Housing and happiness : Subjective well-being and residential environment in Korea

1. Tenure

In general, studies show that homeowners are more satisfied with their housing situation than tenants (e.g., Diaz-Serrano 2009, Elsinga & Hoekstra 2005). Tenure can be expected to influence SWB primarily through personal control.

Homeownership is widely associated with higher levels of personal control (Stone 1993). Under existing law, the property rights associated with homeownership typically allow more freedom to alter, use, alienate, destroy, exclude, and exchange than those rights associated with renters (Kreuckeberg 1995, Stone 1993). This control is reflected in the greater likelihood that homeowners will invest in maintaining and improving their homes, will be involved in local affairs, and will have higher environmental consciousness (Dietz & Haurin 2003).

Due to the large transaction costs associated with selling and buying homes, homeowners tend to be less mobile than renters (Dietz & Haurin 2003). This increased stability is associated with a number of positive impacts on SWB. Dietz and Haurin (2003), for example, cite a number of studies that associate homeownership with improved physical and mental health. Studies also suggest that housing stability beneficially impacts children, particularly in educational attainment (Aaronson 2000, Cattaneo et al. 2009).

2. Size

Increased floor area is also generally associated with increased SWB measures. Studies date back at least to Mitchell (1971), who found that the amount of space available to residents of high-density housing in Hong Kong was positively related to satisfaction. In a review of the literature, Bratt (2002) drew on overcrowding studies to suggest that reduced space diminishes well-being. More recently, Galiani et al. (2015) found similar impacts in Latin America.

The quantity of housing floor area could express itself through each of the four categories. Larger living spaces could offer more privacy for each individual and thus a stronger sense of personal control. The amount of space could also contribute to one's self-esteem, as larger homes often reflect social status. By supplying sufficient space for social activities, larger housing units could also make it easier to foster the social relationships that provide generalized social support and facilitate participation in valued lifestyles.

3. Housing type

The third variable this paper focuses on is housing type, which can also be expressed through any of the four channels. For example, living in an independent house may convey a greater sense of personal control over one's property than living in an apartment complex, since in the former case one has more rights to alter the exterior.

A housing type can also shape one's sense of identity and self-esteem, as housing is an essential component of many individuals' ontological identity (Saunders 1990). Cooper Marcus (2006), for example, has claimed that houses serve as a “mirror of the self” for many individuals who seek to symbolically manifest their personal characteristics in the way they decorate their homes. Others make similar claims. Karsten (2007), for example, claims that urban middle class families' actively choose housing to reflect the identities they want to project.

To a lesser extent, one's type of housing may facilitate social support by increasing the proximity of family and friends, either by providing living space for them or offering spatial proximity through density.

Finally, to the extent that a housing type is associated with a valued lifestyle, like suburbs in the US, it can also contribute to social inclusion, which is the logic underlying housing vouchers for low-income households in the US.

However, due to evolving cultural preference, the type of housing that represents a valued lifestyle is not clear. For many years, high-rise housing was perceived as dangerous and undesirable in Western countries, where it is often associated with low-income social housing (Coleman 1985, Spicker 1987, Newman 1972). Only in recent years has high-rise living become more attractive (e.g., Baxter & Lees 2009). But while suburbia has endured as the presumed norm in the West (Richards 1990), preferences differ in the East in general and in Korea in particular, where rapid urbanization drove high-rise construction (Lee & Hong 1991). Though Koreans were not initially receptive to apartment living (Lee 1971), this has changed dramatically and apartments have come to represent the aspirations of middle-class Asians, including Koreans (Doling 1999, Park & Ferrari 2011). Thus, this paper hypothesizes that living in apartments will be correlated with higher SWB.

4. Applicability of SWB measures to policy making

The inclusion of SWB measures into public policy-making is increasingly being called for these days (IADB 2008, Stiglitz et al. 2009, European Commission 2009, OECD 2015). Being established measures in psychology and sociology, SWB measures have become increasingly prevalent in empirical economics over the past two decades. The measures have provided additional insights into the well-being effects of unemployment, absolute and relative income, health shocks, and changes in marital status. Many studies have further confirmed the validity of SWB measures and their use in statistical models when suitable techniques are applied (Sandvik et al. 1993, Diener 1994, Lepper 1998, Beegle et al. 2012, Clark & Georgellis 2013). SWB has also been identified as a useful goal criterion from a utilitarian standpoint (Layard 2006, Veenhoven 2004).

There are, however, some studies that dismiss the use of SWB as a measure of government policy, primarily from conservative think tanks like the Cato Institute. The critiques adhere primarily to questioning the empirical value of SWB. Wilkinson (2007), for instance, argues that SWB represents a cultural moving target rather than a simple empirical phenomenon. Johns and Ormerod (2007) similarly argue that happiness measures are an insensitive measure of welfare over time because people adapt to their conditions, resulting in such measures containing little information. However, two counterarguments are available. First, in the context of the fixed effects regressions here, such personal adaptability is controlled for and effects are determined through substantive changes in people’s lives. Second, even if SWB represents a moving target, this can function like a relative poverty measure that continually orients government policy to substantive improvements in quality of life (Iceland 2006).

The other objection is that government targeting of SWB measures restricts freedom of choice. Frey (2011), for instance, argues social planners will manipulate SWB indicators, making it impossible to maximize policy on such an indicator. Johns and Ormerod (2007) extend this argument to claim that using such target measures to set policy restricts free choice. However, adopting any measure, including GDP, restricts some choices in favor of others (Harvey 2003, Sen 1999). As trends in the measurement of well-being are moving away from one-dimensional monetary measures and towards multidimensional indicators, overall satisfaction with one’s life serves as an aggregate measure of a multidimensional evaluation of an individual’s well-being. Therefore, SWB measures represent useful complements to traditional measures to inform policy makers.

1. Data and variables

Annual data for our analysis comes from the Korean Labor and Income Panel Study (KLIPS) for the years 2000 to 2012. KLIPS is a nationally representative longitudinal study of urban Korean households, modeled after the US National Longitudinal Surveys (NLS) and Panel Study of Income Dynamics (PSID). It is conducted annually by the Korea Labor Institute, a government- sponsored research institute. The study started in 1998 with 5,000 households and 13,783 individuals aged 15 years or older. KLIPS collects a wide range of information on individuals and the households in which they reside. In addition, it offers broad information on various SWB indicators. The data quality KLIPS provides satisfies the highest international standards. The panel maintains 76.5% of the original sample throughout all waves, which is comparable to the PSID (78%); the German Socio-economic Panel (GSOEP, 79%); and the British Household Panel Survey (BHPS, 77%). Kang (2010) shows that potential bias produced by attrition is negligible in KLIPS data.

For this study we use the years 2000 to 2012. We exclude 1998 and 1999 since household asset information has only been comprehensively collected since 2000. This yields a total number of 20,400 individuals and 135,729 person-year observations with complete information on the variables needed for the following analysis.

In KLIPS’s individual questionnaire, the question on overall life satisfaction is preceded by a set of detailed questions on satisfaction with different aspects of life: household income, leisure life, housing environment, family relations, relations with relatives, and social relations. The exact wording of the overall life satisfaction question is then: “Overall, how satisfied or dissatisfied are you with your life?” Individuals are asked to respond according to a scale ranging from 1 (“very satisfied”) to 5 (“very dissatisfied”). For the sake of easier interpretation, we recoded the scale so that higher numbers correspond to higher levels of satisfaction.

Table 1 presents the distribution of life satisfaction responses across the sample. In fact, it is rather a 3-point scale than a 5-point scale given that the frequencies with which Koreans tick the boundary categories ‘1’ and ‘5’ are minuscule. Out of ten Koreans, approximately one reports to be dissatisfied, three to be satisfied, and six to be neither satisfied nor dissatisfied with their lives. Mean satisfaction is 3.26 and the standard deviation is 0.639.

Table 1.

Distribution of Life Satisfaction

Drawing from KLIPS’s rich household questionnaire, three key housing-related variables are employed in our analysis. First, the tenure status of the household is recorded. Households are grouped into the following four categories: own house; rent (deposit only, ‘jeon-se’); rent (monthly, ‘wol-se’); and other contract. Note that ‘ban-jeon-se’ (wol-se with large deposits) is counted as wol-se but constitutes only a small fraction of households during this time period. Second, the type of residence groups households into one of the following five categories: apartment complex; detached, single-family house; tenements (>660m² and >4 floors); multifamily buildings (<660m² and <5 floors); and other. Third, the size of usable living space in square meters. These have been divided into categories that reflect common sizing in Korea.

Table 2 shows average life satisfaction by type of housing. Individuals living in apartment complexes report the highest levels of satisfaction. In addition, ownership and jeon-se seem to be related to higher life satisfaction. Least happy are people living in detached, single-family houses on a wol-se basis. While the matrix in Table 2 already sheds light on the relationship between housing and happiness in Korea, housing differences might just be a proxy for wealth differences of households. Thus, in the following it will be crucial to control for the net asset position of the household to see whether housing is related to happiness beyond the wealth channel, which is generally overlooked in related studies.

Table 2.

Average life satisfaction by type of housing

2. Methodology

We use a standard empirical strategy to isolate the determinants of life satisfaction. This approach can be considered a mixture of stated and revealed preferences. Contrary to many previous housing studies that employ hedonic pricing models to determine the relative valuation of features of housing from the price paid, the strategy employed here seeks to identify the average contribution of a given housing characteristic on individual life satisfaction, controlling for an array of standard factors affecting individual well-being, such as income, wealth, and marital status. This stated or self-reported level of life satisfaction functions in lieu of prices in the regressions to reveal those characteristics of housing that lead to SWB.

Having multiple waves per individual at our disposal, we can exploit the dynamic nature of the KLIPS panel data set in order to estimate the relationship between housing and life satisfaction. Subjective well-being measures are known to have a strong time-persistent component. These individual fixed effects can be largely attributed to time-invariant differences in personality traits (Lykken & Tellegen, 1996), but also measurement error in the behavior of reporting household wealth. Since not controlling for these differences can lead to biased results, it is essential to use fixed effects estimators when working with SWB measures. Linear fixed effects (FE) estimators are commonly used in the SWB literature and yield consistent results (Ferrer-i-Carbonell & Frijters 2004, Rudolf 2014). In order to see the importance of panel estimates, we will compare them to least squares estimates after pooling all waves.

Life satisfaction of individual i in year t is denoted υ_it, and we use the following basic empirical model to describe how it is determined:

O_it is a vector of ownership/ tenancy- related dummies, R_it a vector controlling for type of residence, and A_it a vector of size dummies. For comparison, this vector is replaced by continuous variables for size in regressions 2, 4, and 6. Ln(sqm) and ln(sqm)² are employed to ensure that the resulting coefficients are large enough to be meaningful. Other individual or household characteristics are grouped in vector X_it, which includes standard controls such as variables related to age, marital status, number of children, individual earnings, household income and wealth, province of residence. (For a detailed description of all variables used in this study, see Table A1 in the appendix.) Finally, υ_t represents time effects. Whether we can estimate equation (1) with least squares depends on the assumptions we make about the error term. Certainly, ordinary least squares (OLS) estimation requires an i.i.d. error term u_it.

There is, however, good reason to believe that the error term of SWB models is non-random, rather that u_it consists of an unobserved individual fixed effect, μ_i, and a random component, ϵ_it.

Intuitively, μ_it corresponds to individual personality traits that vary across people and are largely persistent over time. Since personality traits are obviously related to many forms of behavior, they are usually correlated with the observed characteristics that enter the model in equation (1) on the right side. For example, it has been found that happier people are more likely to get married and also be more successful on the job. Therefore, unbiased estimation of SWB models requires a fixed effects estimator to take care of individual fixed effects.

1. Type of residence

Both OLS and FE results suggests that the type of residence matters for happiness. Ceteris paribus, living in an apartment complex is related to a significantly higher level of life satisfaction as compared to all other types of residence. Following our preferred specification in column 5 of Table 4, moving from a detached, single-family house into an apartment increases life satisfaction by 0.0631 or 10 percent of a standard deviation in life satisfaction. Note that all results control for household income and total household net assets. Thus, the preference of apartments is not due to asset effects, but might instead be related to self-esteem, inclusion, or social support, which might be more favorable in the case of apartments.

It is interesting to note that the happiness gap between different types of housing reduced from OLS to FE estimates. This usually indicates that unobserved heterogeneity plays a role. This heterogeneity might be due to measurement error in the variable total household assets, i.e., richer household tend to underreport their true level of wealth.

Among the other variables included in the model, happiness is further positively related to higher household wealth and incomes and to a higher individual contribution to household income through one’s labor earnings.

2. Size of residence

Size effects are added to the models in columns 2 to 6. In column 2 and 4, residence size is measured in terms of ln(sqm) with a squared term ln(sqm)². The inclusion of size effects does slightly lower the absolute magnitude of the coefficients for type of residence. However, it does not change their highly significant effects. The coefficient for ln(sqm) is about twice as high in OLS models compared to FE models. Again, this might indicate tendencies of richer households to underreport total household wealth. The positive ln(sqm) and negative ln(sqm)² coefficients in column 4 of Table 4 indicate that SWB is positive with diminishing returns and a peak value, implying a threshold effect after which increasing size reduces SWB gains. Similar results are observed in column 6 for the effect of ln(sqm per capita). Instead of using ln(sqm), columns 3 and 5 use a set of size dummies to account for nonlinear effects and to mitigate potential bias caused by outliers. Again, we find that size matters: doubling size from 60-85m² to 135-164m² leads to increased life satisfaction by 0.0878 or approximately 14 percent of a standard deviation in life satisfaction. It is important to note that the greatest increase in SWB occurs in the move to residences from those smaller than 40m² and that satisfaction peaks in apartments between 135m² and 164m², again indicating a threshold effect. Other forms of housing, like self-constructed housing and SROs are associated with decreases in SWB.

The results for household size and number of children reinforce these findings. Since the regressions control for household size, the weak positive results for number of children captures the satisfaction of having children in general and abstracts from the space they consume. Children’s impact on space is incorporated in the household size variable. And indeed, throughout the regression models increases in household size are associated with lower levels of satisfaction. However, the sharp drop in the size of this coefficient in model 6, which measures the effect of space per person, suggest that the coefficients in other models reflect household crowding.

3. Ownership/Tenancy1

Finally, the effect of ownership/tenancy is included in the models presented in columns 4 through 6. Living in one’s own house enhances life satisfaction the most, even if total household wealth and household size are controlled for. (Note that all housing-related assets and liabilities are included in the household’s net asset position.) Jeon-se follows in second place, and wol-se in third. Note that after the inclusion of ownership/tenancy controls, both type of residence and size effects have slightly dropped, indicating for example that ownership is positively correlated with size of housing. Moving from monthly rent (wol-se) to ownership increases life satisfaction by an average of 0.0827 units, or 13 percent of a standard deviation in life satisfaction.