## Abstract

Rapid demographic changes have occurred in Korea, with the number of one-person households almost doubling between 2000 and 2010 in the Seoul metropolitan region. Developed countries experienced these changes previously through the so-called second demographic transition. The purpose of this article is to ascertain how both the socioeconomic attributes and the location characteristics of one-person households at the time of their formation affect the durations of these households under the rapidly changing Korean demography. The spatial distribution of the areal location quotient indexes for one-person households indicated that the concentration of these households is relatively higher in the inner cities of metropolitan areas and the outskirts of the Seoul metropolitan region. Meanwhile, the distribution patterns of the relative concentration levels for one-person households by age group exhibited obvious differences. In the survival analysis for the entire sample of this research, household attributes were the primary determinants. However, the results of the empirical analyses by age group indicated that location characteristics were significant as well, although the significance of the variables varied with the types of one-person households. The duration of households of one person under 40 years old was affected by their access to employment districts and the concentration level of one-person households in the area. In contrast, the duration of households of one person 65 years old and older was influenced by the distribution of affordable housing. The findings of this study provide a framework that is able to make sense of the changing characteristics of the one-person households of nations in transition from developing to developed countries.

## Introduction

Changes in social, economic, and regional characteristics lead to differences in the demographic structures of regions. A fundamental solution by the Korean government to manage the overpopulation that arose from industrialization and economic growth was a birth control program from the 1960s to the mid-1990s: the Korean family planning program. After 40 years, the outcome of the program was an exacerbation of demographic issues, such as low fertility and aging. Growth in the number of one-person households—the number of such households almost doubled between 2000 and 2010 in the Seoul metropolitan region—and the pace of aging in Korea exceed those of developed countries, including Japan (Chae et al. 2014; Kim et al. 2013; Yi and Lee 2014). One of the most remarkable characteristics of household demography in Korea has been a recent reduction in household size. This phenomenon stems from a continuous decline in marital fertility and an increase in small and nontraditional households, such as one-person households and single persons with children; these trends not only share many similarities with the second demographic transition in developed countries, but they have also emerged in developing nations (Buzar et al. 2005; Schmid 1988).

Furthermore, the amount of experience living alone differs from person to person because the reasons for forming a one-person household vary according to each person’s life course (Qu and de Vaus 2011; Schmid 1988). One of the representative reasons for forming a one-person household is to establish an independent household. Other direct causes of living alone include postponing marriage, the death of a spouse, and the collapse of relationships with household members through divorce or separation from a cohabiting relationship (Buzar et al. 2005; Wulff 2007). Although these reasons themselves are not recognized as a social problem, they have varied outcomes, such as an increase in the rate of unmarried and single elderly households, which are considered demographic, social, and economic problems in terms of the expansion of vulnerable social classes in Korea. These problems are accepted as key considerations in spatial policies, such as regarding urban planning, transportation planning, and housing supply; and they are relevant not only to the number of one-person households formed but also to their continuation or duration (Buzar et al. 2005; Ogden and Hall 2004; Wulff 2007). Thus, an analysis of the characteristics and changing patterns of one-person households is important for understanding the socioeconomic changes in contemporary urban society.

One-person households are generally concentrated in specific areas (Hall et al. 1997; Shin 2010; Yi and Lee 2010), and the phenomenon has continually intensified until recently in the metropolitan regions of Korea (Chae et al. 2014). In the early nineteenth century, urban ecologists from the Chicago School argued about the temporal changes to the urban spatial patterns caused by residential locations and the relocation of inhabitants and immigrants. One of the main findings was a segregation of residential areas according to a series of processes, including invasion, occupation, and succession, which appeared among the socioeconomic classes in a city. Moreover, Simmons (1968:650) stated that “households with similar social characteristics but different life styles prefer widely different housing and neighborhood conditions.” These preceding studies argued that the residential concentrations of specific classes are connected to location characteristics and socioeconomic attributes. Therefore, studies on the maintenance periods of households also need to consider the effects of both location characteristics and socioeconomic attributes.

In this context, the purpose of this article is to ascertain how both socioeconomic attributes and location characteristics at the formation junctures of one-person households affect the duration between their creation and extinction given the rapidly changing Korean demography. This article begins with a review of related theories—namely, the second demographic transition, life cycles, and life courses—and literature on the transition between one-person households and other household types, and it examines the aspect of change in and the distribution of the share of one-person households in the Seoul metropolitan region. Next, empirical studies that analyzed the effects of socioeconomic and location factors on the durations of one-person households were conducted using Cox’s proportional hazard regression model. To overcome the limitations from the cross-sectional data used in the previous studies, both census data and time-series labor panel data were used in the analyses, and accessibility and location quotients by one-person household type were calculated to represent location attraction and the opportunity for social relationships. Finally, this study summarizes the outcomes of the empirical analyses and provides the meaning and significance of the results for a better understanding of the rapidly changing Korean demography. The results of this study could contribute to decision making for the demographic and planning policies of countries in transition from developing to developed, such as Korea.

## Literature Review

Since the mid-1990s, most developed countries have witnessed a rapid transformation of household structures. This phenomenon is generally defined as “the second demographic transition,” which is different from the contemporary patterns of household demography in the developed world (Buzar et al. 2005; Lesthaeghe 1995; Ogden and Hall 2004). The deepening of the second demographic transition, as several researchers have argued (Burch and Matthews 1987; Kuijsten 1995, 1996), includes the following: a decline of the birth rate to lower than the replacement rate; postponing or avoiding marriage; delaying childbearing; a rapid increase in divorce; and the rise of small households, such as one-person households, two-person households, and single-parent families. A growing search for individual standing has also been referred to as one of the main factors related to the second demographic transition (Schmid 1988). These behavioral changes led to a decline in household sizes that has continued over the past few decades in line with the expansion of the gap between the population and the number of households—that is, a decrease in the population growth rate and a rapid increase in the number of households—and the deepening complexity and diversity of individual living arrangements (Burch and Matthews 1987; Buzar et al. 2005; Gober 1990; Ogden and Hall 2004; Schmid 1988).

The second demographic transition was a sociodemographic change that brought with it the replacement of the life-cycle model with the concept of life course in terms of a demographic analytical framework in postindustrial countries. The life-cycle model explains that individuals pass through a predetermined order of stages, including pre-marriage (infancy, childhood, adolescence, and free-living adult), marriage, childbearing, child rearing, empty-nesting, and widowhood (Wulff 2007).1 Thus, the traditional life-cycle model overlooks most untraditional and nonfamily household types (e.g., one-person households of middle-aged to older individuals or those who live together but never married) (Wilkes 1995). In contrast, the life course concept highlights the dynamic features and the variety of the timing and durations of demographic events and stages, including marriage, the birth of a child, divorce, and the death of a spouse (Kendig 1990; Willekens 1988). Because demographic events and stages are not equally allocated during a life span (Wulff 2007), the notion of the life course focuses on the occupant probability of individuals’ living arrangements (Richards et al. 1987).

An increase in the number and types of one-person households has been one of the main features of the second demographic transition as it proceeded in developed countries. A one-person household is generally positioned at the beginning and the end of an individual’s life course (Forster 1995). However, the progress of the second demographic transition caused the emergence of one-person households at various points in time (Mutchler and Burr 1991), and occasionally, they have been considered a “superior good” (Burch and Matthews 1987) or a “basic figure of fully developed modernity” in developed countries (Beck 1992:122). The forcing factors behind one-person households were leaving parental homes, relationships with a spouse or cohabitant ending, children leaving the home, and bereavement (Bennett and Dixon 2006; Qu and de Vaus 2011). In previous studies (Burch and Matthews 1987; Heath 1999; Kramarow 1995; Michael et al. 1980; Mutchler 1992; Mutchler and Burr 1991; Schmid 1988), the following were cited as determinants for the formation of one-person households: economic resources, the purchase of privacy and independence, adequate health, the existence of kinship, and demographic characteristics (age, gender, and educational level). In contrast, the precedent studies (Qu and de Vaus 2011; Richards et al. 1987) that were relevant to this one insisted that the transition from one household type to another was related to age, income level, race (nonblack vs. black), and health condition. Further, Mutchler (1992) indicated that kinship also had an effect on the transition of elderly households.

Similarly, previous studies have conducted a variety of analyses on the determinants for forming one-person households and their transitions. However, this study highlights the effects of not only household attributes but also location characteristics at the time of formation (i.e., when one-person households are formed) on their durations. Preceding studies have paid scant attention to the location characteristics. To that end, in this study, accessibility is calculated to investigate the effect of location attraction, and location quotient by one-person household type is calculated to ascertain the influence of the opportunity to join in marital and social relationships. The results of this study represent a significant advancement in terms of the approach from a different viewpoint on the duration and transition of one-person households.

## Change and the Distribution of One-Person Households

Korea has been in a period of rapid demographic change since the 2000s, characterized by a rapid reduction in the average number of people per household and an increase in the number of one-person households. The Seoul metropolitan region (SMR) is the capital area of Korea and a representative region for this rapid demographic transition.

Administrative facilities are concentrated in the SMR, which is recognized as a center of Korean society in the areas of politics, economics, education, and culture. The SMR comprises the city of Seoul and its surrounding areas: its elements in terms of administrative districts are the city of Seoul, the city of Incheon, and Gyeonggi province. The SMR has 23.836 million people living in 8.415 million households, more than one-half of the total population and more than one-half of the total number of households, respectively, in Korea. The total land area of the SMR is 12,071 km2. The SMR includes two metropolitan cities (the cities of Seoul and Incheon) and one province (Gyeonggi), which contains 20 cities (Si), 6 counties (Gun), and 53 boroughs (Gu) (see the black and white boundaries of Fig. 1 and also Table 1).

The number of one-person households in the SMR almost doubled, increasing from 0.937 million in 2000 to 1.823 million in 2010. The share of one-person households among all households increased by an average of 0.767 % annually during the same period. The changes in the share by administrative district were all similar. However, the changing patterns of one-person households differed by age group.2 The rate of increase of one-person households under 30 years has slowed in the SMR, and the shares of those in the city of Incheon and the Gyeonggi province declined between 2005 and 2010. Weakening of Confucian customs, the collapse of the parent-arranged marriage system, improved education standards, and women’s involvement in economic activities were the reasons for the changing marriage patterns in East Asia (Jones 2010). In contrast, in Korea, the differences in the changing patterns in the share of young single by region is interpreted as the accelerated aging phenomenon at both the state and regional levels—the aging trend is even more rapid in the nonmetropolitan regions (Ko and Kim 2012). The increase in the number of one-person households in the range of 40 to 64 years of age has rapidly accelerated (see Table 2)—a reflection of the high divorce rate among households in this age range. According to Statistics Korea (2015), in 2010, the average divorce ages were 45.0 for men and 41.1 for women, and persons in their early 40s had the highest age-specific divorce rate (10.4 %); these trends have been continually increasing. Another significant factor behind the increase in one-person households was the rapid growth trend of the aging population in Korea. The national average age in 2010 was 38.0 years, up from 25.9 years in 1980. That is, the remarkable growth in the number of one-person households aged 40–64 was the result of the complex interaction between the divorce rate and the aging phenomenon.

The spatial distribution of one-person households also appears to differ by regional groups. To investigate the spatial patterns of one-person households, this study uses a location quotient (LQ), which is an analytical methodology that measures an area’s specialization relative to an entire geographic region (Brown and Chung 2006; Chung and Brown 2007).
$LQij=Xij/XjXi/X,$
1
where LQij is the LQ index of area j for household type i; Xij / Xj is the share of household type i in area j; and Xi / X is the share of household type i in the entire region (the SMR).

Figure 2 shows the spatial distribution of the LQ indexes for one-person households, which were calculated using Eq. (1). An LQ index of 1.0 means that the specialization level of the area and that of the entire region are equal. An LQ of less than 1.0 indicates a lower concentration level for the area compared with the entire region, whereas an LQ of more than 1.0 means a relatively higher area concentration level than that of the entire region. Based on these standards, the thematic map shows that a relatively higher concentration of one-person households is in the inner areas of the city of Seoul and the outskirts of Gyeonggi province, with some exceptions: downtown in the large cities (Incheon, Suwon, Seongnam, and Goyang) and the industrial cities (Ansan and Siheong). Moreover, the neighboring areas of the city of Seoul have lower values in terms of the concentration of one-person households.

Meanwhile, the distribution patterns of the relative concentration levels for one-person households by age group exhibit obvious differences (see Fig. 2). These results coincide with the outcomes of previous research analyses based on more detailed spatial units for the city of Seoul (Chae et al. 2014; Yi and Lee 2010) or relative to all of Korea (Lee et al. 2011). One-person households under 30 years and aged 30–39 are mostly concentrated where universities and employment centers are located in the SMR. In contrast, one-person households aged 40–49 and 65 and older are relatively concentrated in the areas where country life can be enjoyed and in the centers of primary industries. Therefore, the results from the LQ analyses show that the spatial distribution between the younger and older generations exhibited a distinct geographical gap. This gap is caused by a combination of factors when choosing the residential (re)location of one-person households and the determinants of their durations. This study focuses on the latter, which, relatively speaking, have not been given due attention.

## Empirical Analysis on Duration of One-Person Households

### Generation of Analysis Data

The spatial scope of this empirical study is the SMR in Korea, and the temporal range is from 2001 to 2012. The objects of the analysis are newly formed, one-person households after 2001, which cover all dissolved households, those that transformed from one type to another prior to the final year, and those maintained prior to and after the cutoff date.

Research related to the duration of one-person households requires a longitudinal data set. The Korean Labor and Income Panel Study (KLIPS), a representative longitudinal survey of Korean households and individuals residing in urban areas, was initiated in 1998 by the Korea Labor Institute and has been conducted annually thereafter. The KLIPS collects not only the attributes of households but also the income, expenditures, education, job training, economic activities, labor movement, and social activities of individuals. There is also a follow-up survey of 6,415 households—1,415 households have been added since 2009 to enhance the representativeness of the samples—and their members. Response rates were 77.3 % in 2001 and was 73.7 % in 2012 (Korea Labor Institute 2013). This empirical analysis used the KLIPS to calculate the durations of one-person households and to verify their survival. The estimating criteria are presented in Fig. 3.

As shown in Fig. 3, the symbols denote the times of formation (●), dissolution and transformation (○), and failure to track (×) one-person households. The solid line indicates the estimated continuation of a one-person household, and the broken line is not contained within that interlude. The horizontal line indicates the passage of time from 2001 to 2012. The letters on the vertical line refer to instances in the process of changing from creation to either the dissolution or continuation of one-person households that are relevant to the duration calculations of the households in this article. Points A, D, and E, respectively, describe the beginning, ending, and duration of each one-person household. B is a one-person household that has continued since its formation, and C is a case that failed to be tracked through the follow-up surveys. Moreover, D1 and E1 indicate the reformation of one-person households because of the complexity and diversity of the relevant individual life courses.

The explanatory variables that affect the durations of one-person households were selected on the basis of the results of previous studies and the hypothesis of the present study. The one-person household attributes were available from the KLIPS, and some of the results from previous research were used as independent variables in this empirical analysis. An investigation of location characteristics that affect the continuation of one-person households is this study’s main focus. Thus, this empirical analysis examined characteristics concerning the residential location choice of one-person households, such as accessibility to employment districts, the concentration of a specific household type, and the supply of small (less than 40 m2) dwelling units.3

The independent variables relating to location characteristics were selected for the following reasons. First, “small dwelling” was selected as an explanatory variable to verify how housing adjustment according to financial conditions affected the durations of one-person households. The financial condition of a one-person household is relatively lower than that of other household types: in the 2010 Household Income & Expenditure Survey of Korea, one-person household income was approximately 34.1 % of the average income of multiple-person households. Additionally, household income is generally low when individuals are young, gradually increasing as they mature. Retirement4 brings a decrease in income levels. Households can adjust the scale of their dwelling units in response to changes in income in order to save on living costs. Second, Korea is changing from family-oriented to individually oriented, which has increased the importance of social relationships in the activities of work, culture, and leisure among all age groups. As seen in the spatial distribution analysis of one-person households, these households by age group tended to be concentrated in specific areas for a variety of reasons (refer to Fig. 2). The “concentration” variable denotes whether the formation probability of social relationships affected the duration of one-person households. Third, young one-person households are located in city centers, where there are core districts of urban activities such as work, shopping, and leisure; aged one-person households who prefer healthy environmental elements are generally located in suburban areas. Further, middle-aged one-person households are located between city centers and outer areas. The indicator of accessibility better represents the availability of spatially accessible opportunities than do the simple indicators of urban land use or activities, such as density patterns, density gradients, homogeneity, concentricity, and sectorality (Kawabata and Shen 2006). Thus, “accessibility” was selected as a variable measuring whether the location advantage of urban activities affected the duration of one-person households.

In Table 3, the explanatory variables can be classified into attributes of household and location characteristics. First, variables related to the attributes of households are “age,” “gender,” “employment status,” “housing tenure,” “education level,” “income,” “asset,” “debt,” and “kinship,” which were measured using the KLIPS from 2001 to 2012. “Age” is defined as the age of the householder. “Gender,” “employment status,” and “housing tenure” are nominal variables. The coding value of the variables is equal to 1 if the householder is male, a worker, and a housing owner; it is otherwise 0. “Education level” is a nominal variable equal to 1 if the householder holds a bachelor’s, master’s degree, or Ph.D.; it is otherwise 0. “Income,” “asset,” and “debt” are variables related to the economic status of households, which are also measured as ratio scales using the KLIPS. “Kinship” is defined as 1 if a separately living parent exists, and it is otherwise 0.

Second, location variables are also included. “Small dwelling” is defined as the supply of dwelling units less than 40 m2. “Concentration” of same household types is represented by the LQ index for one-person households by administrative district (see Eq. (1)). The data source for “small dwelling” and “concentration” is the census in Korea from 2000 to 2010. Moreover, “accessibility” to employment districts was calculated using the methodology mentioned by Hansen (1959) and Wilson (1970), which is based on representing location attraction through Eq. (2). The shortest network distances among administrative districts were calculated using GIS network analysis. The parameters that resulted from an analysis of commuting patterns in the SMR in 2009, which was conducted by the Metropolitan Transport Association, were also reflected in the empirical analysis: 0.364 (α), 0.528 (β), –0.111 (γ).5
$Acci=∑jJobj×αdijβ×expγdij,$
2
where Acci represents the accessibility of administrative district i; Jobj is the number of jobs in destination administrative district j; dij is the shortest network distance between administrative districts i and j; and α,β,γ are the parameters.

### Construction of the Statistical Model

In the present study, a survival analysis was applied to identify the determinants of the duration of one-person households. Survival analysis is a statistical methodology that can review a span of time that continues in a particular state and specify the determinants affecting that period as well as their influences. The survival model is widely used in the medical field to study events such as exposure to disease, pathogenesis, recovery, recurrence, and expiration. The model is also applied to research on periods of employment or unemployment (economics) and studies on product life cycles in the area of manufacturing (Park 2006). In the demographic field, when the goal is to provide a view of the cumulative effect of characteristics and resources, the survival models are more appropriate for an analysis regarding household choice (Mutchler 1992). Therefore, studies on the changing processes of both household life cycles and individual life courses can also usefully employ the model. Additionally, survival analysis can be conducted when data are uncertain about whether an event has occurred (in short, censored data). In other words, although censored data can affect the analysis results, the possibility of system error can be avoided by containing the data in the analysis.

The survival model applied in this study is Cox’s (1972) proportional hazard regression analysis, a semiparametric approach that does not require a formal specific probability distribution (Archer et al. 2010). Cox’s proportional hazard regression model is expressed as follows:
$hit=h0teβ1x1+β2x2+…+βnxn⇔hith0t=eβ1x1+β2x2+…+βnxn,$
3
where hi(t) is the hazard to individual i at time t, h0(t) is an unspecified baseline hazard at time t, xn represents the explanatory variables at the start of the duration (i.e., covariates), and βn is the survival regression coefficient of xn.

The hazard ratio (hi(t) / h0(t)) indicates the relative difference between the hazard to individual i at time t and the baseline hazard at time t. Thus, βn means the variation in the value of a logarithm that transforms the hazard ratio in accordance with the change of one unit of xn.

Cox’s proportional hazard regression model can be used to analyze the survival function’s determinants that explain the difference in the probability of an event according to the attributes of individuals. Applying the hazard model to this empirical study, the event is the dissolution of one-person households, which includes both the extinction of a household and the transition to another household comprising two or more members. The duration is the period from the time of formation of a one-person household to its dissolution or the censoring time. The attributes of the explanatory variables, which were applied on the basis of the information from the points of formation of the one-person households, were entered into the hazard model. The definitions of the attributes for the explanatory variables are related to an assumption in the present research that ascertains a relationship between a one-person household’s duration and its formation conditions.6

### Descriptive Statistics

This empirical analysis contains 711 samples. These data were gathered from the KLIPS using the generating method of analysis data. The descriptive statistics of the analyzed data set are shown in Table 4.

In the data set, the average duration of one-person households is 45.188 months (that is, 3.766 years), and the range of duration is 0 to 154 months. Thus, the durations of one-person households are diverse because of the varied reasons for their formation—for example, leaving the parental home or the ending of relationships with cohabitants (divorce or death)—as well as various reasons for dissolution of a one-person household, such as (re)marriage and living with a parent.7 For the 12 years under study, the occurrence ratio of dissolution was 41.5 %; that is, 295 households dissolved, and the rest of the sample (416 households) continued.

Regarding the attributes of the households, the average age of the sample is approximately 43, and the minimum and maximum ages are 16 and 89 years, respectively. The data set is composed of 47.5 % male and 52.5 % female. The number of working one-person households is more than twice the number of nonworking households: the former is 480 households (67.5 %), and the latter is 231 (32.5 %). The ratio of householders who graduated college is 28.6 %, and the proportion of homeowners is 22.8 %. In terms of the economic resources of the one-person households, the average annual income is 11.745 million KRW ($9,788 USD), and the average assets and debt are 20.762 and 10.865 million KRW ($17,302 and \$9,054 USD), respectively. The standard deviations of the economic variables are the largest, thus implying enormous gaps between the economic levels of the highest and lowest brackets. Furthermore, the ratio of one-person households who have a separately living parent (in short, “kinship”) is 55.7 %.

In terms of the location characteristics, the average of the natural log-transformed “accessibility” (which indicates access to employment districts) is 13.813, with a standard deviation of 0.858. Although local accessibility exhibited some variation every five years, generally, the areas contained in the city of Seoul and the city of Incheon have relatively high values (the dark area in the Fig. 4), whereas the outskirt areas of the SMR generally have low values (see panels a, b, and c in Fig. 4). The range of the concentration index of one-person households is 0.524 to 1.761, and the average value is 1.050. As shown earlier in Fig. 2, the concentration of one-person households is generally higher in the inner areas of Seoul and the outskirts of the Gyeonggi province than it is in other areas (see panels d, e, and f in Fig. 4). Finally, the average number of small dwellings is 10.752 thousand units. Small houses are distributed more in the borderline areas between metropolitan cities and Gyeonggi province, as well as in the outer areas in Seoul and Incheon. The quantities of such houses had increased in the outskirts of the SMR until 2010 (see panels g, h, and i in Fig. 4). The housing supply ratio reached 100 % in 2008. After that, based on the growing demand for small-sized dwellings, the government adopted a small and medium housing–oriented supply policy in the late 2000s. The supply of small dwellings has been concentrated at the station-adjacent areas and the greenbelt zones surrounding Seoul in the SMR (Choi et al. 2013; Ha and Cho 2009).

### Results of the Cox’s Proportional Hazard Regressions

In this research, the empirical analyses were based on two assumptions. First, location characteristics and household attributes influence the duration of one-person households. Second, the influencing factors are differentiated by the types of one-person households. Therefore, the empirical study analyzed the entire sample as well as each one-person household group.8 The interpretation of the results of the hazard regression analysis is based on the hazard probability of the event’s occurrence, with the event defined as the dissolution of a one-person household or the transition to another household type. That is, a positive sign for a coefficient represents a greater probability of the event’s occurrence and indicates a decline in the duration of a one-person household. In contrast, a negative sign can be interpreted as the opposite.

The results of the hazard regression analysis for the entire one-person household sample (711 households) are shown in Table 5. The table shows that the factors influencing the duration of all one-person households were mainly the attributes of the individuals in the households. “Age” had a negative coefficient and was significant at the 99 % level. Thus, the older the one-person householder was at the time of formation, the lower the probability of the event (1.9 % by every one year). “Employment status” also had a negative coefficient and was significant at the 95 % level. Thus, employed one-person householders at the formation juncture were less likely to end this state and change into another household type. “Education level” had a positive effect on one-person household duration at the 99 % level. The results verified that the durations of highly educated, one-person households were relatively shorter than those of others. Both “debt” and “kinship” were also significant determinants at the 90 % level. The coefficient of “debt” was negative, and that of “kinship” was positive. The outcomes indicate that the durations of one-person households with more debt were longer and that the intervals of one-person households with a separately living parent were shorter than those of others.

Among the significant variables, the effects of “age” and “kinship” are consistent with previous studies (Mutchler 1992; Richards et al. 1987). “Income,” which was an important variable in precedent studies, was not a significant variable in this analysis. However, “debt”—another variable that represents the financial state of one-person households—was a significant determinant. Moreover, the “employment status” and “education level” attributes of households were also significant factors that affected the durations of one-person households. In contrast, the results of the empirical analysis of the entire sample of one-person households show that their durations were not affected by location characteristics.

The results of the empirical analyses by one-person household type are shown in Table 6, which displays the findings related to the assumptions of this empirical analysis. Location characteristics and household attributes were influences on the duration of one-person households, and the influencing factors varied by the groups of households.9 However, interpretations of the factors that verged on statistical significance (p < .10) require additional attention.

First, in the results for one-person households under 30 years old, both household attributes (such as “gender,” “employment status,” “housing tenure,” and “debt”) and location characteristics (such as “accessibility” and “concentration”) were significant. Male one-person households under 30 years old were less likely to end that state, which is due to relatively late first marriages for men in Korea. Employed one-person households at the time of formation were less likely to dissolve and transform into another household type. The occurrence probability of the event for homeowners was relatively lower than that for renters. Moreover, the more debt a one-person householder had, the greater the occurrence probability of the event. The first two factors are relevant to the first marriages of young single households, and the others are related to their financial states. These outcomes are in line with the Beresford and Rivlin (1966:254) argument that the formation of one-person household living is related to “purchase privacy.” In terms of location characteristics, “accessibility” had a negative coefficient, and “concentration” had a positive coefficient at a slight trend toward significance (p = .097). Young one-person householders located in areas that were highly accessible to employment districts were less likely to end their states. Additionally, the occurrence probability of the event for young one-person households located in the concentrated areas of relatively more potential spouses was higher.

Second, the hazard regression results for one-person householders aged 30 to 39 were as follows. The significant variables were “age,” “employment status,” “debt,” and “concentration.” The interpretations of the first three variables are analogous to those of the entire sample of one-person households. However, “concentration” had a negative effect, unlike the results for one-person householders under 30 years. Persons older than 30 years old who wanted to stay single forever were less likely to end their states and change to another household type. Their concentration can be understood as replacing a marital relationship with friendship, in line with the notion of the importance of friendship in modern society (Putnam 2000; Watters 2003).

Third, in the empirical analysis for one-person householders aged 40 to 64, “age,” “asset,” “debt,” and “small dwellings” were marginally statistically significant (p = .08). The interpretations for “age” and “debt” are the same as those of the results for the entire sample of one-person households. “Asset” can also be understood as a signifier of economic resources, similar to “debt.” Michael et al. (1980) indicated that growth in economic resources raises the propensity to live alone or marry. Because a main reason for dissolution among single persons aged 40–64 years is (re)marriage, the economic resources were obviously positive variables. Moreover, despite a marginally significant tendency, in terms of location characteristics, the 40- to 64-year-old householders who lived in small housing areas are relatively less likely to change their states because of the availability of affordable housing in residential locations.

Finally, the influence of determinants on the duration of the aging one-person household class was distinctive. The factors were “age,” “debt,” and “small-sized dwellings.” “Small dwellings” was negative and significant at the 95 % level, which could be interpreted as elderly singles choosing to live where (small-sized) housing is more affordable because of their relatively few economic resources. The coefficient of “debt” is positive, the same as in young one-person households. In contrast, “age” had a positive sign, unlike the other groups, which could be related to their health. Previous studies also singled out health as a constraint to living alone for aging persons (Kobrin 1981; Mutchler and Burr 1991; Wolf and Soldo 1988).

One of the main findings of the empirical analyses by age group is that the factors influencing the duration of one-person households are similar to the determinants of residential location choice verified in the previous literature and in the census survey. In the 2012 Korea Housing Survey, the main reasons for the residential location choices of the voluntarily moved one-person households within the SMR are as follows: the reasons mentioned by 122 of 203 one-person householders under age 40 were primarily “relations with workplace” and “convenient residential environment,” whereas the reasons given by 273 of 642 one-person householders over 40 years old were related to affordable housing, such as “modification of housing space” and “low housing price or inexpensive rent.” These results reflect rational decision-making behaviors in the residential location choice process (Brown and Moore 1970). Moreover, the spatial segregation phenomenon in the one-person household group can also be explained as the cumulative effect of the residential location choices of one-person households and their durations.

Another main finding is that the effects of some variables—“age,” “debt,” and “concentration”—on the durations of one-person households change directions across age groups. The coefficients on “age” of the one-person household groups were negative for those aged 30–39 and those aged 40–64, but were positive for those 65 years and older. The average man and woman in Korea in 2010 married at the ages of 32.13 and 29.41, respectively. Therefore, the negative coefficient for those aged 30–64 can be explained as the decrease in the marriage rate after age 30, whereas the opposite coefficient among those 65 years or older can be explained by health issues (e.g., children’s care and support, and death). The coefficients for “debt” were positive for those under 30 and 65 years or older but negative for those aged 30–39 and 40–64. The former outcomes are related to rational decision making for soundness of household economics; the latter is the result of financial problems being one of the representative obstacles to consolidation with separately living family members, both first marriage and remarriage, which are the main events behind household dissolution in this category. Lastly, the coefficient on “concentration” was positive for those under 30 years and negative for those aged 30–39. These results can be interpreted in terms of the influence of the average age at first marriage. The marriages of single households reach their peak at approximately 30 years old and gradually decrease thereafter. Therefore, “concentration” means greater opportunity for contact with relatively more potential spouses for the under-30 singles and for forming friendships with people of socioeconomically similar characteristics for the group aged 30–39.

## Summary and Conclusions

Rapid demographic changes have occurred in Korea. These changes have already been witnessed in developed countries through the so-called second demographic transition; they are related to various socioeconomic changes—low birth rates, aging, the spread of individualism, and improvements in living standards—in nations that are in transition from developing to developed countries. One of the most representative examples is the growing trend of one-person households in Korea. Given this background, the main purpose of this research was to ascertain a relationship between the durations of one-person households and their formation conditions. In particular, this study focused on the influence of the location characteristics of a household’s surrounding area. A review of the distribution patterns of one-person households in the SMR, which is a representative metropolitan area of Korea, revealed that the concentrated areas were divided by age group.

Based on this fundamental result, the outcomes of the empirical analysis on the durations of one-person households according to their formation conditions in the SMR using Cox’s proportional hazard model can be summarized as follows. In the analysis for the entire sample during the temporal scope of this research, household attributes were the main determinants of household continuation; the variables “age,” “employment status,” “education level,” “debt,” and “kinship” were significant, as previous studies have found.

However, the results of the empirical analyses of the household types classified by age group differed from those of previous research. To summarize, in the results for the one-person household types, the location characteristics were significant variables, although the significance of the variables varied according to the type of household. The durations of one-person households under 40 years old were affected by (1) accessibility to employment districts, with one-person households located in more employment-accessible areas maintaining their states longer; and (2) the concentration levels of one-person households in their area—a factor related to the opportunity for contact with relatively more potential spouses or potential for friendship with people of socioeconomically similar characteristics. In contrast, the duration of one-person households 65 years and older was influenced by the distribution of affordable housing.

Moreover, the significance of household attributes also varied according to the types of one-person households. Two peculiar results were those for householders under 30 years and householders aged 65 and older. Households with more debt were more likely to end their states and transform into another household type, which can be interpreted as a rational decision made to avoid financial risk. This outcome is in line with the concept of purchase privacy of one-person households, but the economic problems of one-person householders in the middle-aged group (aged 30–64) suggest a negative influence of having a (re)marriage. In the middle-aged groups, the fact that age is large was a negative factor for ending one-person households. In contrast, in the aging group, the increase in the probability of changing household type according to age can be explained by health issues.

The second demographic transition emerged in the developing countries of Asia, and the demographic changes in Korean society accelerated in the 2000s. This article provides a framework to make sense of the characteristics of one-person households in nations that are in transition from developing to developed countries, such as Korea. Additionally, this article found a significant relationship between the duration of one-person household and formation conditions, which had previously received relatively scant attention compared with the formation factors for one-person households.

The policy implications of this research are as follows. First, the analyses presented here suggest that the determinants of the duration of one-person households function quite differently by age group. Therefore, demographic policy-making designed to address the increases in one-person households should be differentiated by age group. Second, the analyses provide basic knowledge for population policymakers who aim to decrease the number of one-person households, which is one of the national sociodemographic problems (e.g., increases for first marriages of young singles and small households). Third, the influences of the location characteristics on the durations of one-person households are also significant for constructing national or regional policies regarding one-person households, such as creating socially mixed urban spaces and supplies of affordable public housing, which are considered to be one of the representative demographic classes in contemporary Korea. Additionally, the durations of one-person households were affected by both socioeconomic factors and location characteristics. This result implies that a solution to the current sociodemographic problems of Korean society needs to be constructed through interdisciplinary research by demographers and urban planners. If these actions are taken, the problems can be more efficiently solved.

Generalizing the findings of this study will require further empirical research on additional metropolitan regions in developing and developed nations. Furthermore, future research should explore the effects of variables that were not considered in these analyses, such as household attributes (e.g., causes of formation and marital status) and changes in location characteristics resulting from moving and the micro-location characteristics of convenience, safety, and amenities.

## Acknowledgments

This research was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) and funded by the Ministry of Education (2013R1A1A2058091) & the Ministry of Science, ICT and Future Planning (2015R1A2A2A04005886).

## Notes

1

Because a classification of the life-cycle stages varies depending on the field and purpose of research, the stages are not standardized.

2

In previous studies (Bae 1993; Kim and Moon 2009; Stein 1981; Yi and Lee 2010), one-person households were classified by age, gender, income, and willingness, with age a common criterion. One-person households of Korea are best able to be classified by age group: under 30 (generally an unmarried student), 30 to 39 years (a voluntarily unmarried and first-hired person), 40 to 64 years (an involuntary live-alone due to divorce or bereavement), and 65 years or older (commonly an aged and retired person).

3

The residential movement of a household causes changes in its location characteristics. However, changes in the residential location characteristics of a one-person household according to residential mobility were not considered in this study for two reasons. First, this study focused on household formation conditions. Second, 384 (54.0 %) of the analyses’ 711 one-person households were encountered more than once. The total number of their residential movements was 596, but almost all residential movements were internal within the originally located district (392 movements) or external into closely surrounding areas (66 movements). Moreover, the distribution by age group of movements into an area far from their originally located districts (138 movements) is as follows: 76 movements by persons under 30 years, 26 movements by those 30–39 years, 25 movements by those 40–64 years, and 11 movements by those 65 years and older. The share by each age group of movements within the same district or into a closely surrounding area was 70.0 % to 84.1 %. These findings are identical to outcomes of previous research showing short-distance mobility and mobile patterns into a living area with similar socioeconomic attributes (Choi and Cho 2005; Simmons 1968).

4

The official retirement age (state pension age) is currently 60 years, and the normal pension age will be gradually increased to reach 65 by 2033. However, the effective retirement age is 71.1 years for males and 69.8 years for females (OECD 2013). Thus, this study assumes a retirement age of 65 years, which is the average of the official and effective retirement ages.

5

The location characteristics were calculated by administrative spatial unit (see Fig. 1 and Table 1).

6

A previous study (Richards et al. 1987) used the same method based on the evidence that the economic resources of households are correlated from one year to the next and that year-to-year changes are not very meaningful.

7

The reasons for the dissolution of a one-person household were not asked for the KLIPS. In the previous literature, the reasons by age group except for death of the one-person householder were cited as forming a couple or marrying (Qu and de Vaus 2011; Richards et al. 1987), living with parents (Qu and de Vaus 2011), (re)marriage of middle-aged one-person households (Wulff 2007), economic difficulties (Mutchler 1992), and health problems (Mutchler and Burr 1991; Qu and de Vaus 2011; Soldo et al. 1984) in elderly one-person households.

8

See footnote 2 for the classification standards of one-person household groups.

9

The sample size (n), required in an analysis using time-to-event data, was calculated using the methodology proposed by Wang and Chow (2007), and the equation is as follows: n = (zα / 2 + zβ)2 / b2p1p2d. The variables and parameters were assumed in the equation: 1.96 (zα / 2; significance level 0.05), 0.842 (zβ; power 80 %), 0.7 (b: hazard ratio between the standard group and the test group), 0.5 (p1, p2: even distribution between the two groups) and 0.4 (d: ratio of event occurrence before the end of the study). The required sample size is approximately 160. Therefore, the sample sizes of all empirical analyses by the one-person household type except the group over 65 years old (see Table 6) are generally sufficient. However, the significance level of the explanatory variables on location characteristics may be improved by increasing the sample sizes.

## References

Archer, W. R., Ling, D. C., & Smith, B. C. (
2010
).
Ownership duration in the residential housing market: The influence of structure, tenure, household and neighborhood factors
.
Journal of Real Estate Finance and Economics
,
40
,
41
61
. 10.1007/s11146-008-9126-2
Bae, W. O. (
1993
).
A study on one person households in Korea
.
Journal of the Population Association of Korea
,
16
(
2
),
125
139
.
Beck, U. (
1992
).
Risk society: Towards a new modernity
.
London, UK
:
Sage
.
Bennett, J., & Dixon, M. (2006). Single person households and social policy: Looking forwards (Joseph Rowntree Foundation report). York, UK: Joseph Rowntree Foundation.
Beresford, J. C., & Rivlin, A. M. (
1966
).
Privacy, poverty, and old age
.
Demography
,
3
,
247
258
. 10.2307/2060076
Brown, L. A., & Chung, S. Y. (
2006
).
Spatial segregation, segregation indices and the geographical perspective
.
Population, Space and Place
,
12
,
125
143
. 10.1002/psp.403
Brown, L. A., & Moore, E. G. (
1970
).
Migration flows in intraurban space: Place utility considerations
.
Annals of the Association of American Geographers
,
60
,
368
384
. 10.1111/j.1467-8306.1970.tb00726.x
Burch, T. K., & Matthews, B. J. (
1987
).
Household formation in developed societies
.
Population and Development Review
,
13
,
495
511
. 10.2307/1973136
Buzar, S., Ogden, P. E., & Hall, R. (
2005
).
Households matter: The quiet demography or urban transformation
.
Progress in Human Geography
,
29
,
413
436
. 10.1191/0309132505ph558oa
Chae, J., Park, S., & Byu, B. (
2014
).
An analysis of spatial concentrated areas of single person households and concentrating factors in Seoul
.
Seoul Studies
,
15
(
2
),
1
16
.
Choi, E., & Cho, D. (
2005
).
The characteristics in intra-urban migration in Seoul
.
Journal of the Korean Association of Regional Geographers
,
11
,
169
186
.
Choi, S. H., Jeong, S., Kim, Y. T., & Jeong, G. S. (
2013
).
A study of the evaluation & direction of the urban life housing
.
Daejeon, Korea
:
Land & Housing Institute
.
Chung, S. Y., & Brown, L. A. (
2007
).
Racial/ethnic residential sorting in spatial context: Testing the explanatory frameworks
.
Urban Geography
,
28
,
312
339
. 10.2747/0272-3638.28.4.312
Cox, D. R. (
1972
).
Regression models and life-tables
.
Journal of the Royal Statistical Society: Series B (Statistical Methodology)
,
34
,
187
220
.
Forster, C. (
1995
).
Australian cities: Continuity and change
.
Melbourne, Australia
:
Oxford University Press
.
Gober, P. (
1990
).
The urban demographic landscape: A geographic perspective
. In Myers, D. (Ed.),
Housing demography: Linking demographic structure and housing market
(pp.
232
248
).
:
University of Wisconsin Press
.
Ha, S. K., & Cho, S. C. (
2009
).
Suburban development and public housing provision on greenbelt zones in the Seoul metropolitan region
.
Journal of the Korean Urban Management Association
,
22
(
1
),
183
207
.
Hall, R., Ogden, P. E., & Hill, C. (
1997
).
The pattern and structure of one-person households in England and Wales and France
.
International Journal of Population Geography
,
3
,
161
181
. 10.1002/(SICI)1099-1220(199706)3:2<161::AID-IJPG64>3.0.CO;2-2
Hansen, W. G. (
1959
).
How accessibility shapes land use
.
Journal of the American Institute of Planners
,
25
,
73
76
. 10.1080/01944365908978307
Heath, S. (
1999
).
Young adults and household formation in the 1990s
.
British Journal of Sociology of Education
,
20
,
545
561
. 10.1080/01425699995263
Jones, G. W. (
2010
).
Changing marriage patterns in Asia
(
Asia Research Institute Working Paper Series No. 131
).
Singapore
:
Asia Research Institute, National University of Singapore
.
Kawabata, M., & Shen, Q. (
2006
).
Job accessibility as an indicator of auto-oriented urban structure: A comparison of Boston and Los Angeles with Tokyo
.
Environment and Planning B: Planning and Design
,
33
,
115
130
. 10.1068/b31144
Kendig, H. (
1990
).
A life course perspective on housing attainment
. In Myers, D. (Ed.),
Housing demography: Linking demographic structure and housing market
(pp.
133
156
).
:
University of Wisconsin Press
.
Kim, D. S., Huh, M. G., & Lee, S. H. (
2013
).
An analysis on effect of aging on regional economy – Focusing mainly on manufacturing industries
.
KIET Monthly Industrial Economics
,
6
,
52
64
.
Kim, O. Y., & Moon, Y. K. (
2009
).
Housing analysis of one person household
.
Journal of the Residential Environment Institute of Korea
,
7
(
2
),
37
53
.
Ko, P. S., & Kim, D. H. (
2012
).
A study on the future prospects of the housing market in the 2010 census results
.
Journal of the Korea Institute of Electronic Communication Sciences
,
7
,
1117
1124
.
Kobrin, R. E. (
1981
).
Family extension and the elderly: Economic, demographic, and family cycle factors
.
Journal of Gerontology
,
36
,
370
377
. 10.1093/geronj/36.3.370
Korea Labor Institute
. (
2013
).
Korean Labor and Income Panel Study (KLIPS) Waves 1–15 User’s guide
.
Sejong-si
:
Korea Labor Institute
.
Kramarow, E. A. (
1995
).
The elderly who live alone in the United States: Historical perspectives on household change
.
Demography
,
32
,
335
352
. 10.2307/2061684
Kuijsten, A. (
1995
).
Recent trends in household and family structures in Europe: An overview
. In van Imhoff, E., Kuijsten, A., Hooimeijer, P., & van Wilssen, L. (Eds.),
Household demography and household modelling
(pp.
53
84
).
New York, NY
:
Plenum Press
.
Kuijsten, A. (
1996
).
Changing family patterns in Europe: A case of divergence?
.
European Journal of Population
,
12
,
115
143
. 10.1007/BF01797080
Lee, H. Y., Noh, S. C., & Choi, E. Y. (
2011
).
Growth pattern and spatial distribution of one-person households by socio-economic demographic characteristics
.
Journal of the Korean Geographical Society
,
46
,
480
500
.
Lesthaeghe, R. (
1995
).
The second demographic transition in Western countries: An interpretation
. In Mason, K. O., & Jensen, A. M. (Eds.),
Gender and family change in industrialised countries
(pp.
17
82
).
Oxford, UK
:
Clarendon Press
.
Michael, R. T., Fuchs, V. R., & Scott, S. R. (
1980
).
Changes in the propensity to live alone: 1950–1976
.
Demography
,
17
,
39
56
. 10.2307/2060962
2010 statistics of urban planning
. (
2011
).
Gwacheon, Korea
:
Ministry of Land, Transport and Maritime Affairs
.
Mutchler, J. E. (
1992
).
Living arrangements and household transitions among the unmarried in later life
.
Social Science Quarterly
,
73
,
565
580
.
Mutchler, J. E., & Burr, J. A. (
1991
).
A longitudinal analysis of household and nonhousehold living arrangements in later life
.
Demography
,
28
,
375
390
. 10.2307/2061463
Pensions at a glance 2013: OECD and G20 indicators
. (
2013
).
Paris, France
:
OECD Publishing
.
Ogden, P. E., & Hall, R. (
2004
).
The second demographic transition, new household forms and the urban population of France during the 1990s
.
Transaction of the Institute of British Geographers
,
29
,
88
105
. 10.1111/j.0020-2754.2004.00116.x
Park, C. B. (
2006
).
Survival analysis: Theory and practice
.
Seoul, South Korea
:
Shinkwang Publishing Co.
.
Putnam, R. D. (
2000
).
Bowling alone: The collapse and revival of American community
.
New York, NY
:
Simon and Schuster
.
Qu, L., & de Vaus, D. A. (
2011
).
Starting and ending one-person household: A longitudinal analysis
.
Journal of Family Studies
,
17
,
126
145
. 10.5172/jfs.2011.17.2.126
Richards, T., White, M. J., & Tsui, A. (
1987
).
Changing living arrangements: A hazard model of transitions among household types
.
Demography
,
24
,
77
97
. 10.2307/2061509
Schmid, J. (
1988
).
Principles emerging from sociology for definitions and typologies of household structure
. In Keilman, N., Kuijsten, A., & Vossen, A. (Eds.),
Modelling household formation and dissolution
(pp.
13
22
).
Oxford, UK
:
Clarendon Press
.
Shin, S. Y. (
2010
).
A study on the spatial distribution of one person households: The case of Seoul
.
Journal of the Korea Planners Association
,
45
(
4
),
81
95
.
Simmons, J. W. (
1968
).
Changing residence in the city: A review of intraurban mobility
.
Geographical Review
,
58
,
622
651
. 10.2307/212686
Soldo, B. J., Sharma, M., & Campbell, R. T. (
1984
).
Determinants of the community living arrangements of older unmarried women
.
Journal of Gerontology
,
39
,
492
498
. 10.1093/geronj/39.4.492
Statistics Korea
. (
2015
).
The 2014 statistics of marriage & divorce in Korea
(Resource document).
Daejeon, Korea
:
Statistics Korea
Stein, P. (
1981
).
Single life
.
New York, NY
:
St. Martin’s Press
.
Wang, H., & Chow, S.-C. (
2007
).
Sample size calculation for comparing time-to-event data
. In D’Agostino, R. B., Sullivan, L., & Massaro, J. (Eds.),
Wiley encyclopedia of clinical trials
(pp.
1
7
).
New York, NY
:
John Wiley & Sons
. doi:10.1002/9780471462422.eoct007
Watters, E. (
2003
).
Urban tribes: A generation redefines friendship, family, and commitment
.
London, UK and New York, NY
:
St. Bloomsbury
.
Wilkes, R. E. (
1995
).
Household life-cycle stages, transitions, and product expenditures
.
Journal of Consumer Research
,
22
,
27
42
. 10.1086/209433
Willekens, F. (
1988
).
A life course perspective on household dynamics
. In Keilman, N., Kuijsten, A., & Vossen, A. (Eds.),
Modelling household formation and dissolution
(pp.
87
107
).
Oxford, UK
:
Clarendon Press
.
Wilson, A. G. (
1970
).
Entropy in urban and regional modeling
.
London, UK
:
Pion
.
Wolf, D. A., & Soldo, B. (
1988
).
Household composition choices of older unmarried women
.
Demography
,
25
,
387
404
. 10.2307/2061539
Wulff, W. (
2007
).
Growth and change in one person households: Implications for the housing market
.
Urban Policy and Research
,
19
,
467
489
. 10.1080/08111140108727894
Yi, C., & Lee, S. (
2010
).
Analysis of single household areas and evaluation of their residential environment in Seoul
.
Seoul Studies
,
11
(
2
),
69
84
.
Yi, C., & Lee, S. (
2014
).
An empirical analysis of the characteristics of residential location choice in the rapidly changing Korean housing market
.
Cities
,
39
,
156
163
. 10.1016/j.cities.2014.03.002