601	-0.651^**	0.203	-0.360	0.568	-0.203	0.168
	Housing affordability stress * Monthly-paying renter	1.210^**	0.410	-0.259⁺	0.136	1.136⁺	0.608	-0.267	0.218	1.156^*	0.557	-0.271	0.176
	Severe housing affordability stress * Homeowner	-0.886	0.582	-0.180	0.199	-0.323	0.889	-0.493	0.322	-0.852	0.778	0.033	0.258
	Severe housing affordability stress * Monthly-paying renter	1.629^**	0.620	0.162	0.226	2.262^*	0.956	-0.238	0.372	1.388⁺	0.823	0.338	0.290
	*Housing affordability stress Overcrowding and housing quality**
	Housing affordability stress * Overcrowded housing	0.544	0.778	0.435⁺	0.228	-0.017	1.112	-0.123	0.322	1.116	1.092	0.845^**	0.320
	Housing affordability stress * Substandard facility housing	-0.355	0.588	-0.330⁺	0.183	-0.042	0.880	-0.330	0.277	-0.502	0.795	-0.367	0.246
	Housing affordability stress * Substandard construction and environment housing	0.410	0.408	0.185⁺	0.104	0.959⁺	0.581	0.144	0.161	0.057	0.569	0.201	0.136
	Severe housing affordability stress * Overcrowded housing	-1.151	1.202	-1.403^**	0.430	-2.118	1.621	-0.601	0.608	0.142	1.771	-2.037^**	0.606
	Severe housing affordability stress * Substandard facility housing	-0.845	0.952	0.056	0.300	-0.639	1.422	0.250	0.455	-1.041	1.286	-0.118	0.403
	Severe housing affordability stress * Substandard construction and environment housing	1.285^*	0.636	0.547^**	0.166	2.393^**	0.905	-0.013	0.252	0.463	0.891	0.863^***	0.220
Year dummy variables		Included				Included				Included
Intercept		10.422^***	0.379			8.950^***	0.541			11.820^***	0.535
Random coefficient		2.354				2.201				2.463
ICC		0.288				0.295				0.284
N (observations)		153,326				68,127				85,199
Note. p + < 0.1; p* < 0.05; p< 0.01; p* < 0.001

To examine varying relationships between housing insecurity and depressive symptoms across geographical regions, we estimated a series of hybrid models for each region (Model 7–10). The results of estimated between-individual effects derived from Model 7, Model 8, Model 9, and Model 10 are primarily consistent with those in existing research. Thus, we focused on introducing the results related to within-individual effects. CES-D 11 scores increase when individuals experience HAS or live in substandard construction and environmental housing in both Seoul (b = 0.364 and b = 0.422, respectively, in Model 7) and the Capital region surrounding Seoul (b = 0.328 and b = 0.448, respectively, in Model 8). The results of Model 9 and Model 10 indicate that in addition to HAS (b = 0.207, in Model 9, b = 0.193, in Model 10) and living in substandard construction and environmental housing (b = 0.498, in Model 9, b = 0.683, in Model 10), overcrowding works a significant stressor for those in non-Capital regions (b = 0.372, in Model 9, b = 0.551, in Model 10). Furthermore, individuals who experience the tenure transition from Jeonse renters to monthly paying renters in other areas experience higher CES-D 11 scores compared to stable Jeonse renters (b = 0.249).

Table 4

Region-specific models that explain CES-D 11 scores.
Variable		Model 7: Seoul				Model 8: Capital region surrounding Seoul (i.e., Gyeonggi and Incheon)				Model 9: Metropolitan cities in non-Capital region				Model 10: Other areas
		Dependent variable: Depressive symptoms (CES-D 11)
		Between		Within		Between		Within		Between		Within		Between		Within
		Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.	Coeff.	S.E.
Precarious housing condition variables	Housing affordability status (ref. no housing affordability stress)
	Experiencing housing affordability stress	0.938^**	0.347	0.075	0.102	0.481	0.304	0.086	0.090	1.248^***	0.304	0.207^*	0.095	1.661^***	0.248	0.193^**	0.072
	Experiencing severe housing affordability stress	2.000^***	0.450	0.364^*	0.153	2.148^***	0.442	0.328^*	0.138	2.329^***	0.475	0.065	0.148	1.762^***	0.373	0.058	0.103
	Housing tenure (ref. Jeonse renter)
	Homeowner	-0.337^*	0.161	0.207⁺	0.113	-0.234	0.148	-0.078	0.090	-0.148	0.166	0.025	0.115	-0.323^*	0.157	-0.034	0.095
	Monthly paying renter	0.104	0.187	-0.035	0.128	0.420^*	0.174	0.051	0.110	1.147^***	0.183	-0.022	0.124	0.284	0.182	0.249^*	0.112
	Residential mobility (ref. no experiencing residential mobility)	0.056	0.312	0.087	0.081	0.563^*	0.258	-0.064	0.067	-0.100	0.268	-0.019	0.076	0.228	0.215	-0.003	0.061
	Overcrowding
	Living in overcrowded housing (ref. no living in overcrowded housing)	0.427	0.348	0.271⁺	0.149	0.431	0.391	0.015	0.158	-0.684⁺	0.383	0.372^*	0.174	0.320	0.415	0.551^**	0.164
	Housing quality
	Living in substandard facility housing (ref. no living in substandard facility housing)	0.678	0.523	0.342	0.230	1.362^**	0.400	0.168	0.177	0.760^**	0.287	-0.008	0.166	0.505^**	0.177	0.087	0.085
	Living in substandard construction and environmental housing (ref. no living in substandard construction and environmental housing)	1.253^***	0.244	0.422^***	0.072	1.687^***	0.218	0.448^***	0.067	0.980^***	0.211	0.498^***	0.071	1.439^***	0.168	0.683^***	0.050
Sociodemographic variables	Female dummy (ref. male)	0.274^*	0.114			0.503^***	0.097			0.519^***	0.093			0.416^***	0.075
	Age	0.021	0.017			-0.043^**	0.015			-0.044^**	0.015			-0.064^***	0.011
	Age square	0.000	0.000			0.000^*	0.000			0.000^**	0.000			0.001^***	0.000
	Education attainments (ref. below elementary school)
	High school	-0.365^*	0.178			-0.559^***	0.143			-0.037	0.131			-0.426^***	0.099
	University degree or above	-0.491^*	0.211			-0.742^***	0.175			0.009	0.166			-0.476^***	0.133
	Disability status (ref. a respondent without disabilities)
	A respondent with disabilities	0.907^***	0.213			0.737^***	0.176			0.865^***	0.155			1.101^***	0.110
	Health status (ref. a respondent who has no chronic disease)
	A respondent who has a chronic disease	1.468^***	0.193	0.452^***	0.074	1.437^***	0.158	0.309^***	0.064	1.663^***	0.160	0.363^***	0.068	1.404^***	0.123	0.374^***	0.049
	Marital status (ref. married)
	Single	0.690^**	0.217	-0.250	0.245	-0.077	0.188	0.038	0.203	0.093	0.193	0.251	0.225	-0.023	0.159	-0.078	0.182
	Widowed, divorced, or separated	1.255^***	0.180	0.056	0.199	0.971^***	0.147	0.058	0.157	1.071^***	0.136	0.249	0.170	1.143^***	0.098	0.038	0.110
	Number of children within a household	0.175⁺	0.104	-0.050	0.072	-0.056	0.079	0.047	0.053	0.110	0.084	-0.076	0.062	0.171^**	0.065	-0.040	0.043
Economic variables	Log(averaged monthly household income)	-1.105^***	0.110	-0.482^***	0.073	-0.989^***	0.090	-0.261^***	0.060	-1.119^***	0.087	-0.585^***	0.066	-0.968^***	0.064	-0.290^***	0.045
	Employment status (ref. regular worker)
	Temporary worker	0.230	0.221	0.121	0.109	0.172	0.182	0.031	0.092	-0.336⁺	0.185	-0.014	0.103	0.043	0.159	-0.030	0.083
	Stable self-employed	0.276	0.326	0.104	0.188	0.217	0.257	0.201	0.154	0.193	0.252	0.168	0.166	0.140	0.180	-0.042	0.117
	Unstable self-employed	0.856⁺	0.482	0.557^**	0.204	0.393	0.309	0.249	0.155	-0.727^*	0.343	0.260	0.181	-0.295⁺	0.175	0.015	0.105
	Unemployed or economically inactive	1.118^***	0.197	0.275^*	0.115	0.874^***	0.163	0.254^*	0.098	0.457^**	0.164	0.366^**	0.108	0.633^***	0.135	0.335^***	0.085
Year dummy variables		Included				Included				Included				Included
Intercept		8.841^***	0.924			9.913^***	0.770			10.133^***	0.760			9.674^***	0.573
Random coefficient		2.538				2.421				2.195				2.296
ICC		0.320				0.307				0.247				0.285
N (observations)		25,107				33,593				34,189				60,437
Note. p + < 0.1; p* < 0.05; p< 0.01; p* < 0.001

Investigating the relationship between housing insecurity and mental health has gained increasing attention. Notably, as several studies have suggested that housing insecurity should be viewed from multiple dimensions rather than a single dimension [15], we examine the effects of housing insecurity on mental health from multiple perspectives in Korea. Although efforts to elucidate the relationship between multidimensional housing insecurity and depressive symptoms are increasing, few studies have examined the roles of gender and region in shaping depressive symptoms from multiple angles. We identify the heterogeneous effects of multidimensional housing insecurity, indicated by housing affordability, housing tenure, residential mobility, overcrowding, and housing quality, on mental health according to gender and location of residence.

Hybrid models yielded several noteworthy results. First, multidimensional housing insecurity is associated with worsened mental health in a manner that varies by gender and region. We confirm that HAS and housing quality consistently exhibit a negative impact on mental health across all models, consistent with previous studies [14]. In addition, HAS exerts a greater impact on men’s mental health compared to women, while housing quality has a more pronounced influence on women than on men. These gender-specific findings can be attributed to the tendency that women spend a longer time at home due to domestic labor [21], rendering them more sensitive to housing conditions. One interesting point is that the direction of substandard facilities’ effect on mental health differs between men and women. When housing conditions worsened in terms of the presence of substandard facilities, women show an increase in CES-D 11 scores, which is consistent with previous findings. However, in response to the same deterioration, men exhibit a decrease, even though the between-individual effect indicates the positive association between living in units with substandard facilities and CES-D 11 scores. One possible explanation for this counterintuitive finding is that men may experience a decline in CES-D 11 scores due to compromises made by living in substandard facility housing. To explore the potential compromises, we conducted a supplementary analysis to examine the differences between when men resided in units with substandard facilities and when they did not (see Appendix 2). The results revealed that as men live in units with substandard facilities, they tend to have lower rates of residential mobility and a reduced incidence of living with children, which may lead them to lower CES-D 11 scores. Furthermore, for men, a decrease in the level of depressive symptoms was observed when there was a change in the place of residence, particularly when moving from Seoul to metropolitan cities in non-Capital regions or other areas. This finding can be attributed to the increasing availability of affordable housing options in non-Capital regions, resulting in a reduction in housing-related stress.

Gender differences in the relationship between housing insecurity and mental health are further evident in the findings related to the interaction terms. For men, the transition from being a Jeonse renter to a homeowner was associated with a reduction in the effect size of HAS on CES-D 11 scores, while housing conditions such as overcrowding and housing quality had no significant effect on the size of the effect of HAS on depressive symptoms. This suggests that, for men, changes in housing tenure play a crucial role in the mechanism of mental health formation, while housing conditions exert a stronger influence on women. Regarding women, when severe HAS, and overcrowding coexist, CES-D 11 scores decrease. This result suggests that despite suffering from severe HAS and overcrowding, the positive effects provided by the neighborhood in which women reside (e.g., decent educational environments) may mitigate the level of depressive symptoms. The differential impact of housing insecurity could be attributed to the patriarchal society. Within Korean society, a prevailing perception exists that men should take responsibility for homeownership as a symbol of economic and residential stability, while many women tend to leave their jobs and focus on domestic labor after marriage. This unique social atmosphere may serve as a significant factor, with housing tenure playing a prominent role for men and housing conditions exerting a significant influence on women in the mechanism of mental health. These findings underscore the importance of considering gender-specific dynamics and the societal context when examining the relationship between housing insecurity and mental health.

Our study revealed that the mechanism of mental health also exhibits variations depending on the region. Individuals residing in Seoul or the Capital region surrounding Seoul demonstrated an increase in CES-D 11 scores when experiencing severe HAS. In contrast, individuals residing in metropolitan cities in non-Capital regions and other areas exhibited an increase in CES-D 11 scores when facing HAS. This can be attributed to the disparities within the housing market resulting from significant gaps between Seoul, the Capital region surrounding Seoul, metropolitan cities in non-Capital regions, and other areas. Higher housing prices in Seoul and the Capital region surrounding Seoul contribute to a larger proportion of individuals experiencing average levels of HAS. Therefore, individuals residing in the Capital region, including Seoul, tend to exhibit no significant changes in mental health when facing typical levels of HAS. However, it is only when they experience relatively high levels of HAS that depressive symptoms increase.

These differences between housing markets contribute to variations in the mechanisms of mental health formation, which are also reflected in housing tenure and conditions. Individuals residing in metropolitan cities in non-Capital regions exhibited higher levels of depressive symptoms when facing overcrowding, while individuals residing in other areas experienced worsened mental health when facing both overcrowding and changing from Jeonse renters to monthly paying renters. This can be understood as a consequence of individuals who do not reside in the Capital region and are relatively more accessible to becoming homeowners or Jeonse renters due to less tight housing markets, which allows them to purchase spacious, quality housing. In these relatively affordable market environments, downward movements in housing tenure and the deterioration of housing conditions contribute to the worsening of mental health.

This study has two implications for housing policy. First, in addition to reducing HAS and facilitating upward mobility in housing tenure, policymakers should consider the effect of stable housing on mental health from multiple angles. According to our findings, many aspects of housing insecurity negatively affect mental health, which can be followed by other problems, such as job insecurity [32], and lead individuals to a vicious cycle. Second, decentralized housing policies should be established based on the characteristics of each region. When compared to other regions, Seoul may be identified as having the highest prevalence of mental health problems. This result may be attributable to housing situations that vary between Seoul and other regions. According to M. Park et al. [33], while the demand for housing in Seoul and the metropolitan area is high, the supply is low, whereas the demand for housing in non-metropolitan areas is low, and the supply is high. As a result, residents of Seoul and the metropolitan area are gradually placed in substandard housing, which has a negative association with mental health. Moreover, it is anticipated that establishing housing policies tailored to the requirements of each region, taking into consideration the characteristics of each region, will have a positive effect on the mental health of residents.

DeLuca S, Rosen E. Housing Insecurity Among the Poor Today. Annu Rev Sociol Published Online First: 26 April. 2022. 10.1146/annurev-soc-090921-040646.
Cox R, Henwood B, Rice E et al. Roadmap to a Unified Measure of Housing Insecurity. 2017. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2817626.
Kang S. Severe and persistent housing instability: examining low-income households’ residential mobility trajectories in the United States. Hous Stud. 2021;1–26. 10.1080/02673037.2021.1982871.
Beer A, Bentley R, Baker E, et al. Neoliberalism, economic restructuring and policy change: Precarious housing and precarious employment in Australia. Urban Stud. 2016;53:1542–58. 10.1177/0042098015596922.
Wetzstein S. The global urban housing affordability crisis. Urban Stud. 2017;54:3159–77. 10.1177/0042098017711649.
Bhat AC, Almeida DM, Fenelon A, et al. A longitudinal analysis of the relationship between housing insecurity and physical health among midlife and aging adults in the United States. SSM - Popul Heal. 2022;18:101128. 10.1016/j.ssmph.2022.101128.
Holme JJ. Growing Up as Rents Rise: How Housing Affordability Impacts Children. Rev Educ Res. 2022;92:953–95. 10.3102/00346543221079416.
Desmond M, Gershenson C. Housing and Employment Insecurity among the Working Poor. Soc Probl. 2016;63:46–67.
Oishi S. The Psychology of Residential Mobility: Implications for the Self, Social Relationships, and Well-Being. Perspect Psychol Sci. 2010;5:5–21.
Singh A, Daniel L, Baker E, et al. Housing Disadvantage and Poor Mental Health: A Systematic Review. Am J Prev Med. 2019;57:262–72. 10.1016/j.amepre.2019.03.018.
Park GR, Seo BK. Revisiting the relationship among housing tenure, affordability and mental health: Do dwelling conditions matter? Heal Soc Care Community. 2020;28:2225–32. 10.1111/hsc.13035.
Park GR, Jung Y. Housing insecurity and health among people in South Korea: focusing on tenure and affordability. Public Health. 2019;171:116–22. 10.1016/j.puhe.2019.02.017.
Choi H, Oishi S. The psychology of residential mobility: a decade of progress. Curr Opin Psychol. 2020;32:72–5. 10.1016/j.copsyc.2019.07.008.
Seo BK, Park GR. Housing, living arrangements and mental health of young adults in independent living. Int J Environ Res Public Health. 2021;18. 10.3390/ijerph18105250.
Park G-R, Seo BK. Multidimensional housing insecurity and psychological health: how do gender and initial psychological health differentiate the association? Public Health. 2023;214:116–23. https://doi.org/10.1016/j.puhe.2022.11.014.
Bentley R, Baker E, Mason K. Cumulative exposure to poor housing affordability and its association with mental health in men and women. J Epidemiol Community Health 2012;66:761 LP – 766. doi:10.1136/jech-2011-200291.
Boughton S, Street H. Integrated review of the social and psychological gender differences in depression. Aust Psychol. 2007;42:187–97. 10.1080/00050060601139770.
Nolen-Hoekseman S. Gender differences in depression. Curr Dir Psychol Sci. 2001;10:173–6. 10.1007/978-981-10-6580-4_24.
Sloan DM, Sandt AR. Gender differences in depression. Women’s Heal. 2006;2:425–34. 10.1007/978-981-10-6580-4_24.
Mallett S. Understanding home: A critical review of the literature. Sociol Rev. 2004;52:62–89. 10.1111/j.1467-954x.2004.00442.x.
Kolpashnikova K, Kan MY. Gender gap in housework time: how much do individual resources actually matter? Soc Sci J. 2021;1–19. 10.1080/03623319.2021.1997079.
Regoeczi WC. Crowding in context: An examination of the differential responses of men and women to high-density living environments. J Health Soc Behav. 2008;49:254–68. 10.1177/002214650804900302.
Lee JH. Housing quality determinants of depression and suicide ideation by age and gender. Hous Stud. 2022;0:1–27. 10.1080/02673037.2022.2056151.
Green LW, Ottoson JM. Community and population health. 8th ed. Boston: McGraw-Hill; 1999.
OECD. Perspectives on Decentralisation and Rural-Urban Linkages in Korea. Paris:: OECD Publishing; 2021.
Kim HM, Han SS, Seoul. Cities. 2012;29:142–54. 10.1016/j.cities.2011.02.003.
Seo BK, Hwang IH, Sun Y, et al. Homeownership, Depression, and Life Satisfaction in China: The Gender and Urban-Rural Disparities. Int J Environ Res Public Health. 2022;19. 10.3390/ijerph192214833.
Aurand A, Emmanuel D, Threet D, et al. The Gap: A Shortage of Affordable Rental Homes. Washington, DC: National Low Income Housing Coalition; 2021.
Kang S. The Cumulative Relationship between Housing Instability and Mental Health: Findings from the Panel Study of Income Dynamics. J Soc Distress Homelessness. 2021;1–13. 10.1080/10530789.2021.1925038.
Allison PD. Fixed effects regression models. Sage: Publications.; 2009.
Kang S. Why Low-Income Households Become Unstably Housed: Evidence From the Panel Study of Income Dynamics. Hous Policy Debate. 2019;29:1–29.
Frijters P, Johnston DW, Shields MA. The Effect of Mental Health on Employment: Evidence from Australian Panel Data. Health Econ. 2014;23:1058–71. 10.1002/hec.3083.
Park M, Lee H, Cho J et al. Role and Responsibility of Central and Local Government for Enhancement of Housing Right. Sejong: 2019.

AppendixJUH.docx

Heterogeneous Effects of Multidimensional Housing Insecurity on Mental Health by Gender and Region

Status:

Version 1

Abstract

Main text

Data and Methods

Results

Discussion And Conclusion

References

Supplementary Files

Status:

Version 1