EN中文

Journal of Geographical Sciences >

2022 , Vol. 32 >Issue 12: 2560 - 2576

DOI: https://doi.org/10.1007/s11442-022-2061-y

Research Articles

Spatial analyses of factors affecting household structure in Chinese cities: The case of three-generation lineal households

  • LI Ting , 1 ,
  • LIU Tao , 2, 3, * ,
  • LIU Jiajie 2, 3 ,
  • CHENG Tianyi 1
Expand
  • 1. Center for Population and Development Studies, Renmin University of China, Beijing 100872, China
  • 2. College of Urban and Environmental Sciences, Peking University, Beijing 100871, China
  • 3. Center for Urban Future Research, Peking University, Beijing 100871, China
* Liu Tao (1987-), PhD and Assistant Professor, specialized in migration and urbanization. E-mail:

Li Ting (1982-), PhD and Professor, specialized in demography and family studies. E-mail:

Received date: 2021-07-27

  Accepted date: 2021-09-17

  Online published: 2022-12-25

Supported by

National Social Science Foundation of China(20BRK043)

Fold

Abstract

Household structure is an important aspect of family change during China’s modernization process. Existing literature has demonstrated significant associations between various factors and household structure, but the spatial variation in these relationships has not been examined. Using the 2010 Chinese population census data and geographically weighted regression (GWR) model, this study explored the spatial patterns of three-generation lineal households, a functionally important household type in China, and its influencing factors. There was significant heterogeneity in the distribution of three-generation lineal households. Socioeconomic, demographic, and cultural factors were all significantly related to the proportion of three-generation lineal households, but the relationships are place-specific in terms of direction and magnitude. These results suggest that the distribution of Chinese household structures cannot be explained by a single framework of family modernization theory but is determined by the interplay of various local characteristics. Especially, population migration plays an equally important role in affecting household structure than socioeconomic development in China. This work contributes to the family literature by highlighting the spatial heterogeneity in the impact of varying factors on household structure. Beyond the classic modernization theory, it sets a contextualized framework for understanding how Chinese household change in response to the rapid social transformation.

Key words： household structure; three-generation lineal household; geographically weighted regression; migration; China

Cite this article

LI Ting , LIU Tao , LIU Jiajie , CHENG Tianyi . Spatial analyses of factors affecting household structure in Chinese cities: The case of three-generation lineal households[J]. Journal of Geographical Sciences, 2022 , 32(12) : 2560 -2576 . DOI: 10.1007/s11442-022-2061-y

1 Introduction

The past several decades have witnessed a dramatic transformation of Chinese society, during which Chinese families have also experienced an unprecedented impact (Jacka et al., 2013). Understanding how Chinese families change in response to the social transformation is important. On the one hand, the family is a microcosm of society, the alteration of which reflects the larger picture of social transformation. On the other hand, changes in the family have a profound impact on social patterns, economic development, and public policies, which in turn affect the process of social change (Whyte, 1996).
Studying the household structure, reflecting the co-residential pattern of a family, is the first and critical step to understanding family change, given that the structure is easy to measure and determine the functional foundation of a family. Research on household structure dates back to the 19th century, when sociologists began to explore the institution of marriage and family (Engels, 2007). Factors such as socioeconomic development, demographic characteristics, and culture have an impact on the household form people choose (Thornton and Fricke, 1987; Li et al., 2020). The family modernization theory, developed in the mid-20th century, proposes an end point to the evolution of global household structures. It argues that as economies develop and modernization progresses, the nuclear household, consisting of a couple with/without their minor children, would become the dominant household form (Parsons, 1949; Nimkoff, 1961; Goode, 1963).
However, empirical research on the change in Chinese household structures revealed rather complex patterns. Studies based on multi-period census data show that although the overall structure of Chinese households has tended to become smaller and more diversified since the 1980s, the proportion of nuclear households, which have dominated for a long time, exhibited a declining trend in recent years. In contrast, the share of three-generation lineal households (stem households) increased during the same period (Wang, 2008; Wang, 2013; Hu and Peng, 2014; Li et al., 2020). This trend fails to reveal an inverse association between economic development and household complexity, which implies further factors underlying household structure change, and challenges how the process can be understood by the family modernization theory in the Chinese context.
One explanation could be that the effects of the influencing factors on household structure vary spatially, conditional on other local characteristics. In other words, the underlying processes determining household structure are context-specific and vary across space. China is a vast country with significant regional differences in the level of socioeconomic development, cultural traditions, and other characteristics (Yang et al., 2017). These unevenly distributed factors are likely coupled with each other, leading to the spatial variation of household structure. The relationships between household structure and its influencing factors such as socioeconomic development may not be readily summarized by a single global model that ignores spatial heterogeneity. Exploring the possible spatial variation of the effects the influencing factors have on household structure is not only important for understanding how Chinese families respond to the social transition, but may also have theoretical implications. To develop a theory that better describes the changes in Chinese household structure requires incorporating contextual perspectives, and this goes beyond the search for a universal explanation of household structure change implied by the classic family modernization theory.
There are few studies that explicitly explored the spatial variation of either household structure patterns or the effects driving the patterns. The goal of this study was therefore to investigate how the relationships between household structure and a set of socioeconomic, cultural, and demographic factors vary across space in China. Given the complexity and diversity of household structure types, we focused on one special type: the three-generation lineal household. It refers to the household structure consisting of three lineal generations, usually including older parents, one of their adult children with/without that child’s spouse, and their grandchildren. We focused on this household structure type because it has historically played a traditional role in providing child and old-age care (Hareven, 1987). The three-generation lineal household is often contrasted with the nuclear household in a sense that the former is regarded as a symbol of tradition while the latter an indication of modernity (Goode, 1963; Hareven, 1976). Contrary to the expectations of the family modernization theory, the proportion of three-generation lineal households in China has shown a continuous upward trend in recent years (Li et al., 2020). Therefore, studying three-generation lineal households may reveal the unique characteristics of Chinese household structure.

2 Background and literature

2.1 Regional differences of three-generation lineal households in China

Much of the early regional research on Chinese household structures comes from anthropological field surveys. These studies delved into a particular village or community to examine the living arrangements of residents. The classic survey conducted in villages in the lower Yangtze River Basin of south China depicted a relatively stable household structure against the backdrop of a turbulent historical period. The tradition of performing household separation rituals after sons grow to adulthood made the nuclear household the dominant structure type for a long time (Fei, 1983). In contrast, large three-generation lineal and extend households were more commonly observed in villages in north China (Wang, 2003; Huang, 2011). These studies exhibit the diversity of household structure patterns among different locations of China, which sheds light on the cultural variation across space. However, the geographical units examined in these studies are too small and too scattered to constitute meaningful regional patterns.
The more recent studies usually relied on census and household survey data to conduct cross-regional comparisons. It was found that in the early 1990s, provinces in south China as well as Yunnan province, where ethnic minorities were concentrated, had a relatively high proportion of three-generation lineal households (Zeng et al., 1992). Since the 1990s, the proportion of three-generation lineal households has decreased in urban areas but increased in rural areas; the overall proportion of three-generation lineal households was lower in the north than in the south (Wang, 2013). During the first 10 years of the 21st century, there were 14 provinces, mostly in the south-central and southwestern regions, whose proportion of three-generation lineal households increased. In contrast, provinces in the northern region tended to maintain a low proportion of three-generation lineal households (Wang, 2015).
These studies reveal significant geographical differences in the distribution of three-generation lineal households in China. Nevertheless, the cross-regional comparisons are only limited to the provincial level. Lacking analyses at a more nuanced administrative level prevents a thorough understanding of the spatial heterogeneity. More importantly, most of the previous analyses focused on the description of the distribution patterns without a deeper and more rigorous exploration of the mechanisms, which may vary across space.

2.2 Influencing factors of three-generation lineal households

As an important component embedded in society, the distribution of household structures can be influenced by a variety of factors such as socioeconomic, demographic, and cultural factors, which are often identified from different theoretical perspectives.

2.2.1 Socioeconomic factors

Despite controversy, the family modernization theory remains the most influential theory for explaining family changes. It argues that technological developments alter the production mode and economic foundation of society, which polarize family functions and modify individual values, and thus lead to a change in living arrangement (Nimkoff, 1961; Goode, 1963). The theory emphasizes the role of socioeconomic factors in determining household structures.
Socioeconomic factors can be further divided into the dimension of economic institution and development level. The economic institution involves changes in productivity and production relations during the process of modernization. The traditional small-scale agriculture economy is inextricably linked to the living arrangement of extended households. In such a system, individuals’ daily life and work are mostly constrained within the household (Logan and Bian, 1999). With the unfolding of the industrialization process, younger adults have gained greater economic autonomy as independent producers, which caused the decline of patriarchy and led to a preference for nuclear households over three-generation lineal ones.
The impact of socioeconomic development level on household structure can be perceived as promoting the change in individual values. People living in economically more advanced areas tend to form smaller households because of the rise of individualism. Meanwhile, these areas can attract a larger floating population, who usually migrate in the form of small households. This further contributes to the decline of the proportion of three-generation lineal households in these areas (Wang, 2015).
In general, family modernization theory predicts a negative association between socioeconomic factors and the proportion of three-generation lineal households. Nevertheless, the empirical findings were rather mixed. As stated in the Introduction, the proportion of Chinese three-generation lineal households maintained a steady increase trend in recent years, which is incompatible with China’s rapid economic development under the family modernization theory. This trend is not unique to China; a study based on a few developing countries also demonstrated a positive association between economic development and the probability of co-residence with older parents (Ruggles and Heggeness, 2008). They proposed that it could be caused by the interplay between socioeconomic development and local culture. To be specific, the increase of socioeconomic resources would enable people to achieve their ideal family structures. If the region has a strong cultural preference for large families, their people would be more likely to form lineal or more complex households when they can afford the expense of living together. Therefore, we would expect that the impact of socioeconomic development varies across space, in a sense that in some places there are negative relationships between socioeconomic development and the proportion of three-generation lineal families, while in other places the relationships are positive.
Another special socioeconomic factor that can directly affect living arrangements is the housing situation, because housing is an objective reality that constrains the choice of household type (Curtis, 2011). Empirical studies based on Western society often identify a positive relationship between housing prices and the probability of adult children living with their parents, because elevated housing prices prevent the younger generation from buying their own houses (Ermisch, 1999; Curtis, 2011). The situation in China seems more complex. Before 2000, urban housing was mostly distributed by enterprises and institutions. The older generation who worked for a long time had a better ability to obtain housing, and their children had to live with them in order to obtain housing benefits. After the commercialization of housing, the younger generation with better economic conditions could buy their own houses, which promoted the dissolution of three-generation lineal households (Zhong, 2015). However, in recent years the soaring housing prices may induce a similar positive association between housing prices and the proportion of three-generation lineal households as in Western society.
There is a lack of empirical evidence on the impact of housing prices on household structure in China. A recent study failed to discover a significant association between these two elements (Li et al., 2020). It is possible that there is spatial heterogeneity regarding how the living arrangements respond to the change in housing prices. Elevated housing prices tend to impose a stronger adverse effect on the formation of nuclear households in areas where house sales are more sensitive to a change in housing prices. Thus, we would also expect varying coefficients across areas on the relationship between housing prices and the proportion of three-generation lineal households.

2.2.2 Demographic factors

Demographic characteristics are factors that can have a direct impact on household structure (Yi, 1986; Jiang and O’Neill, 2007). Fertility, mortality, and the migration process play a profound role in household formation. A decrease in the mortality rate and a change in mortality patterns mainly affect the household structure by increasing the proportion of the older population. As aging deepens, there are two possible outcomes: an increase in the proportion of large households with three or more generations or an increase in single-person and husband-wife nuclear elderly households. The former may occur in regions with a stronger lineal household tradition, while the latter is more likely to occur in regions where older people have enough resources to live independently. We can expect the effect of the elderly ratio on household structure to be conditional on regional, cultural and/or socioeconomic factors.
The influence of fertility rate on household structure is more complex and presents a non-linear trend. When the fertility rate is above the replacement level (usually regarded as an average of 2.1 children per woman), the delayed effect of fertility decline raises the proportion of lineal households. To be specific, if the same household separation ratio is maintained, the proportion of nuclear families formed by children who are separated from their original household declines simply because of the reduction in the number of children, thereby increasing the proportion of lineal households in an oblique way. Only when the fertility rate is below the replacement level will the decline of the fertility rate lead to the decrease of lineal households (Yi, 1986). In the context of low fertility, fewer children will lead to longer empty-nest periods for elderly parents and increase the number of single or husband-wife households for the elderly.
The level of population migration tends to have a different impact on household structure depending on whether the migration is an inflow or outflow. Due to the restrictions of China’s household registration system, the floating population is usually of a working age (Shi and Dorling, 2020). In a typical emigration area, with an increase in people migrating out, the number of elderly one-person households, single-parent households, and skip-generation households increases, while the number of standard nuclear households decreases (Tang, 2020). Besides, the emigrants usually maintain a relatively simple household structure in the migration destination. The proportion of lineal households among migrants is significantly lower than that of the local residents, which leads to a dilution of lineal households in destination places. In sum, the inflow of population reduces the proportion of three-generation lineal households, while the outflow of population causes a reduction in nuclear households, but its effect on three-generation lineal households is not clear.

2.2.3 Cultural factors

The cultural influence on household structure is often proposed as being opposite to the impact of modernization in conjunction with the discussions on traditionalism and modernity. However, cultural factors have been playing an important and lasting role in shaping the household system before and after the occurrence of modernization (Hajnal, 1982; Li et al., 2020). The influence of culture on household structure is usually realized through customs, morals, and social order, such that it helps to restrain individuals’ behaviors (Wang, 2008).
Chinese history is known to be heavily influenced by Confucian culture, which defines a series of patriarchal social norms and best suits the self-sufficient small-scale peasant economy. In such a society, children are required to be subordinate to parents and accept their arrangements such as marriage and residential settlement, making it easier to form a lineal household. The patriarchal norms also determine clear responsibilities for the family members. Children’s obligation to provide care to their parents ensures the formation of lineal households. Therefore, we can expect that the proportion of three-generation lineal households is higher in a region with a greater degree of the traditional inheritance of small-scale agricultural economy. Some ethnic minorities in China used to primarily adopt a nomadic life mode. Their tradition of lineal household tends to be much weaker.
The discussion above implies that factors affecting the distribution of household structures are diversified and complex. Although we have discussed each factor separately, in reality these factors tend to be commingled. For example, the economic development level of a region directly determines its role as either a place of population inflow or outflow. Socioeconomic advancement also promotes the dissolution of traditional culture. The interaction of multiple factors has resulted in the observed complexity of household structure distribution across regions.
What makes the situation more complicated is the possibly heterogeneous effects of the influencing factors on household structure across regions. First, these factors themselves differ in levels or characteristics at different locations. Chinese regions demonstrate considerable variation in economic development, with a declining gradient from the Southeast coast to the western interior. The floating population is primarily concentrated in the Yangtze River Delta, the Pearl River Delta, and the Beijing-Tianjin-Hebei region. Second, even if two geographical units have similar development levels or demographic characteristics, they could also demonstrate different influential patterns due to unobserved factors. Geographic locations can be good approximations of these unobserved factors, in a sense that the spatial variations can partially reflect the latent heterogeneity resulting from the impact of these factors.
In summary, the spatial heterogeneity of influencing factors and the geographical differentiation of the affecting mechanism jointly lead to the varying patterns of the three-generation lineal household across different regions of the country. Thus, it is necessary to adopt a spatial perspective in order to better understand the underlying mechanisms that determine household structure.

3 Data, measurement, and method

3.1 Data

We used the long-term data from the 2010 Chinese population census, which accounted for about 10% of the total population, to conduct analyses. Based on the answers provided by each household member describing their relationship with the household head, we examined the household type for every sample household. The proportion of three-generation lineal households was then calculated for the prefecture-level administrative units as dependent variables. A total of 332 prefecture-level administrative units (abbreviated as cities hereafter) were included in the sample. Due to the data availability, Taiwan Province, Hong Kong and Macao Special Administrative Regions were excluded in the following analysis, and a couple of cities (Zhongwei, Laibin, Chongzuo, Bayannur, Ordos) were also excluded.

3.2 Measurement

Other independent variables were also collected for prefecture-level cities. According to the theoretical consideration above, we constructed independent variables from the perspectives of socioeconomic development, demographic traits, and cultural inherence to reflect their impact on the distribution of three-generation lineal households.
The socioeconomic development level of a region was measured by four variables: the GDP per capita, the share of agricultural production, the urbanization rate (proportion of residents living in an urban area), and the average education years. Because of the high correlations among these four variables, which would cause serious collinearity issues in the geographically weighted regression (GWR) model, we used factor analysis to reduce the overall dimension. The first common factor can explain 79.74% of the total variance of the four variables, and thus we named the factor as the socioeconomic indicator. Another socioeconomic variable we used is the standardized housing price. We constructed the variable by dividing housing prices per square meters by the per capita disposable income of each area.
In order to reflect the influence of demographic characteristics, the proportion of the population aged 14 and below and the proportion aged 65 and above were included. The former variable indicates the child ratio of a region while the latter measures the old-age ratio of a region. The child ratio indicator can also reflect the long-term fertility of a region. In addition, we used the emigration rate and immigration rate to represent the influence of population flow.
Compared to other factors, culture is difficult to measure directly. We roughly assessed cultural differences across space by adding the share of ethnic minorities of each prefecture-level city in the model. This variable can help detect the differential cultural impact of different ethnic minorities on household structure. Table 1 displays the descriptive statistics of all the variables. The VIF values of independent variables are all smaller than 10, which meets the requirements of multicollinearity.
Table 1 Descriptive statistics of variables
Variable Observations Mean Std. dev. Min Max VIF
Dependent variable
Three-generation lineal household (%) 332 16.89 6.08 3.03 37.54
Independent variables
Socioeconomic indicator 332 0.00 1.00 -2.77 2.66 4.41
Standardized housing price (%) 332 20.98 9.43 1.90 97.54 1.29
Population aged 0-14 (%) 332 17.28 4.67 8.28 31.86 2.78
Population aged 65+ (%) 332 8.65 1.94 1.79 16.50 1.97
Immigration (%) 332 10.49 11.55 0.74 79.92 3.55
Emigration (%) 332 18.34 6.97 1.54 41.77 1.17
Minority (%) 332 15.25 25.79 0.01 97.19 1.77

3.3 Method

We first conducted an OLS regression to assess the global associations between influencing variables and the proportion of three-generation lineal households. To explore the spatial varying effects, the GWR model, which is an extension of the global OLS model, was applied. The model does not uniformly estimate parameters but takes into account the non-stationarity of space and makes local estimates (Brunsdon et al., 1996). It assumes that the relationship between explained variables and explanatory variables changes across space. The basic function of the GWR model is expressed as:
$Y_{i}=\beta_{0}\left(\mu_{i}, v_{i}\right)+\sum \beta_{k}\left(\mu_{i}, v_{i}\right) x_{i k}+\varepsilon_{i}, i=1,2, \ldots, n$
where Yi denotes the proportion of three-generation lineal household at region i; (ui,vi) are the coordinates of the centroid of region i, β0(ui, vi) is the local intercept, and βk(ui,vi) is the local coefficient for predictor k at region i. Stable values of βk(ui,vi) suggest a non-significant spatial variation for the effect of predictor k. εi is the random error, which is assumed to be independent among different regions.
In the GWR model we estimated the regression for each location independently by applying location-specific weighting schemes. The adaptive bandwidth of the bi-square kernel function was adopted for implementing a weighting matrix with calibrations based on the corrected Akaike Information Criterion (Wheeler and Tiefelsdorf, 2005; Wang and Chi, 2017). All the preliminary descriptive spatial analyses were conducted with ArcGIS 10.4, and the GWR model was implemented in GWR4.

4 Results

4.1 Spatial distribution of three-generation lineal households

As indicated by Table 1, the average proportion of three-generation lineal households is around 16.9%, but it varies from 3.0% to 37.5%, implying significant internal heterogeneity. We calculated the Moran’s I index for each variable as well as its Z-score (Table 2). All the variables have a high degree of spatial autocorrelation, suggesting the necessity of adopting spatial regression models.
Table 2 Global spatial autocorrelation analysis results
Variables Moran’s I Z-score P
Three-generation lineal household 0.420 12.431 0.000
Socioeconomic indicator 0.438 12.983 0.000
Standardized housing price 0.147 4.582 0.000
Population aged 0-14 0.633 18.718 0.000
Population aged 65+ 0.608 18.017 0.000
Immigration 0.331 9.955 0.000
Emigration 0.283 8.408 0.000
Minority 0.746 22.109 0.000
Figure 1a demonstrates the spatial distribution of three-generation lineal households, which shows a general pattern of higher values in central-south China and lower values in north China. To shed further light on the pattern and magnitude of spatial clusters, we conducted a hot point analysis using the Getis-Ord Gi* method (Figure 1b). More detailed and complicated patterns of spatial heterogeneity were revealed. On the one hand, hot spots with the highest proportions of three-generation lineal households were found not only in the southwest frontier and the central-southern provinces like Hunan and Jiangxi, but also in parts of northwest China. On the other hand, despite the generally low values in north China, the cold spots with the lowest levels were located mainly in three small regions. These scattered patterns of hot and cold spots further demonstrate the limitation of the global model and the necessity to explore the spatially heterogeneous factors in explaining household structure in different places across the vast territory of China.
Figure 1 Spatial distribution (a) and hotspot analysis (b) of three-generation lineal households

4.2 Global regression results

To examine the relationship between influencing factors and three-generation lineal households preliminarily, we first conducted a global OLS regression, the results of which are shown in Table 3. There is a significant negative relationship between the socioeconomic indicator and the regional proportion of three-generation households. Consistent with the family modernization theory, the higher the degree of modernization in an area, the less likely that it maintains a large proportion of three-generation households. The emigration rate is negatively associated with the proportion of three-generation households, because the floating population usually migrates alone or with small households. Nevertheless, the immigration rate does not have a significant global impact on local household structure. The share of ethnic minorities also has a strong impact on household structure, with a higher regional share of ethnic minorities corresponding to a lower proportion of three-generation lineal households. All the other variables are not significant.
Table 3 Estimation results of OLS regression model
Variable Coefficient Standard error
Socioeconomic indicator -2.813*** 0.593
Standardized housing price 0.018 0.034
Population aged 0-14 0.052 0.101
Population aged 65+ 0.145 0.205
Immigration -0.060 0.046
Emigration -0.123*** 0.044
Minority -0.051*** 0.015
Constant 18.013*** 3.007
Observations 332
Adj. R2 0.287
F 20.043***
AICc 2038.797

Note: *p<0.1; **p<0.05; ***p<0.01

The results of the global OLS regression are largely in line with the previous theoretical considerations (Goode, 1963; Tang, 2020). Socioeconomic development and population migration are the major factors that characterize regional household structure. However, the OLS regression ignores spatial heterogeneity. Even if a variable exhibits a non-significant global effect, it is still possible that in some areas the variable has a large positive effect while in the other areas the effect becomes negative. The offset of positive and negative effects can result in a non-significant global coefficient for the variable. Therefore, it was necessary to implement the GWR model.

4.3 Geographically weighted regression results

Table 4 presents the GWR results, where the adjusted R2 jumps from 0.287 in the OLS regression to 0.547. A big increase in the adjusted R2 value suggests that the consideration of spatial heterogeneity greatly improves the overall model fit. The last column of Table 4 reports the diff-of-criterion test for spatial stationarity (Nakay, 2016). Migration and ethnic minorities have spatially stationary effects, while other covariates are spatially nonstationary. The estimated coefficient maps facilitate a better presentation of the GWR results (Figure 2). Only significant coefficients are colored in each figure.
Table 4 Estimation results of GWR model
Minimum Lower quartile Median Upper quartile Maximum DIFF of
criterion
Socioeconomic indicator -6.476 -1.704 -0.241 1.081 2.448 nonstationary
Standardized housing price -0.386 -0.113 0.040 0.208 0.329 nonstationary
Population aged 0-14 -1.363 -0.650 -0.090 0.256 0.627 nonstationary
Population aged 65+ -1.349 -0.726 -0.473 -0.268 0.355 nonstationary
Immigration -0.563 -0.345 -0.253 -0.192 -0.014 stationary
Emigration -0.312 -0.169 -0.100 -0.058 0.135 stationary
Minority -1.046 -0.012 0.018 0.039 1.004 stationary
Intercept 5.509 20.105 32.019 37.577 47.423 nonstationary
Adjusted R2 0.547
AICc 1930.237
Figure 2 Estimated results of GWR model
As shown in Figure 2a, the coefficients between the socioeconomic index and the proportion of three-generation lineal households have a clear regional gradient, with strong negative values in west China and weak positive values in a small area of China’s east coast. The coefficient gradient is strongly correlated with the development gradient, where west China is the least developed region and China’s east coast is the most developed region. The result suggests that the seemingly global negative association between socioeconomic development and the proportion of three-generation lineal households as implied by the family modernization theory is only valid in the least developed areas of China. In these areas, a small level of increase in socioeconomic development can lead to a significant reduction in complex households. However, with economic and social development, the negative effect becomes insignificant and even reverses. This implies that there may be a certain development threshold beyond which the mechanism of how socioeconomic factors affect household structure changes.
Figure 2b presents the spatially varying relationship between standardized housing price and the proportion of three-generation lineal households. Although the global coefficient is non-significant, the effect of housing prices varies significantly between north and south China. The northern region demonstrates a strong positive association between housing price and the proportion of three-generation lineal households, indicating that higher housing prices prevent the dissolution of lineal households. In contrast, the association was reversed in the middle reaches of the Yangtze River, implying that higher housing prices lead to smaller households. A recent study shows that from 2006 to 2015, China’s commercial housing sales and housing prices presented a negative correlation with a decrease in the absolute values of the coefficient from the north to the south (Wu, 2018). In other words, housing sales in north China are more sensitive to price change. People in the north are more likely to stop buying new houses and maintain a lineal household when the housing prices increase. In the middle Yangtze River, however, housing prices are pushed up largely by rural emigrants who work in coastal mega-city regions. Since the majority of them cannot afford housing in the destination cities, they choose to rent houses there or buy houses in their home cities so that their children are able to access educational resources in the cities that are much better than those in their rural hometowns (Wang et al., 2020). People living in these houses are usually left-behind children and their grandparents. In this sense, a higher housing price in these cities often indicates a larger proportion of split households and correspondingly a lower share of three-generation lineal households. Therefore, we observed a strong positive relationship between housing price and three-generation lineal households in the north but a negative one in the middle Yangtze River.
Figure 2c demonstrates the coefficients of child ratio (the share of population aged 14 and under) across space. Significant local coefficients are found in west and northeast China, but their signs are opposite. These two areas also happen to be respectively the highest and lowest fertility regions in China, and the split has been prevalent for several decades. The GWR result perfectly fits the theoretical model in a sense that when the fertility level is high, a decrease in the fertility rate, corresponding to a decrease in the child ratio, would result in an increase in the proportion of three-generation lineal households; only when fertility is at a low level, a decrease in the child ratio would be associated with a decline in the proportion of three-generation lineal households (Jiang and O’Neill, 2007). The former situation occurs in west China, where the coefficients of the child ratio are negative, and the latter situation occurs in northeast China, where the coefficients of the child ratio are all positive.
Although a non-significant positive coefficient was identified in the global OLS model for old-age ratio, the Beijing-Tianjin-Hebei region and the Pearl River Delta demonstrate a significant negative relationship between old-age ratio and the proportion of three-generation lineal households, as shown in Figure 2d. These two areas are known to be among the most economically advanced regions of China with the lowest dependency ratio, where old people are likely to receive better social security benefits. A high pension standard ensures that the older population can live independently from their children, leading to the dissolution of three-generation lineal households. Therefore, an increase in the old-age ratio can possibly result in a decline in three-generation lineal households in these areas.
Figure 2e and 2f present the local coefficients of the immigration rate and emigration rate respectively. They both have consistent negative impact across space. The coefficients of immigration rate are significantly negative in north China, south China and part of southwest China. Northern Beijing-Tianjin-Hebei region and southern Pearl River Delta region are popular destination places of internal migration where the inflow of individual and couple labors decreases the proportion of local three-generation lineal households. The strong negative association between emigration rate and the proportion of three-generation lineal households is evident in west China. Compared with other outflow places, these areas have less favorable farming conditions and fewer technological investments in agriculture (Zhen et al., 2010). Therefore, adult couples usually need to migrate out together to earn a living for the whole family. The left-behind elderly would either stay in an empty-nest nuclear household or form a skipped generation household with their left-behind grandchildren. Thus, an increase in the emigration rate presents a stronger destructive impact on three-generation lineal households in these areas.
The significant coefficients of the proportion of ethnic minorities are illustrated in Figure 2g. Although the overall coefficient is stationary, we can still observe a few informative differential patterns. For example, Tibet and Xinjiang show a negative relationship between regional share of ethnic minorities and the proportion of three-generation lineal households, while parts of Guizhou and Guangxi exhibit a positive relationship between these two variables. These areas are all minority inhabited districts and the differentiation possibly reflects the cultural differences among these different minority groups. Both Tibetans, primarily residing in Tibet, and Uyghurs, mostly inhabiting Xinjiang, do not have a strong lineal household tradition due to their historical life mode of nomadism. In contrast, the southwest minorities in Guizhou and Guangxi maintain a more traditional agricultural life mode, and thus have a higher proportion of three-generation lineal households.
Figure 3 is a map showing the leading influencing factors for each prefecture-level city based on the absolute values of standardized coefficients and the significant level of 95%. Socioeconomic development is the dominant factor driving the change in three-generation lineal households in northwest China, whereas the child ratio plays a more important role in northeast and southwest China. Except for some areas on the east coast and in central China, the household structures of other majority places are characterized by the immigration rate. These results imply that socioeconomic development and population migration are the two dominant mechanisms shaping household structure in present China. Although economic development leads to a reduction in lineal households, such an effect is only limited to the less developed areas of China. When economic development passes a certain threshold, the impact of socioeconomic factors vanishes, and the effect of population inflow rate becomes more prominent.
Figure 3 Spatial distribution of local leading influencing factors

5 Discussion and conclusion

To explore the spatial variation of household structure in China as well as its determinant mechanisms, we used 2010 prefecture-level census data and conducted a GWR model. After several decades of economic development and social transformation, the patterns of Chinese household structure present a complex picture across space. In some places the variation in household structure can still be explained by the classic theory of family modernization, but in other places the varying patterns of three-generation lineal households imply the emergence of new regimes differing from the conventional modernization framework.
The family modernization theory was once considered a universal framework to explain the change of household structure. While early international studies found an inverse relationship between economic development and family complexity, which was consistent with the theory (De Vos, 1990; Ruggles, 1994; Aykan and Wolf, 2000), more recent works identified mixed findings of how economic development affects family structure especially among developing countries (Bongaarts and Zimmer, 2002; Ruggles and Heggeness, 2008). Although researchers realized that there was spatial heterogeneity regarding effects of development on family structure, seldom studies conducted spatial analyses either among countries or within a single country. Our study contributes to the literature by employing a formal spatial model and our findings deviate from the family modernization theory in two respects. First, the significant negative association between socioeconomic development and household complexity is only valid at a certain development level. An increase in socioeconomic development can effectively reduce the proportion of three-generation lineal households in the less developed western China, but such an impact disappears in its other more developed areas. Although we did not present a tempo relationship between socioeconomic development and household structure change, the spatial variation of their association highly corresponds to the development gradient, providing strong evidence of a threshold beyond which the mechanism alters. These results suggest that on the one hand, family modernization theory is still a useful framework in explaining household change; on the other hand, it can no longer be considered a universal theory.
Second, while family modernization theory emphasizes the importance of socioeconomic development, results demonstrate that other factors, such as demographic and cultural, can play equally important roles in shaping household structure. For example, population immigration rate becomes the dominant influencing factor for many areas of China with an advanced development level. This implies that Chinese household structure is characterized by multiple mechanisms, the importance of which changes with development stages and other social circumstances.
Besides the socioeconomic index, the effects of standardized housing price, old-age ratio, child ratio, immigration rate, and share of ethnic minorities on the proportion of three-generation lineal households also vary spatially in terms of direction and magnitude. The varying effect of housing price probably reflects the sensitivity of local housing markets to a change in housing prices. The distribution of associations between child ratio and the proportion of three-generation lineal household is largely in line with the fertility level gradient, and the place-specific impact of old-age ratio may be a manifestation of differentiated social security levels across space. A high immigration rate in the less developed western region is more likely to impose a significantly negative impact on the proportion of three-generation lineal households. The differentiated impact of the share of ethnic minorities is possibly a reflection of cultural differences among different minority groups.
These findings seem rather complex, but they all point to one conclusion: there is no single framework that can properly explain the household structure pattern in China. The spatial distribution of three-generation lineal households is determined by the joint mechanisms of socioeconomic development, demographic transition, cultural inherence, and their interactions.
Nevertheless, out of these complexities, we can still identify a few characteristics that are unique to Chinese society. One prominent phenomenon in China during the past several decades is the large-scale internal population migration, which greatly improves the efficiency of labor resource allocation. According to our results, internal migration is not only a critical factor that boosts economic growth in China, but also changes the structure of Chinese households. Each household adjusts its living arrangement in response to the migration of household members conditioned on local farming conditions and development level. The varying decisions reflect a pragmatic tendency for Chinese households, who are always pursuing maximum benefits for the whole family.
Another characteristic is the internal cultural variation of China. In this study, we identified varying household structure patterns among different areas inhabited by ethnic minorities, reflecting the cultural influence of ethnic minorities. It is possible that within the majority ethnic group Han, there is considerable cultural variation across space. However, due to the difficulty of measuring culture, we were unable to detect the more nuanced cultural differences as well as their impact on household structure.
In sum, both the classic family modernization theory and the culture and demographic perspective mainly come from the field of sociology and assume homogeneous spatial effect. Through demonstrating place-specific effects of varying factors on family structure, our study highlights the necessity and importance of incorporating geographic aspect in social theory development, and urges to adopting interdisciplinary approaches for better understanding social phenomena.
One limitation of the study is that the cross-sectional data are unable to provide temporal relationship between influencing factors and household structure, such that it is hard to make any causal inferences. In this study, we used the spatial gradient of varying factors to approximate their temporal gradient. For example, west China is treated as being in a less-developed stage, while east China is considered as being in a more advanced stage along the line of modernization process. Future study is encouraged to conduct deeper investigations into the culture variation within the majority ethnic group Han and use longitudinal data to better illustrate the causal relationships.

References

Publishing order | Descend order by publishing year | Descend order by cited within
[1]
Aykan H, Wolf D A, 2000. Traditionality, modernity, and household composition: Parent-child coresidence in contemporary Turkey. Research on Aging, 22(4): 395-421.

DOI

[2]
Bongaarts J, Zimmer Z, 2002. Living arrangements of older adults in the developing world: An analysis of demographic and health survey household surveys. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 57(3): S145-S157.

DOI

[3]
Brunsdon C, Fotheringham A S, Charlton M E, 1996. Geographically weighted regression: A method for exploring spatial nonstationarity. Geographical Analysis, 28(4): 281-298.

DOI

[4]
Curtis M A, 2011. The impact of housing subsidies and prices on mothers’ living arrangements: Evidence from the census. Housing Studies, 26(5): 747-765.

DOI

[5]
De Vos S, 1990. Extended family living among older people in six Latin American countries. Journal of Gerontology: Social Sciences, 45: 87-94.

PMID

[6]
Engels F, 2007. The Origin of the Family, Private Property, and the State. New York: New York University Press.

[7]
Ermisch J, 1999. Prices, parents, and young people’s household formation. Journal of Urban Economics, 45(1): 47-71.

DOI

[8]
Fei X, 1983. The problem of old-age care in the change of family structure: On the change of family structure in China. Journal of Peking University (Philosophy and Social Sciences), (3): 7-16. (in Chinese)

[9]
Goode W J, 1963. World Revolution and Family Patterns. Glencoe, IL: Free Press.

[10]
Hajnal J, 1982. Two kinds of preindustrial household formation system. Population and Development Review, 8(3): 449-494.

DOI

[11]
Hareven T K, 1976. Modernization and family history: Perspectives on social change. Signs: Journal of Women in Culture and Society, 2(1): 190-206.

DOI

[12]
Hareven T K, 1987. Reflections on family research in the People’s Republic of China. Social Research, 54(4): 663-689.

[13]
Hu Z, Peng X, 2014. Household changes in contemporary China: An analysis based on census data. Sociological Studies, (3): 145-166. (in Chinese)

[14]
Huang Z. 2011. Modern Chinese family: Perspective from economic and legal history. Open Times, (5): 82-105. (in Chinese)

[15]
Jacka T, Kipnis A B, Sargeson S, 2013. Contemporary China:Society and Social Change. Cambridge: Cambridge University Press.

[16]
Jiang L, O’Neill B C, 2007. Impacts of demographic trends on US household size and structure. Population and Development Review, 33(3): 567-591.

[17]
Li T, Fan W, Song J, 2020. The household structure transition in China: 1982-2015. Demography, 57(4): 1369-1391.

DOI PMID

[18]
Logan J R, Bian F, 1999. Family values and coresidence with married children in urban China. Social Forces, 77(4): 1253-1282.

DOI

[19]
Nakaya T, 2016. GWR4.09 user manual. Retrieved from

[20]
Nimkoff M F, 1961. Changing family relationships of older people in the United States during the last fifty years. Gerontologist, 1(2): 92-97.

DOI

[21]
Parsons T, 1949. The social structure of the family. In: Anshen R N (ed.). The Family:Its Function and Destiny. NY: Harper.

[22]
Ruggles S, 1994. The transformation of American family structure. American Historical Review, 99(1): 103-128.

DOI

[23]
Ruggles S, Heggeness M, 2008. Intergenerational coresidence in developing countries. Population and Development Review, 34(2): 253-281.

DOI PMID

[24]
Shi Q, Dorling D, 2020. Growing socio-spatial inequality in neo-liberal times? Comparing Beijing and London. Applied Geography, 115: 102139.

DOI

[25]
Tang S, 2020. Determinants of migration and household member arrangement among poor rural households in China: The case of North Jiangsu. Population, Space and Place, 26(1): e2279.

[26]
Thornton A, Fricke T E, 1987. Social change and the family: Comparative perspectives from the west, China, and South Asia. Sociological Forum, 2(4): 746-779.

DOI

[27]
Wang D, Chi G, 2017. Different places, different stories: A study of spatial heterogeneity of county-level fertility in China. Demographic Research, 37(16): 493-526.

DOI

[28]
Wang Y, 2003. A study on the changing family structure in rural north China: Cases in southern Hebei province. Social Sciences in China, (4): 93-108. (in Chinese)

[29]
Wang Y, 2008. Theoretic analysis on transformation and mobility of family structure: Based on historical and realistic experience of Chinese rural areas. Journal of Social Sciences, (7): 90-103. (in Chinese)

[30]
Wang Y, 2013. An analysis of the changes in China’s urban and rural family structures: Based on 2010 census data. Social Sciences in China, (12): 60-77. (in Chinese)

[31]
Wang Y, 2015. A comparative analysis of the family structure in China’s different regions: Based on the 2010 census data. Population & Economics, (1): 34-48. (in Chinese)

[32]
Wang Z, Guo M, Ming J, 2020. Effect of hometown housing investment on the homeownership of rural migrants in urban destinations: Evidence from China. Cities, 105: 102619.

DOI

[33]
Wheeler D, Tiefelsdorf M, 2005. Multicollinearity and correlation among local regression coefficients in geographically weighted regression. Journal of Geographical Systems, 7(2): 161-187.

DOI

[34]
Whyte M K, 1996. The Chinese family and economic development: Obstacle or engine? Economic Development and Cultural Change, 45(1): 1-30.

DOI

[35]
Wu Y, 2018. Research on spatial distribution characteristics of commercial housing prices in China and influencing factors on it[D]. Chongqing, China: Chongqing University. (in Chinese)

[36]
Yang Z, Cai J, Qi W et al., 2017. The influence of income, lifestyle, and green spaces on interregional migration: Policy implications for China. Population, Space and Place, 23(2): e1996.

[37]
Yi Z, 1986. Changes in family structure in China: A simulation study. Population and Development Review, 12(4): 675-703.

DOI

[38]
Zeng Y, Li W, Liang Z, 1992. Current situation, regional differences, and changing trend of family structure in China. Chinese Journal of Population Science, (2): 1-12. (in Chinese)

[39]
Zhen L, Cao S, Cheng S et al., 2010. Arable land requirements based on food consumption patterns: Case study in rural Guyuan District, Western China. Ecological Economics, 69(7): 1443-1453.

DOI

[40]
Zhong X, 2015. Refamilization: Intergenerational cooperation and conflicts in housing consumption among Chinese urban families. Journal of Public Administration, (1): 117-140.

Options
Outlines

/

〈 〉