This study draws on data from the 2022 and 2023 waves of the Korean Labor and Income Panel Study (KLIPS). Conducted by the Korea Labor Institute, a government-funded research organization, KLIPS is a nationally representative longitudinal survey that has been collecting data since 1998. KLIPS tracks the same households and individuals annually, periodically adding refreshment samples (most recently in 2009 and 2018) to compensate for attrition and preserve representativeness. The survey provides detailed information on the socioeconomic characteristics of urban Korean households and their members, including income, housing tenure and type, educational attainment, and employment status. The survey employed a systematic sampling framework and appropriate weights were applied to account for the survey design and ensure national representativeness.

The KLIPS core questionnaire remained largely consistent across the waves, whereas supplemental modules covering different thematic areas were administered annually. In 2023, an additional health module was included to collect information on physical activity, self-rated health, and health service utilization. This study utilized this supplemental health survey to examine the relationship between housing affordability and mental health as measured by levels of stress and depressive symptoms. To strengthen the validity of the causal interpretations, we used the 2022 data points for independent variables and 2023 data for outcome variables. A more detailed rationale for this approach is provided in Sect. "Analytic Strategy" below.

The analytical sample was restricted to private renters. Renters differ from homeowners in terms of housing cost structures; homeowners often pay mortgage interest and property taxes, whereas renters pay regular rent. Moreover, renters face unique stressors, such as lease renewal uncertainty and exposure to rent fluctuations, which may increase their sensitivity to stress and depressive symptoms [31]. Given these distinctions, this study exclusively focused on renters.

To assess mental health outcomes, we used two self-reported indicators: perceived stress and depressive symptoms. In the KLIPS, the head of household was asked to evaluate their psychological condition using a 10-point Likert scale. Stress was measured using the question, "Over the past two weeks, how much stress have you experienced in your daily life?" The response scale ranged from 1 ("Very low") to 10 ("Very high"), with higher scores indicating greater levels of perceived stress. Depressive symptoms were assessed using the question, "Over the past two weeks, how often have you felt sad or depressed in your daily life?" This item also used a 10-point Likert scale, where 1 indicates "Not at all depressed" and 10 indicates "Always depressed."

Although the mental health module in the KLIPS did not include a comprehensive diagnostic instrument such as the Depression Anxiety Stress Scales-21 (DASS-21), it provides a rare opportunity to examine self-reported mental health alongside both objective and subjective measures of housing affordability within a nationally representative dataset.

Housing affordability was measured using objective and subjective indicators. As an objective measure, we calculated the ratio of monthly rent to reported monthly household income. Consistent with prior studies on housing affordability [3, 20], households spending 30% or more of their income on rent were coded as 1 (objectively unaffordable), and those below this threshold were coded as 0 (objectively affordable). For the subjective measure, we used a self-report item asking respondents whether they perceived their housing costs as burdensome. To reduce confusion, we referred to this variable as subjective unaffordability: respondents who reported feeling burdened were coded as 1 (subjectively unaffordable), whereas all others were coded as 0 (subjectively affordable).Footnote 1 Using these two variables, we constructed an interaction term to serve as a key independent variable in our model. This allowed us to examine whether subjective affordability moderates the relationship between objective affordability and mental health outcomes and whether such moderation differs across income groups.

Various socioeconomic characteristics were included as covariates in the regression models. These variables were selected based on prior research and theoretical considerations regarding potential confounding factors in the relationship between housing affordability and mental health [24, 29].Footnote 2

First, a binary variable was constructed for residences in the Seoul metropolitan area, which included Seoul, Incheon, and Gyeonggi Province. Households located in this region were coded as 1, and those outside were coded as 0. The Seoul metropolitan area is home to more than half of South Korea's population and employs approximately 50% of its workforce [32]. Prior studies have shown that living in this region is associated with distinct patterns of housing affordability, stress, and life satisfaction compared to living in non-metropolitan areas [33].

In addition, several household- and housing-level characteristics were adjusted. Educational attainment was coded as 1 if the head of the household had completed education beyond high school and 0 otherwise. Age was included as a continuous variable. Marital status was coded as 1 if the respondent was married, and 0 otherwise. Sex was coded as 1 for female and 0 for male. We also included the number of children and total number of household members as continuous covariates. Finally, housing type was included as a binary variable, coded as 1 if the household resided in an apartment, the predominant housing type in South Korea, and 0 otherwise [34].

Figure 1 presents the variables discussed above and their relationships within the conceptual model of the study. It further illustrates the framework guiding the analysis and highlights the moderating role of subjective housing unaffordability in the association between objective housing costs and mental health outcomes.

Conceptual model of housing affordability and mental health

To examine the potential moderating effect of subjective affordability on mental health outcomes, we included an interaction term between objective and subjective housing affordability. We used temporally lagged independent and dependent variables; mental health outcomes (stress and depressive symptoms) were drawn from the 2023 wave of the KLIPS, while all independent variables, including housing affordability measures and covariates, were taken from the 2022 wave to support a more rigorous temporal alignment of variables. In other words, we established a temporal ordering that helps to mitigate concerns about reverse causation and aligns with prior literature suggesting that housing conditions exert lagged effects on mental health [21]. This approach may help address the concerns of reverse causality, namely, the difficulty of disentangling whether unaffordable housing leads to higher levels of stress or depressive symptoms, or whether individuals experiencing such mental health challenges have a higher likelihood of perceiving their housing as unaffordable or encounter financial strain, although it does not fully eliminate these concerns. One limitation of this approach is that it does not capture changes in housing conditions that may have occurred between the two survey waves. To address this, we restricted our main analytical sample to individuals who did not relocate between 2022 and 2023, thereby increasing the consistency of housing-related exposure across the study period.

Next, we conducted subgroup analyses based on the income strata. Subgroup analyses (Tables 4 and 5) involved stratified models estimated separately for the bottom 40% and top 60% income groups. Rental unaffordability tends to disproportionately influence low-income households, which are also the primary targets of housing affordability policy interventions [17]. Prior research suggests that the relationship between affordability and mental health may vary across income levels [35]. Considering this, we divided the sample into the bottom 40% and top 60% of the household income distribution and compared these subgroup results with those from the full sample.

Although many studies define the objective affordability measure using the conventional 30% RIR threshold, this cutoff is widely acknowledged as a rule of thumb rather than a definitive standard. Thus, to test the robustness of our findings, we re-estimated the models using alternative thresholds of 25%, 35%, and 50% to define objective unaffordability.

All the data used in the regression analyses were weighted to reflect the complex sampling design and ensure generalizability to the national population. Descriptive statistics including means and frequencies were calculated using unweighted data to preserve the observed distribution characteristics of the sample.Footnote 3 All analyses were conducted using R statistical software (version 4.4.3).